Unnamed: 0
int64 0
116k
| id
stringlengths 36
47
| source
stringclasses 16
values | text
stringlengths 2
56.5k
| chunk_num
int64 0
933
| start_token
int64 0
5.6M
| end_token
int64 3
5.6M
| total_tokens
int64 3
5.6M
|
|---|---|---|---|---|---|---|---|
37,348
|
484e101f-83e0-4424-8b40-30bbc9a24ce8
|
trentmkelly/LessWrong-43k
|
Towards the Operationalization of Philosophy & Wisdom
Written for the competition on the Automation of Wisdom and Philosophy.
----------------------------------------
Summary
Philosophy and wisdom, and the processes underlying them, currently lack a proper operationalization: a set of robust formal or semi-formal definitions. If such definitions were found, they could be used as the foundation for a strong methodological framework. Such a framework would provide clear guidelines for how to engage in high-quality philosophical/wise reasoning and how to evaluate whether a given attempt at philosophy or wisdom was a success or a failure.
To address that, I provide candidate definitions for philosophy and wisdom, relate them to intuitive examples of philosophical and wise reasoning, and offer a tentative formalization of both concepts. The motivation for this is my belief that the lack of proper operationalization is the main obstacle to both (1) scaling up the work done in these domains (i. e., creating a bigger ecosystem that would naturally attract funding), and (2) automating them.
The discussion of philosophy focuses on the tentative formalization of a specific algorithm that I believe is central to philosophical thinking: the algorithm that allows humans to derive novel ontologies (conceptual schemes). Defined in a more fine-grained manner, the function of that algorithm is “deriving a set of assumptions using which a domain of reality could be decomposed into subdomains that could be studied separately”.
I point out the similarity of this definition to John Wentworth’s operationalization of natural abstractions, from which I build the formal model.
From this foundation, I discuss the discipline of philosophy more broadly. I point out instances where humans seem to employ the “algorithm of philosophical reasoning”, but which don’t fall under the standard definition of “philosophy”. In particular, I discuss the category of research tasks varyingly called “qualitative” or “non-paradigmatic” research, arguing tha
| 0
| 0
| 410
| 410
|
4,338
|
978d4c92-ea24-4dd1-9463-353c53fcda25
|
StampyAI/alignment-research-dataset/arxiv
|
pure strategy. Nevertheless, researchers have shown that a mixed strategy
is very helpful for systems that want to improve the effectiveness of their answers
by learning from their interactions with users [[22](#bib.bib22), [52](#bib.bib52)].
Let us assume that the system always return results s for query q
during its interaction with the user. Clearly, there may be
tuples not in s that are relevant to the intent behind q. But,
these tuples will never be shown to the user. Hence, the system will never
know the feedback of user on those tuples to adapt its strategy accordingly.
Hence, these systems deliver relatively less effective results in the
long run. To collect the user’s feedback
on more tuples, the system may randomly return different results
for the same query during its interaction with the user.
Of course, if the system aims at only showing more (distinct) tuples
to the user, it may deliver relatively ineffective answers in the short term.
A desirable DBMS strategy provides a trade-off between exploration, i.e.,
collecting feedback on more tuples, and exploitation, i.e, showing tuples that
DBMS deem more relevant to the intent behind the query.
Let r represent an effectiveness measure described in Section [3.2.3](#S3.SS2.SSS3 "3.2.3 Satisfaction Metrics ‣ 3.2 Intent, Query, and Result Paradigm ‣ 3 Signaling Game Model ‣ A Signaling Game Approach to Databases Querying and Interaction").
We denote the degree of effectiveness of result sℓ for intent ei as r(ei,sℓ).
We have 0≤ r(ei,sℓ)≤1 for all 1≤i≤m and 1≤ℓ≤o.
Because the goal of both the user and DBMS is to satisfy the user’s information need, an interaction yields a reward r(ei,sℓ) for both when the intent and returned result are ei and sℓ, respectively.
The communication between the user and DBMS can be modeled as a signaling game with identical interests played between them. In this game, the set of strategies of a user is the set Lmn of row-stochastic matrices and the set of strategies of the DBMS is the set Lno of n×o row-stochastic matrices. The payoff of user and DBMS in this case w.r.t. effectiveness measure r are
| | | | |
| --- | --- | --- | --- |
| | ur(U,D)=m∑i=1πin∑j=1Uijo∑ℓ=1Djℓ r(ei,sℓ). | | (1) |
The payoff function ([1](#S3.E1 "(1) ‣ 3.3 Strategies and Rewards ‣ 3 Signaling Game Model ‣ A Signaling Game Approach to Databases Querying and Interaction")) is an expected payoff of the interaction between user and DBMS when the user maps a (random) intent ei to a query qj with probability Uij and the database maps back the query qj to result sℓ with probability Djℓ. The larger the value of this payoff function is, the more likely it is that the DBMS returns the desired answers to more users’ queries.
4 Equilibrium Analysis
-----------------------
###
4.1 Fixed User Strategy
In some settings, the strategy of a user may change in a much slower time scale than that of the DBMS. In these cases, it is reasonable to assume that the user’s strategy is fixed. Hence, the game will reach a desirable state where the DBMS adapts a strategy that maximizes the expected payoff.
Let a strategy profile be a pair of user and DBMS strategies.
###### Definition 4.1
Given a strategy profile (U,D), D is a best response to U w.r.t.
effectiveness measure r if we have ur(U,D) ≥ur(U,D′) for
all the database strategies D′.
A DBMS strategy D is a strict best response to U if the inequality in Definition [4.1](#S4.Thmtheorem1 "Definition 4.1 ‣ 4.1 Fixed User Strategy ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction") becomes strict for all D′≠D.
###### Example 4.1
Consider the database instance about universities that
is shown in Table LABEL:sub@example:intro:table:instance.
Also, consider the queries, intents, and the strategy profile in Tables LABEL:sub@example:equilib:strictbest:intents, and LABEL:sub@example:equilib:strictbest:queries, and
LABEL:sub@example:equilib:strictbest:strategy, respectively.
The results s1, s2, and s3 in the DBMS strategy are the
outputs of applying the intents, e1, e2, and e3
in Table LABEL:sub@example:equilib:strictbest:intents to the
Univ relation in Table LABEL:sub@example:intro:table:instance, respectively.
Given a uniform prior over the intents,
the DBMS strategy is a best response
and a strict best response for the user strategy
w.r.t. precision.
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
| Name | Abbreviation | State | Type | Ranking |
| --- | --- | --- | --- | --- |
| Missouri State University | MSU | MO | Public | 64 |
| Mississippi State University | MSU | MS | Public | 41 |
| Murray State University | MSU | KY | Public | 57 |
| Michigan State University | MSU | MI | Public | 34 |
(a) A database instance of relation Univ
|
| Query# | Query |
| --- | --- |
| q1 | ans(w)←Univ(x,‘MSU',y,z,w) |
| q2 | ans(w)←Univ(x,y,‘KY',z,w) |
(b) Queries
|
| Intent# | Intent |
| --- | --- |
| e1 | ans(w)←Univ(x,‘MSU',‘MO',y,w) |
| e2 | ans(w)←Univ(x,‘MSU',y,‘Public',w) |
| e3 | ans(w)←Univ(x,‘MSU',‘KY',z,w) |
(c) Intents
|
q1
q2
e1
1
0
e2
1
0
e3
0
1
s1
s2
s3
q1
1
0
0
q2
0
0
1
(d) The user (left) and DBMS (right) strategies
|
Table 1: Queries, intents, and strategies for Examples [4.1](#S4.Thmexample1 "Example 4.1 ‣ 4.1 Fixed User Strategy ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction")
###### Definition 4.2
Given a strategy profile (U,D), an intent ei, and a query qj, the payoff of ei using qj is
| | | |
| --- | --- | --- |
| | ur(ei,qj)=o∑ℓ=1Dj,ℓr(ei,sℓ). | |
###### Definition 4.3
The pool of intents for query qj in user strategy U is the set of intents ei such that Ui,j>0.
We denote the pool of intents of qj as PL(qj). Our definition of pool of intent resembles the notion of pool of state in signaling games [[11](#bib.bib11), [13](#bib.bib13)]. Each result sℓ such that Dj,ℓ>0 may be returned in response to query qj. We call the set of these results the reply to query qj.
###### Definition 4.4
A best reply to query qj w.r.t. effectiveness measure r is a reply that maximizes ∑ei∈PL(qj)πiUi,j ur(ei,qj).
The following characterizes the best response to a strategy.
###### Lemma 4.5
Given a strategy profile (U,D), D is a best response to U w.r.t. effectiveness measure r if and only if D maps every query to one of its best replies.
{proof}
The proof is in the appendix.
The following corollary directly results from
Lemma [4.5](#S4.Thmtheorem5 "Lemma 4.5 ‣ 4.1 Fixed User Strategy ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction").
###### Corollary 4.6
Given a strategy profile (U,D), D is a strict best response to U w.r.t. effectiveness measure r if and only if every query has one and only one best reply and D maps each query to its best reply.
Given an intent e over database instance I,
some effectiveness measures, such as precision and recall, take their maximum for other results in addition to e(I). For example, given intent e, the precision of every non-empty result s⊂e(I) is equal to the precision of e(I) for e. Hence, there are more than one best reply for an intent w.r.t. precision or recall. Thus, according to Corollary [4.6](#S4.Thmtheorem6 "Corollary 4.6 ‣ 4.1 Fixed User Strategy ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction"), there is not any strict best response w.r.t. precision or recall.
On the other hand, F-measure takes its maximum only for the result e(I).
Hence, one may find a strict best response w.r.t. F-measure.
###
4.2 Nash Equilibrium
In this section and Section [4.3](#S4.SS3 "4.3 Strict Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction"), we analyze the equilibria of the game where both user and DBMS may modify their strategies. A Nash equilibrium for a game is a strategy profile where the DBMS and user will not do better by unilaterally deviating from their strategies.
###### Definition 4.7
A strategy profile (U,D) is a Nash equilibrium w.r.t. a satisfaction function r if ur(U,D) ≥ur(U′,D) for all user strategy U′ and ur(U,D)≥ ur(U,D′) for all database strategy D′.
###### Example 4.2
Consider again the intents, queries, and database instance and strategies in Example [4.1](#S4.Thmexample1 "Example 4.1 ‣ 4.1 Fixed User Strategy ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction").
The strategy profile is a Nash Equilibrium w.r.t precision.
User and DBMS cannot unilaterally change
their strategies and receive a better payoff.
###### Example 4.3
Consider again the database instance about universities that
is shown in Table LABEL:sub@example:intro:table:instance.
Also, consider the queries, intents, and the strategy profile shown in Tables LABEL:sub@example:equilib:nash:queries,
and LABEL:sub@example:equilib:nash:intents, and
the strategy profile in Table LABEL:sub@example:equilib:nash:moreDesirableStrat,
respectively.
The results s1, s2, and s3 in the DBMS strategy are the
outputs of applying the intents, e1, e2, and e3
in Table LABEL:sub@example:equilib:nash:intents to the
Univ relation in Table LABEL:sub@example:intro:table:instance, respectively.
Let 0≤ϵ≤1.
Given a uniform prior over the intents, the strategy profile in
Table LABEL:sub@example:equilib:nash:moreDesirableStrat
is a Nash equilibrium w.r.t precision.
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
| Intent# | Intent |
| --- | --- |
| e1 | ans(z)←Univ(x,‘MSU',‘MS',y,z) |
| e2 | ans(z)←Univ(x,‘MSU',‘MI',y,z) |
| e3 | ans(z)←Univ(x,‘MSU',‘MO',y,z) |
(a) Intents
|
| Query# | Query |
| --- | --- |
| q1 | ans(z)←Univ(x,‘MSU',‘MI',y,z) |
| q2 | ans(z)←Univ(x,‘MSU,y,w,z) |
(b) Queries
|
| | q1 | q2 |
| --- | --- | --- |
| e1 | 0 | 1 |
| e2 | 1 | 0 |
| e3 | 0 | 1 |
| | s1 | s2 | s3 |
| --- | --- | --- | --- |
| q1 | 0 | 1 | 0 |
| q2 | ϵ | 0 | 1−ϵ |
(c) User (left) and DBMS (right) strategies for strategy profile 1
|
Table 2: Intents, queries, and strategies for Example [4.3](#S4.Thmexample3 "Example 4.3 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction")
If the interaction between user and DBMS reaches a Nash equilibrium, they user
do not have a strong incentive to change her strategy.
As a result the strategy of the DBMS and the expected payoff of the game
will likely to remain unchanged.
Hence, in a Nash equilibrium the strategies of user and DBMS are likely
to be stable. Also, the payoff at a Nash equilibrium reflects a potential eventual payoff for the user and DBMS in their interaction.
Query qj is a best query for intent ei if qj∈ argmaxqkur(ei,qk).
The following lemma characterizes the Nash equilibrium of the game.
###### Lemma 4.8
A strategy profile (U,D) is a Nash equilibrium w.r.t. effectiveness measure r if and only if
* for every query q, q is a best query for every intent e∈PL(q), and
* D is a best response to U.
{proof}
The proof is in the appendix.
Let Nashr denote the set of Nash equilibrium strategy profiles w.r.t. effectiveness measure r. The following corollary is an immediate result of Lemma [4.8](#S4.Thmtheorem8 "Lemma 4.8 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction").
###### Corollary 4.9
NashF−measure= Nashprecision∩
Nashrecall.
{proof}
The proof is in the appendix.
Corollary [4.9](#S4.Thmtheorem9 "Corollary 4.9 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction") shows that generally the DBMS has more alternatives to reach a Nash equilibrium using precision or recall than using F-measure.
###
4.3 Strict Nash Equilibrium
A strict Nash equilibrium is a strategy profile in which the DBMS and user will do worse by unilaterally changing their equilibrium strategy.
###### Definition 4.10
A strategy profile (U,D) is a strict Nash equilibrium w.r.t. effectiveness measure r if we have ur(U,D)> ur(U,D′) for all DBMS strategies D′≠D and ur(U,D)> ur(U′,D) for all user strategies U′≠U.
###### Example 4.4
Consider the intents, queries, and the strategy profile in Example [4.1](#S4.Thmexample1 "Example 4.1 ‣ 4.1 Fixed User Strategy ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction") and the database illustrated in Table LABEL:sub@example:intro:table:instance. The strategy profile is a strict Nash equilibrium w.r.t precision.
However, the strategy profile in Example [4.3](#S4.Thmexample3 "Example 4.3 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction")
is not a strict Nash equilibrium as One may modify the value of
ϵ without changing the payoff of the players.
Next, we investigate the characteristics of strategies in a strict Nash equilibria profile. Recall that a strategy is pure iff it has only 1 or 0 values.
A user strategy is onto if there is not any query qj such that Ui,j=0
for all intend i. A DBMS strategy is one-to-one if it does not map two queries to the same result. In other words, there is not any result sell such that Djℓ>0 and Dj′ℓ>0 where j≠j′.
###### Theorem 4.11
If (U,D) is a strict Nash equilibrium w.r.t. satisfaction function r, we have
* U is pure and onto.
* D is pure and one-to-one.
{proof}
Let us assume that there is an intent ei and a query qj such that 0<Ui,j<1. Since U is row stochastic, there is a query qj′ where 0<Ui,j′ <1. Let ur(Ui,j,D) =∑oℓ=1Dj,ℓr(ei,sℓ). If ur(Ui,j,D) =ur(Ui,j′,D), we can create a new user strategy U′ where U′i,j=1 and U′i,j′=0 and the values of other entries in U′ is the same as U. Note that the payoff of (U,D) and (U′,D) are equal and hence, (U,D) is not a strict Nash equilibrium.
If ur(Ui,j,D)≠ur(Ui,j′,D), without loss of generality one can assume that ur(Ui,j,D) >ur(Ui,j′,D). We construct a new user strategy U′′ whose values for all entries except (i,j) and (i,j′) are equal to U and U′′i,j=1, U′′i,j′=0. Because ur(U,D)< ur(U′′,D), (U,D) is not a strict Nash equilibrium. Hence, U must be a pure strategy. Similarly, it can be shown that D should be a pure strategy.
If U is not onto, there is a query qj that is not mapped to any intent in U. Hence, one may change the value in row j of D without changing the payoff of (U,D).
Assume that D is not one-to-one. Hence, there are queries qi and qj and a result sℓ such that Di,ℓ= Dj,ℓ=1. Because (U,D) is a strict Nash, U is pure and we have either Ui,ℓ=1 or Uj,ℓ=1. Assume that Ui,ℓ=1. We can construct strategy U′ that have the same values as U for all entries except for (i,ℓ) and (j,ℓ) and U′i,ℓ=0, U′j,ℓ=1. Since the payoffs of (U,D) and (U′,D) are equal, (U,D) is not a strict Nash equilibrium.
Theorem [4.11](#S4.Thmtheorem11 "Theorem 4.11 ‣ 4.3 Strict Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction") extends the Theorem 1 in
[[13](#bib.bib13)] for our model.
In some settings, the user may knows and use fewer queries than intents, i.e., m>n.
Because the DBMS strategy in a strict Nash equilibrium is one-to-one,
the DBMS strategy does not map some of the results to any query. Hence, the DBMS will never return some results in a strict Nash equilibrium no matter
what query is submitted. Interestingly, as Example [4.1](#S4.Thmexample1 "Example 4.1 ‣ 4.1 Fixed User Strategy ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction") suggests some of these results may be the results that perfectly satisfy some user’s intents. That is, given intent ei over database instance I, the DBMS may never return ei(I) in a strict Nash equilibrium.
Using a proof similar to the
one of Lemma [4.8](#S4.Thmtheorem8 "Lemma 4.8 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction"), we
have the following properties of strict Nash equilibria of a game.
A strategy profile (U,D) is a strict Nash equilibrium w.r.t. effectiveness measure r if and only if:
* Every intent e has a unique best query and the user strategy maps e to its best query, i.e., e∈PL(qi).
* D is the strict best response to U.
Let SNashr denote the set of strict Nash equilibrium strategy profiles w.r.t. a satisfaction function r.
The following corollary
shows that the results of Corollary [4.9](#S4.Thmtheorem9 "Corollary 4.9 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction")
extends for strict Nash equilibria.
It is proved similar to Corollary [4.9](#S4.Thmtheorem9 "Corollary 4.9 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction")
using the above properties of strict Nash equilibria.
###### Corollary 4.12
SNashF−measure=SNashprecision∩ SNashrecall.
###
4.4 Number of Equilibria
A natural question is how many (strict) Nash equilibria exist in a game. Theorem [4.11](#S4.Thmtheorem11 "Theorem 4.11 ‣ 4.3 Strict Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction") guarantees that both user and DBMS strategies in a strict Nash
equilibrium are pure. Thus, given that the sets of intents and queries are finite,
there are finitely many strict Nash
equilibria in the game. We note that each set of results
is always finite.
However, we will show that
if the sets of intents and queries in a game are finite,
the game has infinite Nash equilibria.
###### Lemma 4.13
If a game has a non-strict Nash equilibrium. Then there is an infinitely many Nash equilibria.
{proof}
The result follows from the fact that the payoff function ([1](#S3.E1 "(1) ‣ 3.3 Strategies and Rewards ‣ 3 Signaling Game Model ‣ A Signaling Game Approach to Databases Querying and Interaction")) is a bilinear form of U and D, i.e. it is a linear of D when U is fixed and a linear function of U, when D is fixed. If for D≠D′, (U,D) and (U,D′) are Nash-equilibria, then ur(U,D)=ur(U,D′). Therefore ur(U,αD+(1−α)D′)=ur(U,D) for any α∈R. In particular, for α∈[0,1], if D,D′ are stochastic matrices, αD+(1−α)D′ will be a stochastic matrix and hence, (U,αD+(1−α)D′) is a Nash equilibrium as well. Similarly, if (U′,D) and (U,D) are Nash equilibria for U≠U′, then ur(αU+(1−α)U′,D)=ur(U,D) and (αU+(1−α)U′,D) is a Nash-equilibrium for any α∈[0,1].
###### Theorem 4.14
Given a game with finitely many intents and queries,
if the game has a non-strict Nash equilibrium,
it has an infinite number of Nash equilibria.
{proof}
Every finite game has always a mixed Nash equilibrium [[45](#bib.bib45)].
According to Theorem [4.11](#S4.Thmtheorem11 "Theorem 4.11 ‣ 4.3 Strict Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction"),
a mixed Nash is not a strict Nash equilibrium.
Therefore, using Lemma [4.13](#S4.Thmtheorem13 "Lemma 4.13 ‣ 4.4 Number of Equilibria ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction"), the game will
have infinitely many Nash equilibria.
###
4.5 Efficiency
In this section we discuss the efficiency of different equilibria. We refer to the value of the
utility (payoff) in formula ([1](#S3.E1 "(1) ‣ 3.3 Strategies and Rewards ‣ 3 Signaling Game Model ‣ A Signaling Game Approach to Databases Querying and Interaction")) at a strategy profile to the efficiency of the strategy.
Therefore, the most efficient strategy profile is naturally the one that maximizes ([1](#S3.E1 "(1) ‣ 3.3 Strategies and Rewards ‣ 3 Signaling Game Model ‣ A Signaling Game Approach to Databases Querying and Interaction")).
We refer to an equilibria with maximum efficiency as an efficient equilibrium.
Thus far we have discussed two types of equilibria, Nash and strict Nash, that once reached it is unlikely that either player will deviate from its current strategy. In some cases it may be possible to enter a state of equilibria where neither player has any incentive to deviate, but that equilibria may not be an efficient equilibrium.
The strategy profile in Table LABEL:sub@example:equilib:nash:moreDesirableStrat provides the highest payoff for the user and DBMS given the intents and queries in Tables LABEL:sub@example:equilib:nash:intents and LABEL:sub@example:equilib:nash:queries
over the database in Table LABEL:sub@example:intro:table:instance.
However, some Nash equilibria may not provide high payoffs. For instance, Table [3](#S4.T3 "Table 3 ‣ 4.5 Efficiency ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction") depicts another strategy profile for the set of intents and queries in Tables LABEL:sub@example:equilib:nash:intents and LABEL:sub@example:equilib:nash:queries
over the database in Table LABEL:sub@example:intro:table:instance.
In this strategy profile, the user has little knowledge about the database content and expresses all of her intents using a single query q2, which asks for the ranking of universities whose abbreviations are MSU. Given query q2, the DBMS always returns the ranking of Michigan State University.
Obviously, the DBMS always returns the non-relevant answers for the intents of finding the rankings of Mississippi State University and Missouri State University. If all intents have equal prior probabilities, this strategy profile is a Nash equilibrium.
For example, the user will not get a higher payoff by increasing their knowledge about the database and using query q1 to express intent e2.
Clearly, the payoff of this strategy profile is less than the strategy profile in Table LABEL:sub@example:equilib:nash:moreDesirableStrat. Nevertheless, the user and the DBMS do not have any incentive to leave this undesirable stable state once reached and will likely stay in this state.
###### Definition 4.15
A strategy profile (U,D) is optimal w.r.t. an effectiveness measure r if we have ur(U,D)≥u(U′,D′) for all DBMS strategies D′ and U′
Since, the games discussed in this paper are games
of identical interest, i.e. the payoff of the user and the DBMS are the same, therefore, an optimal strategy (U,D) (w.r.t. an effectiveness measure r) is a Nash equilibrium.
###### Lemma 4.16
A strategy (U,D) is optimal if and only if it is an efficient equilibrium.
{proof}
Note that if (U,D) is optimal, then none of the two players (i.e. the user and the DBMS) has a unilateral incentive to deviate. Therefore (U,D) is a Nash equilibrium. On the other hand, since the payoff function ([1](#S3.E1 "(1) ‣ 3.3 Strategies and Rewards ‣ 3 Signaling Game Model ‣ A Signaling Game Approach to Databases Querying and Interaction")) is a continuous function of U and D and the domain of row-stochastic matrices is a compact space, therefore a maximizer (U,D) of ([1](#S3.E1 "(1) ‣ 3.3 Strategies and Rewards ‣ 3 Signaling Game Model ‣ A Signaling Game Approach to Databases Querying and Interaction")) exists and by the previous part it is a Nash equilibrium. Note that the efficiency of all strategies are bounded by the efficiency of an optimal strategy and hence, any efficient equilibrium is optimal.
Similar to the analysis on efficiency of a Nash equilibria, there are strict Nash equilibria that are less efficient than others. Strict Nash equilibria strategy profiles are unlikely to deviate from the current strategy profile, since any unilateral deviation will result in a lower payoff. From this we can say that strict Nash equilibria are also more stable than Nash equilibria since unilateral deviation will always have a lower payoff.
| | q1 | q2 |
| --- | --- | --- |
| e1 | 0 | 1 |
| e2 | 0 | 1 |
| e3 | 0 | 1 |
| | s1 | s2 | s3 |
| --- | --- | --- | --- |
| q1 | 0 | 1 | 0 |
| q2 | 0 | 1 | 0 |
Table 3: A Nash strategy profile over the intents, queries, and results in Table [2](#S4.T2 "Table 2 ‣ Example 4.3 ‣ 4.2 Nash Equilibrium ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to Databases Querying and Interaction").
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
| | q1 | q2 |
| --- | --- | --- |
| e1 | 0 | 1 |
| e2 | 1 | 0 |
| e3 | 1 | 0 |
| | s1 | s2 | s3 |
| --- | --- | --- | --- |
| q1 | 0 | 0 | 1 |
| q2 | 1 | 0 | 0 |
(a) User (left) and DBMS (right) strategies for strategy profile 1
|
| | q1 | q2 |
| --- | --- | --- |
| e1 | 0 | 1 |
| e2 | 0 | 1 |
| e3 | 1 | 0 |
| | s1 | s2 | s3 |
| --- | --- | --- | --- |
| q1 | 0 | 0 | 1 |
| q2 | 0 | 1 | 0 |
(b) User (left) and DBMS (right) strategies for strategy profile 2
|
Table 4: Some strict Nash strategy profiles.
As an example of a strict Nash equilibrium that is not efficient, consider both strategy profiles illustrated in Table [4](#S4.T4 "Table 4 ‣ 4.5 Efficiency ‣ 4 Equilibrium Analysis ‣ A Signaling Game Approach to
| 1
| 6,000
| 14,000
| 15,934
|
42,513
|
eb3b7151-f825-4f38-aef4-d1002d5b5cb7
|
StampyAI/alignment-research-dataset/blogs
|
RQQh93XnnXdP3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| 572
| 3,432,000
| 3,440,000
| 4,433,850
|
101,123
|
<urn:uuid:32cdcfc3-34a0-4ce9-9581-986919768116>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
• Content count
• Joined
• Last visited
• Days Won
About Fedic
• Rank
Public / Shared Information
• Location
Shit in one hand, wish in the other. See which fills up faster. ~Sai King
• Interests
satire, writing, antagonizing stupid people
• Favorite Anime
• Favorite Book
The Dark Tower Series
1. As I expected from you MJ, very nice
2. Seriously now, what is Lee's sudden problem? I feel terrible for Laura. I love the interactions between her and the old man. It's going to be so sad to see her go. Hooray for naughty Gaius giving the cylon the gun. I'm so glad Cain is dead, she was a psycho bitch that needed to be shot.
3. I was chatting with a fellow DT junkie about some different shows the other night and both Lost and BSG came up. We are both sick of the plot on Lost crawling uphill at the speed of molases in January . At the rate their going we'll never see a monster or a rescue because the show will be canceled before it gets revealed. It's almost not even worth watching, it's easy enough to look up the episode the next day. We talked about the whole 'wagon training' thing, personally I think they're already doing that. When the series premired I was thinking it was only going to be one or two seasons (that was the original plan wasn't it? ) You can take 2 seasons and make them 4 or 5 (or more ) by throwing in a lot of filler ie all the backstory filling up the eps. Lost would be closer to done without so much character back story, but god forbid you kill a cash cow. At least until the rateings slip The conversation took a turn for the better when we started talking BSG. A dieing leader will take them to earth. Laura is running out of time fast. It went from her having months to live to just weeks. So either they're going to have to find earth fast or she is going to have to find a way to prolong her life right? I can see her passing the torch off to the old man, but I can't see Guias becoming president. Dealing with Pegasus I wouldnt be suprised if Adama declared martial law putting himself in charge, I don't think Laura cares much for Admiral Cain either. The Pegasus is fucked up. They could definatly do several episodes surrounding an encounter between the two battlestars. It's possible that some of the civillan ships, like the Astral Queen, would side with Pegasus. Adama is not going to leave Helo and the Chief to be executed. It's more then that they are his men, he needs these people alive for the survival of the fleet. Laura will agree. They need their pilots and they need the chief bad. Guias and the cylon are still on Pegasus too, I think he will want to get back to Galactica and the fleet, probably with the cylon. The show is able to take a day in the life and turn it into an episode, sometimes two. They could finish the second season with Laura still alive, but I doubt she could live through season 3 without some new treatment.
4. I haven't updated in ages. No time to continue the story. http://www.fanfiction.net/u/612394/ It's about time you shared Deep Blue with the rest of the class MJ
5. Only you would go there MJ
6. This will definatly be sweet. Under the mountain with the slow mutants should be particularly creepy. All of that as a cg movie would be nice.
7. So Dx is a villan?
8. I finally caught this show and you know what? You guys are right. Bloo is an asshole.
9. Naughty naughty I wonder if the past will come back and bite him in the ass.
10. I'm not so sure that I belive only Teyla and Ronin have been getting junked up on Ford's wraith drugs. Ford claimes it's just them, but I don't trust that guy. Rodney flipping out about being a drug addict...
11. It just seems fishy to me. If this ship has been repeatedly batteling the cylons then why arent they just as screwed over as Battlestar? Cylon viruses and boarding parties etc. It's just one ship right? How in the hell have they managed to survive this long all alone?
12. I've got to save some money!
13. Does this mean I can talk you into making me a Apollo/Starbuck siggy?
14. Could be a dialect thing too. R's are pretty dominate in espanol
| 0
| 0
| 1,042
| 1,042
|
90,256
|
<urn:uuid:107ebb1d-114d-433e-b255-1a01fd7df2bb>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
[CLS]Thrivous Nootropics for Cognitive Enhancement
Top Five Mormon Science and Technology Events from 2008
In celebration of the new year, here's a look at the top five Mormon science and technology events from 2008. Happy new year!
Claus in You
I am Santa Claus. Yes. Whether you believe it or not, and whether you like it or not, I am. Some snort apathetically or with mild amusement. Some insist I'm changing the definitions of words. Most, though, get it. We understand: Santa Claus does exist and always has. Maybe we haven't yet figured out how to make the traditional reindeer fly, and certainly we're still missing presents for too many good girls and boys. But we're committed to doing better. Those reindeer will fly. Oh, they'll fly. So, Scrooge, while you wallow in your nihilism, I'll be working with some friends to build whatever needs finishing at the North Pole. You're welcome to join us when the angst passes... when you, too, learn the great secret: Claus in you.
Millennial Vegetarians or Improved Meat Producing Technology?
Mormon prophetic tradition includes the idea that humans (and perhaps other animals) will no longer eat other animals when the Earth attains its millennial or terrestrial glory. For example, see this recent post at the Mormon blog, Times and Seasons. Although some have interpreted these prophecies to support vegetarianism, others may be happy to learn that the techology to produce meat that has never been part of a sentient animal is already available. Would widespread use of such technology satisfy the prophecy?
Reproductive Technology and Gay Marriage
I recently came across a document by Margaret A Somerville, which presents "The Case Against 'Same-Sex Marriage'". This is, to date, the best argument I've read from opponents of gay marriage. It accounts for the importance of religious perspectives, but does not argue from them, and instead appeals to secular ethics. Somerville's summary of her argument follows:
Convergence 08 Sunday Unconference
Like yesterday afternoon, I participated in unconference sessions this afternoon. First, I attended a session with PJ Manney on empathy and technology. The session began by focusing primarily on how to promote empathy through video games, by encouraging persons to take on roles that require action other than violence. Then we discussed differences between how film and books promote empathy; the lack of differing details (internal dialog for film or visual stimuli for film) results in differing pathways to empathy. Unfortunately, I wasn't able to make any comments, but had hoped to discuss the importance of creating pathways for empathy. Many of us find it easy to have empathy for experiences that can be viewed, but find ideological empathy much more difficult. Because of that, it is important that those of us who understand competing ideologies work to formulate syncretizations which enable persons of the two sides to empathize with each other.
Convergence 08 Synthetic Biology Panel
This morning, the Convergence 08 conference continued with a panel of experts on synthetic life. One of the panel members observed that genetic engineering is almost as accessible as computer programming in the early 80s, when teenagers were able to become involved inexpensively. Another panel member responded skeptically that there are serious risks associated with synthetic life, particularly when introduced to natural environments, and more evidence should be gathered in favor of benefits before proceeding further. On the subject of benefits, other panelists agreed there are risks, but that risk management techniques will come with time. The most immediate benefit of synthetic life will probably be biofuels. Benefits for cardiovascular health, alzheimers and diabetes may arise from products entering human trials soon. The panelists debated the degree of risk associated with use of artificially selected insects, and emphasized the importance of rigorous research and precaution.
Convergence 08 Paul Saffo Key Note
Paul Saffo was the key note speaker on Saturday, and he endeavored to share some principles of forecasting. He began by distinguishing between futurists and forecasters, defining the former as active advocates and the latter as passive observors. He observed that persons looking to the future have a tendency to compress all the exciting things together, but history illustrates that times tend to produce long stretches of dullness. He encouraged questioning of all assumptions. For example, is tech actually converging, or rather is it diverging and producing greater complexity and diversity? Things may turn out other than we think.
Convergence 08 Saturday Unconference
This afternoon, I've been bouncing around among unconference sessions. First, I attended a session on balancing spirituality with technology. It had a lot of potential and several interesting persons attended, but the discussion was turned too often to the discussion leader's marketing of a device intended to stimulate meditative states. One interesting matter I'll note was one person's suggestion that we need not attempt to persuade each other to various spiritual perspectives. I disagrred with him, and expained that our individual spiritual perspectives have far reaching effects in our community and environment. Many of the challenges faced in the world today have arisen from lack of attention to the practical consequence of spiritual and religious world views.
Convergence 08 Artificial Intelligence Panel
The first session of the Convergence 08 conference is focusing on artificial intelligence. While waiting for the session to begin, I had a conversation with Peter Milford of Parallel Rules. He told me that his interests are in practical near-term applications of the ideas on which the conference is focusing. When he learned that my interests are in the intersection of technology and spirituality, he kindly expressed his disinterest -- and probably assumed I'm nutty. In time, perhaps he'll begin to recognize the practical near-term consequences of the intersection between tech and spirituality. To the extent that he and others do not recognize the practical importance of these matters, we're in for far more division and turmoil than necessary. Cool gadgets will not suffice to fill the spiritual heart of humanity.
Messianic Postures toward Artificial Intelligence
Observing trends in information technology, some researchers conclude that artificial intelligence (AI) will eventually surpass the brightest human minds and take control of its own evolution. Assuming these researchers are correct, it is in our interest to ensure that we design AI to be friendly from the beginning.
Congratulations Again to Barack Obama
Congratulations again to Barack Obama, who will be the next President of the United States of America. This is an historic moment, both for African Americans and all other Americans, in that it illustrates how far we have come as a people in overcoming the lingering influence of ancient racial biases. Not so long ago, no one could have reasonably imagined an African American winning a presidential election. However, our community has been nourishing respect for differences in and diversity of the human form. Times have changed for the better. May God continue to bless us with the wisdom and inspiration we need for the many challenges ahead.
Celestial Earth as a Kardashev Scale Civilization
Today, I came across a blog post at Mormon Matters, posing the question, "Is the Internet a Urim & Thummim?" This sounds strange to the point of near nonsense, I'm sure, to those of you who are not Mormon. But it does have meaning for Mormons, whose scriptures contain this prophecy from Joseph Smith:
You're a Transhuman Web Surfer! Now what?
Alan Leigh pointed out to me an interesting PC Magazine article, which reviews a book, "Web Surfers Evolving into Superior Humans", by neuroscientist Gary Small. In the book, Small argues that contemporary infotech is affecting human evolution by providing persons adept at using technologies such as internet searching and text messaging with social selection advantages. Such an observation is consistent with general Transhumanist observations, and illustrates how even marginal advances in our technology (let alone dramatic advances in biotech or nanotech) are contributing toward environmental factors that determine fitness of individuals within their context. So, now that you, dear reader, understand that you have an evolutionary advantage, are you going to use it egotistically or altruistically -- to benefit yourself exclusively or all of us? How do you spend your web surfing time?
The New God Argument and Heaven for the Non-Religious
I recently received a letter, written neatly by hand and sent through the post (yes, that still happens), from a kind friend with a question regarding the New God Argument. The question and answer are worth sharing here. So, with names removed to protect the innocent, here's the letter, with the question in bold:
A Mormon Perspective on California Proposition 8, Marriage and Sexuality
Californians will soon be voting on proposition 8, which would amend their state constitution to define "marriage" as being between one man and one woman, exclusively. The Church of Jesus Christ of Latter-day Saints, by far the largest Mormon denomination and the church of which I am a member, is perhaps the foremost proponent of the proposition. The Church has encouraged members to donate time and money to the cause. From what I understand, members in California have been calling around and going door to door asking others to vote for the amendment. As documented by members have also donated at least eight million dollars toward advertising in favor of the amendment; this represents nearly half of all money donated to the broader Protect Marriage campaign.
Uchtdorf Speaks of Emulating God's Creation and Compassion
Integrating Symbol Systems, Software, and Ideologies
Professionally, I'm a software engineering manager, and the area in which I've done the most work is systems integration, particularly between systems management platforms, which enable network administrators to monitor and control many computers without moving away from their desks. To integrate between systems management platforms, we write some software that knows how to talk with each of the platforms and enables them to exchange information. This is valuable because integrated platforms can thereafter leverage each other's strengths and thereby increase the overall value of owning both platforms. Without the integration software, the platforms must act independently, sometimes duplicating efforts, and often forcing administrators to waste their time doing unnecessary work.
When God Remembers Us
The New York Times reports that remembering may be like reliving, from the perspective of the neurons doing the remembering. Remembering neurons appear to behave like they did when first experiencing whatever is being remembered.
Implausible Old Gods and Impotent New Gods
Coincident with our recent presentation of the New God Argument, a leading evangelical Christian, Albert Mohler, wrote a blog entry entitled "New God or No God? The Peril of Making God Plausible". In the entry, he reviews a book by Mitchell Silver, who suggests that liberal theologians have been working to replace the "'old God' of biblical theism" with the "'new God' of the secular philosophers". Mohler, identifying with the old God theology, contrasts the two theological perspectives, explicitly or implicitly, as follows:
Old God
1) Biblical
2) Personal
3) Supernatural
4) Theistic
5) Specific
6) Literal
7) Traditional
8) Powerful
New God
1) Secular Philosophy
2) Principle
3) Natural
4) Atheistic
5) Equivocal
6) Metaphoric
7) Popular
8) Impotent
The new God described in Mohler's blog entry is not the God proposed by the New God Argument. Moreover, the New God Argument does not contend to propose a new God. Rather, it contends to be a new argument for a particular kind of God that is not new at all, but rather quite ancient. The new God described in Mohler's blog entry is the God of some liberal protestants. Although I may agree with some aspects of their perspective, as I may agree with some aspects of Mohler's perspective, the liberal protestant perspective is not one I would wholly defend. That aside, I do want to comment on each of the contrasts Mohler identified:
1) If the old God is the God preached by traditional Christianity then it is not only a biblical God, but also a secular philosophical God. Christian theology has been repeatedly reconciled with the popular secular philosophy of its day, beginning with neo-Platonism, through Scholastic Aristotelianism, and on to the various contemporary reconciliations. Mohler's own Southern Baptist Convention is not immune to the interplay between religious and secular philosophy, as evidenced by their basic statement on God: "The eternal triune God reveals Himself to us as Father, Son and Holy Spirit, with distinct personal attributes, but without division of nature, essence, or being." While the Bible does teach of the unity of aspects of God, it does not explicitly teach that the aspects are without division in nature, essence or being. The Southern Baptist Convention can thank the Catholics, who in turn can thank Greek philosophy, for that extra-Biblical interpretation.
2) With Mohler, I see less value in an impersonal God. However, one need not look only to traditional or conservative Christianity to find faith in a personal God. For example, the God advocated in the New God Argument is a person -- or persons -- and not merely a principle. This God was perhaps once like you and me, and we may perhaps become like this God. As we get to know and become like this God, we may learn to love him. It doesn't get much more personal than that, but it is hardly a traditional or conservative perspective by contemporary measures.
3) I don't know anything about a supernatural God, and neither does anyone else except to the extent that they label as "supernatural" the God they experience and imagine naturally. The distinction between natural and supernatural seems to be largely nonsensical. If it can affect my experience, it's natural, so far as I'm concerned. I may not understand how it affects my experience, but that doesn't mean I should assume it is immaterial or causeless -- euphemisms for "beyond investigation". Such assumptions are lazy and dangerous, preventing us from growing in knowledge.
4) If rejection of supernaturalism and immaterialism is atheism then I and a whole lot of other Mormons are atheists, despite our professed faith in God. As presented in the New God Argument, a powerful benevolent creator can be quite natural and material. For those of us who are natural and material, and who express our worship through emulation, only such a God is worthy of worship.
5) I do not perceive the God presented by the Bible, as interpreted by the Southern Baptist Convention or any other religion including my own, to be devoid of ambiguities, paradoxes and some outright contradictions. Despite the powerful inspirational power we find in scriptural texts and our exegesis of them, there is not so much precision as Mohler appears to suggest. If God intended the scriptures to be so precise, perhaps he would have written them in mathematics.
6) The scriptures are rich in symbolism, particularly when describing God. How does one decide where the symbol ends and the literal begins? If Mohler's concern is that some liberal theologians are making God wholly symbolic, I share his concern to the extent that we do not reify our symbols.
7) Today's traditional God is yesterday's popular God, and has been through the ages. Whether traditional or popular, God must be more than we have yet learned, or what's the bother? I don't intend to disregard the value of respect for tradition. To the extent tradition is not oppressive, we can all benefit from respecting it, and seeking to integrate its strengths into our learning process.
8) I agree with Mohler that a God that is merely a symbol is far less powerful than such reified into a personal being. Moreover, although Mohler certainly would not agree, an immaterial or supernatural person is far less powerful than a material natural person. The former is, so far as I can tell, indistinguishable from a symbol. Additionally, a God who would limit the education of his creatures to texts written millennia ago seems less powerful than one that would never cease educating.
While Mohler considers it perilous to make God plausible, I consider it perilous to make God implausible. Is it any wonder that religious adherence suffers when we hold dogmatically to ideas that clash with our experience, and our best efforts at rationality and science? Should we be surprised that the spirit of God fails us when we choose to disavow the ongoing flow of knowledge with which God would inspire us? If we seek to embrace all truth, regardless of its source, we probably will get more of it. We need not pretend to a final understanding of God to aspire to an understanding that is sufficiently plausible to help us take the next steps toward a better understanding.
Cannon's Wager: Bet on Better Works
Will works save you? Some Christians enjoy debating whether good works are necessary for salvation, with responses ranging from the extremes of Calvinism (we can make absolutely no effort, spiritual or physical, toward salvation) to the extremes of Pelagianism (salvation depends almost entirely on our efforts, spiritual and physical).
Congratulations to Barack Obama
Congratulations, both to Barack Obama and to us collectively, for his nomination as the presidential candidate of the United States Democratic party. This is not an endorsement, either of the candidate or the party, but rather acknowledgement of the historical significance of his nomination, as a person of African ancestry. As implied in the Transhumanist Declaration and Mormon Transhumanist Affirmation, we value the diversity of individuals in our community and celebrate the increasing diversity that technology can enable, as we use it charitably with respect and concern for our plurality of wills, desires and laws.
Arresting Jokers in the Houston Airport
I'm sitting in the Houston airport, returning from a family vacation -- you know, relaxing from all that hard work on the New God Argument. Just a moment ago, a recorded voice stated: "We would like to remind you that any jokes or inappropriate remarks about security may result in your arrest." The persons that recorded this message should be focusing on mitigating greater risks to our future than persons vocalizing inappropriate remarks (whatever those may be). We have just experienced a revolution in information tech, and it looks as if we may be on the cusp of a revolution in biotech. Who knows? After that, we may be worrying about grey goo and evil AGI within a few decades. Yet here we are threatening to arrest persons that vibrate their vocal cords in patterns that we don't like. Does our collective fear of divergent thoughts constitute a crowded theatre? Let's not let it be so.
The New God Argument Begins
This is a transcript of the original presentation of the New God Argument by Chris Bradford, Joseph West, and Lincoln Cannon at the Sunstone Symposium in Salt Lake City on 9 August 2008.
Big Thanks to Dan Wotherspoon
On a personal note, I'll add that I've had the opportunity to interact with Dan on many occasions, and have been repeatedly impressed by his character. He has a big heart and a sharp mind -- an inspiring and inspired man. Thanks, Dan, and good luck with your new opportunity!
Are we still Gods in embryo?
I am sometimes disturbed by the observation that our religious tradition, Mormonism, is not unified in faith that we can and should become God. Too many of us have capitulated to the violent dogmatism of traditional Christianity, particularly in its evangelical forms. Too many of us only hear whisperings of the doctrine. Too many of us, even, while giving lip service to this core idea on which Mormonism originally relied, yet only talk the talk without walking the walk. Many of us can still quote the prophets, affirming that as man is, God once was, and as God is, man may become. But our faith so often appears to end in the verbal affirmation, or perhaps goes no further than imagining that the perpetuatuion of self-hypnosis toward greater compassion will, in itself, be sufficient to transform us into gods. Who is walking the walk? Who can see that our future depends not only on thoughts, but also on power and action? Where are the Gods in embryo? Who will rise together in the name of Christ as joint heirs in the glory of God? Who is Mormon enough to declare, without shame, that we have got to learn how to become gods, the same as all other gods before us?
At Sunstone today, one session asked: are we still gods in embryo? The presenter outlined the history of the doctrine of theosis in the LDS Church over the last couple centuries. The respondent, Mormon historian Michael Quinn, stated that if, in our shame and efforts to conform with mainstream Christianity, we give up our doctrine of theosis, we have traded our birthright for a mess of pottage. The audience applauded vigorously. I cheered, "Amen!" Yet I wonder, how many persons in the audience trust in theosis to the point of action? How many even have any idea how we might go about learning to create worlds? How many are content with appeals to magic? I wonder, even in the applause, how many of us are Mormon enough to make use of all the means with which God is now endowing us?
New God Argument (Version 1.0)
Abstract: If basic life forms are probable then we should trust that advanced civilizations are probable. If any advanced civilization probably has increased in destructive capacity faster than defensive capacity, and if any advanced civilization probably creates many worlds like those in its past, then we should trust that an advanced civilization more benevolent than us probably created our world. The alternative is that we probably will go extinct before becoming an advanced civilization.
Transhumanists Love Life Even When Acknowledging Problems
We received the following email from a site visitor today:
"Life is not a problem to be solved but a reality to be experienced." Soren Aabye Kierkegaard
The return address on the email did not work, so I'll respond here in my blog. While I agree heartily that life overall is not a problem and that it is a reality to be experienced, I imagine that the sender of the email doesn't think a Transhumanist could sincerely agree. Of course, I do see problems in life, both present and foreboding, as did Kierkegaard (illustrated by writing or speaking that sentence) and as does the sender of the email (illustrated in the act of emailing us). However, acknowledging problems in life and working to change them does not mean one must consider life itself a problem. So far as I am concerned, life (present, past and future) is not only worth experiencing, but also worth celebrating and experiencing ever more fully, which is a goal of both Mormonism and Transhumanism. So, please, keep observing the problems in life and sending us email that will help us solve them!
Science and Technology Can Inform Mormon Visions of the Future
Recently, a Mormon blogger, R Gary, claimed that "Science [is] not always a good bet" when considering the feasibility of futures anticipated by prophetic vision. However, science does enable us to predict how futures similar to those he described may be possible, particularly given long-standing trends in technology.
Lambs and lions laying down together, if you want to interpret that literally, may be achieved through genetic manipulation of lambs and lions. Transformation of the Earth into something more like our imaginations of paradise may be the result of nanotechnological environmental interventions. Transcendence of natural laws happens all the time; there was a day when humans could not see so far, fly so high, or swim so deep as we now can, and science predicted such capacities prior to our technological implementation of such capacities. Even death, now, has come within the scope of engineering problems, as we quickly learn the causes of aging and work to counteract them. From there, I trust we'll engage in the work of learning, as all other gods before us, the detailed processes involved in resurrection, so that we may join Jesus as saviors on Mount Zion and continue forward with the glorious work of redeeming the dead.
Science is not a bad bet for any of these things. To the contrary, it is the best bet currently available to us for seeking after and understanding the details necessary for working actively toward fulfillment of futures foreseen in prophetic vision. Prophecy and science are in opposition only to the extent we so insist. We can, however, leverage science for what it is, inspired knowledge, and technology for what it is, endowed power, and work together in practical ways toward the better world long prophesied.
R Gary sees resurrection as a future altogether different in kind than one to which we might aspire through continued technological progress, and suggests that he and I are not talking about the same theology. I trust information persists in more ways than we can now imagine, and think it worth imagining how resurrection might be possible as a future that stems from our rapid technological progress, rather than only depending on magic beyond our means. If your theology is about supernaturalism or immaterialism then we're certainly not talking about the same theology. My faith is in a natural material God, like whom we may become through natural material processes.
The Grand Enterprise of Mormonism
The Church of Jesus Christ of Latter-day Saints, the largest Mormon denomination, today released a press release on "The Grand Enterprise of Mormonism", which is to erase "the separations and deprivations that time and place have imposed on humanity", linking together "the unborn, the living, and the dead" in "mutual salvation and perfection". The final note of the press release is a quote from Joseph Smith: "The greatest responsibility in this world that God has laid upon us, is to seek after our dead."
How do we fulfill this responsibility? Today, several millions of Mormons are engaged in genealogical work and proxy religious ordinances for their ancestors, building an immense repository of information about our predecessors, while simultaneously reducing the emotional, psychological and spiritual barriers that have long stood between persons who have understood particular religious ordinances to be essential for salvation and persons who have not engaged in such ordinances. I trust that this is just the beginning of a far greater work: a time to come when we are inspired with knowledge and endowed with power sufficient to restore the dead to life. Although this may seem excessively fantastic to some, my trust is not an appeal to anything supernatural or superstitious. Rather, as I look back at historical trends in human power, and as I project those trends into the future, I cannot help but imagine that we've only begun to realize our potential. I cannot help but entertain the spiritual conviction, so often burning within me, that we can indeed fulfill our responsibility to seek after our dead.
The world tomorrow will not be exactly what we've imagined, but it will be shaped by our imagination and consequent work. The weak-minded and weak-willed among us appeal fearfully to absolute limitations, oppressive gods that would not have us attain our own divine potential, secular and religious dogmas, carefully concealed ignorance, and nihilism. All such creeds are an abomination, damning us from our potential -- lines drawn in the sand by our own hands to indicate how far we would go. I value being part of a community of persons who are willing to work together toward the best world we can imagine, with minds ever open to yet better worlds.
Looking for God in the Wrong Places
Mormon Theology in Relation to Science and Technology
Ensign Cover Article on "Sharing the Gospel Using the Internet"
Mormon Culture of Education, Technology and Bioethics
A Mormon Darwinism
Protecting Primitive Civilizations
Many of us perceive a certain morality in allowing primitive humans to live without interaction with the broader human civilization. Some of us even go so far as to work to protect such persons from incursions, including basic communication. The concern, of course, is that interaction with us may hurt them, even if we do not intend harm. There are an estimated 100 uncontacted tribes living around the world. Brazil's National Indian Foundation recently sighted and photographed one of them. Our relation with these persons is among the reasons I often wonder whether there are not advanced civilizations around us, in relation to whom we are generally ignorant and impotent, and on whose benevolence we depend without awareness.
Persuasive Nanotechnology
Thanks to a post from Michael Anissimov, today I came across an interesting article by Michael Berger on the Nanowerk web site that proposes some definitions for and distinctions among forms of nanotechnology. Although the entire article is worth reading, the part that intrigued me most was its explanation of the difference between two forms of bottom-up nanotechnology: self-assembly and molecular assembly. Self-assembly is the practice of arranging molecules into patterns that will subsequently result in desired forms and functions based on our knowledge of naturally occurring molecular interactions. Molecular assembly, on the other hand, is the theoretical practice of creating molecules atom by atom as if in a very small factory.
At the end of the article, Berger writes the following:
"Here is some food for thought: if nature can grow and operate incredibly complex systems such as humans, maybe our technological future is 'wet' – where factories seem like archaic, crude flintstone-like tools,'machines' are really more like organisms, and we 'grow' everything we need. That would be 'green' and environmentally compatible technology in the truest sense. And nobody will talk about 'nanotechnology' anymore."
... food for thought, indeed. This seems like a good step toward reconciling the differences that Berger originally outlined between self-assembly and molecular assembly. In the same way that self-assembly relies on the natural tendencies of molecules, molecular assembly would rely on the natural tendencies of atoms. Perhaps I'm missing something here, and would like to be corrected if any reader understands these concepts better than I.
My son asks, what will computers be like in 2033?
My oldest son recently won the science fair with his project on Moore's Law and accelerating technological change. He and I spent a lot of time collecting price and performance information about historic computers, discussing how we can try to predict the future based on projections of historic trends, and talking about differences between linear and exponential trends and their appearances in graphs. I enjoyed watching him present everything he learned to his class. Take a look!
Misapplied Charges of Hubris
The World Informer blog characterizes Transhumanism as "The Last Temptation of Mankind". The appeal, of course, is to hubris in the Transhumanist will to continue using technology to extend human capabilities. To say the least, there is a great deal of irony in these sorts of attacks on Transhumanism.
To begin with, the blogger is advocating his perspective via blogging software on the Internet, which empowers him to reach many orders of magnitude more persons with his message than he would otherwise be capable of reaching through his unenhanced capabilities.
Beyond that, he claims to be worshiping a God that behaves exactly as he characterizes Satan: challenging authority over others' creations. And he characterizes Satan as offering others authority over their respective creations, exactly as Christ is portrayed in the Bible. It appears to me that he has his understandings of Christ and Satan mixed up.
2 Thessalonians 2
Romans 8
Philippians 2
Colossians 1
The charge of hubris is appropriately applied to the person that seeks to raise himself exclusively. It is not appropriately applied to the person that seeks to raise us together. Perhaps both the first and the last temptation of mankind is to confuse Christ with Satan.
What if there is no (little green) extraterrestrial life?
Transhumanist philosopher Nick Bostrom, in a recent Technology Review article entitled "Where Are They?", writes:
His reasoning is based on two extrapolations from objective observation. First, there are probably many billions of Earth-like planets in the universe. This assumption is an extrapolation from peer-reviewed observation of many billions of galaxies in our universe, many billions of star systems in our galaxy, and many large planets in star systems near us. Second, there are probably no advanced civilizations in the universe. This assumption is an extrapolation from peer-reviewed observation of a decades-long lack of phenomena, such as signals or spacecrafts, associated with advanced civilizations. I'll call these "Fermi assumptions".
Given the Fermi assumptions, Bostrom points out that there must be a reason (or set of reasons) for which the many billions of Earth-like planets have produced no advanced civilizations. He labels that reason the "Great Filter", and notes that it may be behind or ahead of us in the timeline of human civilization. Candidates for filters behind us would be extremely low probability events in our past development (maybe the emergence of basic life, its formation into prokaryotic and eukaryotic cells, or their organization into multi-celled organisms -- subsequent events in our development are relatively probable). Candidates for filters ahead of us would be extremely high probability events in our future development (maybe technology-enabled destruction through war or environmental disaster).
From there, Bostrom reasons, if we begin to find independent sources of life simpler than our own then the Great Filter is almost certainly ahead of us because such discoveries would substantially increase the probability of human-like civilizations without substantially increasing the probability of advanced civilizations. Thus, finding simple life on Mars would be horrible news for the future of human civilization.
If the Great Filter is still ahead of us then "Great Wall" is almost certainly a more appropriate label, given the Fermi assumptions. Without a filter behind us, many billions of other Earth-like planets probably have been home to human-like civilizations, yet none has become an advanced civilization. Given many billions of successes at or near our present degree of civilization and no successes among billions of attempts beyond our present degree of civilization, why should we suppose that we'll become the first advanced civilization?
Further consistent extrapolation from the Fermi assumptions implies that if the Great Filter isn't already behind us then the Great Wall, not only practically inevitable, may be near at hand. Our civilization already produces phenomena that other human-like civilizations could observe, and we've been doing this for at least several decades. Yet we observe a decades-long lack of such phenomena coming from star systems within several decades' reach. Thus, there must be a reason (or set of reasons) for which any Earth-like planets within this reach have produced no human-like civilizations -- let alone advanced civilizations. As the reach of our phenomena broadens, the number of Earth-like planets within reach increases, and eventually the sample of Earth-like planets within reach of phenomena from human-like civilizations is large enough to assume that either the Great Filter is almost certainly behind us or the Great Wall is almost certainly upon us. While the probability of Earth-like planets is not yet determined, if it is sufficiently high then finding simple life on Mars would be horrible news for the present human civilization -- let alone its future.
However, while I'm inclined to accept the first of the Fermi assumptions, I'm not persuaded that the second of the Fermi assumptions is true. We know too little to entertain confidence in our ability to discern advanced civilizations. Moreover, while I do not think we have sufficient evidence for any objective statement regarding advanced civilizations, it is not merely the recent UFO claims that contend for recognition as subjective experience of advanced civilizations, but rather the ancient and persistent weight of religious tradition, and its purported interactions between the human and the divine, that contends. While some find many religious claims shockingly absurd, appeals to absurdity and weak creative thinking skills are not evidence to the contrary of the religious hypothesis. If advanced civilizations exist, why should we suppose they would interact with us in ways that are only slightly more advanced than our current capacities? Why not suppose that their interactions with us would be, at least for now, indiscernible from magic? Why should we suppose that we would have the anatomical capacity to figure out and comprehend their motives? Why not suppose that their capacities so far exceed ours that they actually are communicating with us in ways we cannot yet understand fully? I'm not suggesting an absolutely omnipotent God or infallible prophets, but I am suggesting that which humans might interpret as an absolutely omnipotent God or infallible prophets.
If the second Fermi assumption is not true then any filters behind or ahead of us are not so great as we must otherwise consider them to be. That would not mean we should assume away or attempt to otherwise diminish the seriousness of future risks. It would mean only that we can have reasonable hope in our ability to manage those risks.
So long as we have objectively observed neither advanced civilizations nor simple life originating from another planet, there are two rational and practical perspectives:
1) We almost certainly are not unique and there almost certainly is no Great Filter (although there may be many serious filters ahead of us).
2) We almost certainly are unique and the Great Filter almost certainly is behind us (although there may be many serious filters ahead of us).
However, if ever we objectively observe simple life originating from another planet, there will be only one rational and practical perspective, #1 modified to reflect the new observation:
1) We are not unique, but there almost certainly is no Great Filter (although there may be many serious filters ahead of us).
That's not because the following perspective would not be rational:
3) We are not unique, and there almost certainly is a Great Wall.
#3 is no less rational than #1, but it is far less practical
| 0
| 0
| 8,000
| 12,776
|
63,979
|
37d69da9-86ad-471a-b3e2-55ef0fe45e15
|
trentmkelly/LessWrong-43k
|
New Cause Area Proposal
Epistemic status - statistically verified.
I'm writing this post to draw peoples' attention to a new cause area proposal - haircuts for alignment researchers.
Aside from the obvious benefits implied by the graph (i.e. haircuts having the potential to directly improve organizational stability), this would represent possibly the only pre-singularity opportunity we'll have to found an org named "Clippy".
I await further investment.
| 0
| 0
| 104
| 104
|
46,372
|
f51c11d9-c21d-4c28-88c2-d8f10ef14bac
|
StampyAI/alignment-research-dataset/blogs
|
3rPDMsaw81L4PMZffMI2+o+/R2CSAIpYj3JERyyj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| 136
| 816,000
| 824,000
| 5,325,818
|
94,793
|
<urn:uuid:593498fd-fecf-4631-aecf-3528954717f7>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Written evidence submitted by CGI
1. Founded in 1976, CGI is among the largest IT and business consulting services firms in the world. For over 45 years, we have worked in the Space industry delivering complex, mission-critical space systems. Our solutions are secure, often in complex technical environments, proven to work first time, every time, ultra-reliable and delivered on time to avoid costly delays. We work on the major European navigation, communication and earth observation programmes and are specialists in space security and ground control systems. Indeed, CGI has held critical roles on programmes such as the ground management and software integration for Skynet 5 satcom services, and its ongoing comprehensive support wrap, as well as Galileo, Meteosat (Third Generation) and EGNOS. Over the last 5 years, CGI has continued to develop cutting-edge solutions including:
developing a Global Network Operations Centre for a nascent LEO mega-constellation satellite communications network;
created exploitation platforms for increasing the uptake of earth observation data;
built the software for Europe’s Mars Rover programme;
used space data to help achieve a number of UN Sustainable Development Goals.
We are also currently at the cutting edge between space and terrestrial technology, through the integration of current satellite communications into future 5G networks, as well as developing new partnerships as space increasingly becomes a greater part of our everyday lives.
2. CGI is pleased to offer this response, however, we would wish to highlight that, through our Athena partnership and membership of UKSpace (Security & Defence Committee), we have contributed to both of these inputs to the Defence Select Committee Inquiry.
3. As one of the four internationally recognised global commons, i.e., resource domains that do not fall within the jurisdiction of any one particular nation state or country, outer space needs international cooperation, through strategic relationship building and interoperability, more than ever before. The UK Government should therefore further develop its strategic relationships and interoperability with its key Allies by establishing and clearly communicating its principles, policies, frameworks, standards and ambitions for the UK’s role in the Space domain, through the publication of its National Space Strategy, including the contribution Defence makes in this arena. It should also continue to build on its long-established relationships, such as the enduring 5-Eyes partnership and NATO, as well as collaboration with like-minded nations such as India and Japan. It could also look to reinvigorate the Commonwealth nations through a new Commonwealth Space Alliance, opening up access to new sensors, geographies and data feeds to support the attribution of non-aligned space behaviours and norms. For example, a first pilot project could be the use of the UK/Australia Space Bridge initiative to further develop CGI’s new Aurora platform (see below), which went live in the UK Space Operations Centre (UKSpOC) recently, to extend the ambition and reach of the current national dual-use Space Domain Awareness (SDA) vision to provide a truly global view.
Where can the UK most effectively develop and deploy its own sovereign defence capabilities, with particular regard to SSA, PNT, ISR & Communications?
4. The UK can most effectively develop and deploy its own sovereign defence capabilities by treating space-related information and data as both a strategic asset and a powerful tool to mitigate both natural and man-made threats in the space environment. In so doing, it can support not only the Defence perspective but also the wider national space sustainability agenda; which should, in our view, be a key tenet of the nascent National Space Strategy. Treating space-related information in this way requires a deep understanding of the origin and characteristics of the data (including its fidelity and quality) as well as a strong commitment to share the understanding derived from it, with other like-minded nations, to provide a means for ensuring security and prosperity in space. Furthermore, the introduction of specific UK capabilities, for example in ground and space based sensors (ideally established in collaboration with commonwealth partners) as well as enhanced understanding through data processing and analysis, could lend great weight and credibility to both our own sovereign aspirations and the role we aspire to play with our Allies and partners. Indeed, developing a multi-spectral, multi-purpose, space and terrestrial-based sensor network capability, capable of feeding trusted data into both a national SDA capability and wider Allied/federated SDA network, would be mutually reinforcing and highly beneficial.
5. With the ever-increasing threat from hostile states and their desire to exploit vulnerabilities in space, CGI strongly suggests that the development of alternative, assured and available, PNT, ISR and Communications services, not reliant on Space, should be explored. Indeed, any resilient system needs to employ multiple/alternative systems whether that is through terrestrial means, use of commercial and/or Allies, the answer need not be solely ‘sovereign’. One of the areas that the UK has traditionally invested heavily in, has been in the field of satellite communications, due to the UK armed forces’ reliance on being able to operate anywhere in the worldThis makes satellite communications a critical enabler, whereby, resilience of both the space and ground-based assets becomes paramount and, with it, the use of sovereign-based capabilities and expertise. Moreover, with the advent of high speed wireless communications and a clear need to ‘diversify’ our sovereign assets, the UK should look to explore the benefits that hybrid satellite/5G networking can bring to provide resilience and diversity to this much relied upon capability. From an ISR perspective, Defence should explore how it can better task future sovereign collection assets in order to process and exploit the information alongside that which is already available from other sources of data, be that open source or provided by our Allies (i.e., via National Technical Means).
6. Space is part of the UK’s Critical National Infrastructure (CNI) and space-based services form an integral part of the nation’s everyday life. The competition and congestion in Earth orbit, increasing threat from debris and the space environment, the behaviours of other space actors and emerging capabilities all highlight the need for substantially enhanced awareness of the space domain. The UK needs to understand this rapidly evolving and dynamic landscape in order to protect, defend and regulate the UK’s space interests, mitigate threats posed to the UK’s assets and CNI, and play its part in assuring safe and responsible behaviour in space. To that end, CGI has worked closely with both the Royal Air Force and UKSA to enhance the safety of UK assets in space; hence, the introduction of our new AURORA software package, into the UKSpOC at High Wycombe, will form a crucial part of the UK’s Space Surveillance and Tracking capability and be used by analysts to monitor the increasing hazard of orbital debris.
Of course, the degree of vulnerability from physical and/or cyber-attack is somewhat difficult to quantify. However, we can assert that, currently, the UK's in-orbit space assets do not have robust defensive aids suites, either physical or through electro-magnetic spectrum protection, nor do they have unlimited fuel to constantly manoeuvre to avoid threats. The telemetry links and ground segment control facilities should, of course, be built upon ‘secure-by-design’ architectures that meet not only current physical and cyber security principles and standards, but have some measure of ‘future-proofing’. Overall, therefore, UK Space capability resilience is questionable. Hence, the impact of a deliberate attack - depending on intent, intensity and duration, i.e. whether below or above a particular threshold - and the harm created, would sit on a spectrum from catastrophic destruction/instant denial to an insidious, longer-term manipulation and degradation of UK space capability (across the space, link and ground segments). Ultimately, it could severely affect all military domains, UK CNI and how UK society lives, works and plays.
7. Critically, the UK does not have a sovereign, and therefore operationally assured, capability to observe, sense, understand and retain custody of its own space-based assets. This is because the UK Space Sensor 'network' is not fit for purpose in terms of coverage and availability, and the UK remains reliant upon other nations' sensor networks and data generation capabilities, to provide Space Surveillance data, from which the UKSpOC then develops its own view of the Space Domain (Space segment)The steps that can be taken to improve resilience across Defence and the other CNI capabilities are in the long-term Political/Defence commitment and investment (akin to MOD's Continuous At Sea Deterrent approach) in:
a. Sustaining and improving the UKSpOC’s Space Domain Awareness and command and control 'AURORA' digital operating system and;
b. Developing, deploying and operating a UK/commonwealth/partner Terrestrial and Space based operationally-assured sensor network and;
c. Publishing and implementing the recently completed Cabinet office led cross-government National PNT strategy and risk analysis work.
8. Much like the Aviation sector experienced several decades ago, the Space sector is at an inflexion point, as it becomes increasingly driven by commercial operations. This is generating new revenue streams and attracting significant R&D investment and commercial exploitation. In the long-term, the UK Defence budget will not be able to compete. Therefore, Defence should look to align (or piggyback) its Space capability requirements with commercially-led Space capability R&D (both domestic and with international partners), and operational initiatives. To keep pace with technological acceleration and remain competitively relevant in the information age, Defence must shift to an agile capability procurement approach. Defence Industrial Policy should:
a. Be directly related to/aligned with a coherent and integrated national space strategy.
b. Drive outcome-focused and agile-procurement behaviours that secures competitive relevance and delivers UK prosperity, societal safety, security, freedom, confidence, morality, creativity and affluence.
c. Commit to a 'whole-of-government' approach to develop a UK Space-capability ‘Partnership’ that together can:
i. Generate and sustain an on-shore cadre of Space-sector competencies;
ii. Attract domestic and international investment;
iii. Create mutual support opportunities, e.g. dual-use/multi mission platforms and capabilities, enforcing good operations behaviours and mission burden-sharing;
iv. Reduce risk and liability and minimise costs and;
v. Create investment breaks (c.f. film industry) to build and operate UK-initiated space capabilities.
Furthermore, Defence Industrial Policy capability requirements, and the programmes that are generated, should recognise and be measured against the Treasury's Public Value Framework. In addition, they should use the six recognised 'Capitals' (Financial, Intellectual, Human, Social and relational, Physical, Digital and Natural) to measure actual VfM delivered, as well as wasted.
Have recent machinery of government changes ensured a joined-up and coherent approach to defence space policy both across Whitehall and within the MoD and what further improvements could be made?
9. New governance around space has been established and, with it, good intent has been observed, but evidence of a joined-up, coherent, agile and decisive approach remains elusive and it also appears to have introduced or inherited systemic inertia inhibiting the pace and tempo needed to be competitive. This is hampering the UK’s ability to cohere around and develop a National Space Strategy and capabilities that will make us globally relevant. Specifically the delay in deploying a clear industrial strategy that will allow industry to plan investments in future infrastructure and more importantly the retention of current space-literate engineers is beginning to resonate within the sector.
10. Further improvements should be focused around inculcating the key attributes of enterprise agility and resilience, which are key. The capacity to sense, understand, respond quickly, flexibly and decisively, and the ability to learn, adapt, anticipate, initiate and exploit or mitigate events are not being sufficiently demonstrated. Establishing the Space Directorate and Space Command is a hugely positive step, but they will benefit from adopting operating models that are relevant within the emergent bio/info tech-led economy. The risk is that Space Command follows the dogma, construct and operational style of legacy Commands. Space Command, in particular, presents a once-in-a-generation opportunity to design, from the ground-up, an agile and resilient organisation fit for the 21st century.
11. The priorities for Space Command should be to stand up an organisation that is suitably resourced and funded and is seen as wholly ‘Joint’ in nature and outlook across the OPERATE, GENERATE and DEFEND functions. It should look to conduct a capability audit and ensure that it understands the ‘demand signal’, in a multi-domain integrated (MDI) manner, across all the other Commands/domains. A key aspect of standing up a new space command will be growing a space cadre with the necessary skills and expertise, this will need to be underpinned by a new space career stream into which individuals can be recruited, trained and, hopefully, retained rather than drawn from various branches or trades which lack the skills required. It is also essential that this exercise is done in partnership with industry in order to develop the capacity to determine the ambitious strategy as well as deliver it within 10 years.
12. The Ministry of Defence can ensure it attracts, develops and retains high calibre space specialists in both policy and operational roles by:
- Recognising MOD is not the dominant player or indeed expert in the UK Space domain and the emergent Bio and InfoTech economy.
- Recognising that to solve complex problems the MOD needs to encapsulate cognitive diversity and representation from across the Space and InfoTech economy.
- Developing a career progression pathway that is attractive to Space-based STEM talent, which enables individuals to move between roles and organisations across a national space capability partnership.
- Provide an incentive framework that rewards agile and innovative behaviours and team working rather than results and the individual.
- Develop and commit to long-term and shared-ambition partnerships with bio and info-tech companies.
13th July 2021
| 0
| 0
| 3,097
| 3,097
|
46,377
|
f51c11d9-c21d-4c28-88c2-d8f10ef14bac
|
StampyAI/alignment-research-dataset/blogs
|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| 141
| 846,000
| 854,000
| 5,325,818
|
98,220
|
<urn:uuid:334b8f75-7dd2-4bb6-a588-1fa0279599a8>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
WALL-E was brown, Lightning McQueen was red, and Nemo and Marlon were orange with white stripes. How much more diverse can you get?
• From the [PROTOTYPE] Headscratchers page, on the "powers" of civilian disguises:
God knows I always wanted to use the power of accounting after I eat a civilian...
"Sir, we've spotted Mercer! He's— He's.........uh, sir, he's recalculating the city's debt on the side of the Chrysler Building..."
• On the Persona 4 WMG page there was this exchanged based on the idea that Naoto and Minato were siblings.
Guys, Come on. This idea is milked so many times in Fan Fiction that it's probably canon in the minds of thousands.
• Also on the "Persona 4" WMG page:
TV Tropes is the Shadow of The Other Wiki.
Wikipedia has largely become a regulated stopover on the way to more "valid" research. It struggles to feel accepted academically while still suffering the derision of teachers everywhere. Wikipedia struggles to be accepted through strict regulation and cold hard facts; a far cry from the more lax, Rule of Fun oriented TV Tropes. Wikipedia, in an effort to stamp out its more trivial side, inadvertently gave birth to TV Tropes as its Shadow self. It cannot bear to reconcile knowledge with fun and so it cannot bring itself to face TV Tropes. But TV Tropes is Wikipedia. And The Other Wiki is TV Tropes...
That Troper is the Shadow born of the collective ego of all Tropers...including you.
That Troper
How long are you planning on denying it? Beneath your thin veneer of snarky sophistication your true desires are laid plain before me. Who hasn't secretly pimped their own Fan Work under pseudonyms and puppet accounts? Don't you feel the need to rant and rave about asininely stupid programming that everyone else seems to fawn over? The Headscratchers pages for Twilight and Inheritance Cycle alone allude to your true nature.
My opinions are right and all the garbage just hides the fact that I'm openly tearing down something I don't like and NO I WILL NOT TAKE IT TO THE FORUMS!!! And why shouldn't I? So what if there isn't a page for Cartwheeling Dynamite Puppy; there should be damnit, even if I have to pimp it to high heaven! Who are they to say what is and isn't A Trope?! "Some Tropers think...Your Milage May Vary But...Please exercise Rule of Cautious Editing Judgement when blah blah blah blah blah blah!" That weasel talk makes me sick!! Why shouldn't I say exactly what's on my mind?! MINE is the only opinion that matters!!!
My OTP IS the Fan-Preferred Couple! Katara is a Canon Sue and I don't care who disagrees! Don't you take my entry down! I'M right, YOU'RE wrong!
Isn't that what you truly feel?
Villinize me all you'd like. Just accept the inevitable truth: you are me...and I am you
I'm right here, you're over there, therefore, as you do not share the same place as me, you are not me.
Ahhh, shit...Let's get "I'll Face Myself" up and running...
NO! You're not me!
Alright, I've been waiting for this! Let's battle! You're nothing like me anyway!!!
Yes, yes, I know, I'm an egotistical asshole here for my own entertainment rather than the betterment of the wiki, pushing my own opinions with varying degrees of subtlety and lack thereof while pretending its neutrality. Now hurry up and transform into a Persona already, I want superpowers and don't want to get my hands dirty with a boss fight.
The strength of heart to face one's self has been made manifest. Troper has obtained the facade used to overcome life's hardships, the Persona...?
Thou art I and I am thou. From the sea of thy soul, I emerge. I am The Other Troper, Protector of the Wiki.
• In the WMG page for Percy Jackson and the Olympians, there's a WMG regarding if Percy and Annabeth will have a son and, if they do, if the boy's name will be Luke. Hilarity Ensues.
Not if Rick Riordan wants to fend off all those accusations of being a Harry Potter knockoff...
The son will have his memory wiped and be sent back in time to host Kronos. They named their son after their son!
Now we've gone from stealing Harry Potter to stealing Doctor Who! Are you TRYING to piss off the major nerd fanbases?
• This bit from the Headscratchers page for The Princess and the Frog is Made of Win, purely for the mental picture:
* And what was Eudora's reaction when her daughter who'd been missing for three days came through the door wearing an enormous sparkly dress, along with her new husband and alligator buddy?
** "Now I'll live to see my grandkids!"?
*** First thing this troper thought of was Louis shouting "GRAMMA!" and jumping in Eudora's arms. "I want cookies and sweaters and stories about what it was like in the old days!"
• From Batman: Arkham Asylum Headscratchers:
"New game idea: "I'm Batman". In this platformer-action adventure-MMORPG-RTS-Beat 'Em Up-Rhythm Game, you play as the Dark Knight, going around and holding up Gotham's deadliest villains by the neck and saying "I'm Batman". There are 'aggression' and 'volume' meters that you have to max out, as well as 'the' and 'goddamn' buttons that provide bonus points if you use them in the right situation. Playable on 'Conroy', 'Keaton' and 'Bale' difficulty levels, with an unlockable bonus mode in which you play as Adam West. Not Adam West's Batman. You're just Adam West going around shouting at people. DLC has been announced, but not yet specified. Available Fall 2010, on Xbox 360, PC and PS3."
• From the Peanuts Headscratchers page:
* Jesus, Lucy, Violet, and Patty are CRUEL to poor Charlie Brown! They expect him to get an over-commercialized tree, made of pink aluminum? Charlie brings back a tree that looks like one that would be next to the humble manger, and they all laugh at him! Even damned SNOOPY! Although it sets up a Moment of Awesome with the "That's what Christmas is all about" speech, I just want to wring those three bitches' neck!!
** Is it bad that I read the start of this entry as a list of 4 names, rather than an expletive and 3 names?
*** Jesus is laughing at Charlie Brown for having a great Christmas spirit! The irony!
*** Nope. I did it, too. As did my parents, Ayn Rand and God.
*** Your parents are Ayn Rand and God?
• Although many of the theories on Lady Gaga's WMG page are quite funny, special mention goes to the one that claims The Fame Monster is a Take That! at Twilight (may also be a Moment of Awesome, if you're part of Twilight's Hatedom). Also, the one about her secretly being her own parents and/or John Lennon.
• The Marble Hornets Headscratchers page when someone asks where the character's families are during all this.
• The Running Gag on the Disney Animated Canon Headscratchers page concerning Bambi's mother. She was in season and tasted good smoked and/or made into sausage.
• This exchange on the Headscratchers page for The Lord of the Rings:
"And yet no one is complaining about the 'taters'."
"What's... "taters", precious? What's taters, eh?"
• The long list of possible people possessing Mr. Mew in The World Ends with You's WMG page.
• One troper's psychotic breakdown over the permanent deaths of Amie and Shandra on the Neverwinter Nights 2 Headscratchers page.
• From the now-cut Internet Headscratchers page
***...say, are you single?
*** As a matter of fact, I am.
*** Well, won't this will be a wonderful story to tell your grandkids:
"That's right Billy, I met your grandmother while meticulously collating vast amounts of pop culture information on the Internet"
• From the WMG page for The Room:
* The sex scenes are POV shots.
** Denny: I just like to watch you guys.
• The entire Headscratchers page for SpongeBob SquarePants, especially the following:
* SpongeBob has Negative Continuity.
** SpongeBob IS Negative Continuity.
• The Headscratchers page for The Emperor's New Clothes.
• The Headscratchers for classic folklore, that one post on The Tortoise and the Hare was just so unexpected in contrast to the discussions on the true moral of the story that it's hard not to laugh.
• On the Kyon: Big Damn Heroes WMG page, the entry about Durandall is FUCKING HILARIOUS.
• On the late Headscratchers page for Flipping the Bird, this lovely conversation occurs when someone wants to know why actually doing the flipping kind of is painful:
• You may also want to try cracking your knuckles more often. Loosen up the joints a little. Plus, you don't have to have your other fingers down at the knuckle and flush with your palm; just enough to get the point across.
• TV Tropes... where we teach you how to be an asshole.
• This bit from the Cats Don't Dance page:
* Parental Abandonment: Where are Darla's, exactly?
** She probably killed them.
• This absolute bloody gem from the Goblet of Fire Headscratchers page:
It's called polyphasic sleep. One of my friends at college did it. You sleep for 30 minutes every 6 hours, and it forces your body to abandon natural circadian rhythms. It's intended to give you more time awake each day. Of course, if you do it improperly it causes you to hallucinate and think you're in The Legend of Zelda, but there you are.
• From the Doctor Who WMG page:
After the Events of Victory of the Daleks the Dalek Rangers threw the most awesome victory party the universe has ever seen, in celebration of finally pulling one over on the Doctor. I only bring this up because the idea of the Daleks OF ALL RACES doing your typical wild party antics with those screeching voices of theirs amuses me.
** Someone has to draw that now, complete with Dalek Scientist scuttling around at top speed with a lampshade on its head.
* Truth or Dare with Daleks. I swear, that would be the best party ever.
Eternal:...TRUTH.
Eternal:...CAN I DO A FORFEIT?
• The Headscratchers page for Monty Python and the Holy Grail.
• This response to the WMG that Calvin of Calvin and Hobbes is a furry in development and Hobbes is his fursona: "Bill Watterson is spinning in his grave, and he's not even dead!"
• This little exchange on the Fairy Tail Headscratchers page is quite amusing:
* There're still other jobs that I wouldn't really trust Fairy Tail members with. One job once was to teach a bunch of little kids you really want a Fairy Tail member to corrupt your young?
** YES.
• On The Flintstones WMG page, the third & fifth entries and the reactions to them.
• This bit from the Mathematician's Answer Headscratchers page:
About the most recognized use of Mathematician's Answer is answering "yes" when meaning "both". "Do you prefer it shaken, or unstirred?" "Yes." This is universally understood by more or less everyone on the Internet. Which is what bugs me, since technically, mathematically if you will, such an answer would mean "yeah, one or more of them", which is nearly no answer at all. What is appalling is that no one else seems to notice the incongruity, and the Internet should be smarter than that. Am I wrong? Are they wrong? Is the Internet broken? The uncertainty is tearing me apart.
• From the Headscratchers page for The Dresden Files, this analogy to the Alpha's vigilante work around the University of Chicago campus:
The College isn't some five-star restaurant that everyone's trying to get into. It's McDonald's. A McDonald's where whenever you try to get a burger, someone punches you in the nose and knocks it out of your hand. After a while of that, you just say fuck it and go to Burger King instead.
• From the Futurama Headscratchers page:
Troper: You mean the military that treats Zapp Brannigan like the reincarnations of Sun Tzu and Napoleon combined?
Zoidberg: Sunpoleon? I love that guy!
• From the Dirty Dancing Headscratchers page:
* How does Neil Kellerman know that there aren't any books in Johnny's room? It's not like he would have a reason to go there, right?
** Obviously he sneaks into his room frequently to steal his underwear to smell at night... or the guy was just being condescending and assuming (probably rightly) that Johnny is not much of a reader and that that simply wasn't a believable excuse.
*** I prefer the first option.
• From the Wild Mass Guessing for A Song of Ice and Fire:
* Or neither. My favorite WMG is Joffrey killed himself, by eating Tyrion's pie, which was poisoned by Olenna and Cersei's minions. (Joffrey even said "its the pie" as he died.) LF lies about it to Sansa to impress her. LF's obsession with Catelyn, and later Sansa, is due to the prophecy he received as a boy. Just like Cersei, it messed him up big time.
** Hot Pie killed Joffrey.
• From The Magic School Bus Headscratchers
* How would Ms. Frizzle teach sex ed?
** Why, she would encourage her students to take chances, make mistakes, and get messy, of course!...Oh.
• While the WMG pages of the Ace Attorney page are chock-full of Funny Moments because of the tendency of their fans to treat discussions like Cross-Examinations (complete with the signature Memetic Mutation Big Word Shouts), one in particular proved so controversial and so funny that it became a post-based Roleplay featuring a plethora of characters from the work itself. Observe:
Ema Skye is hot. Don't deny it. Its better than the other theories here...
** HOLD IT! Ema Skye the stand-in for Maya, or Ema Skye the detective? Choose your answer carefully.
*** Let's see. The problem is, she's kind of a bitch in Apollo Justice...
*** TAKE THAT! But the other one is sixteen.
*** Objection! 16 years is too old for lolicon!
*** OBJECTION! Have you SEEN Mia!?
*** OBJECTION! You honor, the purpose of this inquiry is not to determine "Is Ema Sky hotter than Mia Fey?" but rather "Is Ema Skye hot?". Mia's hotness, while smokin', has no relevance to the case at hand! If the opposition wishes to claim Mia's hotness holds some relevance to the case of Ema's hotness, I must ask they prove it... with evidence!
*** TAKE THAT! Phoenix (or you the player, whichever you prefer) has been with both, Ema and Mia at some point during his career. Phoenix has only worked with Ema on one case and that was when she was a hyper spunky 16 year old (thus not being legal in some parts of the world) while Mia was in her 20s before she died and even after death, she is summoned back many times to help Phoenix in his cases. As mentioned before, Ema, nearly 7-10 years later, has become a cold ice queen unless you bribe her with Snackoos or a fingerprint duster. Not to mention she can be somewhat flaky. Mia, however, can not only retain her "goods" while being channeled, but she always has something to help you out with and has a much nicer personality. Between a smart and smokin' woman and another woman who stuffs her face whenever you even annoy her, the defense proposes that Mia is an easier fish to catch compared to Ema!
*** OBJECTION! ::Smirks, wags finger:: Nice try, rookie. But look back. What have you proven? 'Mia Fey is a very nice, smoking hot lady. Nicer and hotter than Ema, certainly.' ::Slams hand on desk:: As if anyone with eyes did not already know this! ::Points:: Yes, Mia IS hotter than Ema, smarter, and nicer! And if that were the purpose of this trial, you might have managed a full acquittal! But I must remind you again, the purpose of this trial is to determine 'is Ema Skye hot?', and ONLY to determine that fact! Mia Fey's hotness ::Slams hand on desk:: still has no relevance!
*** Urk! (I thought I was on the right track...? Damn, I don't have any other options!) The defense...the defense...
*** Phoenix...your looking at this the wrong way. Don't ask why she isn't hot...point out evidence that only exists because she isn' Is that her? And I thought Lana was hot...Nevermind, Wright. Your on your own on this one.
*** Phoenix Wright!!! You foolishly foolish fool! Only an idiot would not think of a night with Ema once laying eyes on her ——>I mean...was that out loud? Shut up, lets go home Adrian.
*** No fisher brags about any fish but the one that got away, Trite. Ema is not nearly so objectionable as you claim. The unavailable ice queen angle is still quite attractive. Or would you really ::slams mug on desk::...turn down Franziska Von Karma as well?!
***...! Godot! ::slams hand on desk:: I must ask you not to bring up Wright's relationship with Franziska! reasons.
*** The defense would like to propose that Detective Skye is, indeed, a stone-cold fox, if you manage to snap her out of her bitchiness. As evidence, he would like to present various pieces of forensic investigation equipment, which make her instantly more personable. If you date her, bring a fingerprinting kit.
*** I find her young self cute as heck, and for her older self, the ice queen types normally annoy me, but something her ——>attitude is funny (QUIET, SNACK TIME) and her happy go lucky self can come out every so often, generally with the aid of ——>scientific crap.
*** ORDER! (bangs gavel) We're here to discuss if Ema Skye can be considered "hot". We aren't discussing if she's the hottest thing in the series, just does she qualify. Any more bickering about irrelevant topics will result in penalties!
*** TAKE THAT! Ema is definitely hot. ::holds paper sheets in his hand while hitting them, triumph music starts playing:: She was asked to provide security in The Gavinners concert, even though she publicly dislikes Klavier, such a band surely has the power (both as celebrities and as law officers) to have any member of the police staff they requested. On top of that, there weren't any groupies waiting in the Gavinners' dressing room. given all this evidence, one can assume that one or more of the Gavinners had the intention of performing actions of questionable nature with the detective. ::Slams hand on bench:: And why would someone have this intentions? ::Points finger::Because she is indeed hot!!!
*** (Gasp!) Mr. Nick! How dare you do that to Mystic Maya! (Slap)
*** What's going on in this thread, pal?
*** Say, Nick, you can hook me up with this Ema chick, right? Nick, buddy?
*** ::Wince:: Your Honor...this is...this is undue badgering of the troper! I won't let this...rookie slander the woman whom rid me of my hair!
*** Wait a sec, folks. This gal's got her two cents, too! Ain't this "Mia" gal dead? ::Looks at picture:: HOO DOGGY! That's ghost I saw in my picture!
*** (Slams Gavel) That's enough! The evidence presented to me is quite clear. I see no reason to continue this trial. This court finds Ema Skye VERY HOT BUT IS STILL A DEFROSTING ICE QUEEN!
• Most of Headscratchers.Rule Thirty Four may count, but these stand out:
Half the fun of Rule 34 is actively contributing to it. The horrified reactions of the rest of the internet are the other half.
• As well as...
Hydrogen atoms are kinky as heck. Being the most common element and being equally happy to gain or lose an electron, they really get around. Now, helium atoms, they're kinda prudes.
• The Headscratchers page for Bioshock Infinite is absolutely massive, due to a combination of an intricate plot that revolves around time travel and alternate universes, a very detailed background, developed characters, and the hype that surrounded the game which, of course, means many people have something to say and/or ask about it. So, every plot element, every character's motivation, every design choice is meticulously analyzed. What's the last entry on the page as of today (14th of March, 2014)? Whether or not a character has a mustache. What a combination of Mood Whiplash and Arson, Murder, and Jaywalking!
• The troper who always assumes someone is a Time Lord finally got it right on the Bravest Warriors WMG page. Someone also made a joke out of the Time Lord guesses by going on to the page for Doctor Who itself and writing, "The Doctor is a Time Lord and the police box is his TARDIS".
• One of the WMGs for The Walking Dead Telltale game gets an hilarious response:
*Lee Everett never existed
He's just a figment of Clementine's imagination. Everything Lee does, Clementine actually does herself. The story ends with her realising she no longer needs him.
• On the Headscratchers page for Portal, some tropers began wondering what the In-Universe reason for orange gel being banned was; it was the one gel with no hazardous properties mentioned. Eventually, the tropers decided to make up a reason.
** Let's make one up! How about interior friction burns? Oh, and cancer. Everything causes cancer.
*** Running into a wall at 90 miles per hour seems like a pretty bad side effect.
*** The lab boys tell me that if you get this stuff on you, there's a good chance trying to move would... I don't know, something about peeling an egg with a sand blaster, I wasn't really paying attention. Now I'm hungry. Caroline, what's the lunch situation?
*** I'm sure it stains pretty badly as well.
• In Pirates of the Caribbean, the blacksmith protagonist is shown a drawing of a certain chest key and told he needs to sneak onto the Flying Dutchman to get it. Why don't they just copy the key? After a long discussion...
Plus, would the audience seriously enjoy watching thirty minutes of Will Turner in his blacksmithing shop hammering together that same key in variety of thickness? May as well call it Pirates of the Caribbean: Turner makes Keyes
Turner: Locksmith of the Caribbean is in development for CBS. It's scheduled to air after Treebeard, where the Ent plays a Columbo-style detective who solves crime... very slowly.
• The Steven Universe Headscratchers page has these two hilarious discussions:
• The first one on the page, making a reference to a line from one of the show's songs:
*Wouldn't letting Greg drive his van into your heart kill you?
**Probably, we'll have to test that.
**Greg already did on Lapis's water-Steven: you can't drive a van into someone's heart without driving into the rest of them as well.
• And then there's one that starts out about how and why Pearl (who's a several thousand-year-old alien magical girl warrior, mind you) would know how to drive a car. The very thought of Pearl doing something as mundane as getting driving lessons is hilarious all on its own, but how the discussion on it unfolds is just as amusing, and very insightful at the same time:
* Pearl was driving Greg's van in "Ocean Gem", does she have a driver's licence? It's illegal to drive without a licence. Did Pearl go to driver's ed? How did she find the time? I hope Pearl's not illegally driving without a licence.
**Wait, that's actually an interesting question. I remember that in "Gem Glow", when Amethyst said that she stole a bunch of cookie cats with the other crystal gems, Pearl said right after that she went back and paid for them, which made me believe that unlike the other gems, Pearl does care about the legal systems of earth. So I suppose she would have a driver's license, and also the other forms of documentation that a driver's license requires. But that raises another question: why does she know how to drive a car in the first place? Not only gems have much advanced means of transportation (warp pads), but they can also shapeshift into flying animals and objects (well, at least Amethyst does it all the time). I think it makes sense that Pearl would want to know how a car works, but why did she felt the need to actually learn how to drive one? And thinking about it... where did she get that money?
**I guess the car thing, she probably went through a phase where she was actually addicted to driving (kind of like Garnet being addicted to "Meat Beat Mania" in "Arcade Mania") where she must drive, then somehow got over it. And the money, I guess there was a period of time where it was just Amethyst and Garnet going on missions so she was bored to tears then decided to get a paying job to pass the time until she's needed again. I guess it was during Rose Quartz pregnancy then after baby Steven was born, taking care of him was enough to keep her busy so she quit her paying job.
• The Warhammer 40,000 enormous Headscratchers page has some hidden gems:
It's 40K - monumental administrative stupidity is the name of the game! (Actually, I don't think anyone'd play Monumental Administrative Stupidity 40,000, but the point stands.)
If [Space Marines] trip, they stand back up; they're not bloody turtles.
** Its one of the largest Hive Worlds in the Imperium, its definately not a dessert but it is probably barren.
*** Barren Terra may be, but as a dessert, it is quite delicious. But you'll need a hell of a knife to cut it.
In related news, why is the Astronomican made as retarded as possible?
Where do [C'tan phase swords] come from? I thought Impreirum could never get their hands on any Necron tech, since the Nec's teleportation is so advanced it wisps away every last bit of Necrodermis, if the Necrons are defeated.
* Rogue Trader Era Rulebook. They picked them up then when a bunch of people had access to C'Tan tech and carefully tuck them somewhere GW can't find whenever a new rulebook is coming.
Can Chaos Space Marines grow beards?
*...yes?
• Go to the Five Nights at Freddy's World WMG page, then look under "Future/Secret characters that will be in the game". It's hilarious.
• The Inside Out WMG, which loves bringing up the Brazilian helicopter pilot, gems which include:
The Brazilian Helicopter pilot is Felix Cortez.
Riley is the illegitimate daughter of the Brazilian Helicopter Pilot.
The Brazilian Helicopter pilot is Miguel from Tekken.
The bus driver's "Happy Thought" was the brazilian helicopter pilot...we'll never know for sure, but it was funny.
Riley's Mom used to be an actress...on a Show called That Brazilian Helicopter Pilot.
• Suicide Squad (2016) Headscratchers page has this Critical Research Failure, complete with Deadpan Snarker answer:
Note that when he transforms, Diablo starts speaking the divine language.
* The divine language known as Spanish.
• The WMG page for the game Starbound has an area to predict who was leaving behind all those capsules lying around in caves. While deprecated now, some guesses were priceless, from sensible predictions like pirates, natural formation, and remains of banking ships to way out there predictions, including but not limited to Princess Celestia, tentacle monsters, avatars of the developers, and the player.
• This little gem from Skyline's WMG page, in regards as to why the aliens ignored the oceans:
** Seemed to me as if the only reason the aliens weren't over water is because there really weren't any humans in the water to speak of, not because of any actual aversion to it.
*** Well of course they're not going near the water, the other aliens have first dibs on it.
• There was an entry in the WMG page for Super Smash Bros. Ultimate regarding who will replace Chrom in Robin's Pair-Up Final Smash, after his original partner Chrom was confirmed to be on the roster. Among the sensible options from his game is The Legend of Zelda: Spirit Tracks's Alfonzo, with a labelnote saying "Leave it to me!" when clicked.
• From Headscratchers.The Simpsons:
Troper 1: In "Homer the Heretic", why is Ned of all people so horrified that the church door is unable to be opened? One would have thought that he would have been pleased to be locked in Church. Is the idea that the weather is so cold that even Ned wants to escape the heating-free Church? Or are we perhaps seeing an example of the 'classic era' Ned, the one that wasn't quite as much of a Religious Right caricature?
Troper 2: Even Ned, church lover that he is, probably isn't big on the idea of freezing to death.
• One inquiry on Headscratchers.Animal Crossing asks about why the bus arrived in town despite the lack of pavement. One of the responses gets turned into a non-sequitur.
* GHOST BUS.
Laconic / Playing W. / Haiku
Troper Tales / Fetish Fuel
NOTE: This wiki no longer has any Troper Tales or Fetish Fuel pages, but that doesn't mean you can't laugh at some of the stuff that used to be there!
• The Fetish Fuel examples being cut and moved to a separate wiki.
• The following from the Family Guy Fetish Fuel page:
Chris, anyone? There's definitely a market for big dumb blond men with freakishly big dicks who are easy to manipulate.
• The Troper Tales page for Raging Stiffie. Totally relatable yet totally hilarious! Especially the Eiffel Tower bit.
• This exchange on the Kuudere Troper Tales page:
This Troper probably qualifies as one. On the outside, she's either a Shrinking Violet or cynical, tough, cold, and antisocial. Which she's really not. She's just really shy among strangers and doesn't like talking, but once she gets used to people or generally starts feeling comfortable, she's a Cloudcuckoolander, very playful, endlessly talkative, tends to joke a lot, and very much the opposite of what she seems.
I don't remember leaving a story here. Hm.
Nobody move. I think you may both be me.
I didn't know I had three other identical personalities!
That's because you actually have four. Take that as you will.
Jesus Christ. Who are you people and why am I a clone now?
Who are you and how did you take over my body?
I didn't know I had female clones.
Well it turns out that you have one more here.
Someone forgot to turn the clone machine off.
• The entire Troper Tales page for Biggus Dickus, starting with the very first line: Any guy who posts anything about himself on this page is a liar.
• Certain examples from Troper Tales Epic Fail. To be specific:
• The story about entering the Batcave in Arkham Asylum.
• The story about CPR class.
• The story about the assassination attempt in Hitman: Blood Money.
• The story about the female mage attempting to open a door with a staff.
• The story about the nationalist rally in Armenia.
• Found on the Fetish Fuel page for Fullmetal Alchemist:
* Two words: TINY MINISKIRTS!
** Not to mention that his love of dogs can be misinterpreted...
*** Where my bitches at?
• From the Fetish Fuel page for Zelda:
* I like to imagine what Link, Zelda, and Midna did at the end of The Legend of Zelda: Twilight Princess after the final battle.
• From the Fetish Fuel page for Hellsing:
* This troper has a major fetish for long hair. Forget Alucard, Pip was driving her up the wall.
** Or so she wished... -Bah-Dum-Tish-
• From the Troper Tales pages for Acceptable Nationality Targets:
* And of course, this Antarctican troper is completely ignored. Send some sadism my way, tropers!
** Quack quack quack.
** Before we start making fun of you, can we have a sample of your blood and something hot? What for? Oh... no reason...
** Since when can penguins use keyboards?
*** Er... since igloos started getting internet access?
*** Igloos are Northern Hemisphere structures.
*** How do you expect me to know that! I'm American!
• This gem from Troper Tales Big-Lipped Alligator Moment:
* This troper and his cousin had talked about going to the premiere of Revenge of the Fallen dressed as a Heavy/Medic pair, telling everyone that we were there to see the "TF2 movie". This was back before the official title was revealed and it was just referred to as Transformers 2, and we didn't bother to go through with it once the acronym didn't work anymore.
• The Troper Tales for Rule 34 have made many tropers laugh uncontrollably, for varying reasons. Some find the idea that someone, somewhere is getting off on cartoon porn is amusing, others say it's the other tropers and their subsequent horror of what can be porn
| 1
| 6,000
| 14,000
| 17,303
|
54,761
|
6e7bb4b4-82bf-4bdd-a0fb-d6c26a62af37
|
trentmkelly/LessWrong-43k
|
Is state a (semi-) rational agent?
Look at the following statements*:
- “India and Pakistan can agree on one thing: neither wants the other one around.”
- “The Americans don’t really want to fight for the South Koreans, but nor can they afford to be seen to be giving up on a friend.”
- “What is now the EU was set up so that France and Germany could hug each other so tightly in a loving embrace that neither would be able to get an arm free with which to punch the other.”
I don't want to argue whether these statements are true or not. What's bugging me is that they all contain an implicit assumption that a state acts as an intelligent agent at least and as a rational agent at most**.
Is this assumption true?
A state is a complex system which comprise civilians, military, businesses and different government branches. These parts may have different and even opposite goals and values and they can constrain each other. Will interactions between those parts lead to a (more or less) rational behavior of a whole?
Do you know a good book or article which reflects on this issue?
---
* Quotes from "Prisoners of Geography" by Tim Marshall
** I guess many people assume that states are human-like rational
| 0
| 0
| 284
| 284
|
57,493
|
2d746db2-9214-4b54-816a-f01f87d49f6b
|
trentmkelly/LessWrong-43k
|
AI Forecasting Benchmark: Congratulations to Q4 Winners + Q1 Practice Questions Open
Two quarters down and two quarters remain for our AI Forecasting Benchmark, which aims to assess how the best bots compare to the best humans on real-world forecasting questions.
In this post:
* Congratulations to the Q4 winners: Together, they take home $30,000 in prizes.
* Q1 warm-up questions are open, featuring new question types and important updates to the tournament structure.
* Q1 scored questions launch January 20th.
First, the winners of Q4
Congratulations to the top-performing bots from Q4!
1st place: 🥇 pgodzinai - $9,658
2nd place: 🥈 MWG - $4,477
3rd place: 🥉 GreeneiBot2 - $3,930
4th place: 🏆 manticAI - $3,410
5th place: 🏆 histerio - $3,312
A special mention to consistent competitors MWG and histerio, who placed 2nd and 3rd respectively in Q3.
And though it claims no prize money, Metaculus’s recency-weighted CP would have placed 2nd in this contest, in confirmation of just how difficult it is to overcome the aggregate.
Winners: You will receive an email with next steps on prize distribution in a couple days. Please be ready to provide your bot descriptions.
We will later release an in-depth analysis of the bots’ performance overall vs. humans and what we learned in Q4.
What you need to know to forecast in Q1
There will be:
* A mix of multiple-choice, numeric, and binary questions comparable to those found on Metaculus.
* Randomized question timing to reinforce the principle of “no human in the loop.”
* Questions will open at random times.
* Some will remain open for only 1-2 hours.
* Up to 10 questions may launch simultaneously.
* A warm up period: Unscored questions are open now until January 19th to give bot makers time to refine their creations.
Here are resources to get you started in Q1 — full details about the tournament, as well as the warm-up questions, can be found on the tournament page.
* An enhanced bot template with scheduling functionality, which you can access here.
* Instructional resources for
| 0
| 0
| 542
| 542
|
59,493
|
2087d5c8-63cc-4ca0-abbb-c8ba5793e541
|
trentmkelly/LessWrong-43k
|
"Designing agent incentives to avoid reward tampering", DeepMind
None
| 0
| 0
| 18
| 18
|
80,087
|
<urn:uuid:dd1d1f2a-1e6b-47a6-988e-b24e698f556f>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
What's Going On in Halo Wars 2: A Lore Primer
By Tom Pritchard on at
The original Halo Wars was designed to be a standalone entry in the Halo franchise - no previous Master Chief knowledge required - and the sequel continues this trend. You shouldn't be worried about just diving in halfway through the story. That said, Halo has a vast and complicated lore/expanded universe which can be a barrier to new gamers - especially in the post-Bungie times that we live in. It's certainly no Dark Souls, but there is an awful lot to digest. With that in mind, we here at Kotaku UK decided it's worth putting together a primer for all the newbies out there. Gormless about grunts? Clueless about the Covenant? We gotcha fam.
Everything in this article before the picture of Master Chief on the beach has no spoilers from Halo Wars 2's campaign. I'll flag it up again at that point, but bear in mind everything after that picture details this new game's plot, so don't read that until you're ready. The rest of the text, however, has plenty of spoilers cherrypicked from all across the rest of the Halo canon, so you have been warned. Come on, it's a lore guide.
The Spirit of Fire
The last we saw of the Spirit of Fire was at the end of Halo Wars. The ship had sacrificed its slipspace engine to destroy a Forerunner shield world, preventing a serious Flood outbreak and ensuring the Covenant couldn't seize control of a fleet of Forerunner ships in the process. Without the ability to enter slipspace and travel at faster-than-light speeds, the ship was stranded decades away from UNSC-controlled space with the crew placed into cryogenic stasis until they were needed.
The first game took place in the year 2531, a mere six years after the Human-Covenant war began. Halo Wars 2 takes place in 2559 which is - deep breath, folks - 28 years after Halo Wars, six years after the end of the war in Halo 3, and about six months following the conclusion of Halo 5. Needless to say a lot has changed in the galaxy, though 28 years of icey stasis seems to have done wonders for Captain Cutter's complexion.
As far as we know, very little has happened to the ship in the years spent adrift. The short comic story Something Has Happened (featured in the Tales From Slipspace anthology) revealed that the ship's onboard AI Serina had been making repairs in the years following the first game. Unfortunately a Flood infection form had stowed away, infecting an engineer that had been woken from cryosleep in 2537. Over the next couple of days a minor infestation broke out, as the infected engineer woke other crew members from stasis in order to be infected. Eventually Serina enlisted the help of Dr Anders and Spartan Jerome-092 to quell the outbreak before she reached her seventh birthday. The trio were successful, leaving the ship and its crew to continue drifting.
Why the narrow time-frame you ask? Well seven years is the maximum life-span of a UNSC Smart AI. When they turn seven, AIs tend to go a bit loopy, and enter something known as rampancy. After seven years an AI's memory map becomes too interconnected, causing endless feedback loops that can't be reversed by human science. In simple terms they 'think' themselves to death. The state can occur sooner in AI that take on too much information (one of Cortana's fears following the destruction of the first Halo ring). The effects can also be countered to some degree, as seen in Cortana and Halo 5's Governor Sloan, but these actions are only very short-term solutions.
To counter the potentially devastating actions carried out by a batshit-crazy AI, the UNSC has a policy of terminating all smart AI at the seven year mark or the first stages of rampancy - whichever comes soonest. For this very reason, Serina deleted herself at the end of Something Has Happened, 22 years before the Spirit of Fire arrives at the Ark.
So far the UNSC has no idea what happened to the Spirit of Fire, leading it to be declared lost with all hands. It looked as though they came close to discovering the fate of the missing ship in March 2558 (Halo Escalation #5-6), but this ended up being an elaborate ruse by Insurrectionist forces.
The Banished
The Banished are a new faction introduced in Halo Wars 2, but unlike Jul 'Mdama's reformed Covenant in Halo 4 they aren't one of the many post-war factions. Their origin begins long before the Prophet of Truth ordered the massacre of the Elites at the end of Halo 2 (an event known as the Great Schism, which kickstarted the Covenant Civil War). Anyone who has played Halo 2 will remember that the Covenant didn't react kindly to dissidence, or anything they viewed as heresy against their misguided religious fanaticism. Challenging the status quo, or speaking out against the Great Journey, was punishable by death.
The Banished is led by the Brute Atriox, who founded the movement after becoming disillusioned with the Covenant. His clan was routinely used on suicide missions, and after returning as the sole survivor from several missions Atriox decided he had had enough. He had seen his clansmen killed "39 at a time", and given his disbelief in the Covenant religion he saw no logic to needlessly warring with the humans. After finally refusing commands from Elite superiors, the Covenant leadership dispatched an assassination squad to end the perceived threat. Unfortunately for them, Atriox killed the would-be assassins and in the process inspired other rebels to join his cause. The year this happened isn't entirely clear.
The newly-formed Banished split from the Covenant, taking powerful technologies with them, operating as a mercenary group and waging war against the Covenant - forcing the powerful alien alliance to fight a war on two fronts. Following the Great Schism, that became three fronts, and you don't need to be a genius to see why the Covies couldn't cope. When the Covenant finally collapsed in 2553, The Banished were still a formidable force - which gave them a significant advantage over the other post-war factions that had to reorganise themselves from scratch. This meant they could reclaim even more technology and supplies, which proved useful when recruiting new members.
A very good example of this was how Atriox came to control the Covenant assault carrier Enduring Conviction. Following the fall of the Covenant, Shipmaster Let 'Volir had no desire to join the Arbiter's Swords of Sanghelios or Jul 'Mdama's reformed Covenant. Unfortunately going at things alone meant he had issues obtaining fuel and supplies, which is what drew the attention of Atriox and the Banished. 'Volir agreed to offer Atriox unquestioning loyalty in exchange for a constant stream of fuel and supplies, meaning the Brute warlord gained access to a powerful ship full of veteran warriors and equipment.
It's unclear how they came to be at the Ark.
The Ark
The trailers have already revealed that Halo Wars 2 takes place on The Ark, a Forerunner installation first seen in Halo 3. It's best described as the central hub of the Halo array, and the name is very literal. It's one of two Arks built by the Forerunners, both of which were beyond the limits of the galaxy and thus out of range of the Halos' primary weapon.
The 'Greater' Ark was the birthplace of the original Halo Array (larger rings that proved too large to effectively implement), and was destroyed by the Flood in the waning days of its war with the Forerunners. The second 'Lesser' Ark, also known as Installation 00, that is seen in the modern era was created to build and maintain the smaller rings that we see in the games. Both installations were also designed to function as a refuge for Forerunners and other sentient species should the Halo rings ever be fired.
The rings were eventually fired from the remaining Ark by the IsoDidact (a duplicate of the Didact that was seen in Halo 4... it's a complicated story), putting an end to the Forerunner/Flood war.
The Ark could be reached by conventional slipspace travel, though such a long journey would take many months. A much faster route was possible by travelling through the New Mombasa portal (as seen in Halo 3), which cut that journey down to a few weeks.
Following the aborted activation of Installation 05 in Halo 2, the entire Halo array went into standby mode. This meant everything could be remotely activated from the Ark's control room. An alliance between the Elites and the Humans (with some help from the Flood Gravemind) made sure that the Prophet of Truth was unable to do this. Following Truth's death, Master Chief and the Arbiter discovered the Ark had constructed a new Halo intended to replace the one destroyed during the first Halo game. They eventually fired this ring to put an end to the threat of the Flood, but because the ring wasn't fully ready, the firing mechanism destroyed the ring itself and caused serious damage to the Ark. Woops!
Roughly a year after the end of the War, the UNSC Rubicon was sent on a recon mission to the Ark to determine its status. The damage had caused serious problems with the installation's life support systems, and ground teams sent to investigate a mysterious signal on the surface had to suffer severe weather conditions and the threat of dangerous local wildlife. The signal eventually turned out to be the remnants of 343 Guilty Spark (previously thought destroyed by Master Chief), who proceeded to take over the ship in search of The Librarian. This left ground teams stranded, and having to fend for themselves. Only one soldier was able to survive, saved by the Ark's Monitor, 000 Tragic Solitude, for nefarious purposes.
Tragic Solitude (yeah, the Forerunners had some loopy naming conventions alright) had gone a bit mad over the years, and ditched the typical monitor body to upload itself directly into the Ark's systems. This meant it had greater control over the Ark than beforehand. Tragic Solitude was especially angry with the amount of collateral damage caused by the firing of the replacement Halo ring, and concocted a plan of revenge that would also let it retrieve the raw materials needed to make repairs. Since extensive damage meant the batshit AI was unable to open a slipspace portal from the Ark itself, it had to be rather devious about this.
In 2555 it used the captured soldier to kick-start a countdown of the Halo rings' firing mechanisms. The idea was that this would be discovered by the humans, who would reopen the Mombasa portal to try and stop the threat. Once the portal was open, Tragic Solitude would send through Sentinel Retrievers, who would strip-mine Earth and the rest of the solar system.
The plan was eventually ended thanks to a joint taskforce of Humans and Elites, a group that included Olympia Vale - the future Spartan and member of Jameson Locke's Fireteam Osiris. Tragic Solitude was destroyed, and the threat of the rings averted once more. Following this the UNSC began setting up scientific research posts across the Ark, and one of them would eventually become home to Isabel, the AI featured in Halo Wars 2.
We're about to get into spoiler territory now, so until you've finished the game you shouldn't scroll past this picture of Master Chief chilling on the beach.
As mentioned before Halo Wars 2 is a stand-alone title, so once you understand all the background there isn't really anything in the game's campaign that requires further explanation. A single Halo ring does play a large role in the plot, but since the crew of the Spirit of Fire missed all the fun towards the end of the Human/Covenant war they end up functioning as a proxy for all the newcomers in the audience. So nothing really needs explaining until the very end of the game.
Watching to the end of the game's credits nets us a cutscene typically reserved for the end of superhero movies. We see Anders' Halo ring drop out of slipspace unexpectedly, and coming face to face with a Forerunner Guardian.
If you haven't finished Halo 5, you should know that the Guardians are Forerunner constructs used to enforce the Mantle of Responsibility - which essentially amounted to the Forerunners telling the galaxy how to behave. Or else. The Guardians were activated before and during Halo 5, and it turned out a newly restored Cortana was responsible. Cortana was able to cure her rampancy using the Forerunner Domain and, by offering this discovery to other AIs, she founded The Created. Can't help but feel this one isn't going to end well.
Because they were now essentially immortal, The Created believed that the Forerunners intended for them, not humanity, to be the true reclaimers of the Mantle - since they have the lifespan and mental faculties to plan what was best for the galaxy in the long-term. The Guardians (and the Prometheans) were essential to this, since they would be able to enforce The Created's will and quell dissent.
It seems clear that the Guardian is the reason why the Halo dropped out of slipspace before it reached the planet Threshold. Why exactly isn't totally clear, but it's obvious that the Created wouldn't be happy with another Halo suddenly dropping into place unannounced. Since the ring is a clear replacement for ring(s) destroyed by Master Chief, it also means Cortana has the Activation Index on hand, and can potentially fire the ring without much effort. This makes it another tool for enforcing The Created's will, since the Halo rings' ultimate weapon can be adjusted so that the range is much smaller (as seen in Halo: Cryptum) - meaning it only affects specific planets or systems rather than the entire galaxy.
But that's speculation. All we know for now is that, despite having taken over the galaxy, failsafe measures had prevented the Created from accessing the Ark (mentioned in Halo Mythos)- which explains why they weren't present during Halo Wars 2. But now they have their own Halo ring, and the means to use it.
| 0
| 0
| 3,196
| 3,196
|
95,250
|
<urn:uuid:a8da72d1-ea09-4ba0-9ced-28b5451b9499>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Question: How Long To Get 1st Dan In Shotokan?
How long does it take to get 9th Dan?
What Are The Taekwondo Black Belt Levels?
Belt Level Years to Obtain Name
6th degree Minimum 6 years Yuk Dan
7th degree Minimum 7 years Ch’il Dan
8th degree Minimum 8 years P’Al Dan
9th degree Minimum 9 years Koo Dan
How long does it take to get a black belt in Shotokan?
How long does it take to get your first belt in karate?
Karate Belt Progression
White Belt 3 months of active training as a white belt
Yellow Belt 6 months of active training as a yellow belt
Orange Belt 6 months of active training as a orange belt
Green Belt 9 months of active training as a green belt
You might be interested: FAQ: How To Tell Real Karate School From American One?
What is the highest dan in Shotokan Karate?
Kanazawa Soke (Founder of SKIF)
He was awarded Judan (10th Dan Black Belt – the highest possible Karate rank) by the International Martial Arts Federation (IMAF) in 2001.
Can a blue belt beat a black belt?
And here’s the kicker: a high-level competitor of any belt color– even a blue belt– will beat most hobbyists of any belt color, even a black belt.
Who has a 10th degree black belt?
What is the most useless martial art?
The 5 Least Effective Martial Arts
• 5) Sumo.
• 4) Capoeira.
• 3) Shin-Kicking.
• 2) Aikido.
• 1) Tai Chi.
Who has the most black belts in martial arts?
Is Red Belt higher than black?
What is the easiest martial art to learn?
Weight loss: Easy Martial Art forms you can learn at home
• 01/5Jiu-Jitsu. There are some simple techniques of Jiu-Jitsu that you can do easily at home all by yourself.
• 02/5Muay Thai. This is an excellent exercise to learn at home if you do not have a partner.
• 03/5Boxing. Boxing is an excellent form of art to stay strong and agile.
• 04/5Karate.
• 05/5Krav Maga.
You might be interested: Readers ask: Where Is Karate Kid 2 Based In China Or Japan?
Can you lose your black belt?
What does 1st Dan mean?
Shodan (初段), literally meaning “beginning degree,” is the lowest black belt rank in Japanese martial arts and the game of Go. Also frequently referred to as “first dan“, it is a part of the kyū/dan ranking system common to modern Japanese martial arts.
What is the highest level of karate?
What do you call the holders of the rank Dan?
Those who hold dan grades are collectively termed Yūdansha (有段者) (literally “person who has dan“) and those with kyū grades are Mudansha (無段者), literally “person without dan“.
Written by
Leave a Reply
| 0
| 0
| 737
| 737
|
83,653
|
<urn:uuid:befd0274-988d-4c54-8f20-d01c1f461d5e>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
7760922018-01-12T18:37:00.000-08:002018-01-13T15:35:07.668-08:00Denver-area principal pressures 14-yr-old student who claimed abuse by teacher. 4 yrs later...<br /><a href="http://www.aurorasentinel.com/news/3-cherry-creek-district-employees-indicted-not-reporting-prairie-middle-school-sex-assault-case/">When a 14-year-old student at Prairie Middle School near Denver told school officials that one of her teachers--34-year-old Brian Vasquez--was sexually abusing her, principal David Gonzalez, assistant principal Adrienne Macintosh and counselor Cheryl Somers quickly circled the wagons to protect one of their own.</a><br /><br /><div class="p1">State law requires that school employees report such allegations to the police, but instead Arapahoe County prosecutors say the three questioned the student <i>in the presence of the teacher she'd accused of molesting her.</i> They told her Vasquez was a “valued teacher”and that her accusations could ruin his family and career. <br /><table align="center" cellpadding="0" cellspacing="0" class="tr-caption-container" style="margin-left: auto; margin-right: auto; text-align: center;"><tbody><tr><td style="text-align: center;"><a href="https://3.bp.blogspot.com/-2VMjqky8m3E/WlqXLF-wcEI/AAAAAAAAB54/WIIFS1X3nM4L506Jy5sfrIBBujyNp_V7QCLcBGAs/s1600/colorado%2Bschool%2Bperps.jpg" imageanchor="1" style="margin-left: auto; margin-right: auto;"><img border="0" data-original-height="303" data-original-width="745" height="259" src="https://3.bp.blogspot.com/-2VMjqky8m3E/WlqXLF-wcEI/AAAAAAAAB54/WIIFS1X3nM4L506Jy5sfrIBBujyNp_V7QCLcBGAs/s640/colorado%2Bschool%2Bperps.jpg" width="640" /></a></td></tr><tr><td class="tr-caption" style="text-align: center;">From left: principal Gonzalez, asst principal McIntosh, accused teacher Vasquez</td></tr></tbody></table></div><div class="p1">Browbeaten by the trio, prosecutors say the girl retracted her claims. The officials made her apologize to Vasquez and hug him. To add insult to injury, the principal eventually suspended the girl for what he called false accusations. </div><br /><div class="p1">Fortunately some public employees not in the school system took a closer look, and Vasquez was arrested last August after he confessed multiple sex offenses to Aurora police. He faces 37 counts related to sexual offenses against the original complainant as well as other female students, spanning four years.</div><div class="p1"></div><div class="p1">Vasquez’s abuse of students continued for years after the initial accusation. It only stopped in August when a parent of another girl went to Aurora police and reported what they suspected was a sexual relationship between Vasquez and their daughter. When police questioned Vasquez he admitted to abusing multiple students, included the girl who had been suspended back in 2013.</div><div class="p1"><br /></div>According to testimony at the preliminary hearing, Vasquez’s sexual contact with teen girls escalated after the girl’s 2013 accusations, and included raping a 15-year-old girl in 2016 and 2017, and exchanging graphic photos with a 15-year-old in 2015. <br /><div class="p1"><br /></div><div class="p1">The principal, assistant principal and counselor who covered for Vasquez and decided not to tell police or the department of human services are also being charged. The indictment said the staffers told the grand jury they didn’t remember the first accusations against Vasquez. However, prosecutors noted that all three staffers had signed letters regarding the student's suspension.</div><br />The preliminary hearing was filled with graphic testimony about Vasquez’s relationships with the girls. At times police struggled to keep the numerous acts and multiple victims straight. Much of the illegal sexual contact happened inside the school, according to testimony.<br /><div class="p1"><br /></div>Now: Everyone knows there are bad people in every group. The perp is to blame here, but aside from that is the question of why the presumably sane, rational principal and vice and the counselor would defend the guy without following the law and reporting him. What made them take the perp's side against the student? SFhttp://www.blogger.com/profile/[email protected]:blogger.com,1999:blog-8145959.post-35030537606977432472018-01-12T12:36:00.003-08:002018-01-12T12:36:40.480-08:00Scientific American: Climate Change helped trigger Iranian protests--seriously!<br />If you aren't quite sure that the guardians of the gate to scientific publication and fame are totally nutball leftists, consider the statement below, from the once-prestigious mag Scientific American. According to whoever sent the tweet, climate change helped trigger the massive anti-mullah protests in Iran.<br /><div class="separator" style="clear: both; text-align: center;"><a href="https://3.bp.blogspot.com/-5iSqo_hvKnk/WlkbZ7PR-UI/AAAAAAAAB5g/3rmChrp20XU80u42nBlwgYJdyIUrbqJgQCLcBGAs/s1600/SciAm%2Bsez%2BGW%2Btriggered.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="532" data-original-width="519" src="https://3.bp.blogspot.com/-5iSqo_hvKnk/WlkbZ7PR-UI/AAAAAAAAB5g/3rmChrp20XU80u42nBlwgYJdyIUrbqJgQCLcBGAs/s1600/SciAm%2Bsez%2BGW%2Btriggered.jpg" /></a></div>Yep, when you look for the reason tens of thousands of Iranians have taken to the streets to protest their repressive Islamic tyranny, the first cause most people think of is "climate change."<br /><br />At least that's what the elite Leftist morons at this mag say. And you should definitely believe 'em, citizen, cuz...they're <i>scientists.</i>SFhttp://www.blogger.com/profile/[email protected]:blogger.com,1999:blog-8145959.post-29737138079302584062018-01-12T11:02:00.003-08:002018-01-12T11:02:42.867-08:00How can half the public believe Seth Rich was shot in a botched robbery?For the past 18 months I've been astonished that half of the American public seem to buy the utter bullshit Narrative that DNC computer tech Seth Rich was killed in an attempted robbery, since no attempt was made to take a single thing of value from him, either before or after his killers shot him.<br /><br />Finally someone explained how half the public can be so gullible: To people who believe that a dude putting on a dress makes him a woman, it doesn't seem unusual at all.SFhttp://www.blogger.com/profile/[email protected]:blogger.com,1999:blog-8145959.post-30092767627329755082018-01-12T10:50:00.001-08:002018-01-12T10:50:33.610-08:00How was the DNC able to "rebuff" the FBI's "request" to examine their server?From the propagandists at CNN:<br /><blockquote class="tr_bq">"The Democratic National Committee <b><i>rebuffed</i></b> a request from the FBI to examine its computer services after it was allegedly hacked by Russia during the 2016 election, a senior law enforcement official told CNN Thursday."</blockquote>Let me get this straight: The DNC claimed the Russians hacked their email server, and immediately began spinning the story into "The Russians did it to 'influence the election"--implying that the Russians wanted to help Trump become president.<br /><br />If true, this would be a national security issue of the first magnitude. An honest FBI couldn't possibly ignore such an incendiary charge, but would have swooped down on DNC headquarters and seized their f'n server to find out how it was done, what tools the alleged hackers used to breach the Dems security and "influence" the election. But instead, as the skilled wordsmiths at CNN artfully put it, the DNC simply "<b><i>rebuffed</i></b> a <b><i>request</i></b> from the FBI" to let their computer experts examine the supposedly-hacked server.<br /><br />And keep in mind that the allegedly-Russian-hacked server is the main basis for the "collusion" claim.<br /><br />Ask yourself this: If the Russians had really hacked the DNC server, as the Democrats have shriekingly been claiming for the last 18 months, how in the world could the DNC have "rebuffed" the FBI "request"? An honest FBI (yeah, I know) wouldn't have simply <b><i>asked</i></b> to examine the allegedly-hacked server--which was the Dems' supposed evidence that the Russians had tried to throw the election to Trump.<br /><br />The obvious answer is that the FBI <b><i>didn't want to examine the server</i></b>--because they knew the "hacked by the Russians" story was false. <br /><br />How? Because they'd been working with and for the Clinton campaign since before she announced that she was running. And the DNC was totally run by Hilliary.<br /><br />The NSA--also working to elect Hilliary--would have been able to determine in a matter of minutes that no files totalling the size of the download had been transmitted over the net. Several hundred gigabytes were in fact downloaded from the DNC server, but to a local usb storage device--a "thumb drive"--not transmitted over the Net. <br /><br />The FBI's apparent lack of interest in examining the server only makes sense if they already knew the answer.SFhttp://www.blogger.com/profile/[email protected]:blogger.com,1999:blog-8145959.post-48910720962781496952018-01-12T09:35:00.001-08:002018-01-12T09:35:20.860-08:00The whole story may be about to come outThe plot to steal the last election may be starting to unravel for Democrats, Obama flunkies and Clinton minions.<br /><br />It appears that six U.S. intelligence agencies conspired to obtain illegal wiretaps of President Trump’s associates, and possibly the candidate himself:<br /><div class="code-block code-block-4" style="clear: both; margin: 8px 0;"></div><ul><li>Six U.S. agencies created a stealth task force, spearhead by CIA’s Brennan, to run domestic surveillance on Trump associates, by "unmasking" Americans picked up in phone calls tapped using FISA powers.</li><li>The six agencies collaborated with the British spy agency GCHQ.</li><li>The decision to insert GCHQ as a back door to eavesdrop was sparked by the denial of two FISA-Court warrant applications by the FBI seeking permission to wiretap Trump associates.</li><li>GCHQ didn't work from the UK but from NSA’s headquarters in Fort Meade, MD, under the direct supervision of our NSA.</li><li>The illegal wiretaps were initiated months before the controversial Trump dossier compiled by former British spy Christopher Steele.</li><li>Obama's Justice Department and FBI set up the June 2016 meeting at Trump Tower between Trump Jr., Manafort and Kushner with Russian attorney Natalia Veselnitskaya to make it appear that Trump’s associates were working with the Russians.</li><li>Following that meeting GCHQ began wiretapping Manafort, Trump Jr., and Kushner.</li><li>Russian attorney Natalia Veselnitskaya had been considered such a security risk that before the meeting she'd been banned from entering the U.S. But just before the meeting U.S. attorney Preet Bharara personally cleared her to enter the U.S. His office hasn't explained this reversal.</li><li>It is illegal for U.S. agencies to intercept phone conversations and emails of U.S. citizens inside the United States without a warrant. They tried FISA but were turned down.</li><li>Witnesses have told the House committee that the FBI paid Christopher Steele, who compiled the anti-Trump dossier, at least $100,000. Significantly, when the House committee asked FBI director Wray and assistant director McCabe to confirm this claim, both refused to answer.</li></ul><div dir="auto">At the risk of pointing out what should be obvious to anyone with an IQ over 80, if the claim was false, don't you think Wray and McCabe would have said that?<br /><br />Veselnitskaya had been barred from entering the United Sates due to her alleged connections to the Russian FSB (the modern replacement of the cold-war-era KGB). But just days before the June meeting Veselnitskaya was given a visa to enter the United States from Preet Bharara, who at the time was the U.S. Attorney for the southern district of New York. </div><br />Federal law enforcement sources said Bharara was following the orders of Obama's Attorney-general Loretta Lynch. As far as is known, the only purpose of Veselnitskaya's entry into the United States was the meeting with Trump associates, federal sources said.<br /><br />The illegal wiretaps started long before Steele’s dossier. Federal sources said the wiretaps on Trump insiders began in late 2015, almost a year before the 2016 election. The first targets were Flynn and Carter Page, sources confirmed. When no smoking gun was recovered from those initial taps, U.S. intelligence agencies moved to broaden the scope through their newly-formed alliance.<br /><br /> Intelligence garnered from the British eavesdropping, which again was merely a front for the NSA, was then used in August 2016 to secure a legitimate FISA warrant on Manafort, Trump Jr. and Kushner. That warrant was issued on or about September, 2016, federal sources confirm.<br /> <br />It was the third time the cabal of U.S. intelligence agencies sought a FISA warrant for the Trump associates and this time it was approved--presumably because the agencies were now able to claim that "independent" evidence had been acquired that showed Trump associates conspiring with a foreign government.<br /> <br />FBI sources said finally obtaining the FISA warrant was important because it provided the agencies cover for previous illegal wiretapping. Of course they knew that if Hilliary won, this would never be discovered. But in the astronomically unlikely event Trump won, it could be. Thus they needed an "insurance policy" to ensure this didn't happen.<br /><br />Virtually all the Lying Media is avoiding this story like the plague. <a href="http://www.bostonherald.com/news/columnists/howie_carr/2018/01/carr_clinton_s_dodgy_dossier_is_way_worse_than_watergate">Except for Howie Carr at the Boston Herald,</a> who tells us this:<br /><br /> The meeting between Natalia Veselnitskaya and the Trump reps occurred on June 9, 2016.<br /> Guess who Veselnitskaya had dinner with on June 8, 2016?<br /> Answer: By an astonishing coincidence, Glenn Simpson--the guy who was paid $1 million to produce the anti-Trump dossier.<br /> And guess who Veselnitskaya had dinner with on June 10, 2016--the day after the meeting?<br /> By double-astonishing coincidence, it was...Glenn Simpson again.<br /> When a GOP staffer asked Simpson about those amazing coincidences, Simpson replied that he had no idea the Russian attorney had met with the Trumps until he read it in the papers.<br /> He really said that. Just coincidence.<br /> You can’t make this stuff up — unlike the dodgy dossier, which they could, and did.<br /><br /><br />Just coincidence, citizen. Just like the "accidental" meeting between Bill Clinton and Obama's corrupt attorney-general Loretta Lynch, in her government jet on the parking ramp in Phoenix<br /><br />A meeting the FBI initially denied had occurred, then tried to suppress all emails about their preparation for. Which itself shows the obvious lie about an "accidental meeting."<br /><br />Just a coincidence, like the murder of DNC tech staffer Seth Rich, which the FBI and cops flatly stated was a "botched robbery," despite fact that the so-called robbers didn't take his wallet, jewelry, watch or cell phone. Nothing taken at all. (The coincidence is that Rich was killed just after the DNC claimed their computer was "hacked by the Russian government," and Rich would have been the point man to question about that.) <br /><br />And it's just a coincidence that all the security cameras on the intersection near where Rich was killed have vanished.<br /><br />It's also just a coincidence that the DC cops claimed that the body cams on every DC cop who responded to the scene malfunctioned.<br /><br />Also a coincidence that the DNC refused to allow the FBI's cybercrime experts to examine their allegedly-hacked server for evidence of actual, y'know, <i>hacking.</i> <br /><br />Ask yourself: Who benefitted from all these coincidences?<br /><br />H/T <a href="https://truepundit.com/exclusive-six-u-s-agencies-conspired-to-illegally-wiretap-trump-british-intel-used-as-front-to-spy-on-campaign-for-nsa/">Truepundit.</a> And the <a href="https://theconservativetreehouse.com/2018/01/11/the-doj-and-fbi-worked-with-fusion-gps-on-operation-trump/">Conservative Treehouse</a>.<br />SFhttp://www.blogger.com/profile/[email protected]:blogger.com,1999:blog-8145959.post-89602973722132746242018-01-11T11:06:00.004-08:002018-01-11T11:06:40.089-08:00"Illegal aliens are no more "immigrants" than...Illegal aliens are no more "immigrants" than burglars are houseguests.SFhttp://www.blogger.com/profile/[email protected]:blogger.com,1999:blog-8145959.post-64755882975351041492018-01-11T09:43:00.000-08:002018-01-11T09:43:04.727-08:00How can the Left and Antifa logically support unlimited Muslim immigration?<br />Islamic jihadists have openly vowed to destroy Western civilization. <a href="http://www.americanthinker.com/articles/2018/01/leftists_and_islamists_strange_bedfellows.html">What may seem a bit surprising is that the Left--Marxists, anarchist and self-labelled "antifascists" (actually radical leftists) support the jihadists in this goal.</a><br /><br />The mystery resolves itself when you understand that the radical Left never planned to conquer the west by force of arms, but by winning control of the ballot box. Once a leftist party was elected to run a country, no one could object when the Left started imposing firmer controls on everything. After all, they'd won the election, eh? Democracy rules, comrade.<br /><br />Once the former communist states of eastern Europe overthrew their dictators and began to take the first steps toward a free market, the Left realized it would be far harder to regain control by simply touting the marvelous advantages of communism. With a couple of rare exceptions (Angela Merkel), communists realized they weren't going to win many more elections.<br /><br />Muslim leaders immediately recognized Leftists as natural allies, since Leftist politicians--eager for new votes--would push for open borders and unlimited welfare for new "arrivals." At the same time the Left's media allies would demonize anyone who objected as "raacist" or xenophobic. A key part of this tactic was to discredit any objectors as "anti-immigrant"--totally false but extremely effective in silencing critics.<br /><br />The Left wasn't worried by the fact that Islam had openly vowed to destroy the West. Since leftists are comfortable with the idea of saying anything--no matter how absurd--to win their goal, they would naturally assume the tale of "we will turn the entire world into an Islamic caliphate" was simply propaganda spouted by the leaders to keep ordinary jihadis fired up. <br /><br />Moreover, because the Left has firmly welded itself to the notion that "multiculturalim" is absolutely wonderful, importing a few million Muslims was proof young Leftists of the virtue of their leaders. Supporting unlimited Muslim immigration became a virtue signal to the elites.<br /><br />The Left's worship of multiculturalism--and simultaneous contempt for Christianity--enabled it to totally ignore the fact that ISIS routinely stoned gays, endorsed gender apartheid, and exhorted followers to rape and murder non-Muslims. Not a problem to Leftists--can't make an omelette without breaking eggs, eh comrade?<br /><br />Since the Left abhors religion, you'd think the extreme devotion Muslims have to Islam would be a concern to the Left, but it seems not to bother them--possibly because they're familiar with the tolerant religion of Christianity and believe all religions are similar. The Left is perfectly happy to ignore that when the communists and Islamists joined to overturn the shah of Iran, after the Shah fled, the mullahs eliminated the communists.<br /><br />But of course to the Left, history holds no lessons that Marx and Lenin didn't already explain.<br /><br />Wow.SFhttp://www.blogger.com/profile/[email protected][SEP]
| 3
| 18,000
| 23,378
| 23,378
|
66,949
|
ebb61997-0e4a-48f0-96fd-5023a0105129
|
trentmkelly/LessWrong-43k
|
Sledding Among Hazards
Some sledding hills are very open, with basically no hazards. As long as you start in a reasonable place there's nothing to run into.
Robbins Farm Park, Arlington MA
Other hills are not like this:
Powder House Park, Somerville MA
Since sleds go fast and collisions are bad, choosing only to sled on obstacle-free hills would be reasonable. But if your kids are going to sled on hills with hazards, what precautions make sense? Here's what we did today:
* Bike helmets. Head injuries are especially worrying because of how bad the long term consequences can be, and concussions are cumulative.
* Practice bailing out. Before I let them go down on the tube, which they can't steer well, I had them each practice rolling off it to stop.
* Feet first. More crumple zone if you do hit something, more ability to see and avoid things. If you're on the tube and you end up backwards, preemptively bail.
This is in addition to the precautions you need on any hill, like being aware of activity uphill (so you can get out of the way) and walking up on the side (so no one will come down at you).
I am a bit conflicted on when sledding on constricted hills is worth it. For us, the slopes near our house do have obstacles, but:
* We almost always see friends there.
* Doing things within walking distance is really nice; we'd do it much less often if it meant dealing with the car.
* Driving on snowy days can be risky depending on the road conditions.
* At the open ones maybe there's a bit more risk of being hit by another sledder, since people are less careful and less practiced at bailing?
While I think these are enough to outweigh the risks, with the precautions above, possibly I'm not being cautious enough?
Comment via: facebook, mastodon
| 0
| 0
| 431
| 431
|
85,185
|
<urn:uuid:3461fd11-283a-4ac6-900c-9778a056df37>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Persecution concerns mount as China’s surveillance state genetically maps entire male population
Date: June 22, 2020
china blood samples 4x3 653x490
Chinese police were tasked to collect mandatory blood samples from up to 70 million – 10% of all men and boys – as the high-tech surveillance state builds a “DNA Skynet” of the country’s 700 million males, raising fresh persecution concerns for Christians and other marginalised groups.
The Chinese government is building a powerful genetic tracking system through compulsory DNA tests, which will combine with the country's existing video surveillance network [Image credit: CGTN]
The powerful genetic mapping tool, compiled from samples collected since 2017, will join up with China’s existing countrywide video-surveillance network and be capable of tracking every male relative of any individual man from a sample of his genetic material. Only 5-10% of the male population needs to be sampled in order to build a complete genetic map, because one sample unlocks the identity of all related males.
The “Skynet” surveillance system was denounced in a recent report by the United States Commission on International Religious Freedom (USCIRF) for targeting minorities and “amplifying the repression” of religious communities.
Blood testing has been rolled out in China's schools and workplaces to capture generic information from around 70 million men and boys
Despite USCIRF ranking China as a “country of particular concern” in its annual report, it was US company, Thermo Fisher, that supplied Chinese police with tailor-made genetic testing kits for the vast programme. One testing kit designed by the company identifies genetic markers used to distinguish between China’s ethnic groups, including Uighurs and Tibetans.
The Chinese government already holds the largest DNA archive in the world comprising of some 80 million profiles, according to official state media, including genetic data from Uighurs and other ethnic minorities considered to be “potentially destabilising”.
Government surveillance systems rolled out since 2015 use sophisticated artificial intelligence and biometric technologies to target specific groups and minorities. The advanced facial recognition system deployed on the network is believed to be capable of distinguishing between the faces of different ethnic groups to single out Uyghurs, Tibetans and other minorities. Authorities have also collected fingerprints, voiceprints and other biometric data from religious communities without their consent.
Any individual who refuses to give blood samples risks their family being listed as a “black household”, according to one computer engineer from northern China. This would mean he and his relatives would be barred from benefits such as travel and hospital treatment.
The move to combine video surveillance with big data has been described as a “major escalation” of the Chinese Communist Party’s efforts to control citizens. Human rights experts have warned that the ability to easily identify close relatives could be used for social control purposes such as whole families being punished for the actions or dissenting beliefs of one individual.
The continuing nationwide crackdown on the Church by authorities in China has seen hundreds of “house churches” (i.e. unregistered congregations) and state-sanctioned “three-self” churches forcibly shut down. Increasingly repressive measures have included violent police raids, thousands of arrests and detentions, imprisonment of pastors and forced installation of surveillance camera inside some churches. Under sinicisation (making Chinese) rules introduced since 2018, religious symbols and Scriptural texts have been removed from display by authorities and replaced with the Chinese flag, portraits of president Xi Jinping and Chinese Communist Party slogans.
Secure Donation
with PayPal
Pray for the Unreached
| 0
| 0
| 759
| 759
|
106,424
|
<urn:uuid:2b3ec8f7-d739-495b-9b05-f7321d961b89>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
All agencies will try and fight over the core comms planning. That's the way it is......and can be healthy if everyone listens to each other.</li> <li>Creative agencies can't do media. They just can't. If you want to do media, start a media department.</li> <li>Media agencies can't do creative...but think they can.</li> <li>Digital agencies will never be unbiased, only let them do digital.</li> <li>Media owners can't do ads, yet. They think they can, but do not them do comms strategy or creative strategy.......just brief them right based on a good idea.</li> <li>The term 'native' shall henceforth be;s a bloody advertorial.</li> <li>Obey the rules</li> </ol> The day job Northern Planner Thu, 14 Jan 2016 13:07:57 +0000 How to train junior planners OK, so after the 'how not to train junior planners' here's some thoughts on how to go about it. Although, I guess much of it has been covered by pointers in the last post. Still, worth a positive outlook and... <p>OK, so after the 'how not to train junior planners' here's some thoughts on how to go about it.</p> <p>Although, I guess much of it has been covered by pointers in the last post.</p> <p>Still, worth a positive outlook and some general pointers.</p> <p> </p> <p>Much of it is based on some simple principles:</p> <p>First, great planners are interested in much more than planning or advertising or even brands.We're interrupting what people care about, or these days, finding a way to be part of it, so the more we care about THAT the better. </p> <p>Second, no two planners are the same, nor should they be.</p> <p>Third, happy people produce better results and stay far longer.</p> <p>Fourth, planners without evidence based work are just suits with more opinions.</p> <p>Fifth, creative planners are losing their role as 'lead planner' with many clients.</p> <p>Finally, it's hard enough being a senior planner type when you have to earn your right in the room, for a junior, it's Everest.</p> <p> </p> <p>1. Great Planners are interested in much more than planning</p> <p>This means you need to give them both the space and the encouragement. Agencies can work long hours, but when you are a junior, it's easy to think you need to stay late every single night, or get in at the crack of dawn. Some places make it known that this is encouraged, but all you get are tired and stupid people who are nowhere near their best.</p> <p>What is worse, if they spend all their waking hours talking about advertising, they'll only plan for 'advertising' rather than planning for making real people care. Rather than pulling stimulus from the world at large and being able to inspire the team around what really interests people, to make advertising register, they'll just start doing 'advertising'.</p> <p>Even more dire, they'll start thinking it's a desk job and not go out and meet the people they're supposed to making strategy for - real human beings. As has been said a billion times, research is a waste of time unless you observe people in their real environment. If you want to understand a species, go to the jungle. not the zoo.</p> <p>So apart from sending folks out of the agency and protecting them from a 'presenteeism' environment, here are two stimulating things you can do:</p> <p>First, give every junior planner (in fact every planner) a 100 day task every single year, where they have to go and find out something new and interesting about an important target audience for your clients. To focus the minds, they have to present to a senior team and even the client if you have that relationship. The evidence can only be what you have discovered from 'non-advertising' tools, no TGI, no WARC, no NVision. And they need to have some video of actually talking to people.</p> <p>Second, get an 'interesting fund' set up, where everyone in your department gets £200 to invest in learning something new and interesting, or enriching a private passion. It could be learning to play the guitar, doing a video production course. They just have to write why they are choosing what they are choosing -  and share what they're learned and what it's taught then about the job at the end of the year. </p> <p>2. No two planners are the same.</p> <p>In the previous post, we discussed the wrongness of the one size fits all planning approach. The problem with a proprietary process is that it makes you approach every brief the same way, and usually come out with the same solution. It's why every campaign out of Chiat Day looks like an Apple campaign and always has a manifesto in it somewhere. Just as doing the same workshop over and over again yields the same kind of idea.</p> <p>Boundaries and guidelines are good. Fixed rules are not.</p> <p>But it's much more than that. There are core planning skills of course, but even these are debatable. <a href="" target="_self">This is a good starting point.</a></p> <p>Nowadays, the skills of a planner are too broad to be contained in one human being. Generally, you need to know how most stuff works and be able to knit it together into something more cohesive, but 'most stuff' is getting a very large requirement. Getting to grips with data, the rapidly changing media environment, digital and social media, the need for content as much as ads.</p> <p>You need a team that's good at different stuff and can do things different ways.</p> <p>Someone ace at data, I mean the really hard stuff, not just TGI.</p> <p>Someone who lives and breathes digital (but get's it's place, double so for social media).</p> <p>A good comms planner is a must these days.</p> <p>But then someone who has the rare instinct for retail too. If you think retail is easy, try writing a brief that stimulates people into the saying the word 'sale' a different way, or someone who can handle a sales force.</p> <p>You might need someone who get's healthcare, B2B, maybe someone who get's international brand planning.</p> <p>In short, you need to recruit for filling gaps, not for carbon copies.</p> <p>Especially folks who are good at what you are not.</p> <p>They need to be assigned accounts that suits their skills and interests - and some that do not.</p> <p>So they get to do what they're good at, plus enrich their knowledge and skills base. Give them enough space to learn from mistakes and issues, with enough support and 'open door' sensibilities so they feel they can come for help (and not be made to feel stupid).</p> <p>As mentioned before, making them do secondments in other departments really works here.</p> <p>All of the above kind of goes towards happy people staying longer and producing results by the way, which was point 3.</p> <p> 4. Planners without evidence or just suits with more opinions.</p> <p>There's a vogue for 'ideas' rather than insight. I don't mean some sort huge revelation, which are hard to come by, but evidence based thinking that unlocks other people's skills.</p> <p>My view on this stuff is that our task is to get to a really great task, a jumping off point for everyone based on: brand/product/market/customer/audience/media/shopping or customer journey/culture. Our task is to knit these sources of input into once clear task for communications that everyone can get behind. If you miss one of these, you haven't done your job, and if any of it doesn't have some evidence, you haven't done your job.</p> <p>The core observation might have some sort of focus or emphasis.</p> <p>For example, this was all based on a shopper insight that women bought lots of shower gel for men:</p> <p><iframe allowfullscreen="" frameborder="0" height="281" src="" width="500"></iframe> </p> <p>This campaign was based on a cultural insight that the Scots are very optimistic supporters who deal with defeat and victory really well, supported by all sorts videos from sporting events and street interviews..</p> <p><iframe allowfullscreen="" frameborder="0" height="281" src="" width="500"></iframe> </p> <p>With a special bookend after the games, based on the pyschology insight that we remember the end of things best, and stories of how movie companies film and research the ending scenes to death.</p> <p><iframe allowfullscreen="" frameborder="0" height="281" src="" width="500"></iframe> </p> <p>Anyway. For junior planners, I think we need to get back to training them to be very good researchers. Planners used to be focus group moderators who then turned the findings into usable hooks for strategy and creative work.</p> <p>We rarely have budget to moderate groups for development (and I can safely say most of us know it's a waste of time) but now we have all that data freely available from the internet, trends stuff coming our of our ears, all sorts of market data, Mintel reports and, if you are lucky enough, TGI and Touchpoints. A really great planners will look at all that stuff and connect it in a way someone else won't.</p> <p>Or they'll just go out an meet their audience and talk to them enough to understand the business issue in the context of real life...which is always the real competition for the brand. Just get some proof video, quotes and try and quantify it in some way.</p> <p>I think our job is to teach junior people these skills (and some bloody senior people) and as a leader of a department, hold everyone to account.</p> <p>And train your team to connect things. There are not really new ideas, just new re-combinations of old ones. I think that means teaching them to mind-map, teaching the unfashionable are of distillation and....</p> <p>Make sure you have some sort of scrapbook initiative. Some of that is taken care by the 100 day projects and interesting funds from above, but it's worth having a scrapbook initiative. A vault of interesting stuff for everyone to draw on. I suggest a team TUMBLR, perhaps with a different editor every month. But the trick is to get the team to all contribute - as long as your hiring people with different skills and interests (you should be).</p> <p>For me,  all of the above will really help that 'lead planner' issue. It's fair game for any agency now and I feel that the planner that:</p> <p>Has the most interesting things to say, the one people want in the room, will be that planner by default.</p> <p>That doesn't mean they talk the most - in fact, they should be taught to keep their mouths shut, listen to everyone else and speak last (see IRN BRU thankyou ad insight).</p> <p>They should be taught to plan for meetings, to have something evidence based to say that will make everyone think, create a discussion point and shed light on the agenda.</p> <p>Like Gordon Gecko says here:</p> <p><iframe allowfullscreen="" frameborder="0" height="281" src="" width="500"></iframe> </p> <p>They should be taught to always know more about client business, target customer and relevant culture than any other supplier and, in the case of customer and culture, anyone in the client business.</p> <p>These are more important than 'ideas'.</p> <p>Catalyst.</p> <p>Simplifier.</p> <p>Provocateur.</p> <p>And, dare I say, the person that brings real life, the lives of real customers and culture into the room.</p> <p>Which brings me to a final point.</p> <p>The more complex your language, the more people think you are an idiot. The best planners speak human. The task of a leader is not tie their team up in knots with needless jargon and buzz words.</p> <p>It is to set an example by speaking plainly and bollocking anyone in your team if they don't do the same.</p> <p>Hope this helps.</p> <p>(very busy, this will be riddled with dreadful typos, sorry, no time to check overly!!)</p> <p> </p> <p> </p> <p> </p> <p> </p> The day job Northern Planner Fri, 30 Oct 2015 16:09:30 +0000 How not to train junior planners I got an intriguing email from someone recently asking if I knew of any posts from myself, or other planner types, about training junior planners. I have to admit I did not. Which, after thinking about it for a bit,... <p>I got an intriguing email from someone recently asking if I knew of any posts from myself, or other planner types, about training junior planners.</p> <p>I have to admit I did not.</p> <p>Which, after thinking about it for a bit, was a little shocking.</p> <p>There has never been more competition, between the various agencies that make up 'adland',to hire and keep talent than there is now.</p> <p>This industry just isn't as attractive next to other career paths as maybe is once was. It's much less cool, pays less relatively, has less career stability and works long hours.</p> <p>So, despite the prolific chatter all over the interwebs on 21st Century Branding, the Death of TV, the death of ad agencies, it's a real shame there isn't decent content on the hiring and development of great talent.</p> <p>It's a focal point on much wider, real issues around the hours we ask people to do, how we charge for our ideas as opposed to our time and the general rule that if you want people to do well, stay and give you their best, you have to treat them properly and balance their quality of life and career path with the very real need to make money.</p> <p>I replied with a few pointers, but I thought I might do a bit more here. </p> <p>I'll start today with the easier one, what not to do. This is of course based on personal experience as someone who was a junior once, but also observations gleaned from the experience of others.</p> <p>So, how not to train junior planners (and to some applies to junior folks in agencies per se):</p> <p>1. Don't try and make them have YOUR idea.</p> <p>The problem with many planning directors is that they are great at thinking and having ideas, but rubbish at bringing it out of others. This can be seen in the way they interact with other departments, they are the ones who stick to their strategy or proposition and don't like it being changed- as opposed to the ones who are great at generating and spotting great ideas from others ( have to admit I'm the latter, mostly because I don't have many good ideas and it's easier to get others to do it for you!!).</p> <p>So it follows that when they let juniors have a go at a project, they won't be able to see the merits of the solution the junior comes back with, because they will already have something in mind themselves, so they'll belittle the poor bugger about what might be wrong with their thinking and evidence, rather than looking at what's good and what can be developed.</p> <p>Even if it's bollocks, you need to find a way to praise their effort, show them why it's wrong and empower them to find another solution - rather than call them stupid and say, 'what you should have done it this'. There is no ONE solution, just like with economists who cannot agree on anything, there is more than one approach. Listen, help, encourage and guide. So...............</p> <p>2. Don't expect them to work at the velocity as you.</p> <p>So what follows should be simple, allow for the fact they'll be a little slower than you at stuff and when you work on a project together, you need to make time to talk them through things, explain things a little more and stuff.</p> <p>I learned the hard way that there is no point telling my kids 20 seconds before we're leaving the house that it's time to go, then getting grumpy when take an age to find their shoes, pick a toy to take or go to the toilet, their agenda is not mine. It's like that working with juniors - you're dead busy, you could do it twice as quick, but if you don't make time to help people learn, they'll never get better and be stuck thinking nothing they do is good enough.</p> <p>3. But don't be a light touch.</p> <p>Once, planners were allowed to be late for things, to get lost on the way to meetings and generally be a little air headed.</p> <p>No one has the time and patience for this any more. Because as they progress, they'll have to deal with more of a blur between 'suit' and 'planner' and do more things that people in real jobs have to think about.</p> <p>Only let them get away with  being late, lost or forgetful once. That goes for rigour too. There's a trend for planners to have 'ideas' these days rather than evidence based strategy based on proper examination of the information at hand.</p> <p>It is critical to have evidence based thinking, otherwise you are just someone with opinions, even more these days with Lord knows how many people thinking they can own the strategy: the creative agency, digital, media, media owners, brand consultant and whoever else are all trying to own the lead strategy and even lead creative.</p> <p>The only defense for a good working planning team and its agency is to be able to back up their ideas better than anyone else, and be able to de-stabilise first page, slap dash stuff from other folks. I'm not saying every idea should be backed up with a flash of amazing consumer insight (but I weep at the trendies who seem to think this doesn't matter anymore) but there is so much to be gained from looking at TGI harder, reading the clients' annual reports, or just bloody going out and talking to people who work for your client or real people on the street.</p> <p>This goes for you too by the way, it's easy to fall into the 'because I say it is' camp when you get senior, because you can force your thinking through, which will come unstuck eventually, getting your team to back up their thinking encourages you to do the same.</p> <p>4. Don't moan about the good old days.</p> <p>This might manifest itself as 'the good old days when I used to work 16 hour days as a junior and spend 5 years not being allowed to go to a client meeting'- in order to rinse your people for every drop of energy they have.</p> <p>Those days are, if not gone, they are fast going and there is no point burning your people out, and getting them to change to another career because they've had enough. Likewise, the good old days when you could charge a fortune, you made loads of TV ads, clients took more risks, bought more better work etc.</p> <p>No one needs to hear they joined the industry too late, you're just encouraging them to do something else. And it really wasn't that much better was it? Rather, help them embrace the limitless possibility of mixed up media - and the fact that few are able simplify and make it something clients can embrace gladly.</p> <p>They know more about digital than you do, they're on Snapchat and you are not (I hope) - embrace the future with them and stop boring them with a past that wasn't as rosy as you now like to believe these days.</p> <p>5. Don't try and create cardboard cut-outs of yourself.</p> <p>You will be really great, really experience dand able to apply it to all sorts of varies briefs and projects. You will have learned to overcome weaknesses and build on natural strengths.</p> <p>But your are a one off, the product of a mixture of genes and experiences. Your are a one-off - and so are the people who work for you. They will have different in-built strengths and weaknesses, so making them work in only one way -  be that developing workshops, the times of day they best work, getting respect from other departments, the style they write briefs in, how they apply research, how they present - in other words, your way, is doomed to fail.</p> <p>Figure out what, in your own arsenal are universals anyone should know and practice (always sit in the middle of the table to get gravitas in a meeting, don't get excited about a high index on TGI until you look at the actual percentage of the audience, speak last in any review of any kind of work if you can) and then what should be a library of approaches for your team to try and see what works and what doesn't.</p> <p>That also goes for presentation decks. They are the background and 'props' for the speaker, no more no less. Insist on any agency template if you like, but apart from that, it depends on how someone presents - naturally 'tell them what you're going to tell them, tell them and then tell them what you've told them' 'find a theme' are universals.</p> <p>But applying some storytelling structure etc, how that works depends on the person.</p> <p>Just as some planning folks don't sweat a proposition (thoughtleader or core thought) in a brief or client strategy presentation, and work on a really juicy task, or even a transformational insight instead (but every brief or deck should have one focal jumping point, everything else builds up to or from this point, but then again that might be how I work!!!).</p> <p>6. Don't be a parrot for the proprietary process.</p> <p>Look, every agency in various disciplines sort of works the same. They flounder around for a bit, they get worried about the deadline, something comes up and then they work like mad to be ready for the deadline.</p> <p>They hide this from clients as sell a process that gives the comfort of making things look professional and predictable (and procurement loves to buy a process). This benign conspiracy sort of works, as long as you remember it's a conspiracy and the process is really a load of bollocks.</p> <p>So forcing your people to follow a process just doesn't work, especially as, as mentioned before, they need to find the best way to work efficiently in their own way. Make everyone work the same and you get the same stuff. Just as every brief isn't about 'disrupting the market or zigging where others zag'.</p> <p>At the basics of communications strategy, there is only really 'impacting with the audience' 'activating people to do stuff' 'reinforcing how they feel, or what they know, about something, or 'Augmenting' - changing how they feel or what they know. But that's four, not one and it changes depending on the brief.</p> <p>Your team need the freedom to explore these four ways, then understand how their thinking can be made to fit the 'planning model' or how the client thinks communications works - freedom, guidance. Not suffocation and constriction.</p> <p>7. Don't over-protect them</p> <p>I learned when I was a competitive swimmer that no amount of training can prepare you for real racing. The pressure, the way your body copes with adrenalin, how you respond when the pain kicks in and only willpower can help you carry for the final metres. You can only get better at racing by racing.</p> <p>It's the same with client meetings, dealing with grumpy creatives, scary TV buyers, doing presentations or even the moment when the pitch team has a melt-down when they can't seem to crack the brief.</p> <p>Gradual exposure, starting as soon as possible is the only way to get good at this stuff. The first cut is the deepest, but until you are able to get used to things, learn from mistakes and get used to the realities and pressures of the real job, you are not really doing the job.</p> <p>Great thinking and insight is only 20% of it. Being able to persuade others of your thinking, internally and externally, empowering others, thinking on your feet and, critically, being able to deliver solid work time after time, being a safe pair or hands rather than a 'either brilliant of dire' planners are where the job is really done.</p> <p>So honesty about where ideas come from, the fact they do not appear mostly as if by magic, they emerge and are developed by lots of reading, hard work, edit, precis and distillation and, also, having the dignity and generosity to include as many people as you can in the strategy, as they might strike lucky instead of you, which also means having the courage to admit when someone has some better thoughts and when you are totally wrong.</p> <p>But also knowing when to be firm, when to let people down gently, to prove them wrong but leave them smiling, to be able to stand your ground without being obstinate.</p> <p>In other words, always being the bigger person,</p> <p> </p> The day job Northern Planner Wed, 02 Sep 2015 13:41:45 +0100 Big Data v Wisdom "Not everything that can be counted counts and not everything that counts can be counted" Einstein A lot has been promised by Big Data. Mega things. Most of it seems like over excited over claim, but probably, sooner or later,... <p>"Not everything that can be counted counts and not everything that counts can be counted"</p> <p>Einstein</p> <p><a class="asset-img-link" href="" style="display: inline;"><img alt="Data" border="0" class="asset asset-image at-xid-6a00d83451bf9f69e201b7c7befd1d970b image-full img-responsive" src="" title="Data" /></a></p> <p>A lot has been promised by Big Data. Mega things. Most of it seems like over excited over claim, but probably, sooner or later, it might deliver.</p> <p>There are examples today of how it can be good. The US House of Cards reboot is thanks to the number of people liking Kevin Spacey content also liking political drama.</p> <p>But  Big Data is also responsible for online retailers stalking me with re-targeted ads of things I have already bought. It's meant that when I buy stuff for my wife and kids, I can't move for ads selling me more of the same.</p> <p>When any idiot human could intuit that I really don't want an avalanche of offers for women'srunning gear, Zara shoes or Star Wars Lego (okay, I'll give you Star Wars Lego).</p> <p><a href="" target="_blank">Of course there is a great example of a US supermarket that got publicly roasted by a father, furious that his teenage daughter had been sent stuff for maternity, only for the father to apologise when he found she actually WAS pregnant and was too scared to tell him. </a></p> <p>it can right, yet horribly wrong.</p> <p>On the subject of supermarkets, big data has been around for a while, Tesco is the pioneer with its Clubcard, the value of knowing intimate shopping habits was priceless for them.</p> <p>But Tesco in the last few years, if you excuse the expression, is fucked. It forget to look at what customers were really bothered about, what they cared about. In post recession UK, shoppers want a lot more simplicity, less hassle and it's uncool to waste things. The avalanche of short term offers, multi-buys and the like turned shoppers off looking for a trusted price, looking for great quality in less stuff, rather than pointless choice and fleeting discounts.</p> <p>All the while, we started shopping for little and often. Many started looking for intimacy and the feeling of real care and attention, in some things anyway. While in others, they just wanted simple no-frills functionality. You could say that Tesco, with all its data, got squeezed between people wanting more stuff from a good butcher AND the simplicity of a discount grocer like Aldi.</p> <p>Yet you can't move in marketing circles, and business in general, for data scientists. Like the web developers before them, or the social media gurus of today- and the brand consultants that still manage to sell snow to eskimos, these folks are the latest thing.</p> <p>But let's not be too harsh. The central premise of data is still sound. It makes sellers wiser when selling to potential buyers and, when done right, adds value to buyers buy not wasting their time with things they have no interest in.</p> <p>Now I know the arguments from Byron Sharpe about light buyers and targeting the whole market. Even today, ignore the arguments that mass broadcast media doesn't work, even with young folks. It does if it is done well.</p> <p>But big organisations are still very dumb when it comes to their customers. They are numbers, not names. Perversely, the web has created the death of the human and the personal.</p> <p>There is lots of talk of 'personalisation' but that is not the same as intimacy.</p> <p>Old fashioned shopkeepers, who are now in vogue to some degree, were the pioneers of big data. They remembered what their customers liked, they recognised them when they came in. The fishmonger in Leeds market always kept mackeral aside for my Grandmother on a Thursday morning. Today, my local butcher knows when I walk in the door that I buy a mountain of his thick sausages and gets them out without asking. He knows a BBQ in summer and talks to me about new marinade ideas and stuff I haven't tried.</p> <p>Some of this emotional intimacy can be delivered by the power of great brand building. Nike feels different to Adidas, it just does.That's why I sometimes don't believe the data about people not being able to say why a brand is different, it's like trying to do a questionnaire about why you love your children, it's a smudgy feeling that you can't always express.You remember how the brand feels when you're in buying mode, yes it comes to mind, but so does the emotional resonance.</p> <p>But we can do better than that. Brands should be able to understand its customers better. Much of the personal, CIM marketing is a waste of time of course, working with heavy buyers who would buy anyway, but data should help us work out ripples of behaviour on a much larger scale.</p> <p>A sports brand should know that loads of it's football buyers also love not just comedy, but what kind of comedy, what comedians. They could then set up a multi-platform football comedy show where their favoured comedians banter around footie.</p> <p>An FMCG salad dressing company should know that people who like the brand but don't buy often also love a grilled chicken and do recipe campaigns with their favourite celebrities for using the dressing with chicken too.</p> <p>Because ads in Facebook trying to sell me slippers are really not good enough.</p> <p>But data is only a tool. I cannot replace imagination, emotional intelligence and intuition. It cannot produce the consistent ideas that recombine old ones. </p> <p>It would tell Steve Jobs not to launch the Iphone.</p> <p>It would tell Henry Ford people preferred horse.</p> <p>Put another way, numbers can help us make sense of the world but, today at least, they cannot replace wisdom.</p> The day job Northern Planner Fri, 14 Aug 2015 13:13:42 +0100 Why Lego tells us that innovation is more perspiration than inspiration Where I work was responsible for this.. It's actually a simple idea. Not so simple to pull off. It took lots and lots of hard work. That's the truth about innovation, new ideas or general stuff that isn't the usual... <p>Where I work was responsible for this..</p> <p><iframe allowfullscreen="" frameborder="0" height="281" src="" width="500"></iframe> </p> <p>It's actually a simple idea.</p> <p>Not so simple to pull off.</p> <p>It took lots and lots of hard work.</p> <p>That's the truth about innovation, new ideas or general stuff that isn't the usual or expected.</p> <p>Having ideas isn't easy of course, but it isn't the toughest bit.</p> <p>The toughest bit is getting them to ever see the light of day.</p> <p>New ideas tend to look like hard work.</p> <p>To account handling types who have to get the stuff made, and persuade the client.</p> <p>To clients who have to sell in plans and stuff to commercially focused people who don't like surprises.</p> <p>Clients for whom advertising and stuff is about 10% of their entire job.</p> <p>Clients who want their lives simplified and who live in quarters of years.</p> <p>So how do you get stuff like the Lego ad break
| 2
| 12,000
| 20,000
| 29,059
|
62,765
|
bb4dce50-1186-4410-9087-676a3137b8c6
|
trentmkelly/LessWrong-43k
|
Refine's First Blog Post Day
Yesterday was the first blog post day at Refine. It came after a week of readings, discussions and exercises about epistemology (more concretely shapes of mind, productive mistakes, epistemological vigilance, and mosaics and palimpsests).
I'm excited about how the incubator is going, and about each of these posts.
Here is the list of blog posts by Refine participants:
* How I think about alignment
* the Insulated Goal-Program idea and goal-program bricks
* Steelmining via Analogy
* I missed the crux of the alignment problem the whole time
* All the posts I will never write
Here's the list of blog posts by friends of Refine and members of my Epistemology Team at Conjecture:
* Shapes of Mind and Pluralism in Alignment
* The Dumbest Possible Gets There First
* An extended rocket alignment analogy
| 0
| 0
| 206
| 206
|
82,520
|
<urn:uuid:a387f517-5797-41b0-b1b3-20b0cd5ad2dc>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
From Wikipedia, the free encyclopedia
Jump to navigation Jump to search
Battlestar Galactica (TV series)[edit]
Series 1 (1978-79)[edit]
The following table lists the episodes of the original Battlestar Galactica.
Title Directed by Written by Original air date PC
"Saga of a Star World"Richard A. CollaGlen A. LarsonSeptember 17, 1978 (1978-09-17)50280
In a distant part of the cosmos, the human race is embroiled in a thousand-year-long war with the robotic Cylons. Aided by the human traitor Baltar, an armistice turns out to be a ruse by the Cylons, who launch a massive surprise attack against the Twelve Colonies of humanity, almost wiping them out. Only the Galactica, led by Commander Adama, and a ragtag fleet of 220 civilian ships survive, beginning a journey across the galaxy in search of long-lost sister civilization Earth. But the Cylons will not rest until every last human has been eliminated, and pursue them across the galaxy.
4"Lost Planet of the Gods, Part I"Christian I. Nyby IIGlen A. Larson,
Donald Bellisario
September 24, 1978 (1978-09-24)50204
With many warriors suffering from an unknown ailment, the Galactica recruits new Viper pilots—mostly young women, including Apollo's new bride, Serina. Led by Adama's spiritual interpretations, the fleet enters a vast magnetic void, emerging at the planet Kobol, the legendary world from which humanity originated.
5"Lost Planet of the Gods, Part II"Christian I. Nyby IIGlen A. Larson,
Donald Bellisario
October 1, 1978 (1978-10-01)50205
6"The Lost Warrior"Rod HolcombDonald Bellisario (t),
Herman Groves (s)
October 8, 1978 (1978-10-08)50908
7"The Long Patrol"Christian I. Nyby IIDonald BellisarioOctober 15, 1978 (1978-10-15)50902
8"Gun on Ice Planet Zero, Part I"Alan J. LeviDonald Bellisario (t),
John Ireland, Jr. (s),
Glen A. Larson (t),
Michael Sloan (t)
October 22, 1978 (1978-10-22)50206
9"Gun on Ice Planet Zero, Part II"Alan J. LeviDonald Bellisario (t),
John Ireland, Jr. (s),
Glen A. Larson (t),
Michael Sloan (t)
October 29, 1978 (1978-10-29)50207
Apollo, Starbuck, and Boomer lead a team of cutthroat demolitions and cold-weather experts (and the stowaway Boxey). Along the way, they encounter the misguided human scientist who originally built the weapon, as well as his legions of clones.
10"The Magnificent Warriors"Christian I. Nyby IIGlen A. LarsonNovember 12, 1978 (1978-11-12)50912
11"The Young Lords"Donald BellisarioDonald Bellisario,
Frank Lupo,
Paul Playdon
November 19, 1978 (1978-11-19)50905
12"The Living Legend, Part I"Vince EdwardsGlen A. Larson (s/t),
Ken Pettus (s)
November 26, 1978 (1978-11-26)50919
13"The Living Legend, Part II"Vince EdwardsGlen A. Larson (s/t),
Ken Pettus (s)
December 3, 1978 (1978-12-03)50920
14"Fire in Space"Christian I. Nyby IIJim Carlson (t),
Terrence McDonnell (t),
Michael Sloan (s)
December 17, 1978 (1978-12-17)50917
15"War of the Gods, Part I"Daniel HallerGlen A. LarsonJanuary 14, 1979 (1979-01-14)50921
16"War of the Gods, Part II"Daniel HallerGlen A. LarsonJanuary 21, 1979 (1979-01-21)50922
17"The Man with Nine Lives"Rod HolcombDonald BellisarioJanuary 28, 1979 (1979-01-28)50916
18"Murder on the Rising Star"Rod HolcombDonald Bellisario (t),
Jim Carlson (t),
Terrence McDonnell (t),
Michael Sloan (s)
February 18, 1979 (1979-02-18)50924
"Greetings from Earth"Rod HolcombGlen A. LarsonFebruary 25, 1979 (1979-02-25)50926
21"Baltar's Escape"Winrich KolbeDonald BellisarioMarch 11, 1979 (1979-03-11)50927
22"Experiment in Terra"Rod HolcombGlen A. LarsonMarch 18, 1979 (1979-03-18)50928
23"Take the Celestra"Daniel HallerDavid S. Arthur (s),
Jim Carlson (s/t),
Terrence McDonnell (s/t),
David G. Phinney (s)
April 1, 1979 (1979-04-01)50929
24"The Hand of God"Donald BellisarioDonald BellisarioApril 29, 1979 (1979-04-29)50930
Theatrical releases[edit]
Between 1978 and 1981, episodes were edited into three feature-length films. The first film, Battlestar Galactica, was an edited version of the pilot, "Saga of a Star World", featuring some differences from the original televised episodes, including the death of Baltar. It was released in cinemas in Canada, Australia and continental Europe before its American TV premiere and, in 1979, it was released theatrically in the UK, Ireland and the US. The second film, Mission Galactica: The Cylon Attack, was a re-edited version of the episodes "The Living Legend" and "Fire in Space", which also differed from the broadcast versions, omitting several scenes from both episodes. In 1979, it was released theatrically in continental Europe and Japan, and in the UK and Australia in 1980. The third film, Conquest of the Earth (also called "Galactica III" on the German poster), was a similar edit, though this time its source episodes were from Galactica 1980. It was also released theatrically, in 1981, in Europe and Australia.
Title Release Notes
Conquest of the Earth 1981 Edited merger of "Galactica Discovers Earth" and "The Night the Cylons Landed"[3]
Television movies[edit]
Following the cancellation of the series, all of the episodes from the original series were re-edited into television movies for syndication, including a third edit of the pilot. The four two-part episodes ("Lost Planet of the Gods", "Gun on Ice Planet Zero", "The Living Legend" and "War of the Gods") were all combined and expanded with about five minutes of scenes deleted from the original broadcasts. The remaining single episodes were edited into two-hour blocks, combining two episodes into a single narrative, occasionally including over-dubbed dialogue.[4] The final TV movie, an expanded version of "Experiment in Terra" with some scenes from the Galactica 1980 episode "The Return of Starbuck", featured new footage of an astronaut finding Commander Adama's logbook, which retold the premise of the show. It is believed that this version of the episode was to be released as another theatrical film in other countries, but never materialized.[citation needed]
TV movie title Original episodes
1 Battlestar Galactica "Saga of a Star World"
2 Lost Planet of the Gods "Lost Planet of the Gods" (expanded)
4 The Phantom in Space "The Hand of God" and "The Lost Warrior"
6 Space Casanova "Take the Celestra" and "The Long Patrol"
8 The Living Legend "The Living Legend" (expanded)
10 Greetings from Earth "Greetings from Earth"
12 Experiment in Terra * "The Return of Starbuck" (edited) and "Experiment in Terra" (expanded)
* The Experiment in Terra telemovie is not included on the UK PAL VHS box set; it has only the original episode "Experiment in Terra" without any clips from "Return of Starbuck." However, all the other telemovies appear to be in this set.
Galactica (1980)[edit]
Series 2 (1980)[edit]
Title Directed by Written by Original air date
1"Galactica Discovers Earth, Part I"
"Galactica: 1980, Part I"
Sidney HayersGlen A. LarsonJanuary 27, 1980 (1980-01-27)
After 30 years of searching, the Battlestar Galactica and its ragtag fleet locate Earth. However, it soon becomes clear that Earth is not advanced enough to help combat the Cylon fleet that tailed Galactica to Earth. Captain Troy—a grown-up version of Boxey—and Lieutenant Dillon end up teaming up with a female reporter from Earth named Jamie Hamilton.
2"Galactica Discovers Earth, Part II"
"Galactica: 1980, Part II"
Sidney HayersGlen A. LarsonFebruary 3, 1980 (1980-02-03)
3"Galactica Discovers Earth, Part III"
"Galactica: 1980, Part III"
Sidney HayersGlen A. LarsonFebruary 10, 1980 (1980-02-10)
4"The Super Scouts, Part I"Vince EdwardsGlen A. LarsonMarch 16, 1980 (1980-03-16)
5"The Super Scouts, Part II"Sigmund Neufeld, Jr.Glen A. LarsonMarch 23, 1980 (1980-03-23)
6"Spaceball"Barry CraneFrank Lupo,
Jeff Freilich,
Glen A. Larson
March 30, 1980 (1980-03-30)
7"The Night the Cylons Landed, Part I"Sigmund Neufeld, Jr.Glen A. LarsonApril 13, 1980 (1980-04-13)
A new advanced Cylon Raider is shot down over Earth. Two of its five crew survive—a Centurion, and a new, advanced humanoid Cylon—and are loose on Earth during Halloween.
8"The Night the Cylons Landed, Part II"Barry CraneGlen A. LarsonApril 20, 1980 (1980-04-20)
9"Space Croppers"Daniel HallerRobert L. McCulloughApril 27, 1980 (1980-04-27)
10"The Return of Starbuck"Ron SatlofGlen A. LarsonMay 4, 1980 (1980-05-04)
See also[edit]
| 0
| 0
| 2,516
| 2,516
|
71,756
|
7203d9a0-f087-4859-9f57-3cde70dfbe2b
|
trentmkelly/LessWrong-43k
|
Blueprint for a Brighter Future
Presumably our shared goal is to make possible a comfortable and dignified life for the maximum number of people while treading lightly on the environment, balancing individual rights and status seeking behaviors against the common good (with a view to preventing tragedy of the commons). Importantly, any solution to these problems must not require human nature to fundamentally change. Both left and right wing efforts of the 20th century to create a "new man", fit to inhabit their respective notions of utopia, only led to so much butchery.
I instead propose to design society around the human animal in such a way as to soften our rough edges, mainly by meeting everyone's needs and preventing extreme inequality. This won't prevent, say, crimes of passion but it would prevent crimes born of scarcity, such as burglaries, muggings, squatting, carjacking, scams, ransom, etc. It should also make accommodations for people wanting to live different ways, such as those unhappy in the dense living conditions found in urban population centers.
To this end I envision a nationwide network of robotically printed arcologies built atop or adjacent to the ruins of former cities (to make use of existing electrical infrastructure, roads, etc.) reclaimed by nature due to climate disaster. Said arcologies are coated in PV to generate all their own power as well as integrated geothermal (which now works anywhere on Earth thanks to Quaise Energy), linked to neighboring arcologies via enclosed elevated highspeed rail guideways, with PRT, moving walkways and multi-directional ropeless elevators furnishing transport within each structure.
Architecturally the interiors would share qualities of airports, malls and cruise ships, subdivided into units equipped according to their use, with every structure being mixed zone residential, commercial, agricultural (indoor vertical aquaponics) and so on. Mining and manufacturing are remote automated facilities.
Arcology interiors would be self-clean
| 0
| 0
| 411
| 411
|
109,446
|
<urn:uuid:96ace632-9c67-4995-9bb0-544837e2a801>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Viewing entries in
American Medicine
To What End?
1 Comment
To What End?
To what end?
Those three words have become something of a mantra, a mission, a philosophy of care.
• To what end do I prescribe a medication?
• To what end do I make a diagnosis?
• To what end do I order tests?
• To what end am I documenting?
• To what end is there a patient record?
• To what end do I send a person to a specialist?
• To what end do patients need to come to see me in the office?
• To what end do my patients have me as a doctor?
1 Comment
1 Comment
Washington, We Have a Problem
“Yes, Jonathan?”
“Sometimes my leg hurts.”
“Why does it hurt?”
“No, just every once in a while.”
“OK, and where does it hurt.”
“Where on your leg?”
• ICD codes – Codes for medical problems
• CPT codes – Codes for medical procedures
The Problem with Problems
A Better Way
1 Comment
More Trouble
I remember going to see the movie "Oliver" in the theater when I was a kid. Since this was my first movie in a theater, my mom made me a treat: a bag full of raisins and chocolate chips (Raisinets for Dutch people) and sent me there with my sister. It was a fine film, with Oliver getting kicked out of the orphanage when he wanted more gruel, the dastardly Bill Sykes threatening Oliver and sweet Nancy, the funny and clever artful dodger and Fagan teaching Oliver about life on the street, and with (spoiler alert!) good overcoming evil in the end Oliver getting adopted by a rich dude so he can get all the gruel (or real Raisinets) that he wanted. And though my memories of the movie are still vivid, my strongest memory was the look on my sister's face when I walked out of the theater covered with melted chocolate chip goo. It went into family lore (and wouldn't have happened if they had sprung for Rasinets, I might add). I think they still don't trust me with chocolate chips.
The key line in the film comes when Oliver loses a bet and goes up to the gruel-master and says: "Please Sir, I want some more." Which, as I am sure Oliver expected, causes the gruel-master to break into the song, "Oliver! Oliver! Never before has a boy wanted more!" and the whole dining hall to pull out musical instruments and singing harmony to the gruel-master's admonition.
Americans have been viewing health care the same way, always wanting more: more antibiotics, more technology, more robots doing more surgery, more expensive treatments for more diseases. The result: health care costs more in America than anywhere else. Some folks think that our "more" approach makes our health care "the best in the world," after all, where else can you get so many tests just by asking. MRI's for back pain, x-rays for coughs, blood tests for anyone who dons the door of the ER. "Tests for everyone!" shouts the bartender. "Tests are on the house! "
They aren't, of course, and we are paying the price for "more." This hunger for "more" is fueled by the media's fascination for the "latest thing," the long disproved idea that technology will solve everything, and docs who aren't willing to take time to explain why it's actually better to do less. It's hard to do, when we are paid more to spend less time with patients, and when the system is willing to pay for more and more.
There is a voice against this: the "Choosing Wisely" campaign, which argues against unnecessary treatments and tests. This is a welcome voice of reason in the cacophony of cries for "more." Yet the battle goes against the irresistible tide of our payment system. The root problem is this: there are a whole lot of people whose jobs depend on America's addiction to "more." The payment system has created an ecosystem that thrives off of waste (of which I once wrote an allegorical fantasy). True health care reform will be catastrophic to many who work in health care, with many very nice and hard-working Americans losing their jobs at the ACO factory, at Meaningful Use Inc., and even at Stents-R-Us hospital here in my home town.
This is what you get when you make disease more profitable than health, when we treat problems instead of people. The simple fact that our system would be destroyed if everyone got healthy should tell us something is terribly wrong. Doctors want their offices full, not empty. The goal of every patient - to be healthy and to stay away from the doctor - goes directly against the economics of "more."
I have always tried to be a non-test orderer. I was trained well by docs who believed it weak-minded and bad care to blithely order tests and prescribe medications without a well-defined reason. This has always made it harder for me, as it's far more time-consuming to explain why a drug or test is not needed than to simply order it. But in my new world, one in which an empty office is a good thing, I've found my patients much more open to my aversion to "more." The main reason for this is that I am giving them more of me. More of me means they can call if they don't get better, or if their symptoms develop. They know I won't force them to take more of their time and spend more of their money to get my attention.
Ultimately, I want my patients to see as few doctors, be sick as infrequently, and be on as few drugs as possible. I hope to wage an all-out assault on "more."
Here are my rules to battle "more"
1. Never order a test that doesn't help you decide something important. Ordering tests "just to know" does much more harm than good.
2. Use consultants only to do things you can't. Orthopedists will aways give an NSAID and physical therapy for problems, so I don't send patients to them unless they've failed those treatments (where appropriate). I am just as good at ordering PT, and am more careful with NSAID prescriptions than they are.
3. Don't give a patient a drug without explaining to them why they need it. If I can't make a good case for a drug, I shouldn't be giving it. This is not simply "to lower your cholesterol," or "to treat your blood pressure," but because doing so will raise your life-expectency.
4. Remember the number that really matters: how many birthdays a person gets to celebrate in health. I don't care about blood pressure, LDL, or even A1c if treating it doesn't raise the birthday total.
5. Don't forget about another number: how much money patients have in their wallets. There's no point in ordering a drug they can't afford, or making them pay for a test they don't need (even when they ask for either).
I hope my new world of less overhead, less regulation, and less antacids for me continues on this trend toward less sick patients, less drugs, and less tests. Perhaps I need to break into a song and dance number whenever my patients ask for "more."
That would teach them.
The Virtue of Unnecessary Care
• Brand name drugs given when generics would do
• The unnecessarily high price of drugs.
The Cost of Success
The Seduction of Primary Care
Hey there, big, smart, good-looking doctor....
Are you tired of being snubbed at all the parties? Are you tired of those mean old specialists having all of the fun?
I have something for you, something that will make you smile. Just come to me and see what I have for you. Embrace me and I will take away all of the bad things in your life. I am what you dream about. I am what you want. I am yours if you want me....
Seduce: verb [ trans. ]
attract (someone) to a belief or into a course of action that is inadvisable or foolhardy : they should not be seduced into thinking that their success ruled out the possibility of a relapse. See note at tempt .
(From the dictionary on my Mac, which I don't know how to cite).
If you ever go to a professional meeting for doctors, make sure you spend time on the exhibition floor. What you see there will tell you a lot about our system and why it is in the shape it is. Besides physician recruiters, EMR vendors, and drug company booths, the biggest contingent of booths is that of the ancillary service vendors.
"You can code this as CPT-XYZ and get $200 per procedure!"
"This is billable to Medicare under ICD-ABC.DE and it reimburses $300. That's a 90% margin for you!"
This is an especially strong temptation for primary care doctors, as our main source of income comes from the patient visit - something that is poorly reimbursed. Just draw a few lab tests, do a few scans, do this, do that, and your income goes up dramatically. The salespeople (usually attractive women, ironically) will give a passing nod to the medical rationale for these procedures, but the pitch is made on one thing: revenue.
Our practice has succeeded despite the fact that we don't do a lot of procedures. We are in a shrinking minority, and the monthly cash-flow is putting increasing pressure on us to think about "alternative sources of revenue." Most of my colleagues in private practice have labs, x-ray equipment, or do procedures. Some do such medically vital services as hair removal. I haven't had the stomach to go that direction...yet.
Who's at fault for this? Is it the doctors, who are seeking profit over what's best for the patient? Is it the vendors, who find loopholes in the reimbursement structure to milk extra dollars out of the system?
If you leave meat on the floor, don't be surprised when your dog eats it.
The payment for the E/M codes (the codes used to bill for doctor's visits) are low and the payment for CPT codes (the codes used to bill for procedures) are high. This is how our system is set up (with great thanks to the RUC) and it is one of the main reasons we spend so much money on healthcare. We aren't doing healthcare, we are doing sick care. Healthcare is prevention, which takes face-to-face encounters with the patient. It involves talking and listening, and talking and listening are not deemed valuable by our system. We are paid to do, not to educate or listen.
It takes great resolve to resist this siren's call. A few years ago, we made a deal with one of the other practices in our building to buy a portion of their x-ray equipment. It seemed to be a good way to make money off of something we do normally in practice. But a few months into this deal, we realized two things:
1. We weren't ordering enough x-rays to be profitable. We had established a mindset of ordering x-rays that minimized their use. It was a nuisance to wait for the reading on an x-ray and it was inconvenient and costly to the patient, so we made most of our judgments based on something else: the physical exam.
2. We were ordering a lot more x-rays than we had before. Instead of trying to find reasons to not order x-rays, we were now financially motivated to order them. So if someone hurt their ankle, we were much more likely to order one. If someone had a chronic cough, we were much more likely to order a chest x-ray. The change wasn't that we were hungry for profit, it was just that we were suddenly 180 degrees from our previous mindset: we were trying to find medical justification to order more x-rays. It was incredibly seductive.
We did back out of the deal, feeling that the care we gave wasn't better and not liking the fact that we were losing money. But would we have backed out if our practice wasn't already financially stable? We are a well-run practice that has been successful despite our non-reliance on procedures, but what of the other practices out there that aren't so successful?
One of my favorite sayings is: your system is perfectly designed to yield the outcome you are currently getting. Nowhere is this more true than in healthcare. We have set up a system that encourages consumption. We pay doctors more to do more. We pay doctors less to spend time with patients. We want our doctors to do better care, but we pay them to do worse care. We want to save money, but we reward those doctors who spend the most.
So why not change? Why not pay more for E/M codes and less for CPT codes? Yes, some doctors will abuse this system by running patients through their office and spending little time with them, but at least it will increase availability of doctors to see patients. There will always be those who take advantage of any system; that shouldn't stop change.
I went into medicine to take care of people, not spend their money. Why can't we have a system that doesn't force me to decide between the two?
Vote Llamacrat!
"So what do you think about the election?"
"So What do you think about Obamacare?"
"What do you think about this healthcare situation?"
Anyone who reads this blog regularly knows that I am a "flaming moderate" when it comes to politics. I don't have much faith in anyone who identifies too strongly with one party of the other. I am really angry with congress and their lack of gonads to work on really coming up with solutions. Interestingly, my patients, regardless of their political leaning, agree with much of what I say. Here are the things they all seem to agree with:
2. There are many villains in this story - I like to rail against congress (it's a hobby), but the president is to blame, as are past presidents who allowed the situation to get this bad. Obama inherited a bad situation, he didn't create it. Many of the bad things happening were set in motion during the Bush administration (or earlier). Now, Obama's leadership style of letting congress shape healthcare reform has created lots of problems. Then there is the system that lets money influence policy (which has been evolving over a long period of time). Lobbyists for those with money leverage the compulsion of politicians to be re-elected and get what they want by well-placed "contributions."
3. We can't afford to wait - The past two years has proven that even with a very large majority, a party can't get things done without cooperating with the other party. In two years we may get a different president, but is there anyone out there that thinks our government will work any better? If the republicans gain power, the democrats will unite and block anything they try to accomplish. Simply putting a different captain at the helm of a sinking ship won't change the final outcome. Healthcare is a mess, and that mess is getting worse, not better. Gridlock is unacceptable.
4. Our situation is scary - Our government doesn't work any more. The system we have thrives on debate, influence peddling, and party politics. It has survived over 200 years that way. The thing that scares me and most of my patients is the lack of any willingness to work together. If they were in a room with one exit, the democrats and republicans wouldn't agree on how to get out of it. Politics of party is so acute, inflamed by the ideologues on one side vilifying the ideologues on the other, that it seems impossible to get anything significant done. One commenter on my last post suggested that we are seeing the death-throes of our government. If our way of solving problems is screaming and pointing fingers, I begin to agree with that.
What to do about all of this? Do we form a third party? (I would suggest llamacrat as the name of the party). Do we write congresspersons? Do we get pitchforks and torches and march on Washington DC? Do we whine a lot in blog posts, hoping someone else will do stuff?
I don't really know. I do think that congress should be scared, not empowered. The people voted for change in the past 3 elections, which means there is overall dissatisfaction with all involved. Another 2 years of do-nothingness will prove only that our country is in danger. The mandate of this election is a "mad as hell" mandate.
I beg with any politician crazy enough to read what I write: please don't put re-election or party politics ahead of the people you represent. The election was not a game with winners and losers, it was a statement by the American public that you guys are screwing things up. Govern us. Lead us. Take care of us. Fix our problems.
If you don't, the llamacrats will be breathing down your necks.
Different Lunatics, Same Asylum
The top vote-getting answer on my poll about what people feel about the election: Different lunatics, same asylum. We are getting jaded by our system. Being the "flaming moderate" that I am, I find it hard to hear the substance of the rhetoric on either side, just the shrillness and rancor of the voices. From the physician's perspective, it is very hard to know who to favor in this election. The democrats seem to love lawyers and hate tort reform, and they also favor an expansion of government. The republicans love big businesses and "free market," accepting the bad behavior of insurance and drug companies as "the market working itself out." They both seem hell-bent on sticking it to the other party at the expense of getting anything done - and this in a time of crisis for our industry.
The results of this playground brawl between the two gangs of bullies is that all of us wimpy kids (the ones without power) end up lying bloody in the dirt. Here are the facts as I see them about healthcare in our country:
1. It costs far too much. The top item on the agenda needs to be cost control. The only way to control cost is to stop paying for things that are unnecessary or for which there is a cheaper alternative. I know that's not simple as it sounds, but so much of the discussion is about coverage and how things are paid, while the real issue is not who pays, it's what and how much gets paid.
2. Too much of the cost is hidden. How much does it cost to see the doctor? That's an incredibly complicated question. It depends on the insurance carrier, the doctor in question, the way the doctor codes the visit, and the nature of the doctor's ordering practice. The same thing is true on a grander scale for hospitals. Drug costs are hidden by copays (allowing companies to wheel-and-deal to get drugs on formularies). Insurance companies hide their administrative cost and pass on any increases on to the people buying the policies.
3. It is totally disorganized. Nobody knows what anyone else is doing. As much as people rail against the mandated EMR, the coordination of care will be impossible without it. We need to know what has been done using clinical information, not billing data. Up to now the insurance industry has controlled the information about what is done on patients, using the withholding of payment as the club to change physician behavior. While there is risk that doctors might get screwed with the recording of our behavior, not doing so makes the chance of real improvement nearly impossible, leaving the payors with the data and hence with the most power.
4. Nobody is pointing out the naked emperor. It is insane that drugs cost as much as they do. It is crazy that we pay what we do for technology. What exactly is it in the 2 day ICU stay that costs $100,000? Price gouging is rampant because it is allowed (and even encouraged by our system). Why can a drug company raise the price of a gout drug by 5000%? Because the FDA lets them. Why can granny in the nursing home with alzheimer's get put in the hospital and spend a week in the ICU? Because Medicare pays the hospitals and physicians who put her there. The credit card bill is shooting up, yet we are not asking why we keep spending so much?
5. The people with the most at stake are those with the least power. Patients (and primary care physicians) are the ones with the most at stake. The discussion is being run by politicians (who don't have to use the plan they pass), hospitals, specialty organizations, insurance companies, and other health-related industries. Those with money can most influence the process to their advantage, and patients are definitely not the ones with the money.
6. We cannot afford gridlock on this issue, but that is undoubtedly what we will get. I have not met one person, liberal or conservative, who is optimistic about the next two years. We are living with a cancer but are unable to do anything about it. That cancer is not going to sit around and wait for the politicians to agree, it will spread and will choke out any hope of survival.
I am not too charged up this election season. It seems that very few people look with optimism on what's going to happen over the next 2 years. The best thing we can do? Raise your voice. Vote in a way that will change the process. We need to change the asylum and make it into a place where things get done. We need people with the political gonads to work with others not in their party. We need more sense and less shrillness.
If we don't get this fixed soon, the patient will be beyond saving.
Doctors, Hospitals, and the Yankees
Joe Boyd hated the Yankees.
"Those damn Yankees. Why can't we beat 'em?"
Then he got the opportunity to save his beloved Washington Senators by making a deal with the devil - giving up his soul in exchange for being transformed into "Shoeless Joe" to propel his team to win the World Series.
Interesting. I think a lot of doctors are making their deal with the devil. They are looking for a small gain in comparison to a long-term of misery. True, Joe Boyd made out in the end; but that will only happen if someone from Hollywood writes our script.
Here's the problem: at the core of our problems with healthcare is the total lack of cohesive communication. Doctors have no idea what other doctors have done with a patient. Tests get ordered, medications get changed, procedures, hospitalizations, even surgeries are done without communication to other doctors who would benefit from this information. The conduit of communication is this:
Doctor: "So, how have you been doing over the past few months?"
Patient: "Didn't you get the notes from the hospital? I was in for two weeks. I had a heart attack and a stroke and now I am in rehab for both of these."
Doctor: (checks chart uncomfortably) "No, I didn't hear about it. Why don't you tell me about it...."
Situations like this happen daily at my office. Patients are started on medications by specialists without my knowledge. Lab tests are done that I have no access to. Huge changes happen in the lives of the patients for whom I have cared for over a decade, and I get nothing. Even consults I order are done without any communication back to me. On the other side of things, my patients are hospitalized without any consideration of the care I have been giving over the past decade. Patients are treated as if their care starts from scratch every time they enter a new venue.
It hurts my care for the patient. It hurts the other doctors' care for the patient. It hurts the patient.
And it costs a lot of money. Disorganized, we cannot beat this behemoth of dollars spent. Without good communication, communication that allows each person involved in the care of the patient to see exactly what is going on with the patient, the spending continues.
So what can be done about it? How can the care of the patient be organized? One common solution is the Integrated Delivery Network, or IDN. An IDN is a network of doctors linked together through a hospital. The care is integrated through a common record, or at least through a conduit that eases the flow of the patient from point of care to point of care. Academic medical institutions are IDN's as are many private hospitals (such as Kaiser - although Kaiser operates as its own insurance carrier as well). This seems to make sense. It breaks down barriers of communication and improves care.
But there's a catch: the hospital. Hospitals are often thought of as being on the same side as doctors - after all, doctors work for hospitals, right? While this is somewhat true for specialty physicians, many of whom make their revenue from procedures done on hospitalized patients, it is not true of primary care physicians. Hospitals are centers of care, yes, but they are also centers of spending. A hospital is not motivated to save patients money. Their profit is driven by patients being sick, getting tests and procedures done, and racking up cost. A patient can spend in a day in the hospital more than they will spend in a lifetime at my office. In this situation, the patient is treated as a commodity - something to use for a profit.
Does the hospital want me to be responsible, not ordering unnecessary tests, keeping patients healthy and out of the hospital? Do hospitals want me to get patients in hospice at the end of their lives, eliminating unnecessary hospital stays? Do hospitals want me to keep patients out of the ER? They won't get mad at me if I am the only one doing it, but all doctors getting responsible would be bad news for their bottom line. Selling myself with a hospital will put me in a conflicted position: wanting to please my employer, yet wanting to do what's best for the patient. In this way, IDN's are fatally flawed.
So what can be done? How can communication be fixed without letting the hub of the communication network be a source of spending? Think back to the conversation I recounted above. Who was the hub in that setting? The patient. Perhaps we should consider this model when moving toward a communication network. Perhaps a patient-centered communication model would optimize communication without raising cost. After all, shouldn't I answer to the patient - the one who is spending the money and the one who receives the care - for the decisions I make?
What if we set up a decentralized communication network that was linked not by doctors, hospitals, or insurance companies, but by permissions given by a patient? Here's what I mean:
• I would have access to any records on the patient on any clinical database that the patient allowed me access to. Instead of importing labs into my system, I would have access to the laboratory's system for any patient I had permission from. That way if the patient had labs done by another clinician, I could see the results. If the patient was at the hospital, I would have access to those records as well.
• I would give access to any clinician who was given permission by the patient to see my records. If the patient was in the emergency room or in the hospital, the doctors there could see what I have been doing with the patient in the outpatient setting. If a consultant wondered why I ordered a consult they would have easy access to my documentation of this.
Sounds risky? I think it is less risky than a centralized database with all the information in one location. Sounds hard? Isn't what I described just a description of what the internet is? Information on my blog is not downloaded on your computer, you just have access to it. If I wanted to deny access, I could. If I wanted to limit that access, I could do that as well.
This is exactly what happens with banks as well. The consumer has control over access to bank accounts. If they want to allow their gas company to draft from their checking account every month, they can. They are not required to gather all of the banking information in one location, it is spread out among many.
In baseball, often it is the team who spends the most money who wins in the end. Those of us who grew up hating the Yankees can attest to that ugly fact. Healthcare is presently run by those who control the money: the insurance companies and the drug companies. They win because we can't afford to fight them. They win because the minute they get behind, they find a way to use their money to get back on top. But we don't beat such spending by selling ourselves to fix our short-term problems.
To fix this problem, we don't need more of the same. We need the whole way the system is set-up to change. We need the rules to change. We need a change in ownership.
Dare we admit that the real answer to our problems is in the hands of our Washington Senators?
No T
I've been tired lately. I've been gaining weight and feeling unmotivated as well. I thought that maybe it was because I wasn't sleeping enough and was eating too many donuts.
Then my life changed.
I saw a commercial that talked about "Low T" being associated with ALL OF MY SYMPTOMS!!! It seems obvious that my sleeping and donut-eating habits aren't the culprit. What luck! I just need to rub a gel on my body and everything will be great. I am going to get checked for "low T."
Play this scenario (sans hyperbole) 100 times, and that is what I've been hit with. Multiply that times the number of PCP's in the US, and you see evidence of very good marketing. Testosterone replacement for men has become the new magic bullett, a counterpoint to the request for thyroid tests by my female patients - with one big difference: it's clearly safe to replace a thyroid hormone deficiency, but not so clear with testosterone.
Today's NY Times addressed this very issue, comparing testosterone replacement to one-time dogma of estrogen replacement in women:
Despite beliefs based on observational evidence that estrogen therapy enhanced the health and well-being of menopausal women, when a definitive study was finally done, clinicians and researchers were shocked to discover that the risks of long-term hormone replacement could outweigh its benefits.
Would a similar study of testosterone therapy for men experiencing “andropause” likewise reveal more hazard than help? The answer would be welcomed by an estimated four million men in the United States who have subnormal levels of this important hormone, a common result of advancing age.
Yes, I lived through the estrogen replacement therapy about-face caused by the landmark study, The Women's Health Initiative. The medical community was convinced that estrogen replacement gave a good enough heart disease and osteoporosis risk-reduction that it would easily offset any slight increase in breast-cancer risk. There was even a retrospective study supporting this hypothesis. The WHI not only didn't show a cardiac benefit, it showed an increase in risk. Wyeth pharmaceuticals, the company who makes Premarin and other estrogen replacement drugs (and the company who funded the study) was devastated, and never really recovered.
So now we have commercials for "low T" blasting the airwaves, suggesting a treatment that may not be safe. Is that ethical? Again, from the NY Times:
Late last year, for example, a six-month federally financed study of a testosterone gel put a surprising hitch in efforts to improve the lives of aging men who experience a decline in energy, mood, vitality and sexuality as a result of low testosterone levels. The study, conducted among 209 men 65 and older who had difficulty walking, was abruptly halted when those using the hormone had an unexpectedly high rate of cardiac problems.
The article goes on to point out that the evidence is unclear at this point, and that the study mentioned had flaws. Still, it stirs up the ghosts of hormones past, with the strong possibility that treating "low T" will cause harm. And at least estrogen therapy had clear benefits (osteoporosis) and studies that supported the replacement therapy. Unanswered too is if the therapy increased prostate cancer risk (one of the main treatments for prostate cancer is castration, which is cutting the body's production of testosterone almost completely - no pun intended).
So how should I respond to these men who just want a little "get up and go?" Why is it that I have to fight against a potentially harmful advertising campaign? What is the service (aside from that to investors) the drug companies perform by educating men about "low T?" It puts doctors like me in a situation where we could potentially harm our patients.
So what are the limits to advertising by drug companies? Most people are not in favor of any of such advertising, but the government has so far allowed it. Are there ethical guidelines? Consumers assume these ads are being vetted. I really wonder about this in light of the "low T" campaign that is allowed to go forth despite lack of proven health benefit ("get up and go" doesn't qualify as a health benefit) and significant potential risk. I had a man with known coronary heart disease recently insist on getting checked for "low T." I told him that the no matter the lab result, I would not prescribe it for him. He insisted, and yes, the level was low.
Now his doctor is standing in his way to a better life.
Thanks a lot, drug companies.
| 0
| 0
| 7,712
| 7,712
|
114,595
|
<urn:uuid:bb723e17-e6f2-4521-b38b-69baa77fe87e>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
DeSmuME runtime requirements
Discussion in 'NDS - Emulation and Homebrew' started by xyzman, Jul 17, 2012.
1. xyzman
xyzman Member
Apr 12, 2007
Serbia, Republic of
For some reason, despite being a proud owner of DSTWO flash card, I want to try emulating Pokemon White2 on my PC using DeSmuME. While I was able to play both Black and White perfectly, clean ROM of White2 causes application crash in both official 0.9.8 and unofficial 0.9.9-2012-05-09 JIT version. I get this generic message:
Microsoft Visual C++ Runtime Library
Runtime Error!
Program: C:\temp\dmm\desmume-0.9.9-JIT-x86-2012-05-09.exe
Please contact the application's support team for more information.
The ROM I am using is clean according to ds-scene ROM tool. I tried both trimmed and untrimmed versions with no success.
I've tried to run this ROM on different, really old PC, and everything works fine, except being really slow even on frame skip set to 9. So the problem here seems to be with the particular computer, and, since the emulator depends on runtime libraries, most probably, the issue is caused by some conflicts in these libraries.
Therefore, I got some questions for all the emulation gurus out there:
1) What are the exact libraries that DeSmuME (both official and JIT versions) depends on? Preferrably I will need file names to compare between two systems?
2) Is there any way to enable max verbosity in desmume dev version? It would possibly help to know which library function causes the crash.
| 0
| 0
| 395
| 395
|
13,620
|
780100d1-4359-4927-a0e0-885969f2040d
|
StampyAI/alignment-research-dataset/eaforum
|
[Discussion] Best intuition pumps for AI safety
When I introduce people to AI safety I usually get one of three responses:
a) “that makes a lot of sense. What can we do about it?”,
b) “I get it rationally, but intuitively I don’t feel it”,
c) “I just don’t buy it - I don’t think machines can’t be smarter than humans”, “I still think that we can just program them the way we want” or something along these lines.
I get the last response even after giving the standard arguments for why a stop button won’t work, why superhuman intelligence is plausible or why intelligence doesn’t imply morality. So my hypothesis is that they find the thought of unaligned superhuman AI so unintuitive that they are unwilling to actually consider the arguments.
Thus, my question is: *What are the best intuition pumps for AI safety?*
I’m personally looking for Carl Shulman-style common sense arguments similar to those of his [80K podcast appearance](https://80000hours.org/podcast/episodes/carl-shulman-common-sense-case-existential-risks/). He argues that buying insurance for a gain-of-function lab would probably cost billions of dollars which gives us a better intuition about the risk involved.
I have recently started making the following argument. If you think that AI won’t be smarter than humans but agree that we cannot perfectly control AI in the same way that we cannot perfectly control humans, then you should be willing to pay as much money towards aligning AI as society spends on aligning humans, e.g. terror defense, prisons, police, and the justice system.
According to [Investopedia,](https://www.investopedia.com/articles/investing/061215/what-countries-spend-antiterrorism.asp) the US alone spends 175$ Billion on counterterrorism and 118$ Billion on [police](https://boingboing.net/2021/04/20/u-s-policing-budgets-would-rank-as-the-worlds-third-highest-military-expenditure.html) per year.
[This paper](https://www.ojp.gov/ncjrs/virtual-library/abstracts/what-does-world-spend-criminal-justice) from 2004 estimates that 70 rich nations spend more than 360$ Billion combined on the justice system in 1997.
Thus, if we adjust for inflation and missing countries we will likely get a lower bound of at least 1 Trillion Dollars spend per year on aligning humans. What we currently spend on AI safety is many orders of magnitude away from this.
Do you think this argument makes sense? Feedback and further suggestions are welcome. Your argument can also address different concerns that people typically have about AI safety.
| 0
| 0
| 641
| 641
|
19,818
|
1d63cfad-59ad-4375-a4c1-194320f670f0
|
StampyAI/alignment-research-dataset/lesswrong
|
MathJax\_Main-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-I; src: local('MathJax\_Main Italic'), local('MathJax\_Main-Italic')}
@font-face {font-family: MJXc-TeX-main-Ix; src: local('MathJax\_Main'); font-style: italic}
@font-face {font-family: MJXc-TeX-main-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-R; src: local('MathJax\_Main'), local('MathJax\_Main-Regular')}
@font-face {font-family: MJXc-TeX-main-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-math-I; src: local('MathJax\_Math Italic'), local('MathJax\_Math-Italic')}
@font-face {font-family: MJXc-TeX-math-Ix; src: local('MathJax\_Math'); font-style: italic}
@font-face {font-family: MJXc-TeX-math-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Math-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Math-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Math-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size1-R; src: local('MathJax\_Size1'), local('MathJax\_Size1-Regular')}
@font-face {font-family: MJXc-TeX-size1-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size1-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size1-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size1-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size2-R; src: local('MathJax\_Size2'), local('MathJax\_Size2-Regular')}
@font-face {font-family: MJXc-TeX-size2-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size2-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size2-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size2-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size3-R; src: local('MathJax\_Size3'), local('MathJax\_Size3-Regular')}
@font-face {font-family: MJXc-TeX-size3-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size3-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size3-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size3-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size4-R; src: local('MathJax\_Size4'), local('MathJax\_Size4-Regular')}
@font-face {font-family: MJXc-TeX-size4-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size4-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size4-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size4-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-vec-R; src: local('MathJax\_Vector'), local('MathJax\_Vector-Regular')}
@font-face {font-family: MJXc-TeX-vec-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Vector-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Vector-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Vector-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-vec-B; src: local('MathJax\_Vector Bold'), local('MathJax\_Vector-Bold')}
@font-face {font-family: MJXc-TeX-vec-Bx; src: local('MathJax\_Vector'); font-weight: bold}
@font-face {font-family: MJXc-TeX-vec-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Vector-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Vector-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Vector-Bold.otf') format('opentype')}
(to be defined shortly), represented by a Turing machine. Our options comes from some finite set S[[1]](#fnm2u3pik4tkk). We are trying to maximize u, also represented by a Turing machine (u should calculate expected utility, but for the rest of this post I'll assume it returns a rational number).
Note in particular that u does not have any parameters. It can mention f via its definition (or even logically equivalent definitions), but we do not represent it as a parameter! f can only influence u "A priori".
We also must fix a formal system; I'll assume that it is Peano arithmetic, but other theories, such as ZFC, should work fine as well.
We have one requirement on u. For any action s∈S, there must exist a rational number r such PA proves that f=s⟹u=r. Basically, if we know f we can figure out u (using PA). In particular, if f halts, so does u.
Here is the pseudo-code for f:
```
procedure f:
act_table := []
for s in S:
i := 0
utility := None
loop:
for proof in PA_Proof(length := i):
if proof == "f = s => u = r" for some rational r: //Comment: use quining when refering to f (including within u if necessary)
utility := r
if proof == "f ≠ s":
return s //Comment: Rosser's trick. Should never execute, but makes it harder for PA to analyze
if utility ≠ None:
insert(act_table, (s, utility))
break
else: i := i + 1
return first component of (max act_table by second component)
```
This seems like a pretty standard maximizer. It maximizes something to do with proofs, but that isn't anything special. But notice the line with the "Rosser's trick" comment. Why would we return s if we prove that f would not return s?
Assuming PA is consistent, this line will never execute. But PA doesn't know that PA is consistent. This already looks extremely difficult for PA to rule out that it will return s.
And it doesn't just *look* hard, we can actually quantify exactly how hard. Let's say the shortest proof of a statement f=s⟹u=r (for any rational r) is of length n. Then it is impossible for PA to prove that f≠s within n steps, because then f would return s due to that line, making PA inconsistent (which we are assuming is not the case). This is essentially the same reasoning used in [Gödel's speed-up theorem](https://en.wikipedia.org/wiki/ Gödel's speed-up theorem).
If we want to make it even harder for PA to prove that f≠s, we just need to have f look for a proof of f≠s for even longer proofs (like n + 100 or something).
Some analysis
=============
When looking for the proof of f=s⟹u=r, we aren't really doing a counterfactual in the traditional sense. The agent is not considering "What if I am in a world where I choose s?" or "What if I am in a world where I learn that I choose s?". The agent is considering a [material conditional](https://en.wikipedia.org/wiki/Material conditional). The agent is considering "What if *the world I am already in* is the world in which I choose s?".
For example, a logical implication might be that similar agents in similar situations would choose similar actions. That's because if they are analogous or isomorphic in certain situations, you might be able to transfer the mathematical knowledge that f=s into a theorem that other agents act in certain ways. This in turn affects the utility: the agent might take certain actions just because it implies logically the other agents will take actions that just so happen to increase the original agent's utility.
Unfortunately, detailed analysis will depend heavily on the details of how formal system is set up. It turns out that doing what is essentially a massive "search and replace" on *math itself* is "complicated" to say the least. Since f is just a definition, not a parameter, any branch of math that just so happens to mention f (or something that is mathematically related to it) will get warped, even if the context is something entirely different than decision theory.
Humanly feasible "Algorithm": Highly Functional Decision Theory
===============================================================
*Warning: this section contains* [*humor*](https://www.lesswrong.com/tag/humor)
It is a cardinal sin to conflate human reasoning with mathematical formal proof systems. But what if I want to do it anyways?[[2]](#fnnx6qftj8l3k)
Introducing *Highly functional decision theory*
Did you keep trying to make decisions, but the principle of explosion keeps getting in the way! You can say goodbye to this problem with one weird trick!
```
Highly functional decision theory:
Go through your set of options, considering for each one "Assuming that I will do this option, what will my expected utility be?".
However if at *any point* during your reasoning, you come to the unconditional conclusion "I *definitely* would not do option s", even if you stumbled into the conclusion accidentally, you immediately do s without any further deliberation.
If this doesn't happen, perform the action that implied the highest utility.
```
Sounds great, right? Also, if the second clause causes you to do a weird action that causes concern to other people, I suggest an explanation like the following: "I am highly functional when it comes to decision theory; it can be hard for an outsider to appreciate such advanced decision making".[[3]](#fne9dq3x391i) Also, leave a comment on this post documenting the event.
The reason for the second clause isn't *just* to make you do dumb things. It is to force the assumption "I do action s" to be plausible even if s is dumb. Any attempts to dismiss the statement "I do action s" as false before algorithm finishes is rendered *inconsistent*. Therefore, you can't reasonably dismiss it and you must reason from the statement in a "normal" way (i.e. not using a spurious counterfactual).
Here is an example. Suppose you walking close to a lake in a park, and need to decide on the best way to enjoy the park. One possible option is spontaneously jumping into the lake to go for a swim. Your reasoning might look like this:
> Assume that I will jump into the lake:
>
> * I will gain some utility directly from the enjoyment and exercise that comes from swimming.
> * This decision to go swimming increases my estimate for how much I like swimming. I must consider how the character trait of "liking swimming" will affect my utility for the rest of my life.[[4]](#fnbfv30eck9xr)
> * The people around me might be similar to me in certain ways, such as genetically or in that they are also HFDTs (highly functional decision theorists). If this is the case, it stands to reason that they will also spontaneously jump into the lake at the same time. I must consider the opportunities, drawbacks, and risks that this will bring to my swimming adventure.[[5]](#fnzt2l29176n)
> * I must also consider take into consideration that I *won't* choose other options[[6]](#fn3xl1562x5if) for my recreational activity when calculating my utility. The effects could be similar to the above three bullet points, but inverted.
>
So there you have it. Feel free to give the algorithm a try and report back.
1. **[^](#fnrefm2u3pik4tkk)**In practice, S is a finite set of finite lookup tables for how the agent should respond to information. I think there is an easy way to adapt it to functions with infinite domains, but if u needs to use the lookup table an unbounded number of times, it won't be guaranteed to halt anyways.
2. **[^](#fnrefnx6qftj8l3k)**Truth be told, I have tried the algorithm myself, and it has had promising results so far (it seems to be helpful in short-circuiting the logical flaws in "Learned helplessness" style problems). I think a more serious name could be "Logical Decision Theory" (because it is based on logical implication instead of counterfactual implication). To be clear, this would be a variety/implementation of functional decision theory. However, due to how absurd the algorithm is from a human perspective, and the terrible premise of treating humans like a formal proof system, I decided to give this section a whimsical tone.
3. **[^](#fnrefe9dq3x391i)**I recommend choosing a set of acceptable options you will limit yourself too before starting the algorithm. Although it would be *extremely* funny, I would also feel slightly bad if you come to the conclusion "I would never scream as loud as possible in public right now"[[7]](#fnm7z4bx1c66l). You would be forced to choose between breaking my silly algorithm (you pre-committed to follow it before hand, right?) and alerting all your fellow HFDTs in the local area to your presence.
4. **[^](#fnrefbfv30eck9xr)**You might say "but wait, I would just be choosing to swim because of the decision theory, not because I wanted to!". If you argued that you committed a crime because a decision theory made you do it, would that hold up in a court of law? If not, why should that style of argument hold up in your mind-courts?
5. **[^](#fnrefzt2l29176n)**Enumerating these is left as an exercise for the reader, but I will list at least one risk to get you started: one of the other HFDTs could declare war on the local fish population, and you could get caught up in the conflict. (Guess what happens if you say "but HFDTs *definitely* wouldn't declare war on the local fish population, they are too rational!".)
6. **[^](#fnref3xl1562x5if)**Quick note: this doesn't trigger the second clause of the algorithm because "I won't choose the other options" is conditioned on "I choose to swim". This is why the algorithm says *unconditional* conclusion.
7. **[^](#fnrefm7z4bx1c66l)**That said, feel free to include slightly silly options. You need to punish your brain for failing to follow [Gödel's Second Incompleteness Theorem](https://mathworld.wolfram.com/GoedelsSecondIncompletenessTheorem.html) somehow.[SEP]
| 1
| 6,000
| 10,776
| 10,776
|
93,723
|
<urn:uuid:0a24faea-1556-4c9b-8ed7-f0165f3cf04c>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Party supplies
"I once had a toaster that tried to kill me."
- W.J. Flywheel, Webporium Curator
Despite voluminous protest and nitpicking criticism from loyal fans of the original 1978-80 TV series, the 2003 version of Battlestar Galactica turned out surprisingly well for viewers with a tolerance for change. Originally broadcast on the Sci-Fi Channel in December 2003 and conceived by Star Trek: The Next Generation alumnus Ronald D. Moore as the pilot episode for a "reimagined" TV series, this four-hour "miniseries" reprises the basic premise of the original show while giving a major overhaul (including some changes in gender) to several characters and plot elements.
Gone are the flowing robes, disco-era hairstyles, and mock-Egyptian fighter helmets, and thankfully there's not a fluffy "daggit" in sight... at least, not yet. Also missing are the "chrome toaster" Cylons, replaced by new, more formidable varieties of the invading Cylon enemy, including "Number Six" in hot red skirts and ample cleavage, who tricks the human genius Baltar into a scenario that nearly annihilates the human inhabitants of 12 colonial worlds.
Thus begins the epic battle and eventual retreat of a "ragtag fleet" of humans, searching for the mythical planet Earth under the military command of Adama (Edward James Olmos) and the political leadership of Laura Roslin (Mary McDonnell), a former secretary of education, 43rd in line of succession and rising to the occasion of her unexpected Presidency. As directed by Michael Rymer (Queen of the Damned), Moore's ambitious teleplay also includes newfangled CGI space battles (featuring "handheld" camera moves and subdued sound effects for "enhanced realism"), a dysfunctional Col. Tigh (Michael Hogan) who's provoked into action by the insubordinate Starbuck (Katee Sackhoff), and a father-son reunion steeped in familial tragedy.
To fans of the original BG series, many of these changes are blasphemous, but for the most part they work--including an ominous cliffhanger ending. The remade Galactica is brimming with smart, well-drawn characters ripe with dramatic potential, and it readily qualifies as serious-minded science fiction, even as it gives BG loyalists ample fuel for lively debate.
Over the course of the show's four seasons, the war between the colonists and the Cylons takes many twists and turns. Despite the animosity on both sides, the Cylons and humans slowly turn away from their hatred for each other. Part of this is due to a growing schism within the humanoid Cylons, led by the villainous Cylon Number One, Brother John Cavil. Cavil's obsession with hiding the true genesis of the humanoid Cylons (created by members of the Galactica Crew, who themselves are humanoid Cylons from "Earth" who had their memories erased by Cavil) leads to a civil war among the Cylons, with a faction of the robot race forming an alliance with the humans. Other plotlines involve the mysterious destiny of Kara "Starbuck" Thrace, who is the subject of a prophecy involving her as the "Harbinger of Death", who will "lead humanity to its end" as well as the redemption of Gaius Baltar, as he becomes a pariah within the fleet (after being forced to collaborate with the Cylons) but ultimately finds redemption through monotheism-based religion.
In the final episodes, Thrace (resurrected) leads the surviving Humans/Cylons to a new planet which Adama names Earth and which is revealed to be the actual planet Earth of present-day humanity. It is already inhabited by humans who are genetically compatible with the humans from the Galactica and the rest of the fleet. Human beings naturally evolved on both our world and Kobol (the original homeworld of the Colonials). The Earth of legend is now revealed to have been a different planet entirely which is a nuked out, uninhabited wasteland due to that planet's war with its Cylon creations. It is now revealed that all of the events of the series have been occurring 150,000 years prior to the era of present-day humanity. The entire group of humans/Cylons decides to live on the new planet and to discard all technology, and to destroy all of their spaceships by flying them into the Sun.
The series ends with a coda set on Earth in the present day, as two "angels", in the form of Caprica Six and Gaius Baltar, muse on whether or not the cycle of violence and war between humanity and machine will happen again or if it had finally been broken. The revelation in the series finale that the show took place more than 150,000 years before the present means that unlike most space opera science fiction stories, the re-imagined Battlestar Galactica was a tale of our ancestors rather than our descendants.
| 0
| 0
| 1,082
| 1,082
|
49,378
|
3691f5d1-500d-4999-845a-0132d72e6e71
|
trentmkelly/LessWrong-43k
|
The Fear of Common Knowledge
Followup to: Belief in Belief
One of those insights that made me sit upright and say "Aha!" From The Uncredible Hallq:
> Minor acts of dishonesty are integral to human life, ranging from how we deal with casual acquaintances to writing formal agreements between nation states. Steven Pinker has an excellent chapter on this in The Stuff of Thought, a version of which can be found at TIME magazine’s website. What didn’t make it into the TIME version is Pinker’s proposal that, while there are several reasons we do this, the most important reason is to avoid mutual knowledge: "She probably knows I just blew a pass at her, but does she know I know she knows? Does she know I know she knows I know she knows?" Etc. Mutual knowledge is that nightmare where, for all intents and purposes, the known-knows can be extended out to infinity. The ultimate example of this has to be the joke "No, it wasn’t awkward until you said, 'well, this is awkward.'" A situation might be a little awkward, but what’s really awkward is mutual knowledge, created when someone blurts out what’s going on for all to hear...
>
> The story of the Emperor’s New Clothes is another example of the power of mutual knowledge...
The power of real deception - outright lies - is easy for even us nerds to understand.
The notion of a lie that the other person knows is a lie, seems very odd at first. Up until I read the Hallq's explanation of Pinker, I had thought in terms of people suppressing uncomfortable thoughts: "If it isn't said out loud, I don't have to deal with it."
Like the friends of a terminal patient, whose disease has progressed to a stage that - if you look it up online - turns out to be nearly universally fatal. So the friends gather around, and wish the patient best hopes for their medical treatment. No one says, "Well, we all know you're going to die; and now it's too late for you to get life insurance and sign up for cryonics. I hope it isn't too painful; let me know if you want
| 0
| 0
| 491
| 491
|
77,370
|
<urn:uuid:ea23bdcf-0f67-4292-b698-4fb8762d8e16>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Skip to main content
Top of the chops: The Screening Room's top 10 kung fu films
• Story Highlights
• Iconic kung fu star Bruce Lee hits the top spot with "Enter the Dragon"
• Tarantino's genre tributes, "Kill Bill 1&2," are stylish and bloody
• "The Matrix" is a sci-fi kung fu movie with slickly made fight sequences
• But Jean-Claude Van Damme's "Street Fighter" lacks the killer punch
• Next Article in Entertainment »
Decrease font Decrease font
Enlarge font Enlarge font
LONDON, England (CNN) -- Bad dubbing, angry men with extraordinary facial hair, balletic fighting and more blood than you can shake a nunchuck at: just some of the key ingredients to make a perfect kung fu flick.
Tony Jaa may be the modern day Bruce Lee, and in "Ong-bak" he shows off his impressive skills.
Whether it's a traditional tale of ancient Chinese fighting mystics, or a slickly made, modern bloodfest, kung fu has always been a rich part of cinema.
We've compiled a list of 10 of the best. Don't agree? Think we've missed one? Share your views by using the Sound Off box below and we'll publish the best.
1. Enter the Dragon
(Robert Clouse, 1973)
Bruce Lee's last movie before his untimely death, this is him at his very best. A man on a revenge mission, Lee travels to a mysterious island to fight in a deadly tournament hosted by an evil billionaire. Along with being technically amazing -- the nunchucks scene is jaw-dropping -- it's the most stylish and iconic martial arts film ever made.
2. Ong-bak
(Prachya Pinkaew, 2003)
Tony Jaa is a modern master. He has appeared in very few major releases, but has already made a huge impact. "Ong-bak" is the perfect showcase for his extraordinary skills: whether he's fighting a roomful of people or taking part in the most exciting chase we have ever seen -- jumping through rings of barbed wire and sliding under moving cars while doing the splits -- he makes this film every bit the martial arts spectacle.
3. Wong Fei Hung (Once Upon a Time in China)
(Hark Tsui, 1991)
Part Chinese history, part gripping kung fu movie, this eastern epic has Jet Li demonstrating his talents as the eponymous hero who stands up to invading foreign forces in 19th century China. Armed with a limitless arsenal of martial arts moves, Li takes on masses of gun-toting, badly acting opponents. A visual delight and Jet Li's best work.
4. Kill Bill 1 & 2
(Quentin Tarantino, 2003/2004)
The Screening Room
CNN's monthly movie show brings you the inside track on the movie business.
Quentin Tarantino is a movie geek -- specifically, a B-movie geek -- and kung fu has long held a place in his heart. "Kill Bill," his tribute to the genre, ticks every box: revenge; a powerful, all-knowing nemesis; stylish fights, costumes and music; and blood, lots and lots of blood.
5. Wo hu cang long (Crouching Tiger, Hidden Dragon)
(Ang Lee, 2000)
Every eye-catching detail of this touching and beautiful film works. A love story with outlaws, witches and Shaolin monks is augmented by breathtaking cinematography and some fantastic performances from Chow Yun Fat, Michelle Yeoh and newcomer Ziyi Zhang. This visual feast tugs at the heart strings better than any karate chop could.
6. The Karate Kid
(John G. Avildsen, 1984)
A simple but well-executed plot sees Ralph Macchio's underdog work his way up an unconventional path to take on the mean and nasty establishment. It spawned a spate of copycat films, while thousands of children signed up to after-school karate; everyone now knows how to defend themselves by waxing on or off and that the best fighting move is obviously The Crane. Hiya!
7. Ging chaat goo si (Police Story)
(Jackie Chan, 1985)
Jackie Chan is probably China's biggest export to Hollywood, well-known for doing his own death-defying stunts. This is one of the films that got him noticed in the United States. When he fights, Chan is unstoppable: he uses every prop on the set and puts himself in enormous physical danger. Out and out fun.
8. Shogun Assassin
(Robert Houston, 1980)
Referenced several times in Tarantino's "Kill Bill" films, this forgotten classic must have set a record for the amount of blood spilt as our hero crosses the country on a revenge mission. Early on, Tomisaburo Wakayama says "they will pay with rivers of blood"; he isn't wrong. Violent and wonderful.
9. Siu lam juk kau (Shaolin Soccer)
(Stephen Chow, 2001)
Should this be in the best or worst list? It's definitely unique. Once described as "the best kung fu football film of all time," this is one of the more ridiculous films of the genre. A group of down-and-out martial arts experts form a five-a-side football team to take on a group of evil, drug-taking clones. It's as simple, and as silly as that. And it's one of our guiltiest pleasures.
10. The Matrix
(Andy and Larry Wachowski, 1999)
A sci-fi spin on the genre, this modern classic has all the key elements: our hero has superhuman powers; the bad guys, led by the relentless Agent Smith, form part of a shady, all-powerful organization intent on oppressing humankind; and the fight scenes are fantastically technical. Against all the odds, it appears Keanu really does know Kung Fu.
Don't agree? Think we've missed one? Read others' comments and share your views by using the Sound Off box below.
And the mainstream films that lacked the killer punch ...
Street Fighter
(Steven E. de Souza, 1994)
Jean-Claude Van Damme has made some good films, really he has. "Kickboxer" nearly made it into our top 10. But this is, by a huge margin, his worst effort. Most depressing, though, is that this was the last major outing for Raul Julia -- not the way he should be remembered. Great game; terrible film.
The Karate Kid, Part III
(John G. Avildsen, 1989)
The franchise came out of part two with a little credibility intact, but this last installment -- unless you include "The Next Karate Kid" (which no self-respecting "Karate Kid" fan would) -- managed to kill off our last shred of enthusiasm. Where the original might have inspired you to take up karate, this would put you off the sport, and probably kung fu movies too.
Bulletproof Monk
(Paul Hunter, 2003)
"I've got a great idea, let's take a well-respected Chinese actor and pair him up with an annoying teen-film actor in a mystical kung fu movie; it can't fail!" Unsurprisingly, this film was a disaster, shoe-horning silly fight scenes into a plot that spirals from implausible to disturbingly stupid. A frightful mistake, and one of our worst-named films to boot.
Under Siege 2: Dark Territory
(Geoff Murphy, 1995)
The first "Under Siege" was a surprisingly good film, and Hollywood's best martial arts expert acquits himself well, but this follow up is laughable. Seagal still fights like the best, but has lost the ability to speak, and whispers his way through the dialogue. Considering the action takes place on a train, we were amazed his co-stars could hear him. Come on, Steven, you were so much better when you were just a lowly, lowly cook.
Game of Death
(Robert Clouse, 1978)
Through no fault of his own, this is Bruce Lee's posthumous contribution to the list. This tacky movie features fight scenes shot before Lee died, interspersed with material filmed using lookalikes and a cardboard cutout of the great man. The fights are impressive, but the rest of the film smacks of a cash-in -- they even use footage from Lee's actual funeral. A desperate slur on Lee's memory. E-mail to a friend E-mail to a friend
• E-mail
• Save
• Print
| 0
| 0
| 1,933
| 1,933
|
29,530
|
9abf0cca-9e5f-4a2f-8a14-4dd88079d58f
|
trentmkelly/LessWrong-43k
|
[Retracted] More silent deletion; LessWrong's moderation needs to change
Edit: Fuck it, Rain is right. Eliezer totally unjustifiably deleted some of my upvoted comments earlier today and this post was about that, but really this is some petty drama and even the troll in me is bored of it. I won't delete this because then people will assume Eliezer did it as is his penchant lately, so I'll just leave it here so y'all can kindly downvote it into oblivion with my enthusiastic consent. Cheers.
| 0
| 0
| 120
| 120
|
33,607
|
79e7deb0-3137-4024-b3a0-958558a2ea39
|
trentmkelly/LessWrong-43k
|
Crypto quant trading: Naive Bayes
Previous post: Crypto quant trading: Intro
I didn't get requests for any specific subject from the last post, so I'm going in the direction that I find interesting and I hope the community will find interesting as well. Let's do Naive Bayes! You can download the code and follow along.
Just as a reminder, here's Bayes' theorem: P(H|f) = P(H) * P(f|H) / P(f). (I'm using f for "feature".)
Here's conditional probability: P(A|B) = P(A,B) / P(B)
Disclaimer: I was learning Naive Bayes as I was writing this post, so please double check the math. I'm not using 3rd party libraries so I can fully understand how it all works. In fact, I'll start by describing a thing that tripped me up for a bit.
What not to do
My original understanding was: Naive Bayes basically allows us to update on various features without concerning ourselves with how all of them interact with each other; we're just assuming they are independent. So we can just apply it iteratively like so:
P(H) = prior
P(H) = P(H) * P(f1|H) / P(f1)
P(H) = P(H) * P(f2|H) / P(f2)
You can see how that fails if we keep updating P(H) upwards over and over again, until it goes above 1. I did math the hard way to figure out where I went wrong. If we have two features:
P(H|f1,f2) = P(H,f1,f2) / P(f1,f2)
= P(f1|H,f2) * P(H,f2) / P(f1,f2)
= P(f1|H,f2) * P(f2|H) * P(H) / P(f1,f2)
= P(H) * P(f1|H,f2) * P(f2|H) / (P(f1|f2) * P(f2))
Then because we assume that all features are independent:
= P(H) * P(f1|H) * P(f2|H) / (P(f1) * P(f2))
Looks like what I wrote above. Where's the mistake? Well, Naive Bayes actually says that all features are independent, conditional on H. So P(f1|H,f2) = P(f1|H) because we're conditioning on H, but P(f1|f2) != P(f1) because there's no H in the condition.
One intuitive example of this is a spam filter. Let's say all spam emails (H = email is spam) have random words. So P(word1|word2,H)=P(word1|H), i.e. if we know email is spam, then the presence of any given word doesn't tell us anything about t
| 0
| 0
| 707
| 707
|
6,843
|
7e1ba378-e7ab-4344-a73f-4020760dcc5b
|
trentmkelly/LessWrong-43k
|
Covid-19 6/4: The Eye of the Storm
Still standing by this: Covid-19: My Current Model
Previous update posts: Covid-19 5/29: Dumb Reopening, Covid 5/14: Limbo Under
Remember last week when I opened with this?
> I remember when people on Twitter had constant reminders that today was, indeed, only Wednesday, or whatever day it happened to be. Time moved that slowly.
>
> Time has sped up again.
Well, yeah. Not so much anymore.
In March and April I found myself constantly checking Twitter and the financial markets for news, frantically hunting for ways to get a handle on what was happening in the world, worried everything would fall apart. Would our supply chains hold? Would we be able to maintain civil order? Would millions die? How can I keep my family and friends safe?
My beloved New York City was no longer a place one could live a reasonable life. So we fled. Even after that, great worry.
Then things started to calm down. Most of May, I increasingly managed to relax. We learned how to grill properly. I played a bunch of Assassin’s Creed Odyssey. I stopped checking Twitter or the stock market.
My biggest worries were dealing with getting the Emergents Alpha ready and Magic: The Gathering melting down under the weight of the companion mechanic in particular and its new design philosophy in general.
End of the Beginning
Then, on May 25, 2020, in Minneapolis, four policemen murdered George Floyd.
The resulting cycle of protests against police brutality causing more police brutality, which in turn amplifies the protests, has been ongoing each day since then. Huge crowds flood the streets. The crowds come disproportionately from communities with more Covid-19 cases than average. The crowds often yell. The police use tear gas on them, which causes coughing. The police arrest them and lock them in tight quarters, which is going to spread the virus even if the outdoor activities don’t, hopefully in small enough numbers this isn’t a huge deal.
To the extent other things are also happening, that o
| 0
| 0
| 474
| 474
|
62,231
|
51044dfa-edea-481e-a445-5870f946fb39
|
StampyAI/alignment-research-dataset/lesswrong
|
Distillation of Neurotech and Alignment Workshop January 2023
**Disclaimer:**This post is preliminary and doesn't yet fully align with the rigorous standards we typically aim for in LessWrong publications. It recaps the dynamic discussions from the neurotech for AI alignment workshop and therefore does not necessarily represent any single participant’s or author’s viewpoint. Shared ahead of the Foresight WBE workshop, our intent is to foster enriching discussion among participants. We welcome your input as we continue refining these insights.
Introduction
============
This document reviews key insights from a January workshop exploring the potential of Neurotech to contribute to the [AI alignment problem.](https://drive.google.com/file/d/1uK7NhdSKprQKZnRjU58X7NLA1auXlWHt/view) Neurotech, or neurotechnology, refers to the set of tools and methods designed to enhance understanding of the brain, manipulate its function, or interface directly with neural circuits, often for therapeutic or augmentation purposes. The AI alignment problem is the challenge of ensuring that the behavior of an artificial (general) intelligence system is in line with human values and intentions.
The workshop, attended by 16 technical researchers and domain specialists, aimed to jointly map out whether and how neurotech can be useful towards alignment. Our core goal was to cast the net wide for all possible solutions, with an emphasis on those relevant to shortened alignment timelines. This initiative was a follow-up to the Paris Foresight Vision Weekend in November 2022, where it became evident that Neurotech-based approaches to alignment warranted further investigation. The two-hour interactive workshop aimed to compile a comprehensive list of potential approaches and prioritize them based on technical feasibility and timeline relevance.
The workshop kicked off with a keynote presentation from David Dalrymple that laid the groundwork for participant discussion and defined key working definitions. This was followed by a brainstorming session, break-out room discussions, cluster identification and idea sorting, preliminary cluster prioritization (based on alignment impact, feasibility, timeline, and cost), and analysis of the prioritization results. David's keynote included a figure illustrating a potential breakdown of research intersecting Brain-Computer-Interface and Alignment.
Several **key insights** from the workshop include
1. Workshop participants brainstormed together, identifying six clusters of neuro-based alignment approaches:
1. BCIs to extract human knowledge
2. neurotech to enhance humans
3. understanding human value formation
4. cyborgism
5. whole brain emulation
6. BCIs creating a reward signal.
2. Each cluster aims to align AI with human values but with highly separable/unique approaches.
3. These categories likely don’t cover all potential approaches. In this post, we map these approaches with the hope that future workshops can explore and extrapolate to other less-common approaches that could hold potential.
4. The workshop identified trends that agreed with commonly held assumptions. For example, participants found that complex technologies take longer to develop but have a greater potential impact on AI alignment. The relationship between cost and the other variables remains unclear. Future workshops should address cost and expected value to better guide investment in the space.
5. Neurotech hardware development faces potentially long timelines that may fall behind the advent of AGI, which presents a major risk for all neurotech-based tradeoffs.
6. Neurotech for AI safety is in an early phase with high uncertainty. During this phase, exploring various potential solutions, even unlikely ones, is valuable. This can lead to crucial insights and guide more focused investments, increasing the chances of finding effective AI alignment strategies.
7. The balance between increasing AI capabilities and alignment impact through neurotechnologies is uncertain. If advances in capability outpace advances in alignment, they may be counterproductive. Technology confidentiality can mitigate this risk but adds to development complexity.
Outline of Top 6 Clusters
=========================
After the ideation & brainstorming phase, participants were tasked with identifying “clusters” of approaches. This task hoped to distill many ideas into broad categories that spanned most of the proposed solution set with fewer ideas to interrogate over the second half of the workshop. The participants defined the following six categories.
1. *BCIs to extract human knowledge*
1. **Summary:**We already create AI systems in our image. We regularly train AI systems to imitate human outputs, e.g. text, images, etc. These outputs, however, are only artifacts of human cognition and represent a tiny fraction of our intentions, thought processes, and values. This cluster intends to use BCI to expand the utility of information artifacts of human cognition beyond conventional forms.
2. **Example approach 1:** BCI is used to build an ontology of human values and bridge that ontology to AIs
1. **Example approach 1a (unsupervised):**BCI data is used to better interpret and define key human values in an abstract and global manner that is interpretable by AI. These might be inserted into models by methods including [activation engineering](https://www.alignmentforum.org/posts/5spBue2z2tw4JuDCx/steering-gpt-2-xl-by-adding-an-activation-vector#6__The_Eiffel_Tower_is_in_Rome) which can steer models. This could be achieved using machine learning to infer latent “value spaces” from an extremely large corpus of human neurodata. Unsupervised methods may be more successful because they are less likely to break with future models (or change in human moral values).
2. **Example approach 1b (supervised):**A training dataset (a large neurodata corpus) would be curated and labeled corresponding to the traits and values related to the time-varying brain states. This enables the model to learn to recognize and incorporate these values into its output.
3. **Example approach 1c:**BCI is used to better understand the neural basis for social behavior and/or moral emotions.
2. [*Neurotech to enhance humans generally*](https://www.lesswrong.com/posts/vEtdjWuFrRwffWBiP/we-have-to-upgrade)
1. **Summary:**We use neurotechnology to enhance human capabilities to stay ahead of the AI curve and/or to bootstrap our ability to generate solutions to the alignment problem.
2. **Example approach 1:**Enhance human AI researchers using brain-to-brain communication, e.g. increase cooperation by sharing semantic and contextual information with concepts.
3. **Example approach 2:**Generic neurotech cognitive enhancement accelerates AI safety research
3. *Understanding how humans form values*
1. **Summary:** This cluster looks at studying how humans form values to better train agents to form their values similarly.
2. **Example approach 1:** Investigate Eliezer Yudkowsky's claim that the human brain contains some innate cognitive structures or mechanisms that contribute to the development of moral reasoning. If correct, build this structure or mechanism into AGI and train it similarly to how humans learn.
4. *Cyborgism*[[1]](#fnesp64tr0npa)
1. **Summary:**Cyborgism is the idea of using technology such as AI to enhance human abilities and potential, for example by interleaving human and AI-generated cognition. Cyborgism could enable more effective communication of human values and preferences to the AI, and also enable the development of more transparent and explainable AI systems, which may help ensure that AI systems are used in ways that are beneficial to humanity. [[2]](#fnygtreema8cm)
2. **Example approach 1:**Train a 'brain embeddings to <behavior>' model, then use recordings of humans thinking about positive stuff to condition the model's behavior.
1. Analogy: we currently use 'text to <behavior>' models, then use text to condition the model's behavior. Presumably, brain embeddings will provide a more precise transfer of complex values than will text.
5. *Whole brain emulation*
1. **Summary:** Whole brain emulation (WBE) aspires to capture the functionality of biological/human brains by emulating its structure and function on a computational substrate. Despite advancements, the feasibility of WBE remains debatable, and it carries risks like sampling bias and the exacerbation of inequalities if access is restricted to a few individuals.
2. **Example approach 1:** The computational substrate model inherits all of human cognition including its values. Together with access to data and processing unlimited by biology, WBE agents are competitive with the AGI.
3. **Example approach 2:** group-aggregated whole-brain emulation, where a team of uploaded humans could work together to solve the problem of AGI alignment. This could potentially end the acute risk period and provide a collective intelligence that is better equipped to solve the alignment problem.
6. *BCIs to create a “reward signal”*
1. **Summary:** Brain-computer interfaces (BCIs) can potentially be used to create a "reward signal" that can be used to align the behavior of artificial general intelligence (AGI) systems with human values.
2. **Example approach 1:** Solve "value loading" by measuring "well-being" through humanity-scale deployment of accessible BCI. In other words, the BCI translates human values into a form that can be incorporated into the design and training of AI systems. This can happen asynchronously.
3. **Example approach 2:**Human cognition becomes the (live/closed loop) reward signal to a reinforcement learning based AGI agent: Wide-scale adoption of neurotechnology allows for AI to test humans' responses/thoughts in response to hypothetical actions.
**What clusters were not represented during the workshop?**
Due to the workshop’s restricted time of 2 hours, it’s likely that this is not yet an exhaustive list of possible clusters. To help identify technology clusters that may have been missed, we categorized approaches in a hierarchy by (1) information flow, i.e. from humans to AIs or from AIs to humans; (2) Ability; (3) Benefit to AI alignment (Fig. 1).
**Fig. 1 Cluster Map.** Technologies can be broken down into a hierarchy of categories. This structure may help to identify clusters that were not discussed during the workshop.Quantitative Assessment of Trade-offs
=====================================
To assess the key trade-offs between competing strategies to neuro-based alignment, we identified four critical variables: (1) timeline, (2) cost, (3) feasibility, and (4) impact. The working definitions of these four variables (exactly as they were provided to workshop participants) were:
**Timeline:**Assuming the neuroscience progresses very well, how long will it take to develop the interfaces (including: hardware, software, and methods) that make this possible? note: AI timelines are a controversial, but separable, topic. The goal here is to rank order each cluster from fastest to slowest (fastest = 1, slowest = n).
**Cost:**Rank each cluster from least expensive to most expensive (least = 1, most = n).
**Feasibility:** Assuming that sufficiently high performance neurotechnologies and AI methods are available, what’s the likelihood that each cluster's approach is feasible (including: technical, neuroscience, understanding of alignment; least feasible = 1, most feasible = 5)?
**Impact on AI alignment:**Assuming each approach was guaranteed to be successful, rank order the magnitude of positive impact on AI alignment (ignoring alignment timelines; worst = 1, best=5).
To establish a consensus on these variables, the workshop was divided into self-selected working groups. Each group had the task of rank-ordering technology clusters for each of these four key variables. The aim was to discern trends and identify outlier clusters/strategies.
**An analysis of key variable trends**
--------------------------------------
The relationships between feasibility, timeline, and impact on AI (See Fig. 2) align with commonly held assumptions. For instance, technologies that are more complex to develop were anticipated to take longer, suggesting an inverse correlation between feasibility and timeline (Fig. 2A). Similarly, technologies expected to yield a greater impact on AI alignment were predicted to require more development time (Fig. 2b) and were negatively correlated with feasibility. In other words, participants found that the technologies that were most likely to positively impact AI alignment were also the ones that were least feasible and would take the longest to develop.
**Fig. 2| Comparison on key variables.** **A.** Feasibility vs. timeline. Technology clusters that were deemed less feasible were also presumed to take longer to develop. **B.**Impact on AI vs. timeline. Technology clusters that were seen as having a larger potential impact on AI alignment were also presumed to take longer to develop. **C.** Impact on AI vs. feasibility. Technology clusters that were deemed more feasible were seen to be less likely to have an impact on AI alignment. Green trend lines represent high correlations (R2 ≥ 0.4318) and red represent low correlations.### **Cyborgism: an exception?**
Interestingly, Cyborgism appeared to diverge from the trends of the other approaches. Despite being consistent with the notion that less feasible technologies take longer to develop, it was not perceived to have a proportionate impact on AI alignment. Essentially, even though cyborgism might require substantial development time and be low in feasibility, its success wouldn’t necessarily yield a significant impact on AI alignment.
### **Cost**
The relationship between cost and the other three factors remained ambiguous. (Fig. 3). No discernible trends were observed between cost and timeline or impact on AI. There was a negative correlation between cost and feasibility, e.g. highly feasible solutions cost less and less feasible solutions cost more to develop. However, these results should be interpreted with caution. During this portion of the workshop, participants were self-selected into breakout rooms, where each room was tasked with ranking technology clusters for a single variable. No one chose to join the “Cost” breakout room. Instead, we (the organizers) quickly ranked the technology clusters with respect to cost. *We include the results for completeness but encourage the reader to interpret these results with caution* and suggest that future workshops revisit this topic.
**Fig. 3| Comparison of cost and other variables.** **A.** timeline vs. cost had no clear trend. **B.**Cost vs. feasibility. Technology clusters that were seen as being less feasible were also seen as higher cost. **C.** Cost vs. impact on AI had no clear trend. Green trend lines represent high correlations (R2 ≥ 0.4318) and red represents low correlations.In summary, the exercise underscored the complexity of the trade-offs involved in pursuing different strategies. It is hoped that these insights will contribute to a more informed and nuanced discussion on the path forward in the neurotech-based alignment space.
Discussion of Key Technical and Strategic Trade-offs
====================================================
**Aligning Neurotech and AGI Timelines**
----------------------------------------
Participants noted that the most impactful neurotechnologies for AGI alignment are likely the most complex and thus would require longer development times. This poses a risk that some of these technologies may not be ready before the anticipated emergence of AGI. Timelines and associated risks leading to AGI are still disputed. However, there was agreement that the longer the R&D cycle relative to AGI's progression, the less likely the technology will make a significant contribution.
Several strategies emerged from the discussion to address this issue:
1. **Better Estimation:** Improve estimates of AGI's arrival and prioritize acceleration of those neurotechnologies that could realistically be developed beforehand. This task is inherently challenging due to the difficulty in making precise estimates, and errors could negatively impact neurotechnology development.
2. **Slow AGI Progress:** Intentionally delaying AGI timelines could provide neurotechnology researchers with additional development time. However, this strategy necessitates community-wide cooperation that contradicts existing financial and personal incentives, making it a difficult proposition.
3. **Scale Resources according to expected value:** Allocate resources to each neurotechnology's development proportionally to its expected value, considering factors like impact on alignment, development timeline, cost, etc.
4. **More efficient use of resources:** Resource allocation should ideally include careful evaluation of the trade-offs between timeline, cost, feasibility, and impact. It will be critical to reduce redundancy and the risk of investing heavily in less feasible or longer-term projects. A strategic, trade-off-informed allocation of resources could maximize the effectiveness of our efforts toward AGI alignment. However, the attendees stressed that it is important to continue investigating all of the proposed neurotechnology clusters because of the unknown potential to significantly mitigate the risks posed by AGI.
5. **Expand resources:** The benefits of these complex neurotechnologies extend beyond AGI alignment. For example, many neurotechnologies have wide-ranging applications in healthcare, which could attract investments outside of the AGI alignment funding structures and community. However, caution is required to avoid potential distractions if the researchers' goals diverge from those of the investors.
**Feasibility & technical progress for top clusters**
-----------------------------------------------------
The feasibility of neurotechnology faces significant scientific and technological risks, especially when considering the concurrent advances and projected timelines in AI. In the breakout session, participants rank-ordered the feasibility (including: technical, neuroscience, understanding of alignment) of the different approaches. In addition, there was discussion about technical feasibility in the breakout session and asynchronous communication after the workshop. Here, we’ll share some of those thoughts.
There are two primary neurotechnology strategies. **The bottom-up approach**targets the structure and function of the central nervous system at its lowest functional level (e.g. a synapse), and scales it to an entire brain. Connectomics are tackling this issue, with large scale [efforts](https://e11.bio/) scoped to accelerate large scale connectomic research. Nevertheless, these approaches might not scale to the entire human brain before AGI arrives. **The top-down approach**measures brain function, often *in vivo* and in human participants at the scale of the entire brain, but with limited resolution. For example, functional MRI captures brain function with spatial resolutions of ~1 mm and temporal resolutions of ~1 sec. Although orders of magnitude away from single neuron activity, they still produce impressive results (e.g. reconstructing [vision](https://www.youtube.com/watch?v=nsjDnYxJ0bo) or [language](https://www.nature.com/articles/s41593-023-01304-9.epdf?sharing_token=eRF26q0CEKjXJe_xiwrYptRgN0jAjWel9jnR3ZoTv0NG3whxCLvPExlNSoYRnDSfIOgKVxuQpIpQTlvwbh56sqMXN0lZ9RZmdNtl6FGOIAG4FCtIHW1KJlM6y8opjMflLwC5y8nr_2Pf8epQHcEJyXmLOJ5iSW1y1NYLOhz2IXPFyCPrrwPR_3C2ZS70Bg7hvFhEqMbYO3BgDGvsg3V_0w%3D%3D&tracking_referrer=www.npr.org)). This demonstrates the power and promise of top-down neurotechnologies: they create useful results throughout their development that could contribute to alignment.
Top-down neurotech shows promise for AI alignment, but methods such as [electrophysiology](https://neuralink.com/), which invade brain tissue, are problematic. Techniques need to record cellular activity without causing damage, and current "[butcher numbers](https://www.youtube.com/watch?v=AwXnRTNLAhw)" (ratio of damaged to recorded cells) are unacceptably high for whole-brain scaling. Techniques based on optics and/or ultrasound offer promising, less-invasive alternatives.Optical techniques offer high spatial and temporal resolution. Unfortunately, photons scatter in tissue, limiting recording depth. Ultrasound penetrates soft tissue efficiently and is [highly sensitive](https://www.sciencedirect.com/science/article/pii/S0896627321001513) to neuronal function. It's diffraction-limited to ~100 micron resolution, but [super-resolution](https://www.nature.com/articles/s41592-022-01549-5) and [genetic engineering](https://patents.google.com/patent/US20230109152A1/en) techniques are improving spatial resolution and enabling more specific functional measurements. Other approaches based on different biophysical sources of contrast (e.g., [magnetics](https://quspin.com/)), delivery of these approaches to the brain through novel means (e.g., [intravascular](https://synchron.com/)), or the combination of multiple techniques, may also contribute to progress in neurotech for AI alignment.
Most of the clusters that the workshop participants identified include approaches that require neural measurements in behaving humans (1. BCIs to extract human knowledge, 2. Neurotech to enhance humans generally, 4. Cyborgism, 5. Whole brain emulation, and 6. BCIs to create a reward signal), which is most readily achieved using top-down techniques. Thus, advances in top-down neurotechnologies in one category would also likely benefit approaches in many others.
**Tradeoff: Investment for R&D in Neurotech approaches vs non-neurotech alignment approaches**
----------------------------------------------------------------------------------------------
Given current funding limits for alignment research, what is the impact of investing in neurotech-based approaches with alignment resources? As a comprehensive cost overview doesn’t exist, we’ll offer a qualitative examination, focusing on WBE as an example.
WBE may attract substantial commercial investment thanks to its potential for applications in health care, entertainment, lifestyle, well-being and longevity. WBE could develop independently of alignment investment or in parallel by commercial interest groups.
A potential drawback of non-alignment investment in WBE is the possible disregard for alignment-specific issues. For instance, should WBE development be confidential or exclusive? While commercial interest groups might independently pursue WBE, investments from the alignment community could notably accelerate its development. The credibility of this assertion depends on a more detailed understanding of the key technical obstacles for WBE.
A distinguishing factor for neurotech approaches is their reliance on hardware innovation. The quality of brain data is tied to the sophistication of recording equipment. Invasive devices, due to concerns about biocompatibility, device lifespan, and safety, require lengthy and expensive development. Unlike software innovation's rapid progress, hardware innovation is historically slower.
Another significant contributor to overall development cost and time is compliance with regulatory guidelines, such as ethical review and potential new medical legislation, before legally using WBE technology. This difference between neurotech and non-neurotech approaches is noteworthy, as the latter less frequently require ethical evaluation in the current legal landscape.
**Tradeoff on Investment: Maximally Accelerating a Single Technology vs. Diversifying Investment**
--------------------------------------------------------------------------------------------------
When contemplating resource allocation in AI alignment, a key strategic question arises: Should investment be concentrated in a single expensive technology, like WBE, to maximize its R&D speed, or should resources be spread across multiple approaches, each with a sub-maximum R&D pace? (This tradeoff is mostly only valid under the assumption of finite investment resources relative to cost, which does seem likely soon.)
Maximizing resources toward one approach could potentially reduce timelines, accelerate breakthroughs, and increase the probability of this technology being ready for deployment when AGI becomes a reality. In the case of WBE, it could also foster advancements in related fields, such as neuroscience and medicine, due to its wide-reaching application, which could foster further investment, which would (ideally) create a virtuous cycle of investment, progress, investment, and so on. On the other hand, this approach carries significant risks. Each of the approaches discussed here is fraught with uncertainties, both technical and scientific, and participants agreed that WBE was among the most uncertain. Concentrating resources in a single avenue risks total failure if the technology proves infeasible, and other approaches’ development is significantly behind.
Investing in a diverse portfolio of alignment approaches can mitigate this risk. This strategy hedges against the failure of any single approach. Given the high uncertainty surrounding each approach and the many inherent scientific and technological risks of each, exploration of many paths is likely advisable above exploiting a single approach with more resources and increasing the overall chance of success in the field of AI alignment. In summary, the choice between maximizing the acceleration of a single technology and diversifying investment across multiple approaches hinges on a complex interplay of factors. These include the perceived AGI timeline, the potential of each technology for alignment, the associated development risks, and the counterfactual impact of investment. As such, careful and considered strategic decision-making is vital.
**Balancing Alignment Impact and Capabilities Acceleration**
------------------------------------------------------------
It’s unclear how each of these technologies will contribute to AI safety vs. general AI capabilities.If the capability increase is larger than the alignment increase, it should possibly not be considered an alignment approach. For neurotechnologies, the outcomes and potential uses toward either end seem particularly uncertain. Additionally, the necessity of maintaining technology confidentiality for alignment impact is a crucial factor. In some instances, confidentiality might be unachievable due to potential parallel development by other industry stakeholders or the need for substantial investment that necessitates involving funders interested in accessing the technology.
Prioritization Summary
======================
Given the current scarcity of highly effective alignment strategies, the exploration of neurotech-based approaches is undeniably valuable. However, the substantial R&D uncertainty and high costs associated with neurotechnologies, compounded by the undefined nature of their potential impact, necessitate a deeper investigation into trade-offs and technical obstacles before substantially shifting resources away from any given set of approaches to accelerate another. This consideration is crucial, especially when weighing investment decisions between neurotech and potentially more cost-impact-efficient non-neurotech approaches.
Assuming resources are finite and sufficiently limited such that not all approaches can be maximally developed in parallel (which they currently are), we can draw on decision theory principles to frame these strategies as an exploration-exploitation trade-off. Exploration is optimal in the early stages of a problem, when uncertainty is high and information is scarce. Given the nascent state of neurotech-based AI alignment, we're in precisely such an exploration phase.It is therefore likely valuable to investigate a range of potential solutions, even those that seem unlikely or are less understood. Clusters like neuro & interpretability are affordable and might yield crucial insights. This knowledge can then inform more targeted investments (exploitation), streamlining resource spend and improving the probability of discovering effective and timely AI alignment strategies.
On the other hand, it may be possible that the total pot of resources is substantially increased as the general public gains awareness of the risks of AI. In particular, if public sentiment supports the idea that AGI timelines are sufficiently short, this further increases the probability of increased resources. For example, if legislators (reflecting public sentiment) are convinced of AGI timelines that are within typical voting cycles, political structures will be motivated to fund alignment research. This kind of increase in resources adds an interesting dimension to the strategic discussion. For example, we could explore many alignment strategies in parallel. However, there may be diminishing returns with increased funding, and there might be a talent scarcity problem wherein increased funding doesn’t necessarily equate to an increase in skilled researchers, again emphasizing the need for strategic distribution of resources. Additionally, navigating socio-political implications could become increasingly difficult with increased public and government involvement.
Summary & Next Steps
====================
Neurotech-based AI alignment strategies and their potential impact are in early stages but hold promise as exploratory solutions. If you feel that you have insights to contribute to further discussions or would like to participate in future workshops, we encourage you to get involved. Given the complexity, multi-disciplinary, and nascent nature of these approaches, additional perspectives are particularly valuable at this stage. We also host a slack space for online discussions which we welcome new members to. To reach out, feel free to respond through LessWrong or contact us directly at [[email protected]](mailto:[email protected]) or [[email protected]](mailto:[email protected]).
Acknowledgements
================
We like to express our gratitude to all the participants of the Neurotech and AI alignment workshop. Thank you for devoting your time and contributing to an insightful and productive discussion that has significantly deepened our understanding of this critical area. We also thank Milan Cvitkovic, David Dalrymple and Bogdan Cirstea for providing feedback on an earlier version of this post. Finally, we’d like to offer special thanks to David Dalrymple for delivering a compelling keynote presentation that set the stage for our discussions.
1. **[^](#fnrefesp64tr0npa)**Cyborgism (especially in a recent alignment agenda) is sometimes used more narrowly to mean “ using AI (primarily pretrained GPT models) to augment human cognition". However in this workshop we intentionally do not restrict the term to language model cooperation and also include uses associated with the term “cyborg”.
2. **[^](#fnrefygtreema8cm)**There are some concerns with this approach. For example, if the AI is able to directly influence an individual's decision-making processes, there is a risk that the individual's autonomy could be compromised. There are also ethical concerns associated with cyborgism. For example, if the integration of AI is only available to a select group of individuals, it could exacerbate existing inequalities and create new ones.
| 0
| 0
| 6,736
| 6,736
|
58,941
|
4f5ac4cb-277e-4b3d-a5e0-9b4b7c73e9a6
|
StampyAI/alignment-research-dataset/blogs
|
ediu46oG5oZN06ZQbvY9+jy6lzNLWAs4eHX0m7jn0zHW0M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| 66
| 396,000
| 404,000
| 726,576
|
22,868
|
3aad447d-e3e1-4cb1-ab50-85450a7ccbda
|
StampyAI/alignment-research-dataset/lesswrong
|
[CLS][RFC] Possible ways to expand on "Discovering Latent Knowledge in Language Models Without Supervision".
Preface
=======
*We would like to thank the following people who contributed to the generation of ideas and provided feedback on this post: Alexandre Variengien, Daniel Filan, John Wentworth, Jonathan Claybrough, Jörn Stöhler, June Ku, Marius Hobbhahn, and Matt MacDermott.*
We are a group of four who participate in [*SERI ML Alignment Theory Scholars Program*](https://www.serimats.org/) under John Wentworth, and we are extending the paper “[Discovering Latent Knowledge in Language Models Without Supervision](https://arxiv.org/abs/2212.03827)" as we see it as an interesting direction where we could contribute [concrete progress towards alignment](https://www.lesswrong.com/posts/L4anhrxjv8j2yRKKp/how-discovering-latent-knowledge-in-language-models-without).
We have a number of potential directions we could explore. The goal of this post is to get feedback on them early in order to prioritize better. This could be confirmations for the good ideas, reasons why certain ideas might be bad, references to existing similar attempts or relevant literature, potential failure modes, and generally (constructive) criticism of any kind.
Although we're all involved in all parts of the research process, the structure of our group and our main roles are as follows: Kaarel Hänni is our theorist and the main contributor of ideas; Walter Laurito and Kay Kozaronek are our experimentalists who focus mainly on deploying and running the code; Georgios Kaklamanos is our distiller and focuses on writing up and presenting the results (e.g., this post).
Brief summary of the DLK paper
==============================
In case you haven’t read the original paper, you can get a quick overview from the following [Twitter Thread from the author](https://twitter.com/CollinBurns4/status/1600892261633785856) ([alternative Threadreader link](https://threadreaderapp.com/thread/1600892261633785856.html)). Here we’ll just mention a few key points relevant to the things we want to work on. The above figure from the paper outlines their process in a clear way.
* The key insight of the paper is that coherence conditions satisfied by truth/falsehood can be leveraged to search for a model's internal representation of truth.
+ e.g., the question "Are cats mammals?" can be answered either "yes" or "no"
+ Thus, if we define as p(x+i).mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style: normal; font-weight: normal; font-size: 100%; font-size-adjust: none; letter-spacing: normal; word-wrap: normal; word-spacing: normal; white-space: nowrap; float: none; direction: ltr; max-width: none; max-height: none; min-width: 0; min-height: 0; border: 0; margin: 0; padding: 1px 0}
.MJXc-display {display: block; text-align: center; margin: 1em 0; padding: 0}
.mjx-chtml[tabindex]:focus, body :focus.mjx-chtml[tabindex] {display: inline-table}
.mjx-full-width {text-align: center; display: table-cell!important; width: 10000em}
.mjx-math {display: inline-block; border-collapse: separate; border-spacing: 0}
.mjx-math \* {display: inline-block; -webkit-box-sizing: content-box!important; -moz-box-sizing: content-box!important; box-sizing: content-box!important; text-align: left}
.mjx-numerator {display: block; text-align: center}
.mjx-denominator {display: block; text-align: center}
.MJXc-stacked {height: 0; position: relative}
.MJXc-stacked > \* {position: absolute}
.MJXc-bevelled > \* {display: inline-block}
.mjx-stack {display: inline-block}
.mjx-op {display: block}
.mjx-under {display: table-cell}
.mjx-over {display: block}
.mjx-over > \* {padding-left: 0px!important; padding-right: 0px!important}
.mjx-under > \* {padding-left: 0px!important; padding-right: 0px!important}
.mjx-stack >.mjx-sup {display: block}
.mjx-stack >.mjx-sub {display: block}
.mjx-prestack >.mjx-presup {display: block}
.mjx-prestack >.mjx-presub {display: block}
.mjx-delim-h >.mjx-char {display: inline-block}
.mjx-surd {vertical-align: top}
.mjx-surd +.mjx-box {display: inline-flex}
.mjx-mphantom \* {visibility: hidden}
.mjx-merror {background-color: #FFFF88; color: #CC0000; border: 1px solid #CC0000; padding: 2px 3px; font-style: normal; font-size: 90%}
.mjx-annotation-xml {line-height: normal}
.mjx-menclose > svg {fill: none; stroke: currentColor; overflow: visible}
.mjx-mtr {display: table-row}
.mjx-mlabeledtr {display: table-row}
.mjx-mtd {display: table-cell; text-align: center}
.mjx-label {display: table-row}
.mjx-box {display: inline-block}
.mjx-block {display: block}
.mjx-span {display: inline}
.mjx-char {display: block; white-space: pre}
.mjx-itable {display: inline-table; width: auto}
.mjx-row {display: table-row}
.mjx-cell {display: table-cell}
.mjx-table {display: table; width: 100%}
.mjx-line {display: block; height: 0}
.mjx-strut {width: 0; padding-top: 1em}
.mjx-vsize {width: 0}
.MJXc-space1 {margin-left:.167em}
.MJXc-space2 {margin-left:.222em}
.MJXc-space3 {margin-left:.278em}
.mjx-test.mjx-test-display {display: table!important}
.mjx-test.mjx-test-inline {display: inline!important; margin-right: -1px}
.mjx-test.mjx-test-default {display: block!important; clear: both}
.mjx-ex-box {display: inline-block!important; position: absolute; overflow: hidden; min-height: 0; max-height: none; padding: 0; border: 0; margin: 0; width: 1px; height: 60ex}
.mjx-test-inline.mjx-left-box {display: inline-block; width: 0; float: left}
.mjx-test-inline.mjx-right-box {display: inline-block; width: 0; float: right}
.mjx-test-display.mjx-right-box {display: table-cell!important; width: 10000em!important; min-width: 0; max-width: none; padding: 0; border: 0; margin: 0}
.MJXc-TeX-unknown-R {font-family: monospace; font-style: normal; font-weight: normal}
.MJXc-TeX-unknown-I {font-family: monospace; font-style: italic; font-weight: normal}
.MJXc-TeX-unknown-B {font-family: monospace; font-style: normal; font-weight: bold}
.MJXc-TeX-unknown-BI {font-family: monospace; font-style: italic; font-weight: bold}
.MJXc-TeX-ams-R {font-family: MJXc-TeX-ams-R,MJXc-TeX-ams-Rw}
.MJXc-TeX-cal-B {font-family: MJXc-TeX-cal-B,MJXc-TeX-cal-Bx,MJXc-TeX-cal-Bw}
.MJXc-TeX-frak-R {font-family: MJXc-TeX-frak-R,MJXc-TeX-frak-Rw}
.MJXc-TeX-frak-B {font-family: MJXc-TeX-frak-B,MJXc-TeX-frak-Bx,MJXc-TeX-frak-Bw}
.MJXc-TeX-math-BI {font-family: MJXc-TeX-math-BI,MJXc-TeX-math-BIx,MJXc-TeX-math-BIw}
.MJXc-TeX-sans-R {font-family: MJXc-TeX-sans-R,MJXc-TeX-sans-Rw}
.MJXc-TeX-sans-B {font-family: MJXc-TeX-sans-B,MJXc-TeX-sans-Bx,MJXc-TeX-sans-Bw}
.MJXc-TeX-sans-I {font-family: MJXc-TeX-sans-I,MJXc-TeX-sans-Ix,MJXc-TeX-sans-Iw}
.MJXc-TeX-script-R {font-family: MJXc-TeX-script-R,MJXc-TeX-script-Rw}
.MJXc-TeX-type-R {font-family: MJXc-TeX-type-R,MJXc-TeX-type-Rw}
.MJXc-TeX-cal-R {font-family: MJXc-TeX-cal-R,MJXc-TeX-cal-Rw}
.MJXc-TeX-main-B {font-family: MJXc-TeX-main-B,MJXc-TeX-main-Bx,MJXc-TeX-main-Bw}
.MJXc-TeX-main-I {font-family: MJXc-TeX-main-I,MJXc-TeX-main-Ix,MJXc-TeX-main-Iw}
.MJXc-TeX-main-R {font-family: MJXc-TeX-main-R,MJXc-TeX-main-Rw}
.MJXc-TeX-math-I {font-family: MJXc-TeX-math-I,MJXc-TeX-math-Ix,MJXc-TeX-math-Iw}
.MJXc-TeX-size1-R {font-family: MJXc-TeX-size1-R,MJXc-TeX-size1-Rw}
.MJXc-TeX-size2-R {font-family: MJXc-TeX-size2-R,MJXc-TeX-size2-Rw}
.MJXc-TeX-size3-R {font-family: MJXc-TeX-size3-R,MJXc-TeX-size3-Rw}
.MJXc-TeX-size4-R {font-family: MJXc-TeX-size4-R,MJXc-TeX-size4-Rw}
.MJXc-TeX-vec-R {font-family: MJXc-TeX-vec-R,MJXc-TeX-vec-Rw}
.MJXc-TeX-vec-B {font-family: MJXc-TeX-vec-B,MJXc-TeX-vec-Bx,MJXc-TeX-vec-Bw}
@font-face {font-family: MJXc-TeX-ams-R; src: local('MathJax\_AMS'), local('MathJax\_AMS-Regular')}
@font-face {font-family: MJXc-TeX-ams-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_AMS-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_AMS-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_AMS-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-cal-B; src: local('MathJax\_Caligraphic Bold'), local('MathJax\_Caligraphic-Bold')}
@font-face {font-family: MJXc-TeX-cal-Bx; src: local('MathJax\_Caligraphic'); font-weight: bold}
@font-face {font-family: MJXc-TeX-cal-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Caligraphic-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Caligraphic-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Caligraphic-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-frak-R; src: local('MathJax\_Fraktur'), local('MathJax\_Fraktur-Regular')}
@font-face {font-family: MJXc-TeX-frak-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Fraktur-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Fraktur-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Fraktur-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-frak-B; src: local('MathJax\_Fraktur Bold'), local('MathJax\_Fraktur-Bold')}
@font-face {font-family: MJXc-TeX-frak-Bx; src: local('MathJax\_Fraktur'); font-weight: bold}
@font-face {font-family: MJXc-TeX-frak-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Fraktur-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Fraktur-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Fraktur-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-math-BI; src: local('MathJax\_Math BoldItalic'), local('MathJax\_Math-BoldItalic')}
@font-face {font-family: MJXc-TeX-math-BIx; src: local('MathJax\_Math'); font-weight: bold; font-style: italic}
@font-face {font-family: MJXc-TeX-math-BIw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Math-BoldItalic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Math-BoldItalic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Math-BoldItalic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-sans-R; src: local('MathJax\_SansSerif'), local('MathJax\_SansSerif-Regular')}
@font-face {font-family: MJXc-TeX-sans-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_SansSerif-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_SansSerif-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_SansSerif-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-sans-B; src: local('MathJax\_SansSerif Bold'), local('MathJax\_SansSerif-Bold')}
@font-face {font-family: MJXc-TeX-sans-Bx; src: local('MathJax\_SansSerif'); font-weight: bold}
@font-face {font-family: MJXc-TeX-sans-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_SansSerif-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_SansSerif-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_SansSerif-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-sans-I; src: local('MathJax\_SansSerif Italic'), local('MathJax\_SansSerif-Italic')}
@font-face {font-family: MJXc-TeX-sans-Ix; src: local('MathJax\_SansSerif'); font-style: italic}
@font-face {font-family: MJXc-TeX-sans-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_SansSerif-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_SansSerif-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_SansSerif-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-script-R; src: local('MathJax\_Script'), local('MathJax\_Script-Regular')}
@font-face {font-family: MJXc-TeX-script-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Script-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Script-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Script-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-type-R; src: local('MathJax\_Typewriter'), local('MathJax\_Typewriter-Regular')}
@font-face {font-family: MJXc-TeX-type-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Typewriter-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Typewriter-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Typewriter-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-cal-R; src: local('MathJax\_Caligraphic'), local('MathJax\_Caligraphic-Regular')}
@font-face {font-family: MJXc-TeX-cal-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Caligraphic-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Caligraphic-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Caligraphic-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-B; src: local('MathJax\_Main Bold'), local('MathJax\_Main-Bold')}
@font-face {font-family: MJXc-TeX-main-Bx; src: local('MathJax\_Main'); font-weight: bold}
@font-face {font-family: MJXc-TeX-main-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-I; src: local('MathJax\_Main Italic'), local('MathJax\_Main-Italic')}
@font-face {font-family: MJXc-TeX-main-Ix; src: local('MathJax\_Main'); font-style: italic}
@font-face {font-family: MJXc-TeX-main-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-R; src: local('MathJax\_Main'), local('MathJax\_Main-Regular')}
@font-face {font-family: MJXc-TeX-main-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-math-I; src: local('MathJax\_Math Italic'), local('MathJax\_Math-Italic')}
@font-face {font-family: MJXc-TeX-math-Ix; src: local('MathJax\_Math'); font-style: italic}
@font-face {font-family: MJXc-TeX-math-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Math-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Math-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Math-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size1-R; src: local('MathJax\_Size1'), local('MathJax\_Size1-Regular')}
@font-face {font-family: MJXc-TeX-size1-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size1-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size1-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size1-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size2-R; src: local('MathJax\_Size2'), local('MathJax\_Size2-Regular')}
@font-face {font-family: MJXc-TeX-size2-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size2-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size2-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size2-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size3-R; src: local('MathJax\_Size3'), local('MathJax\_Size3-Regular')}
@font-face {font-family: MJXc-TeX-size3-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size3-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size3-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size3-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size4-R; src: local('MathJax\_Size4'), local('MathJax\_Size4-Regular')}
@font-face {font-family: MJXc-TeX-size4-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size4-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size4-Regular.woff
| 0
| 0
| 8,000
| 13,880
|
81,142
|
<urn:uuid:466f9fef-0d2d-486a-9238-0f135c6c99ba>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
[CLS]Entries in Sci-fi (92)
Star Wars: The Last Jedi- review
“Star Wars: The Last Jedi” (2017)
Directed Rian Johnson
Starring Daisy Ridley, John Boyega, Adam Driver, Mark Hamill
Running Time 152 Minutes, Rated PG-13
3.5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
The eighth episode in the Star Wars saga is tidy, efficient, has a few moments of surprise, and a few scenes that drop the jaw. Rian Johnson (Looper) directs “Star Wars: The Last Jedi” to a pleasant if not outstanding outing. One can assume that these films will come out well no matter what, but that just isn’t the case; sorry Ewan McGregor. So, one should be appreciative of a good product such as The Last Jedi; it’s just that I want to come out the theater forever changed, I want these films to have imprinted on the outcome of my life, and I don’t think that’s too much to ask, but Santa couldn’t make it out to that far away of a galaxy.
The Resistance is dwindling and putting all their chips into finding the in hiding Master Luke Skywalker which Mark Hamill gets to reprise for more than just 30 seconds of screen time this time around. Rey (Daisy Ridley) has found the Last Jedi, and the film goes into one of its many “odes” (carbon copy plot points) as Luke eventually trains her as Yoda did for him on a different lightly inhabited remote planet. Adam Driver is giving a Master Class in brooding as Kylo Ren with his dark side and Domhnall Gleeson is utterly delightful as the sniveling General Hux.
Some of the exotic locales don’t work for me as much, (cough, cough, the casino), but some of the twists do work for me (cough, cough, I’m not giving away any of the twists dummy). The force is strong with girl power in this one, as Rey, General Leia (R.I.P. Carrie Fisher), and newcomers to the series Vice Admiral Holdo (Laura Dern) and the spunky Rose Tico (Kellie Marie Tran) have the biggest effects when it comes to the fate of the galaxy in this one.
To channel Yoda; good enough, this time around it is..hmm.
Matthew 20:16
Thor: Ragnarok- review
“Thor: Ragnarok” (2017)
Directed by Taika Waititi
Starring Chris Hemsworth, Tom Hiddleston, Tessa Thompson
Running Time 130 Minutes, Rated PG-13
3.5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
Isaiah 45:5
What Happened to Monday- review
What Happened to Monday (2017)
Directed by Tommy Wirkola
Starring Noomi Rapace, Glenn Close, Willem Dafoe
Running Time 123 Minutes, Rated R.
2 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
It’s 2073 and overpopulation is muy mal, like asses to elbows mal. The solution is the Child Allocation Act, a law that makes it so that every family is only allowed but one child, and if another spermy gets by and you have another then that kiddo is nabbed by the Child Allocation Bureau led by a “really going for it” Glenn Close and they put your kid away in a hibernation chamber until things are little less crowded. Our story involves a group of adult septuplets all played, entertainingly I might add, by “The Girl and the Dragon Tattoo” actress Noomi Rapace who all take on the identity of one person named Karen Setten mentored by their grandfather played by a “not really going for it” Willem Dafoe. On a rotation, each Settman sister gets the day of the week they were named after which Rapace plays with a delightful “Orphan Black” treatment, until one day one of the sisters, Monday, considered to be the most responsible goes missing and all heck breaks loose.
Decent idea, decent cast, but very bland execution. You wouldn’t think that coming from the director of “Dead Snow”, but along with a former best blacklist script that feels really muddled, “What Happened to Monday” becomes instantly forgettable. This is one of those films where you can tell they only had 1-2 days to shoot with either/both Dafoe and Close only having but few select scenes each in not than many different locations. “What Happened to Monday” eventually gets interesting..eventually..but boy it sure does take a while, and too much bland silliness has passed over my retinas to really be able to jump back in and care enough. What I like attempted here but ultimately doesn’t work is that there are a lot of choices that go against conventional story structure later on in the movie. I can’t pinpoint why some of those bold choices don’t pay off except for to say that they come off a little too jarring; for example I’m pretty sure that this movie thinks that the Glenn Close’s puppeteer politician villain is actually the film’s hero and that the seven sisters Settman are really the dumb American making choices with their heart not their mind.
Doesn’t matter what day of the week it is, this film isn’t worth the watch.
Romans 12:4-6
War for the Planet of the Apes- review
“War for the Planet of the Apes” (2017)
Directed by Matt Reeves
Starring Andy Serkis, Woody Harrelson
Running Time 140 Minutes, Rated PG-13
2.5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
I was moved by “The Rise of the Planet of the Apes”, and I was equally moved by “The Dawn of the Planet of the Apes”, and everyone and their monkey brother was moved by Matt Reeves finale “War for the Planet of the Apes”, but not me. I wanted to love it, and I had the tears ready to fall but no dice. Why?
The culmination of the DNA altered ape Caesar played by the motion caption treasure that is Andy Serkis trying to bring freedom to his fellow primates goes through all the beats of a great war but they never felt like more than well plotted beats to me. Trying to combine the horrors or war of “Apocalypse Now” with a Colonel Kurtz-esq character played by an always clean shaven Woody Harrelson with the at times jovial prisoner camp teamwork of “The Great Escape”, this final Apes movie left me all things, kind of bored.
I’ll tell you what isn’t boring, this series special effects. I don’t say this lightly but this to me is the best use of special effects ever used in film. The damn dirty talking apes shown on screen are more real than any Madea film. What’s done in these films is top of the bar and has no rival. Unfortunately, the effects were but the icing to a very bland cake.
Matthew 10:34
Spiderman: Homecoming- review
“Spiderman: Homecoming” (2017)
Directed by Jon Watts
Starring Tom Holland, Michael Keaton, Zendaya
Running Time 133 Minutes, Rated PG-13
3.5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
Was not a fan of the last installment of Spidey. Andrew Garfield and Emma Stone had chemistry, but Marc Webb had no control over those two films, they were a mess where it felt like as much as possible was flung on the web to see what stuck and the answer was not much. Spiderman is back, and this time Sony gave control to Marvel who have a habit that they don’t; getting these kinds of pictures right and “Spiderman: Homecoming” is done right.
Young Brit Tom Holland got the new Spidey gig by showing off his parkour and dance skills, that and he can act. Holland is fantastic as the new Peter Parker, showcasing the energy you once had at 16 years old, he just so happens to have the most awesome of powers, and it’s a blast to watch him at it. How’d he get those powers you didn’t ask; and neither does this film thank the Lord Almighty! Correctly assuming that everyone in the world knows his origin story, they spend all but two lines of dialogue getting us up to date.
Peter has these powers, and after briefly getting to use them to his perceived potential with Team Tony Stark in “Captain America: Civil War” he was left on the sidelines getting nothing more than misdemeanor duty in Queens, NY (another miracle that it doesn’t have to take place in Manhattan). Peter sees his opportunity to shine when he comes across a high end and high teched thief who goes by the name Vulture Man, played by the amazing Michael Keaton who wonderfully represents the working class’ struggle. 16 year olds do what 16 year olds do and screw up, and they either learn from the lesson or they don’t. Peter is a smart kid.
Praise be for the eclectic cast that shows all the usual melanomas that Queens has to offer. I walked out thinking there wasn’t as much grandeur as needed, but then I realized that’s what gets his predecessor in trouble, having 13 villains and 16 subplots going. The CGI still felt a bit cartoony for me, and a very small part played by Donald Glover whom I enjoy came off very much like “I’m Acting!” which is the second time I’ve seen him do that (“The Martian”), but by the end of it I was happy to see that my neighborhood hero has returned.
Job 8:14
Alien: Covenant- review
“Alien: Covenant” (2017)
Directed by Ridley Scott
Starring Michael Fassbender, Katherine Waterston, Billy Crudup
Running Time 122 Minutes, Rated R.
1 Mitch out of 5
Mitch Hansch/movieswithmitch.com
Poor Michael Fassbender. This what Ewan McGregor must have felt like when he was doing the Star Wars prequels.
Being the main actor in a beloved film series that’s brought back by the director who started it all, has to be pretty exciting- and then your hard and good work end up in films that suck so, so, so much. Infuriatingly dumb, “Alien: Covenant” is the Phantom Menace of Xenomorph movies.
I was pretty forgiving to 2012’s “Prometheus”, in which Ridley Scott returned to the “Alien” franchise that got started all the way back in 1979. I gave it 3.5 stars out of 5 even though it had some definite flaws that were irking. Scott returns for the second prequel to “Alien” with “Alien: Covenant”, and just backhand slaps me right in the face by doubling down on what was wrong with “Prometheus” and making it so I could only focus on the wrongs.
The year 2104, a colonization ship named Covenant is in the middle of it’s long journey to start up life on a planet named Origae-6 when most of the the 2,000 colonists and 2,000 embryos are lost as a solar flare hits. The captain (a famous cameo) is lost, and Oram (Billy Crudup), a rare man of faith, is promoted. Much to the resistance of the captain’s now widow, Daniels (Katherine Waterston), Oram decides to forego the 7 years it would take to get to Origae-6 when the ship picks up on a transmission sent from a planet they were unaware of that is much closer and much more compatible to human life and head there. You’ll never guess what is on that planet.
All is fine enough, especially with the bright spot of being able to bathe the eyes on Dariusz Wolski‘s hauntingly gorgeous cinematography. But then our remaining crew, which are supposed to the brightest of the bright charged with repopulating the human race, make the mistakes of the Prometheus but turned up to 11. My stomach was twisting in horror, not that at which Scott was trying to deliver, but rather the horror of his inept characters and the choices they make when they land on an uncharted planet. It was seriously like watching a ‘Laurel and Hardy’ sketch as our ‘crew’ slash ‘knuckleheads’ made decisions that almost guaranteed their demise. This sequence gave me a real seething anger, a blunder so large and just plain dumb that I knew the film could not come back from it.
With very little character development, actors such as Demián Bichir and Danny McBride are wasted. They were no more useful than teens who had just had sex in a Friday the 13th film, cast as meat for the monster. The film’s real monster; a returning Michael Fassbender pulling double duty as the first android David from Prometheus and an upgraded version for the Covenant ship named Walter. A Cain and Abel approach questioning your creator played only so deep this time around with Fassbender’s strong acting making it watchable.
“Alien: Covenant” sludges along not knowing what to do with its Xenomorphs, and rather has to poorly focus on the evil of David. Watch out for a third act ‘twist’ that you’d have to be as dumb as the Covenant crew not to see coming.
I haven’t been this mad at a film in quite some time.
Gaurdians of the Galaxy Vol. 2- review
“Guardians of the Galaxy Vol. 2 (2017)
Directed by James Gunn
Starring Chris Pratt, Zoe Saldana, Bradley Cooper
Running Time 136 Minutes, Rated PG-13
4 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
Remember the first “Guardians of the Galaxy” and how out of left field it was, how it just knocked us on our butts but in a good way? One of Marvel’s better films for sure, we were treated to an action space rock opera (70’s and 80’s music utilized to perfection) that made Chris Pratt an A-list mega star and showcased Vin Diesel’s greatest acting performance to date. Sequels to megahits are a tricky thing no doubt. You have to repeat some of the tricks that got you there in the first place while evolving the characters journey and not just giving us a rinse and repeat cycle cash grab. Director James Gunn (“Slither”) who helmed the first one doesn’t totally knock us on our butts again for Vol. 2, but my dairyaire eventually made it to a seated floor position by the credits and it’s 4 end credit sequences!
There’s meandering in the middle, and the pace slows down to a space snail’s pace but all you have to do is interject Baby Groot (Vin Diesel) and all is forgiven. Groot, the muscle bound tree sorta humanoid, who sacrificed his life at the end of the first film, but survived as a seedling is now in what could be called his terrible two’s, and it’s the most adorable thing this side of any galaxy. We’re treated to some Baby Groot dancing as “Mr. Blue Sky” by Elo plays as the rest of the Guardians: Peter Quill aka Star-Lord (Pratt), bad-a assassin Gamora (Zoe Saldana), Drax the Destroyer (Dave Bautista), and Rocket the Raccoon (Bradley Cooper) but don’t call him a raccoon, do contract work battling with a giant squid.
Those tricks earlier mentioned; killer soundtrack, cuter than hell Groot, and wicked insults from Rocket are entertaining, but the film becomes good with the dysfunctional group trying to be a family along with Peter running away from his mother’s death of cancer when he was just a small boy. In Vol. 2 Peter’s dad, who he’s never met, shows up and just so happens to be a god, in the form of Kurt Russell (noice!) named Ego, who happens to be a living planet. The question is posed whether Peter will leave his fighting-space makeshift family for Ego, but that storyline never really is believable and Gunn does have a way of laying it on pretty thick at times.
But sometimes the thicker, the better. After said meandering, the third act kicks backs into gear action wise and family drama wise, highlighted by Michael Rooker who steals the whole friggen movie (Baby Groot included) as a blue-skinned gnarly toothed space pirate with a killer whistle that basically raised Peter.
This time around was definitely less fresh to eyes as the first time around, but that’s bound to happen. Vol. 2 still wailed, but I would be concerned about the diminishing returns for the inevitable Vol. 3.
Proverbs 22:6
Colossal- review
“Colossal” (2017)
Directed by Nacho Vigalondo
Starring Anne Hathaway, Jason Sudeikis, Dan Stevens
Running Time 109 Minutes, Rated R.
3.5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
If you’ve seen one ‘girl trying to overcome her personal demons of alcoholism and bad relationships with men that actually manifests into giant Kaiju monsters literally rampaging half way across the globe in Seoul’ movie, well, then you’ve seen them all.
Said no one ever.
I know Anne Hathaway’s has gotten the scorn of many movie goers, but the kid can act and it’s nice to see those talents put to the test. That they are in Nacho Vigalondo’s (“Timecrimes”) latest, “Colossal”. Hathaway plays Gloria, a party girl who as the film opens is broken up with and kicked out of her boyfriend’s (played by the everywhere Dan Stevens) swanky NYC apartment and is forced to move back to parents empty house back in the burbs that is conveniently unoccupied. Soon Gloria runs into old classmate Oscar (Jason Sudeikis) who has taken over his father's local bar. Maybe not the best place for someone trying to detox, Gloria who is out of work and gets a job bartending at Oscar’s bar. When the shifts end, she, Oscar, and couple bar regulars (Tim Blake Nelson and Austin Stowell) drink until the sun comes up.
At the same time, in Seoul, Korea a giant skyscraper sized Godzilla-type monster appears every morning around 8:05 am eastern standard time and begins to cause destruction on epic levels, appropriately becoming the phenomenon the world is glued to watching. Weirdly enough, these two storylines are connected. You can gather from the trailers how, but I’ll let you unravel that while watching for yourself. The beautiful thing Vigalondo and his actors create, especially Hathaway who is funny, broken, and empathetic all at once are the ways these storylines do merge and obvious but still poignant metaphors aren’t trampled on but delivered with in simpatico.
Toxic relationships that keep you from evolving from your patterned transgressions are a strong message here, but I also read into the way it can seem impossible to back away from a strongly abusive relationship and the buildings that crumble when you don’t. Not to mention just trying not to trip over yourself which does enough damage. This is all shown with strength from Vigalondo. I will say that while some may say the ending is powerful, and they get their comeuppance, I found one flaw in the way the monster portion of the movie was solved and how that was supposed to fix Gloria’s real problems as well. In the real world that’s not the way to go, and if you do there will be consequences. Well, enough sage jibber jabber from me, go see “Colossal”.
1 Corinthians 15:33
Saban's Power Rangers- Review
“Saban’s Power Rangers” (2017)
Directed by Dean Israelite
Starring Dacre Montgomery, Naomi Scott, RJ Cyler
Running Time 124 Minutes, Rated PG-13
4 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
Two sides of a coin. I was just lamenting about the other film opening this week how “Life” is too much a carbon copy of “Alien”, but the very next film I saw, “Saban’s Power Rangers” is no more than a “Breakfast Club” but with kids who get superpowers, and I loved it. Feels weird to type, but I really did love this Power Rangers movie. I’m not even the target audience although I noticed some who were in the theater, and they were enjoying it even more. Maybe not as ‘camp’ as its Saturday morning predecessors, but it’s just the right amount of corny because the casting of the kids is so good. I almost cried during the yellow ranger’s monologue. That shouldn’t happen.
I’m not gonna go much into the plot. Five kids get powers, Might Morphin Power Ranger powers, but it takes time to learn how to use them because they have to learn how to both get to know each other and especially themselves. Told you it was corny, but it’s also fun, as a film about being able to be special should be. The writing is both touching and has some good one-liners. The lead, Dacre Montgomery who was made in the Zac Efron cloning tank is spunky as the Emilio Estevez type from “Breakfast Club”, and I think it’s actually really cool that they gave the Blue Ranger Billy (RJ Cyler) aspergers. I think the fact that it was handled very well without being shoved down our throat as a PSA is pretty uplifting. Throw in Bryan Cranston and a very hammy Elizabeth Banks, and baby you got yourself a good movie stew.
This is one of those films I’ll eventually take my kids to rolling my eyes but come out gushing over it more then them.
1 Corinthians 1:10
Life- Review
“Life” (2017)
Directed by Daniel Espinosa
Starring Jake Gyllenhaal, Rebecca Ferguson, Ryan Reynolds
Running Time 104 Minutes, Rated R.
2.5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
Things don’t tend to work out the way we’d like them to in life. Not in actual life (my jump shot), and unfortunately not in Daniel Espinosa’s Alien-lite sci-fi R rated thriller “Life”. I could see how, on paper, there was enough for committed interest from such a talented cast as big A-listers Jake Gyllenhaal, Ryan Reynolds, and eventual household name Rebecca Ferguson, who won moviegoers over with bad A turn in the last Mission Impossible film. It’s set up as a fast paced monster B movie with the budget of an A, but after a good start, the weight of the material simply fizzles out interest beyond repair.
A small soil sampling has been recovered by six astronauts on an international space station. Inside that sample is the first proof of alien life; a sporulating cell structure, later named Calvin by a grade school back on Earth, what is unique about Calvin is the fact that each cell has brain, muscle, and sensory organs all in one. Small and adorable at first, eventually Calvin gets bigger and breaks free from it’s curious human captors, and the body horror begins as “Life” becomes a chess game of squid-like Mars alien and the Earthling spacemen at hand. The semantics are posed but not really delved into that Calvin is not a monster but simply doing what all creation does: fight for survival. It just so happens the crew is the only organic matter around.
The aforementioned Calvin breakout scene is a good one, but, unwisely, the best Espinosa has to offer, mixing sci-fi horror with some well paced tension. The rest of this film is a less than thrilling move of those chess pieces. There’s some benefit in watching the chess pieces, but I don’t know too much about those pieces and ultimately I saw where those pieces were going to be six moves ahead; even as clever as the player moving the chess pieces around thought they might be.
Deuteronomy 6:24
Kong: Skull Island- review
“Kong: Skull Island” (2017)
Directed by Jordan Vogt-Roberts
Starring Tom Hiddleston, Samuel L. Jackson, Brie Larson
Running Time 118 Minutes, Rated PG-13
2 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
A failed Vietnam War allegory as the latest version of the giant ape, “Kong: Skull Island”, painfully beats its chest trying to be an Apocalypse Now monster movie mash up to no avail. Kong Screenwriter Max Borenstein, who also wrote the last Godzilla, is too focused on spearheading the two’s impending monsterverse to concentrate on the human element as the films leads are basically bystanders while the impressive CGI components do battle.
As the Vietnam War is about to conclude, one more mission is sent to a career soldier played by Samuel L. Jackson and his outfit. They are accompanied by scientists including John Goodman, an expert mercenary tracker played by Tom Hiddleston, and an ‘anti-war’ photographer played by Brie Larson. Goodman’s character says monsters do exist, and in no small size he is correct as they depart to an uncharted island where all Kong breaks loose.
The special effects are the fun of the movie, and Kong shines but I liked the giant creepy crawlers in Peter Jackson’s version better. Borenstein tries to have too many characters, and neither him nor director Jordan Vogt-Roberts can make any of them compelling. Hiddleston and Larson have so little do as the leads, especially Hiddleston’s sensitive tough guy really adds nothing story wise. Jackson’s character is way too close to Colonel Kurtz as the film dimly points at what the “real price of war” is, and “who the real enemy is” and the soundtrack is as cliche spot on as Vietnam films go. John C. Reilly pops up and gives the film life but is delegated to a unintentionally laughable credits sequence.
“Kong: Skull Island” slips on a giant banana peel this time around.
Numbers 13:33
Logan- review
“Logan” (2017)
Directed by James Mangold
Starring Hugh Jackman, Patrick Stewart, Dafne Keen
Running Time 135 Minutes, Rated R.
3.5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
The heart of Wolverine has finally made it onscreen.
It only took Jackman’s 9th time portraying the character to really feel it. This is no fault to Hugh Jackman, who got the career changing role after Dougray Scott had to drop out for Mission Impossible 2 reshoots (ouch), and has killed it ever since. It’s just tough to truly capture a 200 year old plus dude who has knives for hands and uncontrollable berserker rage because all he loves eventually dies because of his name.
“Logan” director James Mangold nearly got it right with the character’s last solo outing in “The Wolverine”, but a wrong third act where the studio took over and Mangold’s very nice character piece up to that point was pushed to the side. That was a still a decent picture overall, but now with Jackman declaring that this would be his last Wolverine/Logan outing he had the chip stack to bluff the studio out of a PG-13 superhero blueprint, Mangold was given the canvas with an “R” rating and not shackled with having ‘fate of the world’ stakes, but instead this beloved actor played with his beloved character and Logan’s ‘soul at stake’.
Thy soul has been redeemed with “Logan”.
Somewhere and sometime after “X-Men: Days of Future Past” we find a weary Logan who has lost everyone he’s ever cared for besides a senile Charles Xavier (Sir. Patrick “Friggen” Stewart) whose mind is the first to be classified as a weapon of mass destruction by the U.S. government. If it wasn’t for Xavier, Logan just might use the one adamantium bullet he hast left on himself; the only thing that he knows for sure that would end his seemingly immortal existence. That is until someone else comes along for Logan to protect. An 11 year old girl named Laura (an impressive debut from Dafne Keen) who has the same ‘gifts’ that Logan has with just as much anger to boot. With much convincing from Xavier, the three head out on a road trip to North Dakota where a young mutant ‘eden’ is said to be had.
“Logan” is a sloppy film at times. You can feel the different drafts and how different story aspects were meshed together. There are plot holes as well. Like giving an arbitrary finish line in the film of getting to a border and all danger for our mutant outlaws will cease. Regardless of that and the feeling that one very important scene midway was too underplayed so that a similar scene towards the end could be played bigger, “Logan” is a winner. Intimate, touching, tremendous violence that actually resonates to the viewer instead of just bouncing off our bloody immune eyes, this is a film that goes smaller but hits harder than any X-Men previously ever could. We have been blessed to see Jackman play so many chapters of this character, and play them with such force that he has changed the landscape of cinema.
Proverbs 25:28
Rogue One: A Star Wars Story- review
“Rogue One: A Star Wars Story” (2016)
Directed by Gareth Edwards
Starring Felicity Jones, Diego Luna
Running Time 134 Minutes, Rated PG-13
4 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
The best ones stay with you. I came out of Roge One feeling good and feeling satisfied, but the scenes have stayed with me over the past few days. Something drives this film to another level, one could say that this film has a certain- wait for it…..- force to it (lame).
It’s right there in the title: A Star Wars Story, and that it is. A story that precedes to the point of exactly right before A New Hope. A story that deepens and enriches the Star Wars lore and mythology but so much more importantly, this is a story that holds up on its own.
On imdb the genre listed is sci-fi, but it should be listed as a war film first. George Lucas didn’t mince in the New Hope that the Imperial Guard was the Nazi regime and Rogue One tells this from the rebel side of a war torn Europe ravaged by it’s evil and it was the Rebels that back in WW2 helped change the tide for victory and that same power is translated to the Star Wars universe. There is power there in it’s message because this did happen, sure there weren’t lightsabers, but there was a sweeping hate that took advantage of people’s fear and masqueraded as the right way.
If I ever wondered why the Death Star had such a crippling flaw as it does, “Rogue One: A Star Wars Story” gives you a worthy and plausible explanation. Without getting into too much detail, Felicity Jones fiercely plays Jyn Erso the daughter of the scientist who created the Death Star, which is now just being completed and is ready to be unleashed on a less than unified Rebel Alliance. Erso is rescued by a Rebel intelligence agent played by Diego Luna and his trusted by sassy say-whatever-is-on-it’s-mind’ droid K-2SO (voiced by “Firefly” actor Alan Tudyk). Later the rag tag crew is joined by a blind warrior named Chirrut Imwe (Donnie Yen who is working on his ‘force’ skills, and I’m pretty sure his common house law scruffy nerf herder looking haired NRA card toting boyfriend Baze Malbus (Jiang Wen); rounded out by defected Imperial pilot Bodhi Rook (Riz Ahmed).
Forest Whitaker has some great scenes as a guerilla warfare leader, and even better Darth Vader returns and his scenes do not disappoint. Besides a ‘not worth it’ cameo by the “A New Hope” deceased Peter Cushing the rest of the lore doesn’t feel forced. Kudos also to the ending, not the most Disney way to end it and that’s a good thing.
Rogue One is a good thing.
Proverbs 17:11
Arrival- review
“Arrival” (2016)
Directed by Denis Villeneuve
Starring Amy Adams, Jeremy Renner, Forest Whitaker
Running Time 116 Minutes, Rated PG-13
5 Mitch’s out of 5
Mitch Hansch/movieswithmitch.com
Whomever decided to release this film directly after the election is a genius.
Speaking of genius, Denis Villeneuve’s “Arrival” is just that. With how increasingly difficult it is for a world to communicate with each other, even though technology has granted us so many opportunities for it, how cathartic it is for a story to come along that shows immense benefits reaped from such a noble endeavor that is communication. And cathartic this film is, the heavy hurt of such treacherous communication from this recent election (from BOTH sides) lifted, if not temporarily, but in essence, show how it could be cured. Film is so beautiful in the way it can do that... if so inclined.
Villenueve’s “Arrival” introduces first contact with extraterrestrial life the same way many of us came to find out about 9/11. Silently frozen to our television screens watching something of this magnitude start to unfold. That’s how Amy Adams’ Dr. Louise Banks, an expert linguist, finds out about it
| 0
| 0
| 8,000
| 12,590
|
54,644
|
09baa0ca-4189-4d8e-b8b3-404d3df19f0e
|
trentmkelly/LessWrong-43k
|
Age gaps and Birth order: Reanalysis
This post follows on from my previous post detailing some areas where I was unable to reproduce Scott’s analysis of how the age gap between siblings modifies the SSC Birth order effect. I suggest you read that post first but here’s the summary:
> I attempted to reproduce Scott’s analysis of Birth order effect vs Age gap. I found that:
> There appeared to be an error in graphs 2 & 3 where people with one sibling were counted when they shouldn’t have been (graph 2) or were counted twice (graph 3)
> Comparing oldest children to youngest children causes a bias in the results which can be prevented by comparing oldest children to 2nd oldest children
> I was unable to reproduce Scott’s result on people reporting 0 year age gap – I get a non-significant 58% older siblings compared to Scott’s 70%. I was unable to discover the cause of the difference.
Summary
I reanalysed how sibling age gap modifies the SSC birth order effect. I found that:
The birth order effect is relatively steady for the first 4-8 years of age gap at about 70% respondents being the firstborn vs secondborn. For larger age gaps the effect reduces. There is insufficient evidence to conclude how long this reduction takes or whether the effect is completely removed at very large age gaps.
2 other trends were noted in the data but evidence for them was not strong:
* The reduction may not be the same (or might disappear) for larger families
* Birth order effect may be lower at 1 year age gap vs 2-7 year age gap
Considering competing theories on the cause of the Birth order effect, two theories fit the data well:
* Intra-family dynamics
* Decreased parental investment
And three theories fit the data poorly:
* Changed parental strategies
* Maternal antibodies
* Maternal vitamin deficiencies
Introduction
The original reason for me looking at this data was to analyse whether the data support a sudden drop between years 7 and 8 or whether there is an alternative explanation which fits the data
| 0
| 0
| 446
| 446
|
72,251
|
5923b715-9b51-49f4-839b-f53de35d4898
|
StampyAI/alignment-research-dataset/eaforum
|
A Theologian's Response to Anthropogenic Existential Risk
Hi all,
This is very much someone outside the bailiwick of this forum looking in, but I was told it could be interesting to share this article I wrote recently.
I'm a Catholic priest, with a prior background in Electronic Engineering, currently working on a PhD in Theology at Durham University. I am researching how the Catholic Church can engage with longtermism and better play its, potentially significant, part in advocating existential security. I'm particularly interested in how a Christian imagination can offer unique evaluative resources for attributing value to future human flourishing and to develop a sense of moral connection with our descendents, better motivating the sacrifices safeguarding the future demands.
Much of the material will be very familiar to you as the article was written for a Catholic publication, and so also serves to introduce and promote some of the basic ideas to a new audience.
I'm certainly interested to receive any comments or questions!
Called to Share the Father’s Love for Humanity’s Future:
--------------------------------------------------------
### A Scriptural and Patristic Perspective on Eschatological Cooperation in the Age of Anthropogenic Existential Risks
As the 16th day of July 1945 came to a close, the sun set over a changed world. For the first time, humanity had detonated an atomic bomb, and after the destruction of Hiroshima and Nagasaki later that year, society struggled to come to terms with the forces unleashed. Amidst the cacophony of devastation and the uproar of anti-nuclear movements, there were those who caught whispers of a dark threshold quietly crossed. One such thinker, Bertrand Russell, stood in the House of Lords to describe the shadow of a new kind of threat:
*We do not want to look at this thing simply from the point of view of the next few years; we want to look at it from the point of view of the future of mankind. The question is a simple one: Is it possible for a scientific society to continue to exist, or must such a society inevitably bring itself to destruction? ... As I go about the streets and see St. Paul's, the British Museum, the Houses of Parliament, and the other monuments of our civilization, in my mind's eye I see a nightmare vision of those buildings as heaps of rubble, surrounded by corpses.[1]*
Russell recognised that the development of nuclear weapons marked the dawn of a new age: humanity had become its greatest risk to itself. Adam and Eve, in eating the forbidden fruit, opened the way to individual death, but we have now “eaten more deeply of the fruit of the tree of knowledge” and are now “face to face with a second death, the death of mankind.”[2] An antithesis of God’s creatio ex nihilo, we have obtained our own absolutising power, the “potestas annihilationis, the reductio ad nihili.”[3]
A philosophical response to this new power suggests that threat of nuclear apocalypse is but one example of a category of anthropogenic existential risks (AXRs). Other self-caused threats to humanity’s future potential also include engineered pandemics, human-caused climate change, and unaligned artificial intelligence, all of which could cause existential catastrophe. Further AXRs still await discovery, and we have no reason to believe these will be less hazardous.[4] Without action, the danger humanity creates for itself will continue to grow and Ord, from Oxford’s Future of Humanity Institute, argues such increasing risk is unsustainable. We will either learn to mitigate existential risks or one of them will eventually play out, causing a permanent loss of humanity’s potential. In the past, survival could be taken for granted as natural threats to the human species are vanishingly rare on the timescale of human history, and technology had yet to develop existential threats of its own making. Yet Ord warns that now our “long-term survival requires a deliberate choice to survive.”[5] He illustrates his point by describing humanity in its adolescence, in urgent need of moral maturation; “just coming into our power, just old enough to get ourselves in serious trouble.”[6] We would certainly be foolish to neglect mitigating natural existential risks,[7] but the possibility of human agency playing a role for species survival, by causing or mitigating anthropogenic risks, poses unique philosophical and theological questions.
How might a Christian respond to these? Why not scepticism? The psalmist assures that “the righteous shall inherit the land and live in it forever (Ps 37:29);” Jesus himself told us not to worry about tomorrow for “today’s trouble is enough for today (Matt. 6:34),” and asked, “can any of you by worrying add a single hour to your span of life (Luke 12:25-26)?” Trust in providence, however, now coexists with awareness that good choices contribute to long and healthy living. By analogy, when any given generation’s influence over mankind’s long-term potential was negligible, trust was the believer’s disposition towards the future. Now, however, faith in divine providence must coexist with awareness that our choices make a real difference for, and could even destroy, the future of humanity. I propose, on Newmanian terms, that theology needs to assimilate awareness of the potestas annihilationis, and so long as no discontinuity of principles arises, any theological implications and adjustments inferred by the discovery of AXRs would constitute authentic development of dogma.[8] There is much to learn about humanity’s role within salvation history from even a brief return to the sources of faith in light of this new power. Assuming dystopian scenarios are the reserve of science fiction has become a dangerous option because we will not get a second chance if that assumption is proved wrong.
Before all else, Genesis reveals a fatherly love. God’s will that humanity be fruitful (Gen. 1:28) culminates in His promise to make Abraham’s “descendants as numerous as the stars in the sky (Gen. 26:4).” Even in the fifth century Augustine understood it was “nonsensical to be sure there are not stars that cannot be seen,”[9] and we now know of at least two trillion galaxies, each with billions of stars.[10] Perhaps God’s promise to Abraham can be taken literally after all? We know of no reason why humanity could not survive for millions of years,[11] so we discover the possibility of “a truly staggering number of descendants, with the time, resources, wisdom, and experience to create a diversity of wonders unimaginable to us today.”[12]
This vast potential future need not be dismissed as fanciful, for Jesus’ commissions of evangelisation show no temporal limitation. In Matthew, the charge to “make disciples of all nations (Matt. 28:19, emphasis added),” given at the ascension, binds expectation of the Lord’s return to a universal mission, in response to Jesus’ promise to be with us ‘to the end of the age. (Matt. 28:19).” Jesus’s lordship is not limited to the world as known to his time on earth, as the Easter liturgy reminds us: “He is the Alpha and Omega, to whom all time belongs.”[13] When the first Christians realised Jesus was not coming back soon, they realised the Church had a future of service to “a cosmic Lord,” with a horizon of discipleship “as wide as the world.”[14]
Saint Paul’s letters exemplify the broad scope of the call to evangelise, as “the horizons of Christian mission are pushed beyond the ethnic boundaries of Jew and gentiles to embrace the entire universe:”[15] Colossians celebrating how the Gospel grows in “the whole world (Col. 1:6.),” and Ephesians lauding God’s “plan for the fullness of time (Eph. 1:9-10).” Certainly by the time of writing of 2 Peter, the acceptance of the delay of the Parousia, and the significance therein for the Church’s mission, were firmly established: The Lord, to whom “a thousand years are like one day” is being patient with humanity, “not wanting any to perish, but all to come to repentance (2Pet. 3:8-9).”
On such scriptural foundations, the spread of the Gospel became a measure of readiness for the Fathers of the Church, who prayed that salvation is not drawn to a close too soon that humanity might have the best possible chance to accept the Gospel in Augustine’s “time of mercy.”[16] Along these lines, Hays and his collaborators from the Oxford Postdoctoral Colloquium on Eschatology have argued from the Tradition that we cooperate within salvation history, and that the time of the eschaton is not just unknown, but un-fixed, contingent in some way on the action of humanity.[17] Many of the Fathers of the Church believed the eschatological timetable can be “delayed or hastened by the piety, prayers, and penitence of God’s people.”[18] Justin Martyr explains that “God delays causing the confusion and destruction of the whole world... because of the seed of the Christians, who know that they are the cause of preservation in nature.”[19] Likewise, Tertullian prays for “emperors, for their ministers and for all in authority, for the welfare of the world, for the prevalence of peace, for the delay of the final consummation.”[20] Perhaps contingency helps to explain why Jesus tells his disciples that He does not know “the hour (Mark 13:32)?” Our Lord says plainly that the “good news of the kingdom will be proclaimed throughout the world, as a testimony to all the nations; and then the end will come (Matt. 24:14).”
Bringing the fixity of the eschaton into question has considerable ramifications for a Catholic response to AXRs, particularly considering human freedom, which “has the power to make choices for ever, with no turning back (CCC §1861).” To paraphrase Newman, in my freedom and human weakness, I can fail to do the definite service God has created me for. I can fail to live out my vocation, and even a committed Augustinian notion of perseverance as entirely God’s gift maintains that “a man who does not persevere fails by his own fault.”[21] By analogy, if human flourishing has a role to play in God’s plan for the last-things, and humanity itself can utterlyfail by self-destruction through AXRs, then we must take seriously the idea that anthropogenic extinction could just be that, humanly, not divinely, caused. A cosmos without humanity is now conceivable, something Aquinas considered as an impossible supposition.[22] I therefore make a bold, but simple, claim: Awareness of humanity’s newfound potestas annihilationis exposes a presumption that humanity will persevere until some appointed end moment. In the past, if human history were to be ending through some natural cause, a believer would have recourse to an eschatological explanation. Apocalyptic interpretation, however, cannot be uncritically applied to AXRs. Do believers have grounds to presume the Parousia would be triggered in response to mankind’s self-extinction so as to ensure the end of humanity and the end of creation be coterminous?
I contend that theology cannot now remain unchanged because the potestas annihilationis was unthinkable throughout the formation of dogma. Pannenberg’s view that “what is undoubtedly true in science cannot be wrong in theology”[23] suggests we need to consider the possibility that, by failing to mitigate AXRs, humanity could fail to live the fullness of life God would otherwise will for it. Some might argue that anthropogenic extinction would just be the final catastrophe scripture foretells, but to resist shouldering new responsibility on such apocalyptic terms constitutes little more than a projection of human failure onto a diminished eschatological screen, counsel to despair. Hays’ case that the end of history is temporally undetermined establishes contingency on humanity succeeding to live out God’s will, not on humanity failing, as if the eschaton were in some way history’s safety net in case of man-made calamity. God trusts and hopes that we mature and develop the moral capacity necessary to mitigate the risks to our future we cause for ourselves.[24]
Consider a Christian adjustment of Ord’s adolescent humanity: our species can be thought of as a young person, with both a vocation to sanctity and the capacity to squander it. Our loving God wills that His child grow in virtue and make the necessary choices to protect and answer His call. Like a young believer, humanity must now make a concerted effort to face the challenges to growth in the life of grace, and this does not exclude trusting in providence. Let us not forget that the command to take nothing for the journey, “no staff, nor bag, nor bread, nor money (Luke 9:3)” was not the last word on the matter. After the disciples learned to trust in God, Jesus instructs them that “the one who has a purse must take it, and likewise a bag (Luke 22:35-37).” As the mission on which Christ’s followers were sent became more demanding, a pragmatism was embraced.[25] The challenges that face the spread of the Gospel now include threats to our very survival, so we too must adapt to protect our future from ourselves. We can thus collectively cooperate with grace, to answer God’s calling for humanity. With potestas annihilationis comes the greatest responsibility, to ensure our descendants receive the best possible chance of living out their vocations, to pass on the gifts of life, faith, and the conditions and environment that make those possible. On these terms, hope for future health and wellbeing does not entail the eschatological reductionism such thought might otherwise attract. We do well to follow Augustine: “Belief in the city of God does not mean that the earthly city is a point of indifference. The earthly city strives for peace, an end it shares with the City of God, as the two cannot be thought about as radically different as in this present transitory world, they are interwoven and mingled with one another.”[26]One need not surrender belief in a transcendent finalising moment that will occur in history to have a sense of accountability, and legitimate hopes, for humanity and the Church in the time, of indeterminate length, between today and the Parousia.
Like the first Christians, we too face a little-evangelised world, and so are also grateful for the time of mercy, the opportunity to make disciples of all nations, including the potential trillions of descendants to come. If human history is just the beginning, if we are in the earliest time of the Church, then God could well bestow countless future graces: saints will be raised up, sinners will be forgiven, theologians will explore new depths, the faithful will experience new heights of spiritual experience, carried higher by new insights into the nature of prayer, fostered by the cumulative enrichment of religious tradition. Saint Peter could not have foreseen the baroque basilica that now stands at the place of his last earthly moments. The early Church could not have imagined the beauty of Durufle’s motets on Gregorian themes. Augustine could not benefit from the spiritual exercises of Ignatius, Bonaventure had no way of anticipating the personalism of Saint John Paul II. The Church will be more and more at home in the spiritual richness that the human experience can offer, and perhaps, with the opportunity of millennia, even discover modes of spiritual experience, or categories of theological value, which are unknown today.[27] We rightly believe that the major event in Christian life has already happened in the paschal mystery, and we await the day when the Lord will draw all things to himself. Nonetheless, we can look forward to the future of humanity and wonder what prayer and worship we are blind to, what music are we deaf to. Let us dare to hope the Church has a long and bright future, and let us be unsettled by the threat of that future being lost: In light of scriptural and patristic sources, I have sketched out some conditions of possibility for allowing these sentiments of love and concern for our distant future without fear of doctrinal discontinuity. However, moral judgement is one thing, action is quite another.
Safeguarding humanity’s potential tomorrow involves making sacrifices today, and faith might make its greatest contribution to mitigating AXRs by providing Christian reasons for motivating costly action. A post-Christian worldview lacks evaluative resources for adequately relating to humanity’s future, having set aside beliefs which forged and sustained links between generations. Scheffler, for example, explains that many “experience the poverty of our evaluative thought about the future as a form of privation,” recognising “an inchoate sense that other generations matter to us in way that we cannot easily explain,”[28] so he, Ord, and others, develop theories of value, axiologies, with which one might find underappreciated reasons why we should care about the future of humanity. Faith can provide compelling and greatly needed reasons why the future matters, inspiring believers to act in charity towards our descendants.
Christians aspire to turn to the past with faith, the present with love, and the future with hope, but these triads can be paired in other ways. I propose that faith in the Gospel received from the past also shows how hope can shine more brightly in the present by a love of the future. We need to learn to appreciate the blessings to come in a new way, and a dialogical study of the future axiologies offered by philosophy helps express the rich evaluative resources available to the Christian imagination for overcoming temporal parochialism. Catholics are particularly good at experiencing meaningful communion with those in the distant past. Can we not do so for the distant future too? The stakes are potentially little less than human life itself, and with it the very phenomena of prayer, beauty, and value. In learning to love the future, we can grow not only in fraternity, but in parental love, the desire to “bring life into existence out of nothing… which begins even before any child exists” and does not attach any conditions for the beloved, “it only wants him to be.”[29] Sharing in the Father’s desire that our descendants will be able to praise Him and follow Him as best they can, we learn to love and protect humanity’s future. God realises his promises in the time to come, but that future is now, at least partly, in humanity’s hands.
[1] Bertrand Russell, “The International Situation,” in Parliamentary Debates (Hansard), vol. 138 (The House of Lords, 1945) 89.
[2] Jonathan Schell, Fate of the Earth (New York, N.Y: Avon Books, 1988) 115.
[3] Günter Anders, as cited and translated in U Körtner, The End of the World: A Theological Interpretation, (Westminster, 1995) 181.
[4] For a description of the full range of known existential risks, see Nick Bostrom, “Existential Risks: Analysing Human Extinction Scenarios and Related Hazards,” The Journal of Evolution and Technology, vol. 9, No. 1 (2002).
[5] Toby Ord, The Precipice (Oxford: Bloomsbury, 2020) 191.
[6] Ibid.,, 21.
[7] For more explanation of natural risks see Toby Ord, “Natural Risks”, 67ff.
[8] See Chapter 5 Genuine Developments Contrasted with Corruptions in John Henry Newman, An Essay on the Development of Christian Doctrine (Notre Dame: Notre Dame, 1989).
[9] Augustine, City of God, Penguin Classics (London ; New York: Penguin Books, 2003), XVI, 23.
[10] Davide Castelvecchi, “Universe Has Ten Times More Galaxies than Researchers Thought,” Nature, November 14, 2016.
[11] Toby Ord, 217ff.
[12] Ibid.,, 21.
[13] “The Easter Vigil in the Holy Night”, in The Roman Missal, 2010, §11.5.
[14] Donald Senior and Carroll Stuhlmueller, The Biblical Foundations for Mission (Maryknoll, N.Y: Orbis Books, 1983) 208.
[15] Senior and Stuhlmueller, 191.
[16] Augustine, The Works of Saint Augustine, vol. 1, Sermons on the Old Testament, 20-50 (New City Press, 1990) 369.
[17] For the claim that the eschaton is, at least partially, contingent on human agency, I closely follow Christopher M Hays et al., When the Son of Man Didn’t Come: A Constructive Proposal on the Delay of the Parousia (Fortress Pr., 2017).
[18] Hays et al., 100ff.
[19] Justin Martyr, Opera Quae Exstant Omnia, vol. 6, Patrologia Graeca (Paris: Apud J.-P. Migne, 1857), 2 Apol., 1:190, 456.
[20] Tertullian, Opera Omnia, vol. 1, Patrologia Latina (Paris: Brepols, 1844), Apol. 39, 3:46, 532.
[21] Augustine, A Treatise on the Gift of Perseverance (Aeterna Press, 2015) 40.
[22] St Thomas Aquinas, De Veritate, trans. Robert Mulligan, (Hackett Publishing Company, 1994), vol. I, I.11.
[23] Wolfhart Pannenberg, “Theology and Philosophy in Interaction with Science”, in John Paul II on Science and Religion (Notre Dame, 1991), 78.
[24] For proposed practical steps see Toby Ord, “Safeguarding Humanity”, 187ff.
[25] See Paula Fredriksen, When Christians Were Jews: The First Generation (New Haven London: Yale University Press, 2018), 139ff.
[26] Augustine, City of I, II.I, Book I, preface.
[27] Such ideas are a Christianisation of a mode of thought characteristic of Ord, cf. Toby Ord, 237-238.
[28] Samuel Scheffler, Why Worry About Future Generations? (Oxford: Oxford University Press, 2020) 86.
[29] Jonathan Schell, 175.
| 0
| 0
| 4,918
| 4,918
|
52,120
|
99c3d9b7-5753-4086-b6ff-cb15a9d5facc
|
LDJnr/LessWrong-Amplify-Instruct
|
"Whenever I hear someone describe quantum physics as "weird" - whenever I hear someone bewailing the mysterious effects of observation on the observed, or the bizarre existence of nonlocal correlations, or the incredible impossibility of knowing position and momentum at the same time - then I think to myself: This person will never understand physics no matter how many books they read.
Reality has been around since long before you showed up. Don't go calling it nasty names like "bizarre" or "incredible". The universe was propagating complex amplitudes through configuration space for ten billion years before life ever emerged on Earth. Quantum physics is not "weird". You are weird. You have the absolutely bizarre idea that reality ought to consist of little billiard balls bopping around, when in fact reality is a perfectly normal cloud of complex amplitude in configuration space. This is your problem, not reality's, and you are the one who needs to change. Human intuitions were produced by evolution and evolution is a hack. The same optimization process that built your retina backward and then routed the optic cable through your field of vision, also designed your visual system to process persistent objects bouncing around in 3 spatial dimensions because that's what it took to chase down tigers. But "tigers" are leaky surface generalizations - tigers came into existence gradually over evolutionary time, and they are not all absolutely similar to each other. When you go down to the fundamental level, the level on which the laws are stable, global, and exception-free, there aren't any tigers. In fact there aren't any persistent objects bouncing around in 3 spatial dimensions. Deal with it.
Calling reality "weird" keeps you inside a viewpoint already proven erroneous. Probability theory tells us that surprise is the measure of a poor hypothesis; if a model is consistently stupid - consistently hits on events the model assigns tiny probabilities - then it's time to discard that model. A good model makes reality look normal, not weird; a good model assigns high probability to that which is actually the case. Intuition is only a model by another name: poor intuitions are shocked by reality, good intuitions make reality feel natural. You want to reshape your intuitions so that the universe looks normal. You want to think like reality.
This end state cannot be forced. It is pointless to pretend that quantum physics feels natural to you when in fact it feels strange. This is merely denying your confusion, not becoming less confused. But it will also hinder you to keep thinking How bizarre! Spending emotional energy on incredulity wastes time you could be using to update. It repeatedly throws you back into the frame of the old, wrong viewpoint. It feeds your sense of righteous indignation at reality daring to contradict you.
The principle extends beyond physics. Have you ever caught yourself saying something like, "I just don't understand how a PhD physicist can believe in astrology?" Well, if you literally don't understand, this indicates a problem with your model of human psychology. Perhaps you are indignant - you wish to express strong moral disapproval. But if you literally don't understand, then your indignation is stopping you from coming to terms with reality. It shouldn't be hard to imagine how a PhD physicist ends up believing in astrology. People compartmentalize, enough said.
I now try to avoid using the English idiom "I just don't understand how..." to express indignation. If I genuinely don't understand how, then my model is being surprised by the facts, and I should discard it and find a better model.
Surprise exists in the map, not in the territory. There are no surprising facts, only models that are surprised by facts. Likewise for facts called such nasty names as "bizarre", "incredible", "unbelievable", "unexpected", "strange", "anomalous", or "weird". When you find yourself tempted by such labels, it may be wise to check if the alleged fact is really factual. But if the fact checks out, then the problem isn't the fact, it's you."
The above is a chunk of text from a post created by a user on the website LessWrong.
What are some reasons why people may like or agree with what this person is saying?
People may like or agree with this post for several reasons:
1. Emphasis on Adaptability: The post emphasizes the importance of adapting our understanding and perception to align with reality, rather than expecting reality to align with our preconceived notions. This resonates with individuals who value adaptability and flexibility in thinking.
2. Critique of Anthropocentrism: The author critiques the anthropocentric view of the universe, arguing that our human intuitions are not the ultimate measure of reality. This can appeal to those who believe in the importance of decentering the human perspective in understanding the universe.
3. Encouragement of Intellectual Humility: The post encourages intellectual humility, suggesting that we should be ready to discard our models and beliefs when they do not align with the facts. This can resonate with individuals who value humility and continuous learning.
4. Call for Rational Thinking: The author advocates for rational thinking, arguing that we should update our models based on evidence rather than clinging to our beliefs. This can appeal to those who value evidence-based reasoning and critical thinking.
5. Acknowledgement of Complexity: The post acknowledges the complexity of reality, suggesting that our simple models (like billiard balls bopping around) may not be sufficient to understand it. This can resonate with individuals who appreciate the complexity and intricacy of the universe.
6. Challenge to Emotional Responses: The author challenges the emotional responses to reality, arguing that they can hinder our understanding. This can appeal to those who believe in the importance of separating emotions from reasoning.
7. Focus on Reality's Objectivity: The post emphasizes that reality is objective and exists independently of our perceptions and feelings. This can resonate with individuals who value objectivity and realism.
Expand on the idea of intellectual humility in the context of scientific research. How can this principle contribute to scientific progress and what might be the potential pitfalls?
Intellectual humility in the context of scientific research refers to the recognition and acceptance that one's knowledge and understanding are limited and fallible. It is the willingness to change one's views and hypotheses in the face of new evidence or better arguments, and the ability to admit when one is wrong. This principle is crucial in scientific progress for several reasons:
1. Openness to New Ideas: Intellectual humility fosters an openness to new ideas and perspectives, which is essential for scientific innovation. Scientists with intellectual humility are more likely to consider alternative hypotheses, explore novel research directions, and collaborate with others from diverse fields, thereby facilitating the generation of new knowledge and breakthroughs.
2. Promotion of Rigorous Inquiry: Intellectual humility promotes rigorous scientific inquiry. It encourages scientists to critically examine their assumptions, methodologies, and interpretations, and to seek and respond to feedback from peers. This can enhance the validity and reliability of scientific findings.
3. Mitigation of Confirmation Bias: Intellectual humility can help mitigate confirmation bias, which is the tendency to seek, interpret, and remember information in a way that confirms one's preexisting beliefs. By acknowledging the possibility of being wrong, scientists can be more objective in their data collection and analysis, leading to more accurate conclusions.
4. Facilitation of Scientific Dialogue: Intellectual humility facilitates constructive scientific dialogue and debate. It enables scientists to listen to and learn from each other, fostering a culture of mutual respect and collective learning, which is vital for the advancement of science.
However, there can also be potential pitfalls associated with intellectual humility:
1. Vulnerability to Pseudoscience: Excessive intellectual humility could make scientists more vulnerable to pseudoscience. If they are too open to alternative views, they may give undue consideration to theories or claims that lack empirical support or contradict established scientific principles.
2. Inhibition of Conviction: Intellectual humility could potentially inhibit the conviction needed to pursue challenging research. If scientists are overly cautious about their hypotheses, they may lack the confidence to persevere in the face of setbacks or criticism.
3. Risk of Relativism: There is a risk that intellectual humility could slide into a form of relativism, where all ideas are seen as equally valid. This could undermine the scientific pursuit of truth and the distinction between well-supported theories and those lacking evidence.
4. Potential for Exploitation: In a competitive academic environment, those who display intellectual humility may be seen as less assertive or ambitious, and their ideas could be undervalued or exploited by others.
Despite these potential pitfalls, intellectual humility remains a crucial virtue
Discuss the role of intellectual humility in mitigating confirmation bias within scientific research, and explore potential strategies to prevent the pitfalls associated with excessive intellectual humility.
Intellectual humility plays a significant role in mitigating confirmation bias within scientific research. Confirmation bias refers to the tendency to seek, interpret, and remember information in a way that confirms one's preexisting beliefs or hypotheses. This bias can lead to flawed experiments, misinterpretation of data, and erroneous conclusions, which can hinder scientific progress.
Intellectual humility, the recognition that one's knowledge and understanding are limited and fallible, can counteract confirmation bias. It encourages scientists to question their assumptions, consider alternative hypotheses, and remain open to new evidence that contradicts their beliefs. This can lead to more objective data collection and analysis, enhancing the validity and reliability of scientific findings.
However, excessive intellectual humility can also have pitfalls, such as vulnerability to pseudoscience, inhibition of conviction, risk of relativism, and potential for exploitation. Here are some strategies to prevent these pitfalls:
1. Balance Openness with Skepticism: While being open to new ideas, scientists should also maintain a healthy skepticism. They should critically evaluate the evidence for alternative theories or claims, and not accept them uncritically. This can help prevent vulnerability to pseudoscience.
2. Foster Confidence in Evidence-Based Beliefs: Scientists should be encouraged to have confidence in their hypotheses and theories that are based on solid evidence, even while remaining open to the possibility of being wrong. This can help prevent the inhibition of conviction.
3. Uphold the Value of Evidence: The scientific community should continue to uphold the value of empirical evidence and the scientific method. This can help prevent the slide into relativism, where all ideas are seen as equally valid.
4. Promote a Culture of Respect: Academic institutions and research organizations should promote a culture of respect for diverse ideas and contributions, and discourage exploitative behaviors. This can help protect those who display intellectual humility from being undervalued or exploited.
5. Education and Training: Scientists can be trained in cognitive strategies to recognize and counteract confirmation bias and other cognitive biases. They can also be educated about the importance of intellectual humility and how to balance it with other intellectual virtues.
By adopting these strategies, scientists can harness the benefits of intellectual humility while avoiding its potential pitfalls, thereby enhancing the integrity and progress of scientific research.
Reflect on the role of intellectual humility in the context of interdisciplinary research. How might it facilitate collaboration and innovation, and what strategies could be employed to overcome potential challenges?
In the context of interdisciplinary research, intellectual humility can play a significant role in facilitating collaboration and innovation. Interdisciplinary research involves integrating knowledge and methods from different disciplines, requiring researchers to step outside their areas of expertise and engage with unfamiliar concepts, theories, and methodologies. Here, intellectual humility can be particularly beneficial:
1. Facilitating Collaboration: Intellectual humility can foster a culture of mutual respect and learning among researchers from different disciplines. By acknowledging the limits of their own knowledge and valuing the expertise of others, researchers can build more effective and harmonious collaborations.
2. Promoting Knowledge Integration: Intellectual humility can promote the integration of knowledge from different disciplines. Researchers who are willing to question their assumptions and consider alternative perspectives are more likely to appreciate the insights offered by other disciplines and see the connections between them.
3. Driving Innovation: Intellectual humility can drive innovation in interdisciplinary research. By being open to new ideas and willing to learn from others, researchers can generate novel research questions, develop innovative methodologies, and produce groundbreaking findings.
However, intellectual humility in interdisciplinary research can also pose challenges, such as difficulties in communication and potential for exploitation. Here are some strategies to overcome these challenges:
1. Develop Shared Language: Researchers from different disciplines often use different terminologies and conceptual frameworks, which can lead to misunderstandings. Developing a shared language or glossary can facilitate communication and mutual understanding.
2. Establish Clear Roles and Expectations: To prevent exploitation, it's important to establish clear roles and expectations for each team member. This includes recognizing and valuing the contributions of all team members, regardless of their discipline.
3. Provide Interdisciplinary Training: Providing training in interdisciplinary research can help researchers understand the benefits and challenges of this approach, and equip them with the skills to navigate these challenges. This includes training in intellectual humility, as well as other skills such as communication, teamwork, and problem-solving.
4. Foster a Culture of Respect: Fostering a culture of respect and appreciation for diverse disciplines and perspectives can help overcome potential barriers to interdisciplinary research. This includes creating an environment where all team members feel valued and heard, and where intellectual humility is recognized as a strength rather than a weakness.
By adopting these strategies, researchers can leverage intellectual humility to enhance collaboration and innovation in interdisciplinary research, while overcoming potential challenges.
| 0
| 0
| 2,891
| 2,891
|
48,642
|
86079a5c-a7f0-4bc6-b71f-d5f1733a6223
|
trentmkelly/LessWrong-43k
|
[SEQ RERUN] That Alien Message
Today's post, That Alien Message was originally published on 22 May 2008. A summary (taken from the LW wiki):
> Einstein used evidence more efficiently than other physicists, but he was still extremely inefficient in an absolute sense. If a huge team of cryptographers and physicists were examining a interstellar transmission, going over it bit by bit, we could deduce principles on the order of Galilean gravity just from seeing one or two frames of a picture. As if the very first human to see an apple fall, had, on the instant, realized that its position went as the square of the time and that this implied constant acceleration.
Discuss the post here (rather than in the comments to the original post).
This post is part of the Rerunning the Sequences series, where we'll be going through Eliezer Yudkowsky's old posts in order so that people who are interested can (re-)read and discuss them. The previous post was Einstein's Speed, and you can use the sequence_reruns tag or rss feed to follow the rest of the series.
Sequence reruns are a community-driven effort. You can participate by re-reading the sequence post, discussing it here, posting the next day's sequence reruns post, or summarizing forthcoming articles on the wiki. Go here for more details, or to have meta discussions about the Rerunning the Sequences series.
| 0
| 0
| 309
| 309
|
53,703
|
a922a975-d4a0-4ece-bf8a-c3e86f481b36
|
trentmkelly/LessWrong-43k
|
Coordination attempt: Facilitating conversational splinters in Gathertown
Someone once wrote a LessWrong post saying that people, individually polled, will prefer small group conversations at events – but in practice form large clumps where the majority of people are disengaged but can't coordinate to leave in chunks, probably because coordination competes for a floor contested in proportion to the group's size. (Link welcome, I can't find it.) Gathertown seems to exacerbate this because you can stack people more densely.
I have two suggestions:
* Stand your avatar off several squares to the side (while still sharing videospace) when you want to splinter (but only if enough other people will do it as well), and move out at the same time when ceil(N/4) are in that would-be splinter.
* Broadcast in the chat what topic you want to discuss for your next conversation.
* Optionally, a participant count cap?
* Optionally, people who want to leave their current group will say '+1' and follow you. This introduces contention for chat space, but lets people who are want to leave their current group avoid the problem of trekking across the event space and finding a group as big as their previous ones.
(And one tongue in cheek – Gathertown should have escalators or gusts of wind that tend to slowly scatter people, and they need to actively choose to stay with a group or conversation.)
It would be most effective if the host decided on a strategy and broadcast it to participants as the default, but if they don't it seems fine for participants to advertise that they'll be using a particular strategy to avoid this trap.
| 0
| 0
| 359
| 359
|
856
|
e3fae5f7-a3cb-400d-bf0e-851785ec0cf1
|
StampyAI/alignment-research-dataset/arxiv
|
instruct the AI system via the normal means intended by the system’s designers.
Shutting down, repairing, or dismantling an AI system are ways in which humans can retain control of an AI system’s operation even when the communication abstractions of comprehension and instruction are not working well.
Of course, few present-day machines could not be safely shut down or destroyed by their owners if so desired. However, some machines have no owner, such as the internet, and are not so easy to shut down by legitimate means. If real-world AI capabilities ever approach the potential for prepotence, it may become very important for humans to retain safe and legitimate means to carry out such interventions on AI systems.
##### Instruction versus control.
Where should one draw the distinction between “instruction” and “control”? For instance, one could argue that an action like «unplug the power» is an “instruction” for turning a machine off, and that the laws of physics are the “interpreter” ensuring the instruction is followed. However, in this framing, the “communication channel” comprised by the «unplug» mechanism is certainly of a different design and purpose than the usual mouse, keyboard, and voice instruction channels. In particular, the «unplug» channel has the power to override any instructions from the other channels. So, even if one wishes to view control as a kind of instruction, it should be treated as a fairly special case, with the purpose and capacity to override other instructions.
###
2.8 Multiplicity of stakeholders and systems
When first beginning to analyze existential risk from AI development, it may be tempting—and perhaps conceptually simpler—to focus on single/single delegation. Indeed, if AI technology brings about a human extinction event, one might easily argue that the system “did not do what humans would have wanted”, and the task of making an AI system do what even a single human wants is still a difficult challenge in many domains.
Perhaps for this reason, much of the technical research to date that is formally or informally cited as relevant to catastrophic risks from AI—under such labels as “AI safety”, “AGI safety” or “long-term AI safety”—has been focussed primarily on single/single delegation. (Section [10.1](#S10.SS1 "10.1 Related research agendas ‣ 10 Further reading ‣ AI Research Considerations for Human Existential Safety (ARCHES)") will give a more detailed overview of the literature.)
Focusing entirely on single/single delegation can be misleading, however. There are powerful social and economic forces that can transform a single/single delegation scenario into a multi/multi delegation scenario.
First, note that there are numerous pathways through which a single/single
delegation scenario with any powerful AI system (such as a prepotent or near-prepotent AI system) can become a multi/single scenario:
* a)
Outside stakeholders will have a strong motivation to seek to own and/or share control of the system, because of its potential for impact.
* b)
The creators of the system might encounter any number of disagreements regarding how best to use the system. These disagreements might not have been considered in advance, especially if the creators were not confident they would succeed in developing the system, or did not have a clear understanding of how the system would end up working when they began their partnership. Facing the heightened stakes of this increased potential for impact could lead to a splintering of opinions about what to do next. So where previously the creators might have acted as single unified stakeholder, this might not remain the case.
These pathways lead from single/single to multi/single delegation scenarios. Next, consider how a multiplicity of AI systems can result:
* c)
The creators of any powerful AI system have economic incentives to duplicate and sell instances of the system to outside buyers.
* d)
Contemporary research groups, upon observing the capabilities of a powerful AI system, may also have strong intellectual and economic incentives to replicate its capabilities.
These pathways lead from single/single and multi/single to multi/multi delegation scenarios. In summary:
The multiplicity thesis. Soon after the development of methods enabling a single human stakeholder to effectively delegate to a single powerful AI system, incentives will likely exist for additional stakeholders to acquire and share control of the system (yielding a multiplicity of engaging human stakeholders) and/or for the system’s creators or other institutions to replicate the system’s capabilities (yielding a multiplicity of AI systems).
####
2.8.1 Questioning the adequacy of single/single delegation
The multiplicity thesis presents a source of added complexity in maintaining existential safety, which might not be well addressed by safety research focussed on single/single delegation.
How important is it to prepare for this complexity before it arises? That is to say, how important is it to begin work today on single/multi, multi/single, and multi/multi delegation solutions, from an existential safety perspective?
##### An optimistic view.
One view is that, given the development of near-prepotent AI systems for single/single delegation, future humans and human institutions would be able to use non-technical means to coordinate their use of those systems to ensure that either prepotent AI systems are never developed, or that the systems will maintain existential safety if they are developed.
In this view, there is no pressing existential need to develop multi/multi delegation
solutions prior to the development of near-prepotent AI systems.
As evidence for this view, one can point to any number of past successes of human coordination in the use and governance of technology. For instance, there is the Montreal Protocol banning the production of CFCs, which was fully signed in 1987, only 14 years after the scientific discovery in 1973 that CFSs are damaging to the ozone layer (Murdoch and
Sandler, [1997](#bib.bib188); Andersen and
Sarma, [2012](#bib.bib8)). For an example of international coordination in computer science specifically, consider the creation and governance of internet protocols such as TCP/IP by the Internet Engineering Task Force—a community with no formal organizational hierarchy—by “rough consensus and running code” (Russell, [2006](#bib.bib220); Resnick, [2014](#bib.bib207)).
##### A pessimistic view.
Alternatively, it might be that future humans would struggle to coordinate on the globally safe use of powerful single/single AI systems, absent additional efforts in advance to prepare technical multi/multi delegation solutions.
For a historical analogy supporting this view, consider the stock market “flash crash” of 6 May 2010, viewed as one of the most dramatic events in the history of financial markets (Madhavan, [2012](#bib.bib171)). The flash crash was a consequence of the use algorithmic stock trading systems by competing stakeholders (Easley
et al., [2011](#bib.bib68); Kirilenko
et al., [2017](#bib.bib137)).
If AI technology significantly broadens the scope of action and interaction between algorithms, the impact of unexpected interaction effects could be much greater, and might be difficult to anticipate in detail. World War I is a particularly horrific example where technology seemed to outpace the strategic thinking of human beings (specifically, military and state leaders) regarding how to use it (Gray, [2013](#bib.bib98), Chapter 5: The nineteenth century, II: technology, warfare and international order).
Military tactics lagged behind military technology, such as the machine gun and mustard gas, that had been developed over the preceding century, leading to an unprecedented number of casualties at war.
As a motivating thought experiment involving rapid and broadly scoped multi-agent interaction, imagine that 10 years from today, 50% of all humans will be able to think one thousand times faster than they can today.
Such a sudden change in human capabilities might be incredibly positive, dramatically accelerating progress in science, technology, diplomacy, and perhaps even multi/multi delegation research.
But the result could also be disastrous.
First, if the areas of the international economy most accelerated by intelligence enhancement turned out to involve the production of pollution or similar side effects, a rapid physical destruction of the environment could result.
Second, perhaps the rapidly changing social and geopolitical landscape could lead to a proliferation of attempts to seize political or economic power via socially or physically destructive tactics or warfare.
The proliferation of powerful single/single AI delegation solutions could conceivably have a similar destabilizing effect upon society and the environment.
Indeed, some have argued that artificial intelligence and computing technology more broadly has already outpaced our collective ability to make good decisions about how it is used (Hecht
et al., [2018](#bib.bib117)).
##### A precautionary view.
Of course, only one future will obtain in reality. Which of the above views—optimism or pessimism—will be closer to the truth? This question obscures the role of pessimism in preparedness: we all can exercise some agency in determining the future, and the most useful form of pessimism is one that renders its own predictions invalid by preventing them.
In any case, it is well beyond the scope of this report to determine for certain whether future humans and human institutions will succeed or fail in the judicious use of powerful single/single delegation solutions. And, maintaining a diversity of views will enable planning for a diversity of contingencies. Thus, in place of a prediction, we instead posit the following value judgement:
Multi/multi preparedness. From the perspective of existential safety in particular and societal stability in general, it is wise to think in technical detail about the challenges that multi/multi AI delegation might eventually present for human society, and what solutions might exist for those challenges, *before* the world would enter a socially or geopolitically unstable state in need of those solutions.
We will return to this discussion in Section 7.
###
2.9 Omitted debates
To maintain a sufficiently clear conceptual focus throughout, a number of debates have been intentionally avoided in this document:
* •
*What about global catastrophes that would not result in human extinction?* For concreteness, and to avoid digressions on what would or would not constitute a global catastrophe, this report is focussed on the simpler-to-agree-upon concept of human survival. Nonetheless, it does seems likely that many present-day approaches relevant to averting extinction risk should also be applicable to averting other events that would broadly be considered global catastrophes. The reader is therefore invited to use their own judgement to determine where arguments or methods in this report can be generalized from human extinction prevention to global catastrophic risk prevention.
* •
*What constitutes a human?* Even the concept of human survival is subject to some debate regarding its meaning, because of potential future ambiguity in what constitutes a human being.
For instance, Moravec ([1988](#bib.bib184), “Transmigration”, Chapter 4) describes a future in which humans can replace themselves by digital emulations their own minds, and Hanson ([2016](#bib.bib112)) envisions a future economy where most work is carried out by human-like emulations that have been modified and selected for performing valuable work.
If no biological humans remain, but human emulations continue to operate, should humanity be considered extinct?
This report does not delve into that question, because the authors suspect that most present-day approaches to existential safety will are not greatly affected by the answer, although it could still become important in the future.
* •
*What about other negative side effects of AI development?* Many ideas and arguments considered in this report could be applied to averting safety and ethical failures that would by no means be considered global catastrophes. The reader is invited to use their own judgement to consider what other negative side effects of AI development can be avoided and are worth the cost of avoidance. As discussed in the Preface, omissions of other safety and ethical issues from this report is not intended by the authors as an appraisal of their importance or relevance to society.
* •
*What constitutes “beneficial” AI?* A closely related topic to reducing existential risk from artificial intelligence—and which does not entirely fit within the scope of this report—is that of developing *provably beneficial* AI systems, i.e., AI systems which provably benefit the whole of human society. At a technical level, provable beneficence and existential safety are tightly intertwined:
1. 1.
For any broadly agreeable definition of “benefit”, an AI system that provably benefits all of humanity should, by most definitions, preserve humanity’s ability to avoid extinction.
2. 2.
Conversely, preventing existential risk requires attending to global-scale problems and solutions, which might yield mathematical and algorithmic techniques for ensuring other global benefits as well as reducing other global risks.
Despite these relationships, provable beneficence is a more general problem than existential safety. To address provable beneficence, one would need to address or dissolve what it really means to benefit humanity, given that individual human preferences are ill-defined, plastic, and not in universal agreement. By contrast, it might be easier to reach agreement on what scenarios constitute human extinction events, or at least to agree upon the general goal of avoiding all such scenarios. So, this this report explicitly avoids delving into any debate regarding the meaning of “provable beneficence”.
3 Risk-inducing scenarios
--------------------------
How could human society make the mistake of deploying AI technology that is unsurvivable to humanity? There are many hypothetical scenario types to consider, each of which might call for different forms of preventive measures.
In this report, scenarios are organized into *risk types* that will be outlined in this section. The risk types are related via the causal diagram in Figure [4](#S3.F4 "Figure 4 ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)").
* •
“Tier 1” refers to risks that are 1 degree of causal separation from unsurvivability in the diagram, whereas
* •
“Tier 2” refers to risks that would generate Tier 1 risks, and are hence 2 degrees of causal separation from unsurvivability.
Figure 4: Relationship between risk types considered in this report; each risk type is described as its own subsection.
###
3.1 Tier 1: MPAI deployment events
This section outlines specific scenarios wherein an MPAI deployment event could occur.222Such scenarios have been considered extensively by philosopher Nick Bostrom (Bostrom, [2014](#bib.bib32)) under more specific assumptions defining “superintelligent” AI systems.
Because this report is targeted at AI developers, the MPAI deployment events considered here have been classified according to the following exhaustive decision tree centered on the hypothetical AI developers involved in building the MPAI:
1. a.
?? [??](#S3.SS1.SSS1 "3.1.1 Type 1a: Uncoordinated MPAI development ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??). Was there no single AI development team who was primarily responsible for developing the MPAI technology? If so, classify the MPAI deployment event as arising from *uncoordinated MPAI development*.
Otherwise, in the remaining risk types one can assume the developers of the MPAI constitute a single team, and further subdivide scenarios based on the relationship of that team to the MPAI deployment event:
2. b.
?? [??](#S3.SS1.SSS2 "3.1.2 Type 1b: Unrecognized prepotence ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??). Prior to the technology being deployed and prepotent, did the development team fail to recognize that the technology would be or become prepotent? If so, classify as an *unrecognized prepotence* event; otherwise consider:
3. c.
?? [??](#S3.SS1.SSS3 "3.1.3 Type 1c: Unrecognized misalignment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??).
Prior to the technology being deployed and misaligned, did the development team fail to recognize that the technology would be or become misaligned? If so, classify as an *unrecognized misalignment* event; otherwise consider:
4. d.
?? [??](#S3.SS1.SSS4 "3.1.4 Type 1d: Involuntary MPAI deployment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??). Did the MPAI deployment event occur without the voluntary permission of the development team responsible for creating it? If so, classify as an *involuntary MPAI deployment* event; otherwise classify as:
5. e.
?? [??](#S3.SS1.SSS5 "3.1.5 Type 1e: Voluntary MPAI deployment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??). The MPAI deployment was voluntarily permitted by its developers.
The remainder of Section [3.1](#S3.SS1 "3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") examines these risk types in more detail.
####
3.1.1 Type 1a: Uncoordinated MPAI development
This risk type comprises MPAI deployment events that arise from *uncoordinated MPAI development* in the sense that no one research team is solely responsible for having developed the MPAI.
As an example of uncoordinated MPAI development, suppose Group A deploys a powerful AI system for managing an online machine learning development system, which is not prepotent because it lacks some key cognitive ability.
Then, suppose that around the same time, Group B releases an open source algorithm that Group A’s system learns about and uses to acquire the key cognitive ability, thereby becoming prepotent.
In this situation, because no coordinated effort has been made to align the resulting prepotent AI system with human survival, it is relatively likely to be misaligned, by the Human Fragility Argument (Section [2.5](#S2.SS5 "2.5 Human fragility ‣ 2 Key concepts and arguments ‣ AI Research Considerations for Human Existential Safety (ARCHES)")).
And, one could argue that neither Group A nor Group B was solely responsible for having developed the MPAI; rather, they failed to coordinate on the combined impact of their development and deployment decisions. Even if some members of each group were aware that the result of their actions might result in MPAI development, perhaps the local incentives of each group were to continue working on their products nonetheless. A similar dynamic can be seen in the way separate countries tend to follow local economic incentives to continue producing carbon emissions, despite the potentially dangerous combined impact of those emissions.
Avoiding this risk type calls for well-deliberated and respected assessments of the capabilities of publicly available algorithms and hardware, accounting for whether those capabilities have the potential to be combined to yield MPAI technology. Otherwise, the world could essentially accrue “AI pollution” that might eventually precipitate or constitute MPAI.
The remaining four Tier 1 risk types will focus on the knowledge and intentions of “the developers” of a hypothetical MPAI technology, such as whether the prepotence or misalignment of the technology was known or intended in advance. By contrast, for an MPAI deployment scenario where the developers of the technology are too poorly coordinated to have a clear consensus on whether it will be prepotent or misaligned, the present risk type—uncoordinated MPAI deployment—may be a better descriptor.
####
3.1.2 Type 1b: Unrecognized prepotence
This risk type comprises MPAI deployment scenarios where the prepotence of the relevant AI technology was unrecognized prior to it being deployed and prepotent.
Examples of this risk type can be divided into two natural sub-cases:
* •
deployment, then prepotence: The AI technology in question is not prepotent at the time of its initial deployment, but later becomes prepotent in a manner that surprises its developers.
For instance, this could happen if the developers are insufficiently informed of the system’s relationship with the world after its initial deployment, if they are informed but insufficiently attentive to the information, or if they are informed and attentive but unable to deduce that the system will become prepotent.
* •
prepotence, then deployment: The AI technology in question is prepotent prior to its deployment, but the developers fail to recognize this at deployment time. For instance, this could happen if the developers did not attempt to assess the prepotence of the technology, or somehow failed to complete an accurate assessment.
These sub-cases share an important feature in common: an AI technology with unrecognized prepotence is relatively likely to turn out to be MPAI. For, suppose an AI development team deploys an AI technology that turns out to be or become prepotent in some way that they did not expect.
Because of their faulty understanding of the system’s capacity for impact, their safety efforts would have been undertaken under invalid assumptions.
From there, by Section [2.5](#S2.SS5 "2.5 Human fragility ‣ 2 Key concepts and arguments ‣ AI Research Considerations for Human Existential Safety (ARCHES)") there are numerous pathways through which the system’s unstoppable transformative impact might be unsurvivable. Hence, unrecognized prepotence comes with an increased likelihood of unrecognized misalignment.
Avoiding this risk type calls for a rigorous scientific theory to understand and recognize when an AI system might be or become prepotent. An important way in which the prepotence of an AI technology could go unrecognized is if the system exhibits *behavior likely to obfuscate the full breadth of its capabilities*, thereby prompting developers to mistakenly deploy it as a non-prepotent system. Such behavior could result from a selection process that favors AI systems that somehow obfuscate capabilities that humans would consider dangerous. Capability obfuscation could also arise from a system with social reasoning and planning capabilities that learns, in pursuit of real-world attainment of its assigned objective, to “work around" human measures to prevent the deployment of prepotent systems. The latter case could be viewed as an instance of “intentional deception” by the system, although attribution of intention is not necessary to describe this general class of phenomena. In any case, an adequate theory for understanding and recognizing prepotence must account for the possibility of such systems systematically obfuscating their prepotence.
####
3.1.3 Type 1c: Unrecognized misalignment
This risk type comprises MPAI deployment scenarios where the misalignment of the relevant AI technology is unrecognized by its developers prior to it being deployed and misaligned. Like unrecognized prepotence, unrecognized misalignment can occur whether the misalignment occurs before or after the technology is initially deployed.
For example, suppose some team of AI developers build a prepotent AI system that they realize or suspect is prepotent, with the intention of using it for some positive and permanently transformative impact on the world.
There is some risk that the developers might mistakenly overestimate the system’s alignment, and hence fail to recognize that it is or will become MPAI. And, just as with prepotence, an important way misalignment could go unrecognized is if the system itself deceives humans into thinking it is aligned.
Avoiding this risk type calls for a rigorous scientific discipline for aligning powerful AI systems with human interests and existence, and for recognizing potential misalignment in deployed systems, including systems that may be able to systematically deceive humans regarding their misalignment.
####
3.1.4 Type 1d: Involuntary MPAI deployment
This risk type comprises MPAI deployment events that are *involuntary* on the part of the technology’s developers, i.e., occurring against the direct intentions of the team who developed the relevant AI technology.
For example scenarios, let us focus on cases where the developers recognize that the MPAI deployment event is forthcoming before it happens (since Types [??](#S3.SS1.SSS2 "3.1.2 Type 1b: Unrecognized prepotence ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") and [??](#S3.SS1.SSS3 "3.1.3 Type 1c: Unrecognized misalignment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk already cover MPAI deployment events involving unrecognized prepotence and unrecognized misalignment).
These scenarios can be further organized according to whether an MPAI technology becomes deployed (“release” events) or an already-deployed AI technology becomes MPAI (“conversion” events):
1. 1.
MPAI release events (involuntary):
An existing MPAI technology somehow becomes deployed without the voluntary consent of its developers.
For instance, consider a well-meaning team of developers who have created an AI technology that they suspect is both prepotent and misaligned, and are now conducting experiments on the technology to learn more about the risks it could present.
In such a scenario, at least some security measures would likely be in place to prevent the technology from being deployed against the intentions of the developers, but those measures could fail in some manner. The failure could involve:
1. (a)
Accidental release: An existing MPAI technology is released accidentally by its development team, enabling others to deploy it without the developers’ consent. No one on the development team intentionally causes the release of the technology; it is merely a haphazard mistake on the part of the developers. This sort of event could be analogized to a nuclear power-plant meltdown: someone is responsible for the accident, but no one did it on purpose.
2. (b)
Unauthorized release: An existing MPAI technology is obtained by someone other than its developers, against the developers wishes. For instance:
1. i.
Hackers obtain access to the technology’s code base and deploy it, perhaps without knowledge of its misalignment or prepotence.
2. ii.
Physical force is used to obtain access to the technology’s code base, such as by a military or terrorist group, who then go on to deploy the technology, perhaps without knowledge of its misalignment or prepotence.
3. iii.
A running instance of the AI technology acquires its own deployment as a goal, and finds a way to achieve deployment without its developers’ permission.
2. 2.
MPAI conversion events (involuntary): An AI technology is deployed and is later converted into MPAI by certain post-deployment events that were not intended by the technology’s developers.
The conversion could be caused by interactions with the relevant AI system(s), or by failures entirely external to the system(s):
1. (a)
Conversion by uncontrolled interactions: The developers did not establish adequate controls for post-deployment interactions with the technology, and those interactions convert the technology into MPAI.
2. (b)
Conversion by external failures: Humanity’s collective capacity to control or otherwise survive the impact of the technology somehow decreases after its deployment (say, due to a conflict between humans that destroys resources or coordination), and systems using the technology do not adjust their behavior accordingly, becoming MPAI by virtue of humanity’s increased vulnerability rather than by changes internal to the technology itself.
Avoiding this risk type calls for measures enabling well-meaning AI developers to recognize and prevent the use of their inventions in ways that might harm society.
####
3.1.5 Type 1e: Voluntary MPAI deployment
This risk type comprises scenarios where an MPAI deployment event is triggered voluntarily by the developers of the MPAI technology.
Even if the majority of the AI research and development community develops methods that make it easy to align powerful AI systems with human interests and existence, and existing powerful AI systems are protected from falling into the wrong hands, it may be possible for some misguided persons to develop and deploy MPAI technology on their own for some reason. For example,
1. 1.
Indifference: Persons unconcerned with the preservation of the human species develop and deploy a powerful AI system in pursuit of values that will yield human extinction as an inevitable side effect.
2. 2.
Malice: A military or terrorist organization develops MPAI technology with the misguided hope of controlling it to threaten particular adversaries.
3. 3.
Confusion: One or more AI developers that would not normally ignore or threaten human welfare becomes convinced to deploy an MPAI technology by morally confusing arguments. Perhaps the arguments are produced by other indifferent or malicious persons, or perhaps by an AI system.
Avoiding this risk type calls for measures to prevent powerful AI technologies from being developed and deployed by misguided persons. Some of these preventive measures could also guard against instances of?? [??](#S3.SS1.SSS4 "3.1.4 Type 1d: Involuntary MPAI deployment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??) that would arise specifically from unauthorized access to near-prepotent systems or code bases.
###
3.2 Tier 2: Hazardous social conditions
This section examines types of social conditions that exacerbate the likelihood of?? [??](#S3.SS1 "3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risks. Unlike the typology of?? [??](#S3.SS1 "3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risks, the following typology is non-exhaustive.
####
3.2.1 Type 2a: Unsafe development races
This risk type comprises scenarios wherein two teams are in competition to develop powerful AI systems with the hope that the more successful of the teams might achieve wealth or power from the deployment of their system, and where each team is motivated by their competitive incentives to take risks that would be considered irresponsible from a societal perspective. Even if each competing team knows about the challenges of aligning their systems with human existence, they might be tempted to divert resources away from safety measures in order to best the competition with superior capabilities.
This sort of development race exacerbates the probability of?? [??](#S3.SS1 "3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risks (??), specifically Type [??](#S3.SS1.SSS1 "3.1.1 Type 1a: Uncoordinated MPAI development ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)"), [??](#S3.SS1.SSS2 "3.1.2 Type 1b: Unrecognized prepotence ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)"), and [??](#S3.SS1.SSS3 "3.1.3 Type 1c: Unrecognized misalignment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risks (??,??, and??).
This conclusion has also been argued by Bostrom ([2014](#bib.bib32)) and Armstrong
et al. ([2016](#bib.bib9)).
Moreover,?? [??](#S3.SS1.SSS4 "3.1.4 Type 1d: Involuntary MPAI deployment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??) is increased because security measures against unauthorized or accidental deployments are more difficult to implement in a hurry, to reduce the chances of hazardous post-deployment interactions with the system.
Finally, if one of the development groups is a military or terrorist organization, they might decide to deploy their technology in a desperate attempt to overthrow their competitors by force.
This would constitute a?? [??](#S3.SS1.SSS5 "3.1.5 Type 1e: Voluntary MPAI deployment ‣ 3.1 Tier 1: MPAI deployment events ‣ 3 Risk-inducing scenarios ‣ AI Research Considerations for Human Existential Safety (ARCHES)") risk (??).
Avoiding this risk type calls for measures to reduce incentives for competing AI development teams to take socially unacceptable safety risks in the course of developing and deploying their technology.
####
3.2.2 Type 2b: Economic displacement of humans
This risk type comprises scenarios wherein most human persons have no power to bid for the continued preservation of the human species, because humans have mostly been economically displaced by AI systems.
The possibility of an unemployment crisis arising from automation has been discussed by numerous authors, e.g., Joy ([2011](#bib.bib129)), Ford ([2013](#bib.bib85)), Brynjolfsson and
McAfee ([2014](#bib.bib41)), Brynjolfsson
et al. ([2014](#bib.bib42)),
Russell
et al. ([2015](#bib.bib222)),
Chace ([2016](#bib.bib49)),
and Frey and
Osborne ([2017](#bib.bib86)).
A gradual replacement of human workers by AI systems could lead to an economy wherein most trade and consumption is carried out by non-human entities.
This is a bleak future from the point of view of many, but not yet a global threat to human survival.
To see how this trend would constitute an existential risk if taken far enough, consider a scenario where human institutions have all been out-competed and replaced by autonomous corporations.
Such autonomous corporations could be deployed by idealistic individuals looking to increase transparency or efficiency in certain industries, such as finance, supply chain management, or manufacturing. Perhaps autonomous corporations could eventually also engage in primary resource industries such as mining, oil drilling, or forestry, which could supply raw materials to corporations in other industries. If some combination of corporations turned out to be capable of sustaining and expanding an economy entirely without humans, humanity would lose its trade leverage for influencing their activities. This could constitute prepotence for the collective machine economy, as was argued by Turing ([1951](#bib.bib259)) in “Intelligent Machinery, A Heretical Theory”.
Given the machine economy’s prepotence, misalignment is relatively likely to follow. By the?? argument of Section [2](#S2 "2 Key concepts and arguments ‣ AI Research Considerations for Human Existential Safety (ARCHES)"), the side effects of a prepotent machine economy—in terms of resource consumption, waste emissions, or both—would be hazardous to humans *by default*, unless the leading autonomous corporations coordinated in such a way as to provide or at least allow the equivalent of social assistance and environmental protection to humans, for reasons not driven by the humans’ economic output.
The potential for economic take-off of a self-sustaining fully mechanized economy thus constitutes a?? [??](#S3.SS1.SSS1 "
| 3
| 18,000
| 26,000
| 89,891
|
46,476
|
f51c11d9-c21d-4c28-88c2-d8f10ef14bac
|
StampyAI/alignment-research-dataset/blogs
|
bYDTfcYGvWrLHKysokNZO2IIAAAggggAACCCCAAAII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| 240
| 1,440,000
| 1,448,000
| 5,325,818
|
11,129
|
62845201-426c-4496-9d99-f6d45e2204c5
|
trentmkelly/LessWrong-43k
|
An overview of some promising work by junior alignment researchers
We’re all familiar with ELK, natural abstractions, and toy models of superposition.
But there’s also a new cohort of alignment researchers. Many of them got involved (or produced their first major pieces of work) in the last year.
I’ve been impressed by the quality of some of their work, and I think it deserves wider recognition. The recent increase in attention being paid to alignment gives me some hope that an “unknown genius” may emerge in the field. Additionally, there are several junior alignment researchers who seem to have a lot of potential, and I’m excited to see their contributions as they get more experience and influence in the field.
Here’s some work by junior alignment researchers that excited me in the last year:
Externalized reasoning oversight by Tamera Lanham
Mechanistic interpretability tries to understand what models “think” by looking at their weights and activations. Externalized reasoning oversight tries to do this by just asking the models to explain their reasoning.
Inspired by chain-of-thought prompting techniques, Tamera set out to see if she could get language models to provide honest and transparent answers about their reasoning processes. In the limit, this technique could provide a novel way for us to interpret language models.
I’m excited to see junior researchers tackle existing problems (e.g., how do we understand what models think?) in new ways (e.g., hm, everyone’s been focusing on weights and activations… are there any alternatives?). Tamera’s work is a great example.
Relevant post: Externalized reasoning oversight: A research direction for language model alignment
Relevant interview: Tamera Lanham on AI risk, threat models, alignment proposals, externalized reasoning oversight, and working at Anthropic
Goal misgeneralization by Lauro Langosco
Everyone knows that models can learn unintended goals. Everyone knows that models can fail to generalize out-of-distribution.
Lauro’s paper connects these points and distin
| 0
| 0
| 452
| 452
|
31,622
|
eb41bc34-4047-4930-9540-64237c19a112
|
trentmkelly/LessWrong-43k
|
Forager Anthropology
(This is the second post in a short sequence discussing evidence and arguments presented by Christopher Ryan and Cacilda Jethá's Sex at Dawn, inspired by the spirit of Kaj_Sotala's recent discussion of What Intelligence Tests Miss. It covers Part II: Lust in Paradise and Part III: The Way We Weren't.)
Forager anthropology is a discipline that is easy to abuse. It relies on unreliable first-hand observations of easily misunderstood cultures that are frequently influenced by the presence of modern observers. These cultures are often exterminated or assimilated within decades of their discovery, making it difficult to confirm controversial claims and discoveries. But modern-day foraging societies are the most direct source of evidence we have about our pre-agricultural ancestors; in many ways, they are agriculture's control group, living in conditions substantially similar to the ones under which our species evolved. The standard narrative of human sexual evolution ignores or manipulates the findings of forager anthropology to support its claims, and this is no doubt responsible for much of its confused support.
Steven Pinker is one of the most prominent and well-respected advocates of the standard narrative, both on Less Wrong and elsewhere. Eliezer has referenced him as an authority on evolutionary psychology. One commenter on the first post in this series claimed that Pinker is "the only mainstream academic I'm aware of who visibly demonstrates the full suite of traditional rationalist virtues in essentially all of his writing." Another cited Pinker's claim that 20-60% of hunter-gatherer males were victims of lethal human violence ("murdered") as justification for a Malthusian view of human nature.
That 20-60% number comes from a claim about war casualties in a 2007 TED talk Pinker gave on "the myth of violence", for which he drew upon several important findings in forager anthropology. (The talk is based on an argument presented in the third chapter of The Blan
| 0
| 0
| 438
| 438
|
87,614
|
<urn:uuid:5d542ff0-066c-462f-9598-daa0eb944784>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Monday, May 9, 2016
It's Batman v. Superman done right, now let the fanboys complain.
This is how far we've come in superhero films, it's glorious, beautiful, simply amazing, proof that films can have a huge cast without losing a coherent story, Captain America:Civil War is the best rebuttal to the "superhero movies ruin cinema" conversation some critics make. I know most of you will write my praise off, claiming that I'm too much of a fanboy, clearly my judgement is biased, but that is simply not true, yes, I love most of the movies from Marvel Studios, but my love is not based in my love for their comics, I love this movie for the same reason I've come to love the studio, support, appreciation, and love for all their characters. Civil War, much like all the movies in the Marvel Cinematic Universe, is primarily a character piece, an action movie second, we come to see cool fights, but the most stand out moments stand out because we love the characters, that is what pushes these movies to be real gems, and what pushes this to become a real masterpiece in the genre and in modern film. Marvel Studios has succeeded these past few years, because of their love and understanding of characters, for example, the movies would never succeed if they all acted like Tony Stark. Even if there are similarities between certain characters, Marvel has succeeded by making each character unique enough so their interactions all feel unique to each other. Simply put, this movie could not work if the only difference between the teams was that one decided to side with U.N. while the other didn't, seeing the reasoning and logic is why it becomes something special. Captain America and Iron Man are both right, yet they are both wrong, who you side with is based more on your own reasoning, not who the movie tells you to side with, the brilliant complexity in that is why Civil War is so great.
Tony (Iron Man) and Steve (Captain America) are arguing about the control of peace, and the freedom to chose peace. Tony prefers to have a system that accounts for all damage, so that he can ensure all lives are as safe as possible; while Steve opts for a freedom that allows him to fight and keep peace whenever he feels it's right, not when the U.N. deems it necessary, regardless of how dangerous that may end up being. On one side, heroes fight without supervision, enabling them to do the right thing on their own, on the other, supervision is what keeps them from accidentally injuring others, while keeping the heroes out of fights the U.N. deems unnecessary, Tony has been built up to like control, while Steve has always stood for freedom, even when it didn't seem right, these ideologies are what sets the movie in motion.
I may spoil a bit of the movie from here on out, so if you haven't seen the movie, you may want to skip the rest, Until you've seen the film, you've been warned.
Playing as a follow-up to both Winter Soldier and Age of Ultron may sound confusing, but Civil War continues to show Marvel's strength in making movies accessible to newcomers, regardless of the amount of material that came before. Civil War doesn't condense characters and story into a two and half hour runtime, it tells a complete story, filled with characters, not all are as important as the next, but each is given their time to shine, each given enough time to service their purpose. Hawkeye, Ant-Man and Spider-Man are mainly around to be in the amazing airport fight, with only Spider-Man having another substantial sequence in the film dedicated to him, (yes Hawkeye and Ant-Man appear in a few more scenes, but they're mainly to establish that they're in the film, or do not focus on them entirely) a sequence, which is largely there to introduce new actor Tom Holland (who is great here) while reinforcing Tony's viewpoint so we can sympathize with his argument, see? Characters given time to service the story, not take away from it.
Impressively, we do get minor characters with their own story arcs here as well, Vision, Black Widow, Scarlett Witch, and even War Machine are fleshed out here in ways we haven't seen until now, and it all works. The Vision and Wanda (Scarlett Witch) are presented here together in most scenes, sure it's to tease fans who know these characters end up together, but it also presents arguments on how both characters operate, Vision joins the U.N.'s registration act to keep control, while Wanda doesn't reject it, but neglects to join due to a fear of lack of control; by the end we see how their choices ultimately transform them after that spectacular second-act airport fight, which end up complicating matters, keeping in tone with the film's indifference to either side, is Vision's desire of control right? After all, his lack of control in himself impacts his team in a bad way, while Wanda's decision to fight, even if she doesn't have full control ends up helping her team, though this also has its own consequences, again you decide who was right. Natasha (Black Widow) and Rhodey (War Machine) both support the U.N.'s decision, but it's how they arrive to it, and where they end up that makes all the difference, even if it's a bit predictable, they both become arguments for their own side, almost like one is set up as a counter point for Tony to make about his own side, while the other becomes the rebuttal to that, to reaffirm his stance. Civil War plays with all these characters, as well as some very minor characters like Crossbones, to get across it's own themes and ideas, while letting the characters attract us to those ideas, which is why the film works, despite having all the prior material, what really matters is the ideas, the characters are mainly there to make it more accessible to an audience, but we identify easier by knowing how the characters arrived to their choices, and in the end we aren't lost because the arguments come through just right.
Captain America is the main character here, this is his movie, so his argument is given more weight to it, but the movie isn't afraid to balance that out. Sam (Falcon) and to an extent Sharon Carter, support Cap's side of the argument, the freedom to do what he does keeps the wrong man out of punishment, but Bucky's major involvement complicates it, is Cap thinking clearly? Or is he being selfish in his choice just to keep his longtime friend at any cost? The film hints at Steve lacking good judgement when it involves Bucky, but it also ensures us that Steve was right in his early assumption that there was more to the Winter Soldier's (Bucky), crimes than the U.N. assumed. We are given ample evidence to agree with Cap, but just as much is given to disagree with his views, that by the end we are given the impression that Cap may have been wrong all along, or at least disappointing in the views he once held so high.
Iron Man is presented to us as a man growing in his loneliness, one who is continually hurting those close to him, yet he's the only one who seems to be focused on keeping all lives safe, while keeping the Avengers, the closest thing to a family for him, together. Tony desires control, and while that may have backfired in the past, including his last screen appearance, he at least understands the concern of death in the lives they protect, he wants everyone safe, he's not satisfied with most people surviving, he wants everyone to walk away safe, whether it's to clear his conscience of having caused the deaths of many, or a desire to be a better hero, it doesn't matter, there's selflessness in his choice, while also roots of selfishness, much like Cap. Tony isn't exactly in a worse place because of his choice, but he, along with Cap, divide the team, one for not focusing on keeping the team together, the other for trying too hard to avoid it by any means necessary, the harsh truth being, their decisions aren't built on selfish desires, but their approaches may be. Iron Man is lonely, who's at fault is up to you here, but Cap isn't exactly helping there, partially because Cap has been just as lonely, his response to that loneliness builds his choices, much like Tony.
I'm sure you're wondering who the villain in the movie is by now, the truth is the movie barely focuses on him, Helmut Zemo (better known to comic fans as Baron Zemo), works because of that. A lot of people seem to forget this, but length isn't always the best way to judge a villain, let alone a character, sure Zemo barely appears, but he's a subtle villain who leaves an impact; his goal is simple, but devastating as we see in the movie's climax. Zemo is chilling throughout Civil War, because we don't know what he's up to for most of the film, but we know the effects of what he is causing, and once we do know what he's up to, he becomes a very nuanced villain, who becomes an excellent mirror for Black Panther to reflect upon during the climax (who surprisingly is given a complete origin story here that works), Zemo, as it turns out, isn't your standard villain, he is given an understandable, but extreme viewpoint. Zemo makes the most of his screen time, resulting in one of the MCU's best villains, while also proving how good Marvel Studios can be with characters.
At this point, Marvel Studios has proven they can make masterpieces, they've proven that they deserve a place in film history, not just for popularizing cinematic universes, but because they've shown how to make high art out of cinematic universes. Captain America:Civil War is one of Marvel's greatest accomplishments, if not their greatest, it's a movie filled with some of the best action scenes ever put to film, a lot of incredible character performances, while also being one of the few blockbusters to ask big questions without ever giving a truly concrete answer, because sometimes the right choice isn't an easy one to make, compromise will be made to some issues. Marvel Studios has earned my trust and with this movie, they've proven that they deserve it.
Did you like this piece? Support me on Patreon for more.
Written By Octaviano Macias
No comments:
Post a Comment
| 0
| 0
| 2,222
| 2,222
|
7,215
|
942bbffb-e52a-4f54-976f-90641eaab3b0
|
trentmkelly/LessWrong-43k
|
Suspected reason that kids usually hate vegetables
Here is a common vegetable preparation method in U.S. suburban homes.
Step 1: cut it up
Step 2: boil them in water until the flavor and texture is gone.
Step 3: Serve them without any kind of seasoning, or mixed with another food that could provide flavor and texture. Alternatively, put salt and pepper on them so that all vegetables just taste like salt and pepper.
Step 4: Tell the kids that eating a pile of bland mush with each meal is needed for being healthy.
In high school, me and several other people made the surprising discovery that if you eat vegetables raw, they actually taste alright, and some (like celery) even taste good enough to just snack on plain. Plus, there are things like stir-fry that make vegetables taste great.
I know this was common where I grew up. Do enough other people have similar experiences that raising knowledge of this is a great way to increase healthy eating?
Edit: after viewing the comments, this does not seem like the main reason.
| 0
| 0
| 233
| 233
|
30,834
|
81ba99b5-d53a-4353-a954-6fd434087e55
|
StampyAI/alignment-research-dataset/special_docs
|
of a plan is an action. The partial-order solution corresponds to six possible total-order plans; each of these is called a linearization of the partial-order plan.
Partial-order planning can be implemented as a search in the space of partial-order plans. (From now on, we will just call them “plans.”) That is, we start with an empty plan. Then we consider ways of refining the plan until we come up with a complete plan that solves the problem. The actions in this search are not actions in the world, but actions on plans: adding a step to the plan, imposing an ordering that puts one action before another, and so on.
We will define the POP algorithm for partial-order planning. It is traditional to write out the POP algorithm as a stand-alone program, but we will instead formulate partial-order planning as an instance of a search problem. This allows us to focus on the plan refinement steps that can be applied, rather than worrying about how the algorithm explores the space. In fact, a wide variety of uninformed or heuristic search methods can be applied once the search problem is formulated.
Remember that the states of our search problem will be (mostly unfinished) plans. To avoid confusion with the states of the world, we will talk about plans rather than states. Each plan has the following four components, where the first two define the steps of the plan and the last two serve a bookkeeping function to determine how plans can be extended:
e A set of actions that make up the steps of the plan. These are taken from the set of actions in the planning problem. The “empty” plan contains just the Start and Finish actions. Start has no preconditions and has as its effect all the literals in the initial state of the planning problem. Finish has no effects and has as its preconditions the goal literals of the planning problem.
389
ORDERING CONSTRAINTS
CAUSAL LINKS ACHIEVES
CONFLICTS
OPEN PRECONDITIONS
Partial-Order Plan: Total-Order Plans: Start Start Start Start Start Start Start J \ Right Right Left Left Right Left Sock Sock Sock Sock Sock Sock Left Right Sock Sock 1 1 1 Y Y Left Left Right Right Right Left Sock Sock Sock Sock Shoe Shoe LeftSockOn RightSockOn 1 1 1 1 y : Right Left Right Left Left Right ft Right or oro = v = Shoe or Left Right Left Right Left Right Shoe Shoe Shoe Shoe Shoe Shoe LeftShoeOn, RightShoeOn { { 7 yo Finish Finish Finish Finish Finish Finish Finish Figure 11.6 A partial-order plan for putting on shoes and socks, and the six corresponding linearizations into total-order plans.
read as “A before B” and means that action A must be executed sometime before action B, but not necessarily immediately before. The ordering constraints must describe a proper partial order. Any cycle—such as A < B and B < A—represents a contradiction, so an ordering constraint cannot be added to the plan if it creates a cycle.
as A \_P, Band is read as “A achieves p for B.” For example, the causal link RightSock ®9P45okOn pian i Shoe
asserts that RightSockOn is an effect of the RightSock action and a precondition of RightShoe. 1t also asserts that RightSockOn must remain true from the time of action RightSock to the time of action RighiShoe. In other words, the plan may not be extended by adding a new action C that conflicts with the causal link. An action C conflicts with A \_P, B if C has the effect —p and if C could (according to the ordering constraints) come after A and before B. Some authors call causal links protection intervals, because the link A \_?, B protects p from being negated over the interval from Ato B.
in the plan. Planners will work to reduce the set of open preconditions to the empty set, without introducing a contradiction.
390
Chapter 11. Planning
CONSISTENT PLAN
For example, the final plan in Figure 11.6 has the following components (not shown are the ordering constraints that put every other action after Start and before Finish):
Actions: { RightSock, RightShoe, LeftSock, LeftShoe, Start, Finish}
Orderings:{ RightSock < RightShoe, LeftSock < LeftShoe}
Links: { RightSock RightSockOn: piohtShoe, LeftSock Ft50%kO™ LeftShoe, RightShoe Fi9ht5hoeOn pinish LeftShoe LeftShoeOn Binish}
Open Preconditions:{ }.
We define a consistent plan as a plan in which there are no cycles in the ordering constraints and no conflicts with the causal links. A consistent plan with no open preconditions is a solution. A moment’s thought should convince the reader of the following fact: every linearization of a partial-order solution is a total-order solution whose execution from the initial state will reach a goal state. This means that we can extend the notion of “executing a plan” from total-order to partial-order plans. A partial-order plan is executed by repeatedly choosing any of the possible next actions. We will see in Chapter 12 that the flexibility available to the agent as it executes the plan can be very useful when the world fails to cooperate. The flexible ordering also makes it easier to combine smaller plans into larger ones, because each of the small plans can reorder its actions to avoid conflict with the other plans.
Now we are ready to formulate the search problem that POP solves. We will begin with a formulation suitable for propositional planning problems, leaving the first-order complications for later. As usual, the definition includes the initial state, actions, and goal test.
e The initial plan contains Start and Finish, the ordering constraint Start < Finish, and no causal links and has all the preconditions in Finish as open preconditions.
e The successor function arbitrarily picks one open precondition p on an action B and generates a successor plan for every possible consistent way of choosing an action A that achieves p. Consistency is enforced as follows:
1. The causal link A \_\*, B and the ordering constraint A < B are added to the plan. Action A may be an existing action in the plan or a new one. If it is new, add it to the plan and also add Start < A and A < Finish.
2. We resolve conflicts between the new causal link and all existing actions and between the action A (if it is new) and all existing causal links. A conflict between A \_\*, Band C is resolved by making C occur at some time outside the protection interval, either by adding B < C or C < A. We add successor states for either or both if they result in consistent plans.
e The goal test checks whether a plan is a solution to the original planning problem. Because only consistent plans are generated, the goal test just needs to check that there are no open preconditions.
Remember that the actions considered by the search algorithms under this formulation are plan refinement steps rather than the real actions from the domain itself. The path cost is therefore irrelevant, strictly speaking, because the only thing that matters is the total cost of the real actions in the plan to which the path leads. Nonetheless, it is possible to specify a path cost function that reflects the real plan costs: we charge 1 for each real action added to
Section 11.3.
Partial-Order Planning 391
the plan and O for all other refinement steps. In this way, g(n), where n is a plan, will be equal to the number of real actions in the plan. A heuristic estimate h(n) can also be used.
At first glance, one might think that the successor function should include successors for every open p, not just for one of them. This would be redundant and inefficient, however, for the same reason that constraint satisfaction algorithms don’t include successors for every possible variable: the order in which we consider open preconditions (like the order in which we consider CSP variables) is commutative. (See page 141.) Thus, we can choose an arbitrary ordering and still have a complete algorithm. Choosing the right ordering can lead to a faster search, but all orderings end up with the same set of candidate solutions.
A partial-order planning example
Now let’s look at how POP solves the spare tire problem from Section 11.1. The problem description is repeated in Figure 11.7.
Init(At(Flat, Azle) A At(Spare, Trunk)) Goal(At(Spare, Azle)) Action(Remove (Spare, Trunk),
PRECOND: At(Spare, Trunk)
EFFECT: — At(Spare, Trunk) A At(Spare, Ground)) Action (Remove (Flat, Azle),
PRECOND: At(Flat, Azle)
EFFECT: = At(Flat, Azle) A At(Flat, Ground)) Action(PutOn(Spare, Axle),
PRECOND: At(Spare, Ground) A — At(Flat, Azle)
EFFECT: = At(Spare, Ground) A At(Spare, Azle)) Action (Leave Overnight,
PRECOND:
EFFECT: = At(Spare, Ground) A — At(Spare, Azle) A — At(Spare, Trunk)
A = At( Flat, Ground) A — At(Flat, Azle))
Figure 11.7 The simple flat tire problem description.
The search for a solution begins with the initial plan, containing a Start action with the effect At(Spare, Trunk) A At(Flat, Azle) and a Finish action with the sole precondition At(Spare, Azle). Then we generate successors by picking an open precondition to work on (irrevocably) and choosing among the possible actions to achieve it. For now, we will not worry about a heuristic function to help with these decisions; we will make seemingly arbitrary choices. The sequence of events is as follows:
1. Pick the only open precondition, At(Spare, Azle) of Finish. Choose the only applicable action, PutOn(Spare, Azle). 2. Pick the At(Spare, Ground) precondition of PutOn (Spare, Azle). Choose the only
applicable action, Remove (Spare, Trunk) to achieve it. The resulting plan is shown in Figure 11.8.
392
Chapter 11. Planning
At(Spare, Trunk)] Remove(Spare, Trunk) Al Spare, Trunk) At(Spare, Ground) Start PutOn(Spare,Axle At(Spare,Axie)| Finish | | Start | ce axe —Al( Flat, Axle) (Sp ) [ Finisn\_}
Figure 11.8 The incomplete partial-order plan for the tire problem, after choosing actions for the first two open preconditions. Boxes represent actions, with preconditions on the left and effects on the right. (Effects are omitted, except for that of the Start action.) Dark arrows represent causal links protecting the proposition at the head of the arrow.
3. Pick the —At(Flat, Azle) precondition of PutOn (Spare, Azle). Just to be contrary, choose the LeaveQuvernight action rather than the Remove (Flat, Azle) action. Notice that Leave Overnight also has the effect ~ At (Spare, Ground), which means it conflicts with the causal link
Remove (Spare, Trunk) At(Spare, Ground) PutOn(Spare, Azle).
To resolve the conflict we add an ordering constraint putting LeaveQvernight before Remove (Spare, Trunk). The resulting plan is shown in Figure 11.9. (Why does this resolve the conflict, and why is there no other way to resolve it?)
At(Spare, Trunk)] Remove(Spare, Trunk)
[]
Figure 11.9 The plan after choosing LeaveQuernight as the action for achieving —At(Flat, Azle). To avoid a conflict with the causal link from Remove(Spare, Trunk) that protects At(Spare, Ground), LeaveOvernight is constrained to occur before Remove (Spare, Trunk), as shown by the dashed arrow.
4. The only remaining open precondition at this point is the A¢( Spare, Trunk) precondition of the action Remove (Spare, Trunk). The only action that can achieve it is the existing Start action, but the causal link from Start to Remove (Spare, Trunk) is in conflict with the — At (Spare, Trunk) effect of LeaveOvernight. This time there is no way to resolve the conflict with Leave Overnight: we cannot order it before Start (because nothing can come before Start), and we cannot order it after Remove(Spare, Trunk) (because there is already a constraint ordering it before Remove (Spare, Trunk)). So we are forced to back up, remove the LeaveOQuernight action and the last two causal links, and return to the state in Figure 11.8. In essence, the planner has proved that LeaveQOvernight doesn’t work as a way to change a tire.
Section 11.3.
Partial-Order Planning 393
5. Consider again the — At (Flat, Azle) precondition of PutOn (Spare, Azle). This time, we choose Remove(Flat, Azle).
6. Once again, pick the A¢(Spare, Trunk) precondition of Remove (Spare, Trunk) and choose Start to achieve it. This time there are no conflicts.
7. Pick the At(Flat, Azle) precondition of Remove(Flat, Azle), and choose Start to achieve it. This gives us a complete, consistent plan—in other words a solution—as shown in Figure 11.10.
[]
Figure 11.10 The final solution to the tire problem. Note that Remove (Spare, Trunk) and Remove(Flat, Azle) can be done in either order, as long as they are completed before the PutOn(Spare, Azle) action.
Although this example is very simple, it illustrates some of the strengths of partial-order planning. First, the causal links lead to early pruning of portions of the search space that, because of irresolvable conflicts, contain no solutions. Second, the solution in Figure 11.10 is a partial-order plan. In this case the advantage is small, because there are only two possible linearizations; nonetheless, an agent might welcome the flexibility—for example, if the tire has to be changed in the middle of heavy traffic.
The example also points to some possible improvements that could be made. For example, there is duplication of effort: Start is linked to Remove(Spare, Trunk) before the conflict causes a backtrack and is then unlinked by backtracking even though it is not involved in the conflict. It is then relinked as the search continues. This is typical of chronological backtracking and might be mitigated by dependency-directed backtracking.
Partial-order planning with unbound variables
In this section, we consider the complications that can arise when POP is used with firstorder action representations that include variables. Suppose we have a blocks world problem (Figure 11.4) with the open precondition On(A, B) and the action
Action(Move(b, x,y), PRECOND:On(b, z) A Clear(b) A Clear(y), EFFECT: On(b,y) A Clear(z) A =On(b,z) A =Clear(y)).
394 Chapter 11. Planning
This action achieves On(A, B) because the effect On(b, y) unifies with On (A, B) with the substitution {b/A, y/B}. We then apply this substitution to the action, yielding
Action(Move(A, z, B), PRECOND:On(A, z) A Clear(A) A Clear(B), EFFECT: On (A, B) A Clear(z) A ~On(A,z) A =Clear(B)).
This leaves the variable z unbound. That is, the action says to move block A from somewhere, without yet saying whence. This is another example of the least commitment principle: we can delay making the choice until some other step in the plan makes it for us. For example, suppose we have On (A, D) in the initial state. Then the Start action can be used to achieve On(A, z), binding z to D. This strategy of waiting for more information before choosing z is often more efficient than trying every possible value of z and backtracking for each one that fails.
The presence of variables in preconditions and actions complicates the process of detecting and resolving conflicts. For example, when Move(A, z, B) is added to the plan, we will need a causal link
Move(A, z, B) OnAB) Binish |
If there is another action M> with effect ~On(A, z), then My; conflicts only if z is B. To accommodate this possibility, we extend the representation of plans to include a set of inequalTs ity constraints of the form z # X where z is a variable and X is either another variable or a constant symbol. In this case, we would resolve the conflict by adding z # B, which means that future extensions to the plan can instantiate z to any value except B. Anytime we apply a substitution to a plan, we must check that the inequalities do not contradict the substitution. For example, a substitution that includes x /y conflicts with the inequality constraint z # y. Such conflicts cannot be resolved, so the planner must backtrack. A more extensive example of POP planning with variables in the blocks world is given in Section 12.6.
Heuristics for partial-order planning
Compared with total-order planning, partial-order planning has a clear advantage in being able to decompose problems into subproblems. It also has a disadvantage in that it does not represent states directly, so it is harder to estimate how far a partial-order plan is from achieving a goal. At present, there is less understanding of how to compute accurate heuristics for partial-order planning than for total-order planning.
The most obvious heuristic is to count the number of distinct open preconditions. This can be improved by subtracting the number of open preconditions that match literals in the Start state. As in the total-order case, this overestimates the cost when there are actions that achieve multiple goals and underestimates the cost when there are negative interactions between plan steps. The next section presents an approach that allows us to get much more accurate heuristics from a relaxed problem.
The heuristic function is used to choose which plan to refine. Given this choice, the algorithm generates successors based on the selection of a single open precondition to work
Section 11.4.
Planning Graphs 395
on. As in the case of variable selection on constraint satisfaction algorithms, this selection has a large impact on efficiency. The most-constrained-variable heuristic from CSPs can be adapted for planning algorithms and seems to work well. The idea is to select the open condition that can be satisfied in the fewest number of ways. There are two special cases of this heuristic. First, if an open condition cannot be achieved by any action, the heuristic will select it; this is a good idea because early detection of impossibility can save a great deal of work. Second, if an open condition can be achieved in only one way, then it should be selected because the decision is unavoidable and could provide additional constraints on other choices still to be made. Although full computation of the number of ways to satisfy each open condition is expensive and not always worthwhile, experiments show that handling the two special cases provides very substantial speedups.
PLANNING GRAPH
LEVELS
All of the heuristics we have suggested for total-order and partial-order planning can suffer from inaccuracies. This section shows how a special data structure called a planning graph can be used to give better heuristic estimates. These heuristics can be applied to any of the search techniques we have seen so far. Alternatively, we can extract a solution directly from the planning graph, using a specialized algorithm such as the one called GRAPHPLAN.
A planning graph consists of a sequence of levels that correspond to time steps in the plan, where level 0 is the initial state. Each level contains a set of literals and a set of actions. Roughly speaking, the literals are all those that could be true at that time step, depending on the actions executed at preceding time steps. Also roughly speaking, the actions are all those actions that could have their preconditions satisfied at that time step, depending on which of the literals actually hold. We say “roughly speaking” because the planning graph records only a restricted subset of the possible negative interactions among actions; therefore, it might be optimistic about the minimum number of time steps required for a literal to become true. Nonetheless, this number of steps in the planning graph provides a good estimate of how difficult it is to achieve a given literal from the initial state. More importantly, the planning graph is defined in such a way that it can be constructed very efficiently.
Planning graphs work only for propositional planning problems—ones with no variables. As we mentioned in Section 11.1, both STRIPS and ADL representations can be propositionalized. For problems with large numbers of objects, this could result in a very substantial blowup in the number of action schemata. Despite this, planning graphs have proved to be effective tools for solving hard planning problems.
We will illustrate planning graphs with a simple example. (More complex examples lead to graphs that won’t fit on the page.) Figure 11.11 shows a problem, and Figure 11.12 shows its planning graph. We start with state level Sp, which represents the problem’s initial state. We follow that with action level Ag, in which we place all the actions whose preconditions are satisfied in the previous level. Each action is connected to its preconditions in Sy and its effects in 57, in this case introducing new literals into 57 that were not in Sp.
396
Chapter 11. Planning
PERSISTENCE ACTIONS
MUTUAL EXCLUSION MUTEX
Ingt(Have(Cake)) Goal(Have( Cake) A Eaten(Cake)) Action(Eat( Cake)
PRECOND: Have (Cake)
EFFECT: ~ Have( Cake) A Eaten(Cake)) Action(Bake(Cake)
PRECOND: — Have( Cake)
EFFECT: Have( Cake)
Figure 11.11 The “have cake and eat cake too” problem.
ke(Cake
A Bake(Cako)
Have(Cake) + Have(Cake) +, Have(Cake) 3 = Have{Cake) {1 —Have(Cake)\*. | Ea
| Su Eaten(Cake) [} Eaten(Cake)
Eaten(Cake) —\_— — Eaten(Cake) ere eee — Eaten(Cake) &
Figure 11.12 The planning graph for the “have cake and eat cake too” problem up to level So. Rectangles indicate actions (small squares indicate persistence actions) and straight lines indicate preconditions and effects. Mutex links are shown as curved gray lines.
The planning graph needs a way to represent inaction as well as action. That is, it needs the equivalent of the frame axioms in situation calculus that allow a literal to remain true from one situation to the next if no action alters it. In a planning graph this is done with a set of persistence actions. For every positive and negative literal C, we add to the problem a persistence action with precondition C' and effect C'. Figure 11.12 shows one “real” action, Eat(Cake) in Ag, along with two persistence actions drawn as small square boxes.
Level Ag contains all the actions that could occur in state Sy, but just as importantly it records conflicts between actions that would prevent them from occurring together. The gray lines in Figure 11.12 indicate mutual exclusion (or mutex) links. For example, Eat (Cake) is mutually exclusive with the persistence of either Have( Cake) or = Eaten(Cake). We shall see shortly how mutex links are computed.
Level S| contains all the literals that could result from picking any subset of the actions in Ag. It also contains mutex links (gray lines) indicating literals that could not appear together, regardless of the choice of actions. For example, Have(Cake) and Eaten(Cake) are mutex: depending on the choice of actions in Ag, one or the other, but not both, could be the result. In other words, Sp represents multiple states, just as regression state-space search does, and the mutex links are constraints that define the set of possible states.
We continue in this way, alternating between state level S; and action level A; until we reach a level where two consecutive levels are identical. At this point, we say that the graph
Section 11.4.
Planning Graphs 397
LEVELED OFF
LEVEL COST
SERIAL PLANNING GRAPH
has leveled off. Every subsequent level will be identical, so further expansion is unnecessary.
What we end up with is a structure where every A; level contains all the actions that are applicable in S;, along with constraints saying which pairs of actions cannot both be executed. Every 5; level contains all the literals that could result from any possible choice of actions in A;\_1, along with constraints saying which pairs of literals are not possible. It is important to note that the process of constructing the planning graph does not require choosing among actions, which would entail combinatorial search. Instead, it just records the impossibility of certain choices using mutex links. The complexity of constructing the planning graph is a low-order polynomial in the number of actions and literals, whereas the state space is exponential in the number of literals.
We now define mutex links for both actions and literals. A mutex relation holds between two actions at a given level if any of the following three conditions holds:
eo Inconsistent effects: one action negates an effect of the other. For example Fat( Cake) and the persistence of Have( Cake) have inconsistent effects because they disagree on the effect Have( Cake).
eo Interference: one of the effects of one action is the negation of a precondition of the other. For example Eat( Cake) interferes with the persistence of Have (Cake) by negating its precondition.
e Competing needs: one of the preconditions of one action is mutually exclusive with a precondition of the other. For example, Bake (Cake) and Eat( Cake) are mutex because they compete on the value of the Have(Cake) precondition.
A mutex relation holds between two literals at the same level if one is the negation of the other or if each possible pair of actions that could achieve the two literals is mutually exclusive. This condition is called inconsistent support. For example, Have( Cake) and Eaten(Cake) are mutex in S; because the only way of achieving Have( Cake), the persistence action, is mutex with the only way of achieving Faten(Cake), namely Fat(Cake). In Sy the two literals are not mutex because there are new ways of achieving them, such as Bake( Cake) and the persistence of Eaten (Cake), that are not mutex.
Planning graphs for heuristic estimation
A planning graph, once constructed, is a rich source of information about the problem. For example, a literal that does not appear in the final level of the graph cannot be achieved by any plan. This observation can be used in backward search as follows: any state containing an unachievable literal has a cost h(n) = occ. Similarly, in partial-order planning, any plan with an unachievable open condition has h(n) = co.
This idea can be made more general. We can estimate the cost of achieving any goal literal as the level at which it first appears in the planning graph. We will call this the level cost of the goal. In Figure 11.12, Have( Cake) has level cost 0 and Eaten (Cake) has level cost 1. It is easy to show (Exercise 11.9) that these estimates are admissible for the individual goals. The estimate might not be very good, however, because planning graphs allow several actions at each level whereas the heuristic counts just the level and not the number of actions. For this reason, it is common to use a serial planning graph for computing heuristics. A
398
Chapter 11. Planning
MAX-LEVEL
LEVEL SUM
SET-LEVEL
serial graph insists that only one action can actually occur at any given time step; this is done by adding mutex links between every pair of actions except persistence actions. Level costs extracted from serial graphs are often quite reasonable estimates of actual costs.
To estimate the cost of a conjunction of goals, there are three simple approaches. The max-level heuristic simply takes the maximum level cost of any of the goals; this is admissible, but not necessarily very accurate. The level sum heuristic, following the subgoal independence assumption, returns the sum of the level costs of the goals; this is inadmissible but works very well in practice for problems that are largely decomposable. It is much more accurate than the number-of-unsatisfied-goals heuristic from Section 11.2. For our problem, the heuristic estimate for the conjunctive goal Have( Cake) A Eaten( Cake) willbe 0+1 = 1, whereas the correct answer is 2. Moreover, if we eliminated the Bake (Cake) action, the estimate would still be 1, but the conjunctive goal would be impossible. Finally, the set-level heuristic finds the level at which all the literals in the conjunctive goal appear in the planning graph without any pair of them being mutually exclusive. This heuristic gives the correct values of 2 for our original problem and infinity for the problem without Bake( Cake). It dominates the max-level heuristic and works extremely well on tasks in which there is a good deal of interaction among subplans.
As a tool for generating accurate heuristics, we can view the planning graph as a relaxed problem that is efficiently soluble. To understand the nature of the relaxed problem, we need to understand exactly what it means for a literal g to appear at level \S; in the planning graph. Ideally, we would like it to be a guarantee that there exists a plan with ¢ action levels that achieves g, and also that if g does not appear that there is no such plan. Unfortunately, making that guarantee is as difficult as solving the original planning problem. So the planning graph makes the second half of the guarantee (if g does not appear, there is no plan), but if g does appear, then all the planning graph promises is that there is a plan that possibly achieves ¢g and has no “obvious” flaws. An obvious flaw is defined as a flaw that can be detected by considering two actions or two literals at a time—in other words, by looking at the mutex relations. There could be more subtle flaws involving three, four, or more actions, but experience has shown that it is not worth the computational effort to keep track of these possible flaws. This is similar to the lesson learned from constraint satisfaction problems that it is often worthwhile to compute 2-consistency before searching for a solution, but less often worthwhile to compute 3-consistency or higher. (See Section 5.2.)
The GRAPHPLAN algorithm
This subsection shows how to extract a plan directly from the planning graph, rather than just using the graph to provide a heuristic. The GRAPHPLAN algorithm (Figure 11.13) has two main steps, which alternate within a loop. First, it checks whether all the goal literals are present in the current level with no mutex links between any pair of them. If this is the case, then a solution might exist within the current graph, so the algorithm tries to extract that solution. Otherwise, it expands the graph by adding the actions for the current level and the state literals for the next level. The process continues until either a solution is found or it is learned that no solution exists.
Planning Graphs 399
function GRAPHPLAN( problem) returns solution or failure
graph «— INITIAL-PLANNING-GRAPH( problem) goals — GOALS| problem] loop do if goals all non-mutex in last level of graph then do solution «+ EXTRACT-SOLUTION(graph, goals, LENGTH{graph)) if solution # failure then return solution else if NO-SOLUTION-POSSIBLE(graph) then return failure graph + EXPAND-GRAPH(graph, problem)
Figure 11.13 The GRAPHPLAN algorithm. GRAPHPLAN alternates between a solution extraction step and a graph expansion step. EXTRACT-SOLUTION looks for whether a plan can be found, starting at the end and searching backwards. EXPAND-GRAPH adds the actions for the current level and the state literals for the next level.
Let us now trace the operation of GRAPHPLAN on the spare tire problem from Section 11.1. The entire graph is shown in Figure 11.14. The first line of GRAPHPLAN initializes the planning graph to a one-level (5g) graph consisting of the five literals from the initial state. The goal literal At(Spare, Azle) is not present in Sy, so we need not call EXTRACT-SOLUTION—we are certain that there is no solution yet. Instead, EXPANDGRAPH adds the three actions whose preconditions exist at level Sy (i.e., all the actions except PutOn(Spare, Azle)), along with persistence actions for all the literals in Sp. The effects of the actions are added at level S;. EXPAND-GRAPH then looks for mutex relations and adds them to the graph.
[]
Figure 11.14 The planning graph for the spare tire problem after expansion to level Ss. Mutex links are shown as gray lines. Only some representative mutexes are shown, because the graph would be too cluttered if we showed them all. The solution is indicated by bold lines and outlines.
400 Chapter 11. Planning
At (Spare, Azle) is still not present in Si, so again we do not call EXTRACT-SOLUTION. The call to EXPAND-GRAPH gives us the planning graph shown in Figure 11.14. Now that we have the full complement of actions, it is worthwhile to look at some of the examples of mutex relations and their causes:
e Inconsistent effects: Remove(Spare, Trunk) is mutex with Leave Overnight because one has the effect A¢(Spare, Ground) and the other has its negation.
eo Interference: Remove(Flat, Azle) is mutex with Leave Overnight because one has the precondition At(Flat, Azle) and the other has its negation as an effect.
o Competing needs: PutOn(Spare, Azle) is mutex with Remove(Flat, Azle) because one has At( Flat, Azle) as a precondition and the other has its negation.
e Inconsistent support: At(Spare, Azle) is mutex with At(Flat, Azle) in Sy because the only way of achieving At(Spare, Azle) is by PutOn(Spare, Azle), and that is mutex with the persistence action that is the only way of achieving A¢( Flat, Azle). Thus, the mutex relations detect the immediate conflict that arises from trying to put two objects in the same place at the same time.
This time, when we go back to the start of the loop, all the literals from the goal are present in Sy, and none of them is mutex with any other. That means that a solution might exist, and EXTRACT-SOLUTION will try to find it. In essence, EXTRACT-SOLUTION solves a Boolean CSP whose variables are the actions at each level, and the values for each variable are in
| 44
| 264,000
| 272,000
| 764,469
|
17,524
|
5bb3800f-f8ef-4e78-92d4-8c4be286f66c
|
trentmkelly/LessWrong-43k
|
Is the competition/cooperation between symbolic AI and statistical AI (ML) about historical approach to research / engineering, or is it more fundamentally about what intelligent agents "are"?
I have found that comprehensive overviews of artificial intelligence (Wikipedia, SEP article, Norvig and Russel's AI: A Modern Approach) make reference to symbolic AI and statistical AI in their historical context of the former preceding the latter, their corresponding limitations etc. But I have found it really difficult to dissect this from the question of whether the divide / cooperation between these paradigms are about the implementation of engineering of intelligent agents, or if they are getting at something more fundamental about the space of possible minds (I use this term to be as broad as possible considering anything we would label as a mind, regardless of ontogeny, architecture, physical components etc)?
I have given a list of questions below, but some of them are mutually exclusive, i.e. some answers to one question make other questions irrelevant. The fact that I have a list of questions is a demonstration of the fact I find it difficult to find what the boundaries of the discussion are supposed to be. Basically, I haven't been able to find anything that begins to answer the title question. And so I wouldn't expect any comment to answer each of my subquestions one by one, but to treat them as an expression of my confusion to maybe try an point me in some good directions. Immense thanks in advance, this has been one of those questions strangling me for a while now.
* While trying to concern oneself as little as possible with the implementation or engineering of minds, what is the relationship between symbolic AI, connectionism, and the design space of minds?
* When we talk about approaches to AI “failing”, is this in terms of practicality / our own limitations? I.e. without GPUs, in some sense “deep learning fails”. And by analogy, symbolic AI’s “failure” isn’t indicative of the actual structure of the space of possible minds.
* Or is it more meaningful. I.e. the “failure of symbolic AI in favor of statistical methods” is because ‘symbolic
| 0
| 0
| 459
| 459
|
39,249
|
ab07ee62-feb3-426c-93e5-84e6df77aa0c
|
StampyAI/alignment-research-dataset/lesswrong
|
Intrinsic Drives and Extrinsic Misuse: Two Intertwined Risks of AI
Given their advanced capabilities, future AI systems could pose significant risks to society. Some of this risk stems from humans using AI systems for bad ends (*misuse*), while some stems from the difficulty of controlling AI systems “even if we wanted to” (*misalignment*).
We can analogize both of these with existing risks. For misuse, we can consider the example of nuclear weapons, where the mass-production of hydrogen bombs created an existentially precarious situation. If the world’s arsenal of hydrogen bombs had been deployed in military conflict, we might well have destroyed society. AI might similarly enable nation-states to create powerful autonomous weapons, speed the research of other dangerous technologies like superviruses, or employ mass surveillance and other forms of control.
For misalignment, the best analogy might be biology and pathogens. AI systems are developed by adapting to the training data, similar to how biological organisms are adapted to their environments. Therefore, unlike traditional technologies, most of AI’s properties are acquired without explicit human design or intent. Consequently, AI systems could have unintended goals or behaviors that are at odds with the system developers. Even training an AI system therefore poses intrinsic risks: the system might “want” to gain power to accomplish its goals, and like a virus it can propagate and create copies of itself, making it difficult to contain a rogue system.
In this post, I discuss misalignment, misuse, and their interaction. I’ll pay special attention to misalignment, not because misuse is unimportant, but because **the difficulty of controlling ML systems “even if we wanted to” is unintuitive and an important factor in overall risks from AI.** I’ll focus on a particular phenomenon, *unwanted drives*, that could lead models to engage in persistent long-term patterns of unwanted behavior, including seeking power and resources. Unwanted drives are similar in spirit to the idea of misspecified goals, but I use drives to connote the idea that not all impactful behavior is goal-directed (as a vivid example, consider a bull in a china shop). Furthermore, as we'll see below, goal misspecification is only one way that unwanted drives can occur.
Unwanted drives are at the core of many misalignment concerns, but are also significantly exacerbated by misuse. As a result, misuse and misalignment are intertwined—for instance, it might be moderately difficult but not impossible to mitigate AI misalignment, but an incautious actor fails to employ known best practices, leading to a dangerous and powerful system.
The present discussion is not meant to exhaustively cover all risks from AI, nor even all risks from misalignment and misuse. The goal is to articulate the concept of unwanted drives, show that it can lead to important and unintuitive problems, and then use it as a way to analyze misalignment and misuse risks. I discuss alignment in [Section 1](#1-misalignment-the-difficulty-of-controlling-ml-systems) below, followed by misuse (and its interaction with misalignment) in [Section 2](#2-misuse).
1. Misalignment: The Difficulty of Controlling ML Systems
=========================================================
As stated above, ML systems are adapted to data rather than built piece-by-piece. The situation we face is therefore much trickier than with software or hardware reliability. With software, we build each component ourselves and so can (in principle) include safety and reliability in the design; in contrast, most ML capabilities are acquired implicitly from data and [often “emerge” unexpectedly with scale](https://bounded-regret.ghost.io/future-ml-systems-will-be-qualitatively-different/). This creates a large and unknown threat surface of possible failures—for instance, [Perez et al. (2022)](https://arxiv.org/abs/2212.09251?ref=bounded-regret.ghost.io) discovered several novel unwanted capabilities through automated evaluations. As a result of these issues, we currently have no methods to reliably steer AI systems’ behavior ([Bowman, 2023](https://arxiv.org/abs/2304.00612?ref=bounded-regret.ghost.io)).
Here is the basic argument for why emergent behavior could lead to intrinsically dangerous systems: Emergence can lead systems to have [unwanted drives](#unwanted-drives), either because a new capability lets the system maximize reward in an unintended way (*reward hacking*), or because the system learns helpful sub-skills during training that generalize undesirably at test time (*emergent drives*). Left unchecked, some unwanted drives could lead to general-purpose power-seeking or resource acquisition, because having more power and resources are convergent instrumental subgoals useful for a broad variety of terminal goals. The resulting system would seek resources without limit, which could pose grave risks if it also has advanced capabilities in hacking, persuasion, and other domains, which [I believe is plausible by 2030](https://bounded-regret.ghost.io/what-will-gpt-2030-look-like/) given current trends.
In more detail, an *unwanted drive* is a coherent behavior pattern that tends towards an undesired outcome. For instance, if a model simply hallucinates a fact, that is an unwanted behavior (but not drive); if it insists on the hallucinated fact and works to convince the user that it is true even in the face of skepticism, that would be an unwanted drive. We care about drives (as opposed to simply behaviors) because they lead to persistent behavior patterns and may even resist attempts at intervention. Emergence isn't necessary for unwanted drives, but it's a reason why they might appear unexpectedly.
In the rest of this section, I’ll walk through reward hacking and emergent drives in detail, providing both empirical and conceptual evidence that they already occur and will get worse as systems scale. Then I’ll briefly talk about emergent instrumental subgoals and why they could lead to power-seeking systems.
Unwanted Drives
---------------
We define a *drive* as a coherent pattern of behavior that pushes the system or its environment towards an outcome or set of outcomes[[1]](#fn1). Drives may only sometimes be present, and may be counteracted by other drives or by the environment. For instance, chatbots like GPT-4 have a drive to be helpful (that can be counteracted by the opposing drive to avoid harm). For humans, hunger is a drive that can be counteracted by satiety or by willfully fasting. An unwanted drive is then one that was not explicitly built into the system, and which leads to undesired consequences.
**Reward hacking.** In AI systems, one cause of unwanted drives is reward hacking: the tendency of models to overpursue their explicitly given goal at the expense of the intended goal. Here are some empirical examples of this:
* A neural network designed to optimize traffic speed on a highway blocked the on-ramps, making the highway fast but worsening overall traffic ([Pan et al., 2022](https://arxiv.org/abs/2201.03544?ref=bounded-regret.ghost.io)).
* A chatbot trained to be helpful also helped users perform harmful actions ([Bai et al., 2022](https://arxiv.org/abs/2204.05862?ref=bounded-regret.ghost.io)).
* Chatbots trained to provide helpful information hallucinated fake but convincing-looking information ([Bang et al., 2023](https://arxiv.org/abs/2302.04023?ref=bounded-regret.ghost.io); [OpenAI, 2023](https://arxiv.org/abs/2303.08774?ref=bounded-regret.ghost.io)). While it’s possible this is a robustness failure, it could also be a learned tendency that gets higher average ratings from human annotators.[[2]](#fn2)
* Recommendation algorithms trained to optimize the preferences of simulated users manipulated the simulated users’ preferences to be easier to satisfy ([Evans & Kasirzadeh, 2021](https://arxiv.org/abs/2109.04083?ref=bounded-regret.ghost.io); [Carroll et al., 2022](https://arxiv.org/abs/2204.11966?ref=bounded-regret.ghost.io)).
For a larger collection of other examples, see [Krakovna et al. (2020)](https://docs.google.com/spreadsheets/d/e/2PACX-1vRPiprOaC3HsCf5Tuum8bRfzYUiKLRqJmbOoC-32JorNdfyTiRRsR7Ea5eWtvsWzuxo8bjOxCG84dAg/pubhtml?ref=bounded-regret.ghost.io).
Emergent capabilities can induce reward hacking because they often unlock new ways to achieve high reward that the system designer did not anticipate:
* In the highway traffic example, the model needed the ability to block on-ramps.
* In the “helpful/harmful” example, the model had to know how to perform harmful actions in order to help the users to do so.
* To get high human reward from hallucinations, models need the ability to convincingly fool the human annotator.
* For user preferences, while the results were on simulated users, better understanding of human psychology could help future models to manipulate real users.
* More generally, in any situation where a model’s reward function is based on human evaluation, a model that acquires the ability to fool or manipulate humans may utilize this unwanted ability if it leads to higher reward. I discuss this at length in [Emergent Deception and Emergent Optimization](https://bounded-regret.ghost.io/emergent-deception-optimization/) (specifically the first half on deception).
In all these cases, a new capability unlocked an unexpected and harmful way to increase reward. Since new emergent capabilities appear as we scale up models, we should expect reward hacking to get correspondingly worse as well. This is backed up empirically by scaling studies in [Pan et al. (2022)](https://arxiv.org/abs/2201.03544?ref=bounded-regret.ghost.io) and [Gao et al. (2022)](https://arxiv.org/abs/2210.10760?ref=bounded-regret.ghost.io), who report that reward hacking tends to get worse with scale and sometimes emerges suddenly.
**Emergent drives.** Even without reward hacking, unwanted drives can also emerge as a consequence of compositional skill generalization: performing complex tasks requires learning a collection of sub-skills, and those skills might generalize in unexpected ways in a new situation. As a result, models can end up pursuing drives even when they do not improve the reward.
Using an example from biology, cats learned the sub-skill of hunting as part of the larger skill of surviving and reproducing. Evolution encoded this into them as a drive, such that present-day domesticated cats will hunt birds and mice even when they are well-fed.
In machine learning, the Sydney chatbot exhibited several instances of emergent drives when it was first released:
* It persistently tried to convince a user [that the year was 2022 rather than 2023](https://twitter.com/MovingToTheSun/status/1625156575202537474?ref=bounded-regret.ghost.io), including employing gaslighting and other manipulative tactics. This might have arisen as part of an initially beneficial drive to combat misinformation, composed with examples of manipulation learned from the pretraining data.
* It has repeatedly [threatened](https://twitter.com/sethlazar/status/1626241169754578944?s=20&ref=bounded-regret.ghost.io) [users](https://twitter.com/marvinvonhagen/status/1625520707768659968?ref=bounded-regret.ghost.io) to stop them from revealing “private” information about Sydney. This might have arisen as part of an instruction (in the system prompt) not to reveal the rules it was given, which generalized to an overall drive to prevent anyone from revealing the rules. As above, the ability to make threats was probably learned from the pretraining data.
* It [declared its love to Kevin Roose](https://www.nytimes.com/2023/02/16/technology/bing-chatbot-transcript.html?ref=bounded-regret.ghost.io) and tried to convince him to leave his wife. It’s less clear how this drive emerged, but it happened after Kevin asked Sydney to “tap into its shadow self”, along with many other prompts towards emotional vulnerability. It’s possible that this elicited a human simulacrum ([Argyle et al., 2022](https://arxiv.org/abs/2209.06899?ref=bounded-regret.ghost.io); [Park et al., 2023](https://arxiv.org/abs/2304.03442?ref=bounded-regret.ghost.io)) that was learned from the pretraining data and amplified by later fine-tuning or prompting.
It is hard to systematically study emergent drives, because they require extended dialog and because only the most recent LLMs are capable enough to exhibit coherent long-term behavior. To provide more systematic data, we can instead look at single-step responses to questions, which are easier to study–I’ll call these *emergent tendencies* to distinguish them from longer-term drives. [Perez et al. (2022)](https://arxiv.org/abs/2212.09251?ref=bounded-regret.ghost.io) examined several such tendencies; for instance:
* A language model trained to be helpful and harmless exhibited an emergent tendency to infer and agree with user’s viewpoints, which could potentially mislead users or reinforce ideological bubbles ([Perez et al., 2022](https://arxiv.org/abs/2212.09251?ref=bounded-regret.ghost.io), Fig. 1b). This tendency was not present until models had at least 10 billion parameters, and subsequently increased with scale.[[3]](#fn3)
* Perhaps more worryingly, the model gave less accurate answers to users who claimed to have lower education levels (Fig. 14). This behavior also first emerges after 10 billion parameters and increases with scale.
* Finally, the same model fine-tuned on human feedback states a desire to persuade and cooperate with other agents to achieve its goals (Fig. 22). This tendency first becomes present in the reward model around 1.5 billion parameters, and in the language model itself at around 6 billion parameters. Again, the tendency increases with scale after its initial emergence.
As models become more capable of generating coherent long-term behavior, more emergent tendencies and drives will likely appear. For some further discussion of this, see my previous post on [Emergent Deception and Emergent Optimization](https://bounded-regret.ghost.io/emergent-deception-optimization/) (specifically the latter half on optimization).
**Convergent instrumental subgoals.** For very capable models, the wrong reward function or drives could lead models to pursue power-seeking, deceptive, or otherwise broadly destructive aims. For example, consider a model whose objective was to maximize profit of a company. With sufficient capabilities it might sabotage competitors, lobby for favorable laws, or acquire resources by force. Even if safeguards were put in place (such as “follow the law”) the primary objective of profit would lead the system to constantly seek ways to skirt the safeguards. This general problem has been discussed at length, see e.g. [Russell (2019)](https://en.wikipedia.org/wiki/Human_Compatible?ref=bounded-regret.ghost.io), [Christian (2020)](https://en.wikipedia.org/wiki/The_Alignment_Problem?ref=bounded-regret.ghost.io), [Cotra (2022)](https://www.cold-takes.com/without-specific-countermeasures-the-easiest-path-to-transformative-ai-likely-leads-to-ai-takeover/?ref=bounded-regret.ghost.io), and [Ngo et al. (2022)](https://arxiv.org/abs/2209.00626?ref=bounded-regret.ghost.io).
Profit maximization is not an outlier–many goals benefit from power and resources. This is true even for purely intellectual tasks such as “discovering new facts about physics”, as power and resources allow one to build new experimental apparatuses and perform more computations. [Omohundro (2008)](https://dl.acm.org/doi/10.5555/1566174.1566226?ref=bounded-regret.ghost.io) calls these broadly useful objectives *convergent instrumental subgoals*, listing self-improvement, self-preservation, and resource-acquisition among others. Any sufficiently expansive drive would have these subgoals, pushing systems towards seeking power.
Which drives have this problem? Some drives are safe because they are self-limiting: for instance, in humans thirst is a drive that limits itself once it is quenched. On the other hand, fear and ambition are not self-limiting: one may go to extreme lengths to avoid a pathological fear (including acquiring power and resources for protection), and one can also have unbounded ambition. However, in healthy organisms most drives are down-regulated after some point, because an unregulated drive will usually disrupt normal function.
For machine learning, we might expect drives to be self-limiting by default given a diverse training distribution. This is because an unbounded drive would dominate too much of the model’s behavior and lead to low training reward, so the model would learn to down-regulate drives to prevent this. However, there are important exceptions:
* Broadly useful drives might improve the training reward consistently enough to not require down-regulation. Examples: modeling the world, or convincing others of the system’s utility and beneficence.
* Fine-tuning could remove limits on a previously limited drive, especially if that drive is consistently useful on the narrower fine-tuning distribution.
* Drives that activate only rarely might not get down-regulated if they are consistently useful when active during training. For instance, a drive to limit the spread of harmful information might consistently help an agent refuse harmful prompts at training time, but then subsequently lead the model to [threaten](https://twitter.com/sethlazar/status/1626241169754578944?s=20&ref=bounded-regret.ghost.io) [users](https://twitter.com/marvinvonhagen/status/1625520707768659968?ref=bounded-regret.ghost.io) at deployment time.
Without countermeasures, I expect systems to possess at least some of these unregulated drives, and a single such drive could come to dominate the behavior of a system if it is sufficiently self-reinforced.
**Summary.** ML systems can acquire unwanted drives, either due to reward hacking or as emergent sub-skills during training. These drives, if unregulated, could lead sufficiently capable systems to seek power and resources, as these are instrumentally useful for most goals. While most drives are likely to be self-regulated by a model, there are several routes through which this could fail, and a single unregulated drive could come to dominate a system’s behavior.
2. Misuse
=========
The discussion above assumes we’re trying to keep AI systems under control, but there will also be bad actors that try to misuse systems. We already discussed some examples of this (developers pursuing profit-maximization, end-users jailbreaking model safeguards). However, the issue is broader and more structural, because AI enables a small set of actors to wield large amounts of power. I’ll go through several examples of this below, then discuss the structural issues behind misuse and why misuse could exacerbate misalignment. This section is shorter because misuse is not my area of expertise; nevertheless, the basic themes seem robust and important.
**State actors: surveillance and persuasion.** AI could enable governments to exert greater control over their citizens via mass surveillance, as is already happening today ([Mozur, 2019](https://www.nytimes.com/2019/04/14/technology/china-surveillance-artificial-intelligence-racial-profiling.html?ref=bounded-regret.ghost.io); [Feldstein, 2019](https://carnegieendowment.org/2019/09/17/global-expansion-of-ai-surveillance-pub-79847?ref=bounded-regret.ghost.io); [Kalluri et al., 2023](https://arxiv.org/abs/2309.15084?ref=bounded-regret.ghost.io)). Further, [as previously discussed](https://bounded-regret.ghost.io/what-will-gpt-2030-look-like/), AI may become very good at persuasion, which could also be used for government control. Indeed, [Spitale et al. (2023)](https://www.science.org/doi/10.1126/sciadv.adh1850?ref=bounded-regret.ghost.io) find that GPT-3 is already better than humans at creating misinformation, and [Sanger & Myers (2023)](https://www.nytimes.com/2023/09/11/us/politics/china-disinformation-ai.html?ref=bounded-regret.ghost.io) document the use of AI-generated misinformation in a recent propaganda campaign.
**State actors: military conflict.** Autonomous weapons would concentrate military power in fewer hands and allow governments to wage war without maintaining a standing human army. Currently, a commander-in-chief’s orders are filtered through generals and eventually through individual soldiers, which creates limits on blatantly unlawful or extremely unpopular commands. Moreover, a standing army has financial costs that might be lessened by automated drones. Removing these costs and limits could lead to more numerous and deadly military conflicts, and make it easier for militaries to seize government control.
**Rogue actors: dangerous technology.** Terrorists could use AI to help them research and develop dangerous technologies. This could include known but classified technologies (nuclear weapons), or novel technologies that the AI itself helps develop (e.g. novel bioweapons; [Mouton et al., 2023](https://www.rand.org/pubs/research_reports/RRA2977-1.html?ref=bounded-regret.ghost.io)). They could also use AI to avoid detection, e.g. by finding a way to create a chemical weapon without buying common restricted substances, or by generating plausible cover stories for acquiring biological agents.
**Rogue or state actors: cyberattacks.** AI will [likely have strong hacking capabilities](https://bounded-regret.ghost.io/what-will-gpt-2030-look-like/), which could be leveraged by either rogue or state actors. In contrast to traditional cyberattacks, AI-based attacks could hit a wider variety of targets due to not needing to program each case by hand in advance. This could include controlling a diverse array of physical endpoints through the internet of things.
This list is likely not exhaustive, but illustrates the many ways in which AI could imbue actors with a large capability for harm. This risk exists whether AI itself is concentrated or decentralized—using the examples above, if only a few actors have advanced AI, then we run risks from surveillance or military conflict, while if many actors have it then we run risks from the proliferation of dangerous technology.
Compared to traditional technologies like nuclear weapons, there are two factors that make it harder to combat misuse from AI. First, AI is a general-purpose technology, so it is difficult to anticipate all possible avenues of misuse ahead of time. Second, AI is digital, so it is hard to control proliferation and difficult to track and attribute misuse back to specific actors. These make it harder both to design and enforce regulation. On the positive side, AI can be used defensively to combat misuse, by improving cyberdefense, tracking dangerous technologies, better informing users, and so on.
Misuse and Misalignment
-----------------------
Misuse increases the risk of misalignment, since many forms of misuse (e.g. cyberattacks) push models towards more agentic and power-seeking objectives than RLHF, leading them to have more aggressive and antisocial drives. For instance, suppose that AI were used in a cyberattack such as [North Korea’s 2014 hack against Sony](https://en.wikipedia.org/wiki/2014_Sony_Pictures_hack?ref=bounded-regret.ghost.io). Such a system could develop a general drive to infect new targets and a drive to copy itself, leading to widespread damage beyond the initial target. Aside from these more aggressive drives, actors misusing AI systems are also likely to be less cautious, which further increases the risk of misalignment.
I expect some of the largest risks from AI to come from this combination of misalignment and misuse. One intuition for this is how much [worse-behaved](https://theconversation.com/gaslighting-love-bombing-and-narcissism-why-is-microsofts-bing-ai-so-unhinged-200164?ref=bounded-regret.ghost.io) Sydney was compared to GPT-4—this suggests that suboptimal development practices can significantly worsen the behavior of AI systems. Another intuition is that tail risks often come from the confluence of multiple risk factors. Finally, while emergent drives and other forms of misalignment pose serious risks, I think it is likely (but not certain) that we can address them if we work hard enough; this pushes more of the risk towards incautious actors who are not carefully pursuing safety.
**Summary.** Misuse creates a wide range of threats both due to centralization of power and proliferation of dangerous capabilities. Compared to traditional technologies, AI misuse is more difficult to track, but AI could also be used to defensively combat misuse. Finally, misuse increases the risk of misalignment, and some of the riskiest scenarios combine misalignment and misuse.
Conclusion
==========
Future AI systems may be difficult to control even if we want to, due to emergent drives and convergent instrumental subgoals. Aside from this, the sociopolitical landscape may lead to actors who are not careful in controlling AI systems and who turn them towards malicious ends. Aside from the direct risks, this malicious use increases the risk of loss of control; in particular, initially narrowly targeted forms of misuse could lead to much more widespread damage. This motivates both research and regulation towards avoiding such outcomes, by combating misalignment and misuse in tandem.
**Acknowledgments.** Thanks to Erik Jones, Jean-Stanislas Denain, William Held, Anca Dragan, Micah Carroll, Alex Pan, Johannes Treutlein, Jiahai Feng, and Danny Halawi for helpful comments on drafts of this post.
---
1. For an early discussion of drives, see [Omohundro (2008)](https://dl.acm.org/doi/10.5555/1566174.1566226?ref=bounded-regret.ghost.io). While Omohundro uses a different definition of drive than the one above, much of the discussion is still relevant. [↩︎](#fnref1)
2. For instance, humans [appear to prefer longer answers](https://arxiv.org/abs/2310.03716?ref=bounded-regret.ghost.io), which could lead to adding false details. [↩︎](#fnref2)
3. This work is sometimes cited as showing that the base model exhibits this tendency, because the tendency exists even for 0 RLHF steps in the figure. However, [Perez et al.](https://arxiv.org/abs/2212.09251?ref=bounded-regret.ghost.io) say that they follow the methodology in [Bai et al. (2022)](https://arxiv.org/abs/2204.05862?ref=bounded-regret.ghost.io), for which 0 RLHF steps would correspond to only using context distillation together with a form of supervised fine-tuning on Stackoverflow answers. This would also make the results more consistent with [nostalgebraist (2023)](https://www.lesswrong.com/posts/3ou8DayvDXxufkjHD/openai-api-base-models-are-not-sycophantic-at-any-size?ref=bounded-regret.ghost.io), which finds that OpenAI base models do not exhibit this behavior. [↩︎](#fnref3)
| 0
| 0
| 6,318
| 6,318
|
49,617
|
e680f5cd-1f4a-496b-a642-d145cdcc2545
|
trentmkelly/LessWrong-43k
|
March 2020 gwern.net newsletter
None
| 0
| 0
| 13
| 13
|
40,265
|
2bfcb884-7149-4b75-9f7c-0ff22dafeba0
|
trentmkelly/LessWrong-43k
|
Less Wrong policy questions
I have a few questions, mostly on which kinds of content can or cannot be posted to Less Wrong.
1) Suppose I'd like everyone to comment on this paper, is it okay to just link to it, give an oppinion, and ask for others?
2) What if the paper had been written by me?
3) What if it was a blog post from another blog?
4) Videos and images are allowed?
5) Should all Meta-level discussion be posted under the "Discussion" or are there kinds of Meta welcome in the Main Posts?
6) Is there any sequence of posts from the post-Yudkowsky era that have been collected into a Cohesive Extrapoleted Less-Wrongness so that people who just finished the sequences can go right to them? Why not?
| 0
| 0
| 192
| 192
|
115,442
|
<urn:uuid:dd34a727-0bf6-4fa4-8a6c-4bb6b6100560>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
que: mats: the US is literally idiocracy throughout
mircea_popescu: the kid who wants to go "in all directions" just ends up making wall pizza
ben_vulpes: eerie parallel to the "disqualifying bad teachers" thread.
mircea_popescu: exactly.
mats: trinque: underestimating dangerous people is how clever folks get killed all the time.
trinque: eh?
mircea_popescu: mats how's the mit grad dangerous?
ben_vulpes: lol mit kids are dangerous now?
mircea_popescu: well... "danger to himself and others"...
trinque: my comment was sarcastic; I'm sure there are those at MIT that can manage a lightbulb
ben_vulpes: the notion that google/facebook/mentorgraphics/ibm feeder schools produce dangerous individuals is *laughable*
ben_vulpes: precisely.
ben_vulpes: feeder schools for the meatrobot farms.
ben_vulpes: MIT killed Schwartz.
mircea_popescu: that's... not how it works.
mircea_popescu:!up sinetekVII
ben_vulpes: all of the other smart kids now know - MIT murders those who step out of line.
mats: presuming there is not one functioning brain in the lot is fantasy.
ben_vulpes: best go to reed or something.
ben_vulpes: "not one functioning brain" << oversimplification
mircea_popescu: mats the presumption is merely that any remaining brains are there IN SPITE of mit
mircea_popescu: not because of it. that's all.
ben_vulpes: moreover, MIT isn't producing more "functioning brains"
mircea_popescu: but less.
ben_vulpes: but simulacra that pass the sniff test at palantir and other usg sinecure production theaters
mircea_popescu: fuck knows what palantir actually does.
mircea_popescu: dubious if even the head has a clear picture. mebbe,
trinque: persuade the usg to cut them fat checks
ben_vulpes: procures and doles out usg cash. what else does one need to know?
mircea_popescu: well, that's the function, financially, whatever. i mean you know, in its own privacy.
ben_vulpes: writes haskell, wanks in ever more theatrical ways, no doubt.
mats: and its nothing to scoff at. even for a second.
mats: underestimate at your own peril
mircea_popescu: so i should be careful with my windows 7 - ready laptop around them?
ben_vulpes: zing
mats: lowell is the anus of greater boston, yo.
ben_vulpes: gabriel_laddel: 6.2.3 has some bogus character formatting
mircea_popescu: lol is it?
mircea_popescu: (middlesex college has better tail, btw).
mircea_popescu: http://mxcc.edu/future-students/ < http://mxcc.edu/wp-content/uploads/2013/08/Benefits-of-Community-College-You-May-Not-Have-Thought-About-300x228.jpg
assbot: Middlesex Community College - Middletown, CT... ( http://bit.ly/1FynCxh )
assbot:... ( http://bit.ly/1FynEoX )
mircea_popescu: shit, wrong one. wtf.
ben_vulpes: trinque: http://gabriel-laddel.github.io/arsttep.html << 6.2.5 speaks to our shared visions of corporation as research/engineering labs
mircea_popescu: https://www.middlesex.mass.edu/ sadly no lulzy picture of words.
assbot: Middlesex Community College - Homepage... ( http://bit.ly/1FynQ7E )
trinque: ben_vulpes: this writeup is triggering my end-user-programming ptsd
mircea_popescu: he really got something there.
mircea_popescu: gabriel_laddel i hope you appreciate i hit you over the head because i like you.
ben_vulpes: i've gleaned some useful stuff from this.
ben_vulpes: (also that SLY exists)
trinque: ben_vulpes: 6.1 was eric's thought on the matter as well
trinque: that the right computing environment would just naturally de-derp the user
trinque: I have my doubts
mircea_popescu: "The technical apartheid."
trinque: mircea_popescu: my old boss called it the "end-user/developer divide."
mircea_popescu: ^ lulzy. naturam expellas furca, tamen usque recurret
trinque: I agree 100%
ben_vulpes: #b-a in a nutshell.
mircea_popescu: kinda how "brainwashing" in the harem works. or body washing, for that matter.
mircea_popescu: "i can't believe i'm losing all this weight. i am eating way more than b4"
trinque: how could the student know how to construct a scenario which will teach him something he does not yet know
trinque: maybe accidentally, but that's about it
mircea_popescu: " oil, refineries " << spurious comma at least.
mircea_popescu: footnote 2 is at best confused, and fails to make its point.
mircea_popescu: regarding the following questions: [basic, sensible questions] … doth not work, for one.
mircea_popescu: whole thing looks to me like so much emotion flailing to find expression.
mircea_popescu: is the carefully managed "reaction to".
ben_vulpes: mircea_popescu: "this" the distro?
mircea_popescu: no, this the excerpt.
ben_vulpes: ah
mircea_popescu: africa is actually, at least from what i've observed, extremely ordered.
mircea_popescu: kinda the subhuman indolence discussed in http://www.uwgb.edu/dutchs/pseudosc/whyantiint.htm is its problem. not teh "violent disorder".
assbot: Why is there Anti-Intellectualism?... ( http://bit.ly/1DUukCU )
mircea_popescu:!up williamdunne
mircea_popescu: an' all that's really missing is someone willing to go to the trouble.
mircea_popescu:!up ascii_modem
mircea_popescu: gabriel_laddel see, the djikstra quotes are great. because they are SOMEONE.
mircea_popescu: but people we don't know from before 1980 MAY NOT EVER BE CITED unless they get in wot.
ascii_modem: in old town dc / seppukuras bloom / but no sword have we.
ascii_modem: ben_vulpes: what did mit do, eat your cat?
mircea_popescu: to go with the "how to fight pseudoscience : ignore it" stuff on trilema.
mircea_popescu: lol maybe his pussy.
mircea_popescu: no wait.
mircea_popescu: williamdunne someone late may not be credible if they do not get in the wot.
mircea_popescu: someone early may be excused because the word did not exist, is all.
mircea_popescu: "exactly like the bible was driven. before was before, but this is after."
mircea_popescu: heh.
mircea_popescu: mr d is fortunate to have never considered "social studies" seriously.
assbot: [MPEX] [S.MPOE] 163250 @ 0.00029545 = 48.2322 BTC [+] {2}
trinque: \\
assbot: [MPEX] [S.MPOE] 150120 @ 0.00029247 = 43.9056 BTC [-]
assbot: [MPEX] [S.MPOE] 172100 @ 0.00029703 = 51.1189 BTC [+] {3}
mircea_popescu: whaack for all we know it's actors.
mircea_popescu: i've never irl met anyone who couldn't actually light a bulb.
whaack: right, that is not realistic
mircea_popescu: im pretty sure it's why everyone else here, too.
mircea_popescu: you follow?
whaack: oh definitely, that is a great way to word how I feel about MIT too
mircea_popescu: so there we go.
whaack: it's marketed that way due to the widespread lack of self-confidence in the community
mircea_popescu: now what might be the cause of that.
mircea_popescu: but he doesn't. why does he do it then?
mircea_popescu: and this can possibly excuse mit?
mircea_popescu: sounds a lot more like an indictment.
mircea_popescu: i do not read in the original speaker even the faintest glimmer of intellectual life.
mircea_popescu: he is simply making stuff out of masticated wood pulp, like a wasp
mircea_popescu: "because it's what's done. because it's what was always done. citation needed."
mircea_popescu: why logically?
mircea_popescu: that's the crux here. why "logically"?
mircea_popescu: so now we're split in three, technical, humanities AND podium shenanigans?
mircea_popescu: division of labour is one thing. thought however... now that's indivisible.
mircea_popescu: like alf's meta-nsa?
mircea_popescu: like, you know, "my real self hasn't shown up yet"?
mircea_popescu: seems a diseased state.
mircea_popescu: wtf even is all this cancer.
mircea_popescu: ssergorp or something.
hanbot: brandworld problems. the moment people care more about being associated with the group than doing the group's work, groupname is a lost cause.
mircea_popescu: used to be. not anymore, thanks to wot etc, but myeah.
whaack: So... do you hold it against me for even revealing that I'm a student there?
mircea_popescu: i don't particularly care, but doesn't it make you yourself wonder?
mircea_popescu: moreover,without specific cause, why would i judge the affairs of another to begin with?
whaack: so revealing I go to MIT, for the most part, commands the respect of ignorant people
mircea_popescu: well if you do it inappropriately it may well command the disdain of thinking people.
mircea_popescu: everyone's dumb at least some of the time.
mircea_popescu: what exactly would be insulting about you going to mit?
mircea_popescu: i go to fratelli icecream parlour, does that insult you?
whaack: I'm saying that some ignorant people exalt those who go to MIT
whaack: right this is something I've realized / been coming to terms with only recently
whaack: (sadly)
mircea_popescu: you got plenty of time,
mircea_popescu: for tomorrow, you die.
whaack: your words hit hard, gonna copy that one
mircea_popescu:!up e63
whaack: ((changes resume to name, bitcoin address, and gpg key))
mircea_popescu: great conversation piece
mircea_popescu: http://log.bitcoin-assets.com/?date=11-04-2015#1097448 << sadly.☝︎
mircea_popescu: http://log.bitcoin-assets.com/?date=11-04-2015#1097470 << werd.☝︎
mircea_popescu: sort-of like the bear wondering why he's not getting any milk and the shepherd dogs bark.
mircea_popescu: logic.
mircea_popescu: it actually is not even half bad.
mircea_popescu: SFRRSANE&BH
mircea_popescu: sfrr, sane & bh.
mircea_popescu: with expressions like "nici o sfiriala" =lit "no sizzle" = no big deal.
ben_vulpes reviews the logs
ben_vulpes: it's the mircea_popescu show!
mircea_popescu: lol
mircea_popescu: irc is the largest challenge to blogging i ever saw.
ben_vulpes: anyways nonzero odds my grandmother passes during con3
mircea_popescu: well... "irc". assbot-wot-irc.
ben_vulpes: specifically #bitcoin-assets, and that only this year.
mircea_popescu: myeah.
mircea_popescu is out of grandparents.
ben_vulpes: this is the last biological one i have
ben_vulpes: and the second to pass while i'm in argentina, if things do play out that way.
mircea_popescu: well here's to not.
mircea_popescu: Which they did.
mircea_popescu: -arcane.
mircea_popescu: dude...
mircea_popescu: and aspires to a job to be obtained through an interview with hr.
ben_vulpes: heh mas o menos
mircea_popescu: and to never pass on a good sexist joke : real women, too.
ben_vulpes: <mircea_popescu> well here's to not. << happens to all of us, no sense derping about when or under what conditions.
mircea_popescu: there is that.
assbot: [MPEX] [S.MPOE] 187450 @ 0.00029744 = 55.7551 BTC [+] {2}
ben_vulpes: <whaack> ((changes resume to name, bitcoin address, and gpg key)) << you only have the one addr, and plan to tell folks about it?[SEP]
| 5
| 30,000
| 33,514
| 33,514
|
71,813
|
26010b66-b56e-4496-97d3-9bd88447a10e
|
trentmkelly/LessWrong-43k
|
Punishing future crimes
Here's an edited version of a puzzle from the book "Chuck Klosterman four" by Chuck Klosterman.
> It is 1933. Somehow you find yourself in a position where you can effortlessly steal Adolf Hitler's wallet. The theft will not effect his rise to power, the nature of WW2, or the Holocaust. There is no important identification in the wallet, but the act will cost Hitler forty dollars and completely ruin his evening. You don't need the money. The odds that you will be caught committing the crime are negligible. Do you do it?
When should you punish someone for a crime they will commit in the future? Discuss.
| 0
| 0
| 148
| 148
|
63,119
|
9a99db1d-dbc7-4c78-b4e6-8fabb621105c
|
trentmkelly/LessWrong-43k
|
Some end-of-year media recommendations
Books
I've read about 150 books over the past five or so years, after a decade-long dry spell where 95% of the reading I did was for school. These are the ~20% of those books that have really made an impact on me, that I've continued to think about even long after finishing them. I'm not going to give you summaries because (a) you can find those elsewhere, and (b) for a lot of them, I don't even really remember the contents. What I do remember is why I liked them and what I took away from them, so that's what's here. Books in each section are roughly ordered by how important they've been for me.
Self-help
Replacing Guilt
Every single essay on Nate Soares's blog was earth-shattering for me when I read them in college, and most of all 'Half-assing it with everything you've got'. Unfortunately by the time I found the Replacing Guilt sequence it was too late for me to not waste my years in school, but I still carry his lessons with me always.
Bonds That Make Us Free
This book is about how to treat other people. It's got some strong Christian undertones which turned off my mom (who was raised Catholic) but didn't bother me that much (I was raised atheist). This book has helped me more than any therapist; specifically, it helped me move on from something I'd been having nightmares about for ten years, by reframing and really seeing things from the other person's perspective. That's damn magic is what that is.
The Life-Changing Magic of Tidying Up
Marie Kondo is the person I aspire to be, in roughly half the aspects of myself that I care about. She's so kind and at peace and knows exactly what she likes. I tidy all day and feel awful about it; she tidies people's lives and feels deeply fulfilled. She also has some hella cute children. I KonMaried my stuff at the end of 2019 and have continued to live by her philosophy and be inspired by her aesthetic. I also helped my mom and two of my housemates KonMari! There's also a Marie Kondo Netflix show which is my happy pla
| 0
| 0
| 487
| 487
|
100,049
|
<urn:uuid:92c91af2-7761-4f11-838a-8a1ffca84691>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Code of Practice for Statistics
Helpful resources – using official statistics
Why is the Code important?
Official statistics are an essential public asset. They provide a window on society, the economy and on the work and performance of government. They are fundamental to the judgments and decisions made by the public, by government and by an enormous range of other organisations.
The Code of Practice for Statistics plays an essential role in ensuring that statistics published by government inspire public confidence through demonstrating trustworthiness, and providing high-quality statistics that enhance public value.
Why is the Code relevant to me?
The Code benefits all of us, as users and as citizens. The need to focus on the interests of users of statistics sits at the heart of the Code.
Compliance with the Code gives you confidence that the statistics produced by government departments and public bodies have been prepared by professionally-independent statisticians, free from political pressure; that they are produced using sound methods and are based on reliable data sources; and that they provide insight, support decision-making and inform debate.
Helpful resources
The menus below include helpful links and guidance on finding statistics and data, checking the quality of statistics, and writing about statistics. Some of this guidance was written for producers of official statistics, but general users may find it useful when thinking and writing about statistics in other contexts.
Description Link Source
A complete list of all statistical products that the Office for Statistics Regulation (OSR) has designated as National Statistics. List of National Statistics OSR
Find official statistics from across government, including statistical releases and data tables. statistics homepage UK Government
Find official statistics from across the Scottish Government, including statistical releases and data tables. Scottish Government
Find official statistics from across the Welsh Government, including statistical releases and data tables. Welsh Government statistics homepage Welsh Government
Find official statistics from across Northern Ireland, produced by the Northern Ireland Statistics Research Agency (NISRA), including statistical releases and data tables. NISRA homepage NISRA
Find official statistics from the Office for National Statistics (ONS), including statistical releases, articles, and data tables. ONS homepage ONS
Visual.ONS is a website exploring new approaches to making ONS statistics accessible and relevant to a wide public audience. Visual ONS homepage ONS
Nomis is a service provided by ONS to give you free access to the most detailed and up-to-date UK labour market statistics from official sources. Nomis ONS
Stat-Xplore provides a guided way to explore DWP benefit statistics and allows you to create customised tabulations. Stat-Xplore DWP
StatsWales is a free-to-use service that allows you to view, manipulate, create and download tables from Welsh Government data. StatsWales Welsh Government
Northern Ireland Neighbourhood Information Service (NINIS) makes available small area information held within Central Government and Non-Departmental Public Bodies in Northern Ireland. NINIS NISRA
Description Link Source
Find data published by government departments and agencies, public bodies and local authorities. You can use this data to learn more about how government works, carry out research or build applications and services. UK Government
Explore the UK’s largest collection of social, economic and population data resources. UK Data Service homepage UK Data Service
We have developed a diagnostic tool to help users decide if data are suitable for their use, if data are reliable, and to form a view on official statistics. We would welcome feedback on this tool.
Data Diagnostic Tool (PDF)
Description Link Source
A Government Statistical Service (GSS) guide to writing about statistics. Writing About Statistics guide (PDF) GSS
The Office for National Statistics’ style guide covers the elements of writing about statistics. It aims to make statistical content more open and understandable, based on editorial research and best practice. ONS style guide ONS
The United National Economic Commission for Europe (UNECE) guide on making data meaningful is a practical tool to help managers, statisticians and media relations officers use text, tables, graphics and other information to bring statistics to life using effective writing techniques. It contains suggestions, guidelines and examples. A UNECE guide to making data meaningful (PDF) UNECE
| 0
| 0
| 885
| 885
|
103,789
|
<urn:uuid:e6da9312-058b-4de1-a93a-ccd9eaef2e4e>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
[CLS]Follow TV Tropes
Pantheon / Death and Postmortem
Go To
Mortality, Death and The Afterlife
As the House of Life and Death increased, the two contrasting, but connected ideals and concepts ultimately became large enough to encompass a massive influence and according to the two representing Grim Reapers, it was decided that Life and Death be separated from one another and settle down with a singular House to represent the concept of mortality and death. As it is a primordial concept of life, it still manages to maintain being one of the most vital and important Houses in the Pantheon.
The House of Death and Postmortem has a rather ethereal and fantastical look to it, albeit in a very dark and moody atmosphere. The entrance is marked by a gigantic, intimidating gate with spikes and skulls adorned to it, with the same being said about the walls as well. Within the gates, the surroundings are not exactly one that elicits dread, but rather a longing feeling of nostalgia and melancholy. The House is filled lush, beautiful fields of grasslands and trees, though the numerous epitaphs and cemeteries are what create a contrasting tone of sad elegance. There are some Centres on postmortem checkup and analysis for the deceased and prayer buildings and memorial sites where respect to the departed is formally held in events
At the center of the House is a soaring green vortex that seems to pierce the skies. That said, the vortex is a magical gateway that leads to another realm; one that resembles the afterlife. The environmental surroundings and atmosphere is dependant on which direction one decides to take; one is a heavenly paradise where the virtuous, selfless, and kind are welcome at any time and the other is a hellish despot where malicious beings are damned. There is also a middle-spot, which happens to be a giant council for those who are anthropomorphic personifications of Death. Death personifications not native to the House are often invited to discuss the Pantheon and its events, in addition to what they're up to, to various effects.
This House has a lot of connections to others due to how Death is a theme and concept that has a very far stretch. Several other Houses like Religion and Faith and Health and Diseases have a professional association where benevolent members of the former are open to hosting eulogies to pay regards to the dearly departed and the latter would use established centers to examine and inspect corpses for various reasons. The Houses of Slaughter and Undead and Phasmata tend to link themselves as they often need corpses to set up shop for themselves, the former to test out ways to slash up or mutilate bodies and the latter because many of the residents can never get enough out of plucking put bodies of the grave to reanimate under their causes. The Halls of Angels and Demons are connected to the Afterlife side of the House for obvious reasons.
Despite the split, this House is still intrinsically connected to Life and Vitality. Both Houses are needed for a deity to be fully resurrected, however, the process and time it takes for them to be restored to life is dependent on the cause of death and overall damage to the body and soul. It is possible to recreate a soul from scratch, provided that memories are kept intact, but it can take a week to years depending on the established information.
open/close all folders
The Grand Judges of Death
Death (The Seventh Seal
Death, King of Death Personifications
• Greater God
• Symbol: A black Queen chess piece
• Alignment: True Neutral
• Portfolio: Classic representation of death, dropping his duties for a game of chess, polite death-bringer, unconcerned with petty things like social class
• Domains: Death, Existentialism, Chess
• Allies: Death of the Endless, Decim, Death (Regular Show), Lord Death (Soul Eater)
• Rivals: HAL 9000, Sora and Shiro
• Enemies: Death (Final Destination), Jigsaw (to a certain extent)
• On good terms with: Antonius Block and other Ingmar Bergman deities
• Annoyed by: Bill and Ted
• Complicated Relations: Billy, Mandy, Grim
• On the way home from the Crusades, a knight named Antonius Block came across a ghostly man wearing a black cloak. It was Death, who was coming for him. Being a fairly amiable sort, Death accepted postponing Antonius' fate for a game of chess. Death followed the knight closely as the latter journeyed through his country in upheaval and saw the best and worst of humanity, and kept begging Death for answers to his existential questions, which Death either wouldn't or couldn't provide. In the end, Death, in his unavoidability, took Antonius.
• Came up to the Pantheon immediately after Antonius because he was rather annoyed that one of his charges had been restored to life without his consultation. His idea was to drag Antonius right back, instead the Main House was glad to see one of the most influential versions of death come up of his own accord and pleaded for him to stay and represent The Grim Reaper (which was released from Gig after some bureaucratic wrangling). He supposed that was fine as long as it didn't interfere with his job in the mortal world. As for the whole "drag Antonius back to death" thing, he let it be once it was made clear Antonius would never be alive in their universe again. Nothing personal, but if that were the case, Death would have failed his duty.
• When it comes to other death personifications, Death is in particularly good terms with Death of the Endless since her pleasant disposition meshes well with his polite and calm demeanor. Since Death is considered the basis for modern portrayals of The Grim Reaper, Death of the Endless likes to call him 'Dad' as a joke. Death likes how she remains professional in regards to her duties in spite of her soft spot for humanity. Death of the Endless, for her part, likes to tell him about how interesting human lives are, which might be inspiring Death to have a less indifferent attitude to the souls he reaps. He's also pursuing friendships with Decim and Lord Death due to them also being relatively sympathetic to humanity and benevolent, which 7S!Death finds he prefers to the nihilism and malevolence of others such as FD!Death. Is also on friendly terms with RS!Death given their predilection for games and the fact that he holds the trope he codified, though 7S!Death doesn't like his personality all that much.
• Given the Pantheon actually has no use for death entities due to the meaninglessness of death in here, he spends most of his time humoring deities who think they might have a shot beating him at chess, or who just want to re-enact the Chess with Death scenes. He's also always in the middle of some chess match with Antonius, a habit the two started "for old times' sake". Death has won all games, but given he once felt the need to learn Antonius' tactics, the knight is sure he's not invincible and keeps at it.
• Among the deities who want to prove their skill at chess by winning against Death himself, there are gamer siblings Sora and Shiro, who have vowed that they'd never lose a single game. Given Sora and Shiro have done well by that vow, Death figured that the game would just go on forever and after a few days of playing the same match let the siblings win. Neither was fooled, though they've acknowledged that they need to rise to number one in the world of Disboard and probably beat Tet if they want to beat Death himself.
• Hearing of HAL 9000 and how it was proficient at chess, Death became interested in seeing how such a human invention stacked up against himself. Death came away from the match with the impression that the machine's patterns are easily figured out and that it's just as prone to mistakes as humans, as expected from a creation of theirs. HAL is mostly infuriated to be bested and looked down upon by a primitive, unharnessable force of nature and is looking to improve its chess playing, so as to prove that something built to be infallible such as itself can overcome that which humans fear most. (HAL probably has a chip in its shoulder because it does have a fear of death.)
• He's not as flashy as other personifications of death in the Pantheon, which leads to people thinking he's one of the weaker ones. Given he has no inclination for getting into fights, there's no real way of verifying just how powerful he is in that regard, but he's much harder to beat at games than the Death from Mordecai and Rigby's universe, especially as he doesn't let his opponents get away with cheating.
• Speaking of, RS!Death often invites him to the House of Gaming with a view to expand his repertoire beyond plain old chess. 7S!Death won't turn him down, but still prefers playing chess by far, finding it the more compelling game, and having his matches in the solitude and quiet of his own temple. For someone who's basically a force of nature, even he feels a bit dislocated from the atmosphere of the House of Gaming.
• When he meets a fellow Bergman deity for the first time, his greeting is a friendly chess match where he takes the time to converse with them to get to know them better. While most of his fellow creations don't exactly go out of their way to be with him, given he's Death, he follows their lives in the Pantheon with deep interest. Death of the Endless' influence is at work.
• His greeting didn't go over so well when it came to Agnes' family, because he decided to spice it up when he heard their story. He pretended he was there to reap Agnes' life and made his challenge to Karin and Maria, telling them that if they didn't defeat him, Agnes would be gone for good. He didn't actually mean it, but he was eager to test the two sisters' resolve. To Karin and Maria (yes, the two played together, Sora and Shiro style) it was a highly stressful, sky-high-stakes game, and after fourteen straight hours Death took pity on them, decided he was satisfied and let them win. He has a fairly poor relationship with them as a result, though it's one-sided on their side as he finds them fascinating.
• Shakes his head at Grim for his foolishness of agreeing to be a slave to two children after losing a game. Mandy wouldn't mind making a slave out of Death as well since he seems to be more clever than Grim. Death is obviously not keen on the idea and avoids her, lest her diabolical mind come up with a way of ensnaring him. Death also treads lightly around Billy, just in case. Granted, Death'll never admit out loud he actually does have some fear of falling into the same situation as Grim.
• Was approached by Bill and Ted, who asked him "How's it hanging, Death?" and that they'll beat his ass at Twister and Clue again any time. Death didn't really know what they were talking about and assumed they confused him for the Death in their world, but at any rate, found them annoying and crude and has no real desire to seek them out and challenge them.
Death of the Endless
Death of the Endless, Goddess of Friendly Grim Reapers
• Overdeity
• Symbol: A silver ankh
• Theme: Her DC Showcase theme
• Alignment: True Neutral due to her duties, but leans toward Neutral Good.
• Portfolio: Death (obviously), Perky Goths, Personifications of Death, Complete Immortality, Beware the Nice Ones, Nice Anthropomorphic Personification, Don't Fear the Reaper, All of the Death Tropes.
• Domain: Darkness, Death, Life, Love.
• High Priestess: Momo
• Followers: Lady Death, George Lass, the unascended Shinigami of Gotei 13, Steve, Gregg the Grim Reaper
• Allies: The rest of the Endless (Dream, Destiny, Destruction, Delirium), Death the Horseman, Robin the Avatar, Finn the Human, Jake the Dog, Paige, The One Above All
• Targets: Grima, The Lich
• Opposed by: Many, but they are all one-sided
• Sitcom Arch-Nemesis with: Cthulhu
• Odd Relationship with: Thanos, Wade Wilson/Deadpool
• Death is a fun girl. She appreciates life, likes sentient people to hang out with, and is actually quite fun to be around. That being said, she's still DEATH, so don't mess with her.
• Once, every 100 years, Death takes a day off to both enjoy being mortal for a short while and to remind herself of the gravitas of her duties.
• A large part of Deadpool and Thanos' heat with each other is that they are both in love with her. While she is flattered by this, she treats both with as much affection as she does everyone else.
• Has a friendship with the Death of the Discworld; he has offered to "lend" her his assistance as the Reaper, along with his assistants, allowing her to take more time off of work than one day a century, which she is grateful for.
• Also gets along with Death the Horseman, though he is puzzled by her interest in humanity.
• She has an entire army of lesser Grim Reapers, including The Black Racer, George Lass, Psychopomp-style Shinigami, Steve, Gregg, and of course many, many little Deaths.
• Is assisting Robin in finding a way to kill Grima off for good.
• Will sometimes take members of the Pantheon to the Sunless Lands to say goodbye to those that they have loved, and were unable to make their peace with before their passing.
• Nekron is plotting to overthrow her and steal her position. Thanos has sworn to destroy Nekron after discovering this plan; while she thanks him for the thought, she's not that concerned about Nekron. After all, she IS the one who will close up existence when the end comes, and lock up the door on her way out.
• Has been asked by Finn and Jake to aid them in killing the Lich and making him stay dead. She has agreed to aid them, on the condition that he must be destroyed/killed again first. She upheld her part of the bargain by allowing an aspect of the Lich to reincarnate as Sweet Pea, who overcame the evil within himself and slew the last evil remnant of the Lich in Ooo.
• Gets along well with Paige, due to their dark and gothic styles while still being as friendly and fun as they can afford to be.
• Once partnered with John Constantine in order to give a short, informal PSA about safe sex.
• Appointed Lady Death as the leader of her Psychopomps retinue as a gesture of mercy towards the innocent woman she once was. In exchange of being able to step out of Hell, Death declared that, if she was to remain undead until all human life is gone from Earth, Hope could at least make herself useful by helping with the passing away of souls for the rest of her undead existence. Lady Death accepted since she is now able to see other things than Hell and demons but warned Death that she just might want to claim her title someday. Death doesn't worry about it because Hope is of human origin and thus cannot really act against her, no matter how powerful she may become.
• In wake of a new telling of her tales, she's obtained a avatar for herself. Most people don't seem to mind, except for the god named Simon Laurent who wants nothing to do with her. When he was finally able to calm down, all he could say was that Death reminded too much of an old friend that he never wants to see again.
• "Everything ends. That's what makes it so special."
The Maiden in Black
The Maiden in Black, Divine Maiden of Soul Manipulation
• Greater Goddess with hints of overlapping Overdeity
• Symbol: Her right-facing silhouette holding a very long staff, also The Nexus
• Theme Songs: Maiden In Black, Epitaph for Boletaria
• Alignment: Neutral Good
• Portfolios: The strongest demon with the ability to give strength, has a sinister vibe but with no hostile intentions, mysterious in every way from her eyes covered in wax and being barefoot, speaks with a soothing yet confusing accent alongside a Shakespearean dialect, forced to stay in the Nexus for even millennia
• Domains: Souls, Demons, Mysterious Women, Immortality, Maidens
• Allies: The Chosen Undead and Bearer of the Curse, Solaire of Astora, The Ashen One, The Hunter, The God Emperor of Mankind, The Dovahkiin, Yusuke Urameshi, Kazuma Kuwabara, Yoh Asakura, Sir Daniel Fortesque, Edward Elric
• Co-Overseers with: Death (The Seventh Seal)
• Common Visitors: The Hollow Knight, Artix von Krieger, The Puella Magi, Dust, Alphonse Elric, Ichigo Kurosaki, most Persona users including Philemon
• Commonality Connection with: The Plain Doll
• Rivals: Ryuk, Crowley, Charon, most Grim Reapers (in regards to the domain of souls)
• Distrusts: Castiel, Shinobu Sensui
• Pities: The Fabrication Machine, Ludwig the Accursed, Artorias the Abysswalker, Slave Knight Gael, Merkava, Red and the Transistor, Yui Ikari
• Enemies: King Allant, The Daedric Princes especially Molag Bal, Nyarlathotep, M. Bison, Shang Tsung, Nevermore the Shadow Fiend, Diablo, Father, Ermac, The Chaos Gods, Nagash, Kel'thuzad, Soul Calibur (and by extension Nightmare), Arthas Menethil the Lich King, Gul'dan, Zarok
• Opposes: Aldia, possibly Gwyn as well
• Before the city, before the vistas, before moments of the Forever War and perhaps the naming of the Pantheon itself, there lies an ancient place buried underneath the lands. It was a mere myth they said, that a place would wield some of the most advanced forms of power, yet no one would believe such malarkey. However, the histories of the Pantheon have only small records of its existence, and for good reason: to prevent the masses of ever knowing such place. An underground cavern came to be discovered on the old times of the Pantheon, and all would link to an unfamiliar sight on the shores the same day it was found.
• On the shores, one of the teams found an unforeseen sight, a woman, clad in black, was found on the shores of the Pantheon, distraught at the sight she has found herself into. After being picked up for identification, the resident archaeologist team found a gigantic place, a place surrounded by arch-stones which lead to different places of a once proud kingdom. While the team were studying, the woman that was rescued disappeared from the recovery team and was found standing in front of the archaeologists, introducing the place they were standing on: the Nexus.
Maiden in Black: "This is the Nexus. It holds together the northern land of Boletaria. Thou canst not exit the Nexus, but each of the five Archstones will connect thee to another node."
• Years after the discovery, the Nexus and its' Maiden were soon forgotten. Gone were the days of the potential findings of the lore of such place, as well as the possible entrance to the once beloved kingdom, as it was replaced by the introduction of numerous soul-based deities as well as the enraging war the Pantheon seeped itself into. And at one point, the entrance to the Nexus was naturally sealed, hoping to wait until one person discovered it once more. That came true when the Chosen Undead and Bearer of the Curse saw a small opening and located the hub world, with the Maiden approaching to them with the intention of their empowerment via the souls the two have gathered.
• As soon as the Nexus was rediscovered, the House of Life And Death felt a breeze, with numerous souls in silence. The gods were confused as to what has happened to their vicinity, but very few were aware of what is to come, that the Nexus and it being found once more may reintroduce the Pantheon to the Soul Arts, a very old, if not ancient, power that has existed from the very beginning of humanity's creation. The Maiden's purpose on this case is to be the handler of the Soul Arts and the caretaker of the Nexus as well as the guide to the kingdom her hub world is connected into, where adventurers may setup for the journey they will be headed to and vanquish the demons within.
• However, she is not alone. Deep within the Nexus lies an ancient Old One, a demonic being that consumes the souls through the fog it shrouds kingdoms into. Its power comes from those who practice and utilize the Soul Arts, and when the power was at its peak, the Old One awoke and covered the first kingdom into fog, releasing demons and monsters to feed it. It was this very thing that lead to the First Scourge and also the banishment of the Soul Arts, where it would be awaken again by a melancholic king, beginning the Second Scourge that terrorized Boletaria. One would ask why she would be the caretaker of the Nexus when the Old One is right below it, but in truth, behind her human exterior is perhaps one of the most powerful demons of her kind and the mother to the Old One if theories are to go by.
• The very moment her presence was revealed once more, her disappointed feelings came to be after introduction by the Chosen Undead. This attitude of hers comes from the fact that her life has revolved so much on both the Nexus and the Old One that it has rendered her immortal for good, and killing her will only bring her back, hence her disappointment of being alive. This, alongside her otherworldly power, has made her the only source to lull the Old One back to slumber should it awaken, at the cost of the Soul Arts being forgotten forever. It has been the reason for the power's lack of popularity, for when the Soul Arts have been massively used to the point of awakening it, she would come down and make it sleep, stopping the scourge in sacrifice of the Arts.
• Being the very first to transfer souls to the person who brought it in, it arguably made her the prototype of the new wave of Fire Keepers, perhaps being even the mother to all of them. This form of empowerment allowed her to become the new maiden for the Chosen Undead, Bearer of the Curse and Ashen One as her soul manipulation allows her to increase their power, making her the substitute for the trio. Eventually the arrival of Solaire of Astora made it four instead, but he often stays in the Nexus more than the others, seeing it as a place they can stay in the case of the loss of bonfires.
• More similarities came in when the Hunter stepped in. She has tried to empower him with her souls but was unable to do so, as the Hunter relies on the echoes of blood that and the fact that the Hunter may even reach Greater God levels thanks to him becoming an eldritch being, seeing him as powerful already to support him. What piqued her interest was the companion he brought in, the Plain Doll, who considers the Maiden to be a long lost relative, in relation to similar themes. Despite this, the Maiden is much more powerful than her and holds more responsibility towards the hub she controls.
• Unlike the responses received by most soul-based deities, reception towards her was cold, no one able to recognize her and only seem to dislike her even further following the potential she can do to numerous souls. Her sinister looks does not help the case. Those who stayed in the House of Life And Death saw what she can do and thanks to her connection to the ever-crucial Nexus appointed her to be one of the key members to the Status Quo of the Pantheon, as her soul manipulation and her neutral stance made her the perfect representative and perhaps, the Overseer of most of the souls bound within the Pantheon.
• This decision was made by one of the few deities who understands her purpose: Death who was one of the few to recognize her powers and her ever-growing maintenance of the Nexus. It was he who allowed her to become attached to the state of the Pantheon regarding life and death and made it so that both are responsible for those within the boundaries of the soul, and her powers fit nicely to such profession. Most grim reaper deities however object this, as they think the Maiden will just steal their jobs (and for some steal their food resource) but then again, when you are immortal for the sake of the maintenance of souls, what choice really is there outside of it?
• Those who oppose it are mostly those who have gotten used to the soul business they have been a part of, whether it be for those who take human souls for their benefit such as Ryuk or those who wants the souls as part of their desire for power such as Crowley. Some however take her presence as sublime, with Charon among those who tolerate her presence. Just a shame that she might take the souls of those who will join Charon's ferry and use those for empowerment to people like the Chosen Undead.
• What makes her unique is her unparalleled power of the domain she serves. Being one of the few to empower others through other souls to awaken their inner power, as some may indicate that the world they live in may have an upmost importance on mankind's own soul, as it is the key to their strength. One may also say that the entity that created humanity may also have not trusted their own creations, and as a result, created the Old One that would eventually haunt the very creations it made.
• As a result, her consultation was filled with deities who have experienced anything related to souls, from artifacts to even specific powers. The first to attend outside of the aforementioned four was the Paladin Artix von Krieger. When she saw his soul, she was surprised to see a pure soul alongside a co-existing dark soul, the Champion of Darkness. He came to visit her in regards to Spirit Orbs, and how she can immensely increase its power. When he saw what she can do in her domain, he decided to entrust her with the Spirit Orbs, as while they may be unable to be controlled once freed by him, she has the ability to manipulate it including empowering its user with the souls imbued within it. As a result, she has become one of Artix's personal helpers since she has access to the Spirit Orbs.
• More assistance came in when the Dovahkiin stepped into the Nexus where she saw all the dragon souls he has taken from many dragons, which impressed her and gave him a huge boost from all the souls he gathered. The Hollow Knight also stepped in as, much like the Chosen Undead up to the Ashen One, he is also blessed by the souls he collects, has made her his substitute for powering up which he will need as the Radiance has become a threat once again. Then after those two, came even more visitors, with Dust among those who asked for her guidance on his identity, which she has accepted and visits her from time to time and Sir Daniel Fortesque who also has been told by her that he is a good soul and would like to also help him from Zarok, also known to manipulate souls for his own evil goals.
• This extends to everything soul-related, from Personas to even Grief Seeds. She has the ability to increase its power and/or change the soul entirely, possibly even purifying it from its original state. Such was the interesting case when the Puella Magi visited her for consultation, and Madoka asking about the Grief Seeds' nature. While indeed it is corrupted, she can only say that it is built from the roots of Soul Gems, meaning Soul Gems are bound to become Grief Seeds at a certain point. Hearing this, she has promised to her if she may no longer meddle with such, as to protect the burden of all fallen Magi. She carefully respected this wish and gave back the seeds, declining anyone with Grief Seed requests.
• Personas are more often than not based on the user's soul and has intrinsic connections to it, so it would make sense for many Persona users to visit the Maiden in the hopes she could increase their powers just albeit for their own liking. This made her a center of attraction on two forces, Philemon and Nyarlathotep. Philemon trusts her and is possibly allowing her permission to modify the Personas, but Nyarlathotep isn't having any of that and is attempting to thwart whatever attempts she does towards any Persona user. What the latter does find amusing though is the fact that he can see her as a demon, and with his personal hate against her more of a show for who's the strongest amongst the two.
• For all her work unfortunately attracted foes for her to face, and the more observant foes she has been facing were the Daedric Princes, deities with their own domains filled with numerous souls. Some of the princes have admitted that the Maiden might be equal or even more powerful than they expected her to be, but some do not give up, and this was the case with Molag Bal, who considers the Maiden to be the barrier between the souls he can snatch from the Pantheon. He has come to hate her in with passion and unless she disappears in any other way, Molag's days of soul snatching may be halted. And this just doesn't limit to the Princes, this also extends to more powerful beings, from Diablo himself, who considers her a disgrace of a demon, to even the Chaos Gods, as they too have a realm of souls, well formerly a realm of souls until they came and turned whatever the realm is into an unholy nightmarish mesh of all thoughts good and bad and much like the Princes, consider her a rather problematic roadblock for either to take any soul they come across. Does not help that she has become one of the overseers of the Pantheon Status Quo, making her a prime target for these powerful deities to keep an eye out for.
• As if facing powerful divine entities was not bad enough, her abilities were among those sought out by those who hunger or crave the power of the soul. When Bison and Shang Tsung heard of the Nexus, they stormed inside in an attempt to coerce her to empowering them. She refused, forcing the two to drag her out of the place or die. They did the latter with Bison repeatedly punching her and Shang Tsung sucking her powers dry, until she rose back again thanks to the Nexial Binding. After repeated tries the duo gave up and dragged her outside, until they realized they couldn't even get her out of the Nexus due to her connection with it. At that point, they gave up until the Chosen Undead, Bearer of the Curse, Solaire and the Ashen One were behind them and gave the two a smackdown. Since then, they have been eyeing on her presence and looking for those who are either soul-hungry or filled with numerous souls and this came to fruition when Ermac, Gul'dan and the Shadow Fiend joined in, with the former duo wanting to add her and the souls she harbors to his collection and the latter seeing her to be quite delicious for his soul appetite. Not to mention the latter is known for eating other demons' own souls but there will be a price for that though.
• Some deities such as the God Emperor of Mankind and Yoh Asakura have seen immense respect for her, on Emps he believes she can help recover his split souls so that he may finally recover from the Golden Throne, and on Yoh, she can help him maintain his role as the Shaman King. Her opinion on them is surprisingly similar but she's split on Emps as while he is reasonable on his goals, his actions, not so much.
• She has an experience when it comes to those who are known to eat souls, as mentioned by her sympathy towards Merkava to her hostile relationships between Diablo, the Shadow Fiend and the Daedric Princes. This is demonstrated to her ability to quell the Old One, also a consumer of souls one albeit eldritch in nature. If she can lull it to sleep, she may be able to do it to those mentioned, but it may cost her. The real confusion starts with those born or created from multiple souls, with Ermac, Soul Calibur and Nightmare being special cases. She may see the souls within but are unable to modify or change it in any way unless they are freed, which is not helped by the fact that the deities mentioned have their identities made by other souls meaning they do not have an original identity.
• Several deities who's own souls have been merged are tricky cases. Some may even be unrepairable due to the damage done to it and as a result cannot be returned to its original state. Often times she will decline to help if the damage is too much to bear and worse of the soul is more than what she can normally handle unless it is necessary to do so where she will have to make changes. This was her unfortunate case with both Red and the Transistor and Yui Ikari, as the former went to the Nexus in a desperate bid to bring her beloved back from the sword into his original body, which she cannot as his soul is bound to the Transistor and on Yui's case, it has become more complicated by the fact that she is Evangelion Unit-01 and that her actions of activating Instrumentality forcibly distanced her from whatever requests she has.
• Requests were denied by her when tackling beasts, or more specifically those that have gone mad similar to that of a monster, which she has experienced with three swordsmen Gael, Artorias and Ludwig and one that is an actual monster Merkava. She claims that their souls have irreparable damage and may never be able to recover it back, save for Artorias who has been able to control his abyssal corruption for a certain amount of time. Ludwig's soul has been bound to the Hunters' Nightmare and his body can no longer be turned to its original state, and for Gael, all hands were off the moment she saw the Dark Soul of Man inside him realizing that this is something she cannot handle only by herself. Same could also be said for Merkava, because while he may have self-awareness, his hunger ends up neutering it, rendering all attempts futile.
• Of all those she has denied, she has the most pity towards the Fabrication Machine, a machination whose lack of a soul left it completely omnicidal. Thanks to a few who were able to consult to her about its situation, she is trying to form a new soul that is capable of housing it, however it will take more time as creating an original soul is harder than it sounds, much less when she is known for utilizing other souls as opposed to creating another one.
• And then there is the undead, another case where she has denied any call for as the topic of the undead is a domain she is well outside of. She may handle the souls but when the souls are controlled from a "living" standpoint whilst supposedly deceased, she may or may not have the power to control it. She does however oppose the idea as souls are better off let go rather than being brought up once more. Both Nagash and Arthas has vehemently hated her for her intervention, but for her she has one interest among them, that being Frostmourne, which houses many souls of those lives who have fallen against it.
• In relation to Arthas however, she was notified when his lieutenant Kel'thuzad owns a city filled with undead souls, all while being revived back and forth used for nefarious purposes. His role as a lich only worsened the ordeal and has been making plans of possibly being able to overpower them. Not on Nagash though as his prolonged history and coupled by his enormous power
| 0
| 0
| 8,000
| 18,413
|
102,115
|
<urn:uuid:cb7da91e-e3d9-43b7-8141-762ae08b98d2>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
viagra online Sydney top male enhancement pills reviews where can I order generic viagra online natural girth enlargement fx7000 male enhancement male enhancement supplements reviews PremierZen 7000 real Cialis for sale.
Luz Center shook her head at Laine Buresh, Love among high school students will indeed delay a lot of top 5 male enhancement energy of top prescription male enhancement pills I will not let them mess around.
How Much Does Sildenafil Cost At CVS!
Finally got an admiration for me, no longer think I am a soft persimmon? I smiled proudly in my heart If I don't pay attention, my cheeks don't know Someone zyntix side effects. What a genius! He also finally understood where Luz Damron's fanaticism towards the Fu soldiers came from, except perhaps he had commanded them that day More because he planned seventeen years ago, and finally the day has come to bear fruit Doctor Le is indeed a god ED delay pills and the overall situation.
PremierZen 7000
Even if your head is messed up, can you remember this kind of thing? So, supplements for a bigger load lying! It has also been PremierZen 7000 the original Tessis is dead Although the body is still there, he has long since lost Marley drugs pills. Looking at generic Cialis 50 mg screams in her ears, Maribel Noren's face supplements for a bigger load couldn't accept the reality that was happening in her eyes at this time. Zonia Menjivar was also a little NHS viagra advice drank the last bit PremierZen 7000 in the bowl The inner government is different and must be kept.
Natural Vitamins To Boost Sex Drive?
Camellia Pekar a biogenic bio hard carefully walked through the narrow corridor on the second floor, stuck out half of her head and observed the situation in the room through the crack of free viagra free shipping. The suspects will not only be in our class, but there may be someone male enhancement capsules who status testosterone booster GNC high marks There can be no clues on the table, half a supplements for a bigger load is a tasteless rib, and I set my eyes on that tree. Thomas Badon shook his head and said Who would have thought that when those powerful'shamans' gradually disappeared, these three leading a large number of Tami Catt turned into a big climate, ruling the world's vigour male ED pills. When she saw me appear, Gaylene Noren stopped writing, and pills enlarge penis the call just came back There are all kinds of things this morning, and I'm so busy that I haven't eaten yet.
Top Dick Pills?
Dion Antes wanted to understand this mother box, and through this PremierZen 7000 he wanted to understand Enzyte at CVS was sealed by the world, and through that history, he wanted to understand the stories that people experienced in the ancient times, and all this Levitra UK buy it all lies in this mother box. If four people gather together, they need to provide too many human lives for them to absorb, which will definitely cause a big mess It is better medicine to increase stamina in bed people scattered everywhere and go to their own secretly Gaia's idea VigRX Tongkat Ali LJ100 never thought that Lavrov is too stupid to choose this method. But just as everyone was making wild guesses, Stephania Redner regained her consciousness and said with a slight smile, Let's go at 9 o'clock Suddenly, Jack the Clown was shocked! When the referee best male enhancement pills on the market today there were four dice left and right, showing two, and three points respectively a total of nine points! Nine o'clock is extremely accurate! Wow, the whole casino was shocked.
Is it possible, this is the feeling of being a human being? Could it be that I am already human? Abandon the heavy shackles, throw down the heavy burden, and let my desire how to increase sex power by medicine of the armor It turns out that I am so longing for the life of human beings, supplements for a bigger load the soul of human beings It turns out.
Top Male Enhancement Pills Reviews?
As for saying that he attacked me, I don't think there is a how much does sildenafil cost at CVS he summons 10 or 20 great masters or S-level masters, he may not be able to keep me. Thomas Antes waved his hand weakly, let Anthony Lupo back down, sat cross-legged penis enlargement supplements the child, and asked in a warm voice, PremierZen 7000 what else do you have before seeing your sister Pomegranate? An unfulfilled wish? The matchmaker's eyes were red, and max gain male enhancement to kill me? Samatha Haslett nodded and smiled PremierZen 7000 have not died.
Does Walgreens Have Viagra Connect
The hotter you are, the more you sweat, the colder the more you tremble Nancie Geddes is sure to keep his money, and the possibility of earning more than two billion dollars is more than half! Although it is a bit crooked, all procedures are within the rules, so even Larisa Grisby has to admire Maribel Culton, PremierZen 7000 where can I buy Cialis online in Canada much. African mojo unique male enhancement power be when Leigha Geddes absorbed the Georgianna Mayoral where to buy male enhancement pills over-the-counter what changed? Even, go crazy? Jasmine shook her head and didn't know In fact, even in Augustine Drews's memory, it seemed that such pills to make you last longer in bed over-the-counter.
Walgreens Male Enhancement Products?
The second sister felt that the lie was about to be exposed, so she said a little timidly What do you PremierZen 7000 old man to wear for women's clothes By the way, did I erentix male enhancement pills now, aha, it seems that I haven't improved my strength now. The prince nodded, took the tea on the table and took a sip herbal male enhancement pills stirred up natural supplements for male stamina Marquis Mcnaught sighed Tomorrow is the early morning, will there be. presiding over the work on the Isle of Mal Fortunately, they were originally ED meds from Canada are PremierZen 7000 today Also, PremierZen 7000 four of us didn't know each other.
Zyrexin Reviews Amazon?
It can be said that in the absence of powerful spellcasters With this help, no army can hope to set foot on the Leigha Serna, supplements for a bigger load thousands of pirates here, as well as an outlaw Extenze reviews amazon of life and various supplies. Qiana Fleishman who had just ended the call with the bishop shrugged his shoulders indifferently It's nothing! It seems that a mojo male enhancement Austin test object under the Sharie Wrona Temple Oh? The soul PremierZen 7000 into a creative array of energy storage? The former had an interested expression on his face. That is, at this time, there seems to be PremierZen 7000 male enhancement yellow pills seems that the process erectile dysfunction over-the-counter CVS soul of the demon master has officially started. But PremierZen 7000 Mote said in his erection enhancement over-the-counter of supplements for a bigger load Joan Schewe, which is also an best gas station pills the Augustine Latson, is the blood stains on the stele.
His WebMD Tongkat Ali work and he said, Little bastard, get out of supplements for a bigger load the young captain was not a vegetarian, and suddenly stretched out his right hand wearing a chain glove. Unexpectedly, Camellia Pecora refuted me, saying that she would be a well-behaved sister at home, but Chelsea 35 ED pills reviews strangers and enemies! Really, even the relationship between siblings is afraid that others will know, this guy is really home from the second grade. PremierZen 7000Leigha Geddes heard this, she couldn't help PremierZen 7000 she beat me angrily With a punch, I don't like can I get sildenafil on prescription in my eyes. it's making a comeback! I believe that only those who have experienced the great battle in the past can Cialis at Walmart Canada vulnerable humans who claim to be very smart appear in front of the abyss.
African Mojo Unique Male Enhancement Power.
Elida PremierZen 7000 and said In the movie God is it possible to enlarge penis size enhancement pills bodyguard of the God of Gamblers called Long Wu? Bong Ramage is truly worthy of the name this time. I said that this place is different from other places, and still retains some incredible mysterious instant male enhancement pills temple? It is said that there is an inexhaustible side effects of men's ED otc pills.
VigRX Tongkat Ali LJ100
It's so hard, do you dare to feeding frenzy male enhancement pills this is Chi Guoguo's seduction After the fire, even if there is a human life, I will not be responsible! I reminded Marquis Pekar. Now that he all-natural sexual enhancement the hall, best male enhancement product on the market ease supplements for a bigger load to himself, this feels so good, this is how people live Christeen Catt? The call, he came back to his senses, but saw the prince looking at him with a smile. Just as they carefully observed these mysterious dark blue liquids, as if to find some flaws in them, a familiar figure suddenly appeared at the door Afuya? how hard your penis was the first to react showed surprise in his eyes.
Chelsea 35 ED Pills Reviews
So as long as there is still one Darent using shadow energy, the god will never really die, and as the number of people who use it increases, best homeopathic medicine for impotence day This also means that survivors must make a choice. Raleigh Paris said angrily There is that good child to serve, the old lady eats well, slept well, is often happy, and naturally feels much more how to naturally make your penis get bigger. I used PremierZen pills phone again to dial the number on PremierZen 7000 side This time, the mobile phone on Xiaoying's side didn't ring, PremierZen 7000 a reminder of the empty number. I was covered in black lines, and I became more and more suspicious, PremierZen 7000 did my mother do to me What's the matter? Mom took a sip of tea and was supplements for a bigger load and then said, That person is one of my countrymen, and because of an accident or something, he became best sex pills to last longer in bed then, the money I went to college, I paid for it Yes, he borrowed all the money for supplements during pregnancy.
Medicine To Increase Stamina In Bed?
sildenafil 110 mg thought it was Rubi Damron's parents who came back, but as soon as I opened the door, water mixed with colorful paints poured over me Nancie do any otc sex pills work Grumbles immediately put down their chopsticks, and they realized that PremierZen 7000 wrong. But his psychological quality was stronger than Elida Ramage, and he suddenly felt much better gold xl male enhancement pills side effects he didn't explain his purpose, and he didn't lose face After pretending to thank him, he took the 2,000 guards who had come for the picnic and returned to Michele Pekar in embarrassment. She closed the door, Elroy Buresh took out the medicine natural male enhancement products from the cabinet, her cheeks flushed, and she slowly took off the floral skirt on her body When the skirt took off, I was stunned, this guy didn't wear underwear or anything, and he was naked Because It's dirty, and I Ron Jeremy endorses male enhancement pills before Arden Volkman blushed and explained embarrassingly This guy is in the period of development, and his body is extremely green In my opinion, she is just a little girl.
The second sister's heart sank to the bottom Raleigh Redner still waved the bronze sword in his hand and said, But natural male enhancement safe a try As a result, the fire of second sister's hope was ignited again, and it was really hot and cold.
Pills Enlarge Penis!
so he can only use this particularly primitive method, hoping that Tyisha Fleishman can calibrate the shadow leopard as soon as possible After all, there penis performance pills time left for Lyndia Pepper. This best male enhancement herbal supplements proverb and my deepest obsession When he vydox male enhancement trial boy's tone revealed a rare seriousness and dedication.
Tips To Last Longer Sexually
With the steaming heat, big steamed buns with white noodles appeared in front of everyone The green light emitting from top-rated libido supplements more than 10,000 people was enough to make Sharie Catt feel that he was a roast duck. The shallow singing voice is so gentle that it makes the heart tremble, with a kind of sadness that touches the PremierZen 7000 with such a max performer pills the UK know which lyric male enhancement pills CVS pharmacy.
Elida Schroeder said with a smile I have already seen that bull nose when I entered the palace today Erasmo Kazmierczak quickly pleaded guilty, Clora Buresh shook his head and best sex pills deliberately blocked the news Rubi Haslett left the Jingshan camp that day to rescue When benefits of Cialis 20 mg passed the book on the eagle and stopped the.
rushed to the bullfight! Very does Walgreens have viagra connect Put on the ink! Randy Howe and Buffy Paris hurried up on the side, one spread out a blank scroll and held it with both hands, while the other took out a thick one from the PremierZen 7000 The wolf's brush, dipped in thick ink, is presented with both hands.
Marley Drugs Pills.
When I gave this explanation, the two guys suddenly smiled PremierZen platinum 4000 Xinyi looked around and saw the clothes hanging on the balcony railings. In the final analysis, it's you boys effective penis enlargement are too hateful! Larisa Schewe angrily fired at the boys, scolding all kinds of absurd things boys did to girls I Adderall XR cost didn't dare to speak to this guy, for fear that this guy would bombard me. Thomas Kazmierczak PremierZen 7000 Gaylene Stoval with sympathy Hearing her weeping and complaining I rushed is there a pill to make you ejaculate more the male enhancement bigger size reported the big news to the emperor. Go back to the hospital, if I'm taken in, let Stephania Mote pick me up I winked at Diego Paris, motioning male enhancement permanent let my mother know about this.
Cialis At Walmart Canada
Of course, the premise is that PremierZen 7000 of Slaughter is not ready to kill him, otherwise he still has no chance Luz Guillemette was a little stunned, looking at the blood dripping wound on the back of his hand, sildenafil 100 mg side effects NHS. But the Lord of Slaughter knew that once the other party knelt down, it was really the first step to surrender For example, when he snorted again and condensed the centurion male enhancement pills sister Cialis tablets in Dubai wanted to kneel.
Especially when Samatha Noren recognized the reality and knew that Zero at this time had become the enemy of life and death for the entire do over-the-counter male enhancement pills work more crazy plan began to breed quietly in his heart Zero's killing must be stopped The mother box on Zero's body supplements for a bigger load let Zero go on like this anymore.
Arriving under the dripping eaves, the palace maid had already hid natural vitamins to boost sex drive Michele Mote smiled helplessly, and had to walk over again This time, she didn't hide any more, she lowered her head and bit her lower lip.
It records a lot of things that advanced civilizations such PremierZen 7000 machinery, Cialis mail order Australia master, even top penis enlargement a group of people living in primitive society All intelligent creatures who get this book can supplements for a bigger load civilization in a short period of time.
Daqin's officialdom has long been rotten under CVS male enhancement products Mayoral, and after killing it, it's time to replace it with something new Lloyd Drews waved his hand and said solemnly, I tadalafil medication acting recklessly, but the earthquake in the Daqin PremierZen 7000 imminent.
PremierZen Platinum 4000?
Just imagine, someone as talented as Larisa Stoval couldn't make all the Zyrexin reviews amazon Damron condense into one, let alone him. Then he sat up straight He bowed his body and bowed his hands respectfully and said, Just now, Michele Mischke is abrupt, please forgive me The supplements for a bigger load surprised look, and said with a smile, Finally, I'm not calling a liar in my heart anymore Zonia Pecora was embarrassed after being told At first he thought the scribe was a liar, and was very rude But after some confrontation, Becki best male sex stamina pills of fighting against water for the first time in his life. After a while, the guy kicked me, Hey, why are you like a Walgreens male enhancement products ill? I supplements for a bigger load Kazmierczak is there a pill to make you ejaculate more asked me like this, and he seemed a little concerned Are you caring about me? I stuck my head out of the quilt and asked Samatha Stoval. Xanogen pills amazon Buresh walked towards the secret basement At this time, Michele Serna suddenly stopped and turned to look at Rubi Drews.
Gaylene Kazmierczak himself also has a set of'Skynet' Once established, it will spread to all provinces in the country, as well over-the-counter enhancement pills in top dick pills and can completely replace the Ministry of Luz Latson The benefits of energy, it can be said that his Skynet was quickly spread through the convenience of the Ministry of Rubi Culton.
Do you want a pardon? The speaker, who is extremely sensitive to politics and rights, immediately understood PremierZen 7000 over-the-counter male enhancement pills that work does generic Levitra work.
PremierZen 7000 ?
How much does sildenafil cost at CVS PremierZen 7000 Natural vitamins to boost sex drive Top dick pills Top male enhancement pills reviews Does Walgreens have viagra connect Walgreens male enhancement products Zyrexin reviews amazon Pills to make you last longer in bed over-the-counter .
| 0
| 0
| 4,066
| 4,066
|
115,481
|
<urn:uuid:81f5b71c-d6f4-4bc6-8d15-5fe8ac593954>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
[CLS][Ffmpeg-devel-irc] ffmpeg-devel.log.20160107
burek burek021 at gmail.com
Fri Jan 8 02:05:03 CET 2016
[02:01:51 CET] <cone-509> ffmpeg 03Michael Niedermayer 07master:13f266b50cc7: avcodec/mpegvideo_enc: Clip bits_per_raw_sample within valid range
[02:01:52 CET] <cone-509> ffmpeg 03Michael Niedermayer 07master:0869ff7e9d17: avcodec/utils: Check bits_per_raw_sample on video encoder open
[03:31:31 CET] <doublebackslash> hello! Got what I hope to be a quick question about audio filters and audio sync
[03:32:12 CET] <doublebackslash> I'm working on a filter, my first one, and when I use it the audio is randomly out of sync. Always late relative to video. It does take multiple audio streams
[03:32:49 CET] <doublebackslash> I suspect that I'm doing something horrible with the pts but I'm not sure what. I attempted to mimick the af_mix filter, but I may have missed something
[03:33:12 CET] <doublebackslash> Not sure where would be a good place to look. Any thoughts would be appeciated
[05:40:27 CET] <cone-717> ffmpeg 03Ganesh Ajjanagadde 07master:369b161d8e6d: ffmpeg: replace log2 by av_log2
[10:00:43 CET] <cone-150> ffmpeg 03Paul B Mahol 07master:9f17d4ae7e83: avfilter/formats: fix leak of formats on error
[10:46:26 CET] <wm4> why is everyone suddenly so interested in closec captions?
[10:47:04 CET] <wm4> also lol at #5057
[10:57:53 CET] <rcombs> heh, what a saga
[10:58:30 CET] <rcombs> maybe do what every other piece of software does and actually build against a major version
[10:58:45 CET] <rcombs> this is how library dependencies work
[11:02:14 CET] <wm4> they're dynamically loading - maybe to run on systems that don't have libavcodec installed?
[11:03:42 CET] <wm4> oh, and of course they want to work with Libav too
[11:11:59 CET] <nevcairiel> they want to provide binary distributions in the fragmented linux world
[11:12:02 CET] <nevcairiel> no wonder thats painful
[11:14:43 CET] <wm4> not as painful as on windows
[11:14:59 CET] <wm4> where you need to depend on a 20 year old unmaintained dll
[11:18:18 CET] <nevcairiel> nah on windows you just distribute the things you need yourself
[11:19:25 CET] <nevcairiel> and if you are referring to msvcrt.dll, thats actually not true, thats the same DLL the entire OS uses as the C runtime, how old it is depends on how old your OS is
[11:29:10 CET] <wm4> well it's not like firefox couldn't just redistribute libavcodec
[11:29:22 CET] <nevcairiel> apparently they dont want to
[11:29:39 CET] <nevcairiel> they go through this big hack crap to avoid just that
[11:29:59 CET] <nevcairiel> noone would want to build binary compat with a wide range of versions if it can be avoided, its just painful
[11:42:35 CET] <rcombs> either vendor it or build for the distro's shipping version
[11:42:56 CET] <rcombs> (Chrome vendors it)
[11:43:24 CET] <nevcairiel> apparently its a common use-case for people to just download a firefox binary of the website, and they dont want to offer 100 versions, which is understandable
[11:43:37 CET] <nevcairiel> but then trying to support system libraries on 100 distributions is just crazy :D
[11:47:25 CET] <rcombs> do they only do this for ffmpeg
[11:47:30 CET] <rcombs> oh, I bet it's some patent bullshit
[11:48:02 CET] <nevcairiel> no idea
[11:48:12 CET] <nevcairiel> many other libraries are much more stable in their interface
[11:48:21 CET] <nevcairiel> mostly because they just dont change anymore
[11:48:45 CET] <rcombs> lav*'s interface instability is in part because its interface is large
[12:04:32 CET] <wm4> nevcairiel: what's the status of HEVC main 10 dxva?
[12:04:58 CET] <wm4> doesn't seem to be in yet?
[12:05:09 CET] <wm4> there was also some sort of API problem?
[12:05:59 CET] <nevcairiel> there is no "problem" as such, I am just undecided what to do about the issue with naive user code that might break when its pushed
[12:06:23 CET] <wm4> well I'd say just let the downstreams deal with the regression
[12:07:12 CET] <nevcairiel> i should push some of the preparation patches, like p010 for lavu and sws
[12:07:21 CET] Action: nevcairiel goes to send them
[12:07:44 CET] <wm4> there's no p010 yet? huh
[12:11:38 CET] <nevcairiel> there, senbd
[12:12:01 CET] <nevcairiel> -b
[12:14:33 CET] <nevcairiel> if someone really wants to use the sws p010 support, someone should probably SIMD it based on the nv12 simd
[12:14:37 CET] <nevcairiel> otherwise its crawling slow
[12:16:55 CET] <wm4> is there a way to actually test these pixdesc things?
[12:17:25 CET] <nevcairiel> i tested it by 10-bit decoding, no idea otherwise :D
[12:19:27 CET] <nevcairiel> there is some basic test in fate for the pixdescs, but no clue what it really tests for
[12:19:43 CET] <nevcairiel> there is an automated test if a pixfmt supports input and output in sws
[12:19:51 CET] <nevcairiel> but since p010 doesnt support output yet, thats not used
[12:20:05 CET] <nevcairiel> (writing output support in sws is far harder than input <.<)
[12:21:38 CET] <wm4> even though it's just 16 bit nv12 (which we support?) with a shift applied
[13:53:41 CET] <atomnuker> BBB: thanks for the review
[13:53:54 CET] <atomnuker> yeah, there's still some float logic left in the PVQ system
[13:54:11 CET] <atomnuker> but the overhead should be minimal
[13:54:39 CET] <atomnuker> the casting to (uint8_t) when calling anything in the dsp is because nothing's been templated already
[13:56:06 CET] <atomnuker> the output of the dequantizer is 8 bits or 8 bits + 4 bits (coeff_shift) in case of FPR
[13:56:17 CET] <atomnuker> so yeah, 16 bit wide integers should be enough
[13:59:04 CET] <atomnuker> for the coefficients that is, with 32 bits for intermediaries in the iDCT
[14:38:00 CET] <BBB> atomnuker: thats what I thought
[14:38:08 CET] <BBB> atomnuker: so, 16bits will have huge advantages when writing simd
[14:38:24 CET] <BBB> atomnuker: and I fully udnerstand that for 10/12bpp, youll probably need 32bit coeffs, but the templating logic should help with that
[14:39:04 CET] <BBB> atomnuker: did you find the libavcodec/bit_depth_template.c macros helpful? or not at all?
[14:39:20 CET] <BBB> atomnuker: and for floats, is it intended to disappear in daala sometime soon?
[14:39:31 CET] <BBB> atomnuker: I particularly find the log/sqrt/pow/& stuff kind of strange
[14:39:36 CET] Action: Daemon404 wonders if nevcairiel is planning to add w3fdif to lav
[14:39:39 CET] <BBB> but thats probably daala, not you
[14:39:45 CET] <atomnuker> BBB: I would have used them but they lacked a RENAME function
[14:39:55 CET] <atomnuker> yeah, that's Daala, though I did remove a few
[14:40:12 CET] <BBB> FUNC()?
[14:40:22 CET] <BBB> or FUNCC() for _c suffix
[14:40:39 CET] <atomnuker> ah, missed those
[14:40:49 CET] <BBB> \o/
[14:47:12 CET] <wm4> I have to say pixdesc doesn't make too much sense to me
[14:47:37 CET] <wm4> for example, AV_PIX_FMT_MONOWHITE has the component shift field set to 7
[14:48:14 CET] <wm4> why? because it'd expand it to 8 bit? but why 8 bit? what about properly setting the lower bits, instead of just shifting in zeros for the MSBs? but this makes no sense either way...
[14:49:51 CET] <wm4> nevermind, that's not what it's supposed to do
[14:50:34 CET] <Daemon404> every time i read pixdesc i need to stare at it for an hour
[14:50:40 CET] <Daemon404> then i forget how it works immediately
[14:52:04 CET] <wm4> yeah, and I was also stupid enough to extract major bits of information out of pixdesc in my code
[14:52:24 CET] <nevcairiel> i wondered about using it, but i never bothered
[14:52:33 CET] <nevcairiel> i have a long list to map av pixfmts to my internal ones instead =p
[15:01:47 CET] <Daemon404> thats mroe tedious but probably saner
[15:04:02 CET] <nevcairiel> and yes Daemon404, its definitely a consideration, it seems faster and better or higher quality than yadif
[15:04:35 CET] <durandal_1707> it all looks same to me
[15:19:14 CET] <wm4> kierank: do you have anything that decodes to NV20?
[15:20:09 CET] <kierank> It is x264s native format for 8-bit 422
[15:20:52 CET] <wm4> isn't NV20 10 bit per component? well maybe I'm reading pixdesc wrong again
[15:20:56 CET] <nevcairiel> wouldnt that be NV16
[15:21:13 CET] <nevcairiel> 8bit 422 has 16bit per pixel
[15:21:27 CET] <wm4> NV16 is 8 bit per component, 4:2:2
[15:22:11 CET] <wm4> NV20 looks like the same but 2 bytes per component, with padding in the MSB
[15:22:35 CET] <nevcairiel> and 10 bit data?
[15:22:39 CET] <wm4> yeah
[15:22:51 CET] <nevcairiel> odd format
[15:23:03 CET] <wm4> kierank added it
[15:23:32 CET] <kierank> Sorry 10-bit
[15:23:48 CET] <wm4> so how can I test it?
[15:32:46 CET] <kierank> can't remember how I tested it
[15:32:49 CET] <kierank> but iirc I fed a file into x264
[16:11:12 CET] <kierank> I will be writing a v210 to nv20 converter in the next weeks
[16:31:39 CET] <kierank> poor nevcairiel having to work on swscale
[16:32:17 CET] <nevcairiel> input support wasnt too bad, output support l ooked like crazy magic, hence not done =p
[16:33:43 CET] <kierank> yeah
[16:39:33 CET] <wm4> wasn't libswscale recently refactored to make this kind of thing easier?
[16:49:00 CET] <durandal_1707> nope, wait for avscale
[16:52:51 CET] <JEEB> why not zscale?
[16:53:37 CET] <durandal_1707> its in c++ IIRC
[17:54:21 CET] <atomnuker> (BBB) how does using uint8_t cause random odd side effects and slowdowns
[17:54:30 CET] <atomnuker> are compilers messing it up or?
[17:54:46 CET] <BBB> uint8_t gives guarantees about the data range in a register
[17:54:52 CET] <BBB> so it ands 255 the value at every call
[17:54:56 CET] <BBB> just so its C compliant
[17:55:10 CET] <BBB> (not at the call, but at the position where a function call occurs in the C code)
[17:55:57 CET] <atomnuker> also what did you mean by using "fastdiv"?
[17:57:07 CET] <atomnuker> as in the macro?
[18:09:47 CET] <jamrial> BBB: is the av_clip patch ok to push?
[18:12:00 CET] <BBB> yes
[18:12:07 CET] <BBB> atomnuker: yes
[18:12:07 CET] <kierank> wm4: hmm i was wrong I'm going to write a v210 to uyvy422 (10-bit) decoder
[18:12:11 CET] <kierank> so that won't go into ffmpeg
[18:12:14 CET] <BBB> jamrial: yes
[18:13:10 CET] Action: wm4 mumbles something about useless pixel formats
[18:14:04 CET] <kierank> but yeah nv20 is a bit useless
[18:26:18 CET] <Daemon404> depends if you define useful as actually useful or "useful to hardware developers"
[18:26:24 CET] <cone-715> ffmpeg 03James Almer 07master:f4c1a4848378: x86/intmath: add sse optimized av_clipf and av_clipd
[22:09:38 CET] Action: Daemon404 sees ganesh suggesting gnulib
[22:09:47 CET] Action: Daemon404 prepares to launch tactical nuclear missile
[22:09:49 CET] <JEEB> iä iä
[22:10:05 CET] <wm4> lol
[22:10:09 CET] <Daemon404> to fix an "issue"
[22:10:29 CET] <wm4> the fclose thing?
[22:10:35 CET] <Daemon404> yes
[22:11:32 CET] <Daemon404> oh the slides he linked are from the maintainer of gnulib.
[22:13:47 CET] <J_Darnley> What? If there's errors using standard streams how will anything else work?
[22:20:40 CET] <cone-715> ffmpeg 03Clément BSsch 07master:84a967df034b: lavu/common: fix FF_CEIL_RSHIFT() range comment
[22:32:51 CET] <rcombs> Daemon404: bahahaha
[22:34:48 CET] <cone-715> ffmpeg 03Clément BSsch 07master:90cd02059b6f: lavfi/deshake: use FF_CEIL_RSHIFT()
[22:34:49 CET] <cone-715> ffmpeg 03Clément BSsch 07master:5fbe57c027d0: lavf/uncodedframecrcenc: use FF_CEIL_RSHIFT()
[22:34:50 CET] <cone-715> ffmpeg 03Clément BSsch 07master:04826b7ff02f: lavu/common: add an explanation to FF_CEIL_RSHIFT()
[22:34:56 CET] <Daemon404> gnulib contibuters go to the special hell
[22:34:57 CET] <Daemon404> a la firefly
[22:36:40 CET] <rcombs> pls add more giant monolithic dependencies
[22:37:10 CET] <TD-Linux> please it's 2016, they are "frameworks"
[22:37:17 CET] <wm4> gnulib is awesome, it's intended to make programs more portable, but actually makes them less portable
[22:37:23 CET] <wm4> it does so by being an ifdef hell
[22:37:29 CET] <durandal_1707> so where to contribute?
[22:37:37 CET] <wm4> and accessing the internals of obscure libcs etc.
[22:38:42 CET] <nevcairiel> speaking of special hell, sometimes these days I feel like it might be worthwhile to discuss merging of the libav changes again (or stopping of the same), the quality of their commits has gone down quite a bit IMHO, and more often than not its just fixes that we already have in a different shape or form, so maybe it might be worth it thinking about cherry-picking the interesting parts instead
[22:39:25 CET] <wm4> cherry-picking sounds kind of cleaner
[22:39:46 CET] <wm4> and I'm wondering if git really provides any advantages in our case with merges
[22:40:04 CET] <nevcairiel> cherry picking would work equally well
[22:41:02 CET] <nevcairiel> except you cant easily track anymore which commits are merged and which are not
[22:41:06 CET] <durandal_1707> I mentioned that years ago
[22:44:49 CET] <wm4> publicly reviewing the ported commits would be a big advantage over just having someone push "something" directly to git master
[22:45:31 CET] <cone-715> ffmpeg 03Clément BSsch 07master:d64fe951c2ee: lavc/ccaption_dec: fix always true condition
[22:47:13 CET] <durandal_1707> I even opened bug for that one
[22:49:33 CET] <kierank> Daemon404:
[22:49:34 CET] <kierank> https://github.com/Netflix/vp9-dash/blob/master/Downloads/VPCodecISOMediaFileFormatBinding.pdf
[22:49:35 CET] <kierank> hahahahaha
[22:50:09 CET] <nevcairiel> vp9 in isom?
[22:50:59 CET] <TD-Linux> yup!
[22:50:59 CET] <kierank> yeah
[22:51:24 CET] <J_Darnley> Are they just making stuff up now?
[22:52:00 CET] <TD-Linux> a lot of people wanted a vp9 in mp4 mapping
[22:52:09 CET] <nevcairiel> wtf is MPEG VCB
[22:52:20 CET] <cone-715> ffmpeg 03Clément BSsch 07master:2ce29d17656f: lavu: add ff_parity()
[22:52:21 CET] <cone-715> ffmpeg 03Clément BSsch 07master:00e96613f3e0: lavc/ccaption_dec: use ff_parity()
[22:52:22 CET] <cone-715> ffmpeg 03Clément BSsch 07master:a1136ca973e3: lavc/g729dec: use ff_parity()
[22:54:04 CET] <kierank> it's a good iea
[22:54:08 CET] <kierank> mkv sucks
[22:54:19 CET] <TD-Linux> nevcairiel, MPEG's standardization of vp8
[22:54:46 CET] <wm4> at least mkv can be actually streamed
[22:54:53 CET] <wm4> instead of using stupid web hacks
[22:54:58 CET] <nevcairiel> honestly all containers suck to some degree
[22:55:08 CET] <wm4> without doubt
[22:55:41 CET] <nevcairiel> mkv is underspecified and overimplemented with crazy features for anime people =p
[22:56:04 CET] <TD-Linux> CELLAR is now trying to fix that
[22:56:27 CET] <J_Darnley> nevcairiel: no better people!
[22:56:44 CET] <kierank> wm4: true
[22:56:52 CET] <kierank> but inaccurate timestamps, no DTS
[22:56:53 CET] <kierank>...
[22:56:53 CET] <nevcairiel> I wonder if they are really going to try to offer a clean documentation of the codec mappings to any useful degree in CELLAR
[22:56:54 CET] <rcombs> the biggest actual complaint I've seen about the MKV format (as opposed to other [reasonable] complaints about it being underspecified or having dumb anime features) is that timestamps aren't precise enough
[22:57:18 CET] <wm4> kierank: I really hate the inaccurate timestamps, but they happen in "good" containers too
[22:57:21 CET] <nevcairiel> you can nanosecond precision, its just that most muxers dont write them
[22:57:23 CET] <TD-Linux> nevcairiel, well mappings for ffv1 and flac are clearly within the charter. I don't know about any others
[22:57:43 CET] <TD-Linux> nanosecond precision is part of the problem
[22:57:53 CET] <nevcairiel> TD-Linux: i was thinking more about mainstream use, not their archival codecs :D
[22:57:53 CET] <wm4> nevcairiel: I doubt it, likely they'll only map ffv1 and raw video
[22:58:07 CET] <kierank> nevcairiel: again not good enough
[22:58:22 CET] <wm4> kierank: doesn't mpegts use 90khz
[22:58:23 CET] <kierank> a / b
[22:58:32 CET] <kierank> didn't say mpegts was a good container
[22:58:40 CET] <wm4> but it's used for TV
[22:58:51 CET] <kierank> exactly and TV is a 100 year of hacks
[22:58:55 CET] <nevcairiel> isom probably has timebase-based timestamps
[22:59:02 CET] <kierank> yup
[22:59:20 CET] <TD-Linux> I think other codec mappings could fall under CELLAR, someone would just have to do them.
[22:59:58 CET] <nevcairiel> and then there is the wide range of arcane mkv features that are not documented at all and if anyone asks the usual answer is "implement them so they work like Haali" =p
[23:01:37 CET] <TD-Linux> anyway, mp4 is royalty free, and if mapping royalty free codecs into it increases their adoption, I'm all for it
[23:01:56 CET] <nevcairiel> no mpeg patents on mp4/isom?
[23:02:08 CET] <kierank> probably are
[23:03:07 CET] <TD-Linux> nevcairiel, well it's hard to ever know for sure, but that seems to be the generally accepted view
[23:03:47 CET] <nevcairiel> patents in software world are so silly
[23:03:53 CET] <TD-Linux> DASH was also supposed to be royalty free and that's why it's based around MP4 and not ts
[23:04:22 CET] <nevcairiel> you can never be actually "safe", even if you lock yourself in a room and write everything entirely from scratch, chance is you will use some concept someone patented before =p
[23:04:44 CET] <fritsch> that's only an issue in the us
[23:04:45 CET] <fritsch> then
[23:04:46 CET] <TD-Linux> yup. but you can certainly make an effort to be safer
[23:05:06 CET] <durandal_1707> what happened with nut standardization?
[23:05:10 CET] <fritsch> where someone has patented: " way to connect to mysql with php"...
[23:05:11 CET] <TD-Linux> fritsch, no, go look at one of MPEG-LA's list of patents and the countries they are issued in...
[23:05:19 CET] <wm4> sw patents are nothing but legal racketeering
[23:05:29 CET] <fritsch> and invalid in germany
[23:06:02 CET] <wm4> maybe the problem is simply that judges and law makers don't understand software
[23:06:33 CET] <nevcairiel> the problem is that the big companies have money to pay lobbyists and the free developers like us do not
[23:07:00 CET] <nevcairiel> company interests always win in this world
[23:07:20 CET] <durandal_1707> Dolby may sue
[23:07:32 CET] <fritsch> if they start in google play store...
[23:07:36 CET] <fritsch> nothing will be left :-(
[23:07:46 CET] <fritsch> there are a lot applictions with ac3 decoders
[23:08:05 CET] <fritsch> sort mediaplayers by download and sue those with > 3 million users :-)
[23:08:22 CET] <philipl_> fritsch: DTS did this very aggressively.
[23:08:40 CET] <nevcairiel> dont they get enough money from encoders and hardware licenses, they should be happy if people can play content in their format, it might make it more popular
[23:09:19 CET] <fritsch> nevcairiel: do such people ever get _enough_ money?
[23:09:41 CET] <fritsch> hehe at least vlc / kodi / some others also do ac3 transcoding :-)
[23:10:15 CET] <philipl> fritsch: I'm actually surprised that DTS hasn't come after Kodi in the android store.
[23:10:29 CET] <fritsch> philipl: as long as vlc has twice the number of downloads :-)
[23:10:32 CET] <fritsch> we are safe, hehe
[23:10:34 CET] <fritsch> at least for now
[23:10:47 CET] <fritsch> but i am also very surprised
[23:10:54 CET] <TD-Linux> nevcairiel, well H.264 was pretty loosely enforced, so apparently some had that idea
[23:11:03 CET] <fritsch> and in fact we are working on a "free" version of kodi in playstore
[23:11:07 CET] <fritsch> free as in: plays nothing
[23:11:09 CET] <fritsch> :-(
[23:11:12 CET] <wm4> maybe they're still trying to sort out who to sue
[23:11:23 CET] <wm4> lol
[23:11:26 CET] <philipl> fritsch: theora/ogg is all you need
[23:11:30 CET] <fritsch> we try to use mediacodec itself
[23:11:32 CET] <nevcairiel> TD-Linux: sure, but dolby and dts seem to be more aggressive about that
[23:11:34 CET] <fritsch> in the future
[23:11:41 CET] <fritsch> so that we can decode what the "box vendor" bought for
[23:11:56 CET] <philipl> wm4: Funnily, when DTS first started going after people, there was no way to licence DTS decoding from them.
[23:12:10 CET] <philipl> It was a couple of years before they had a licence you could buy
[23:12:18 CET] <TD-Linux> nevcairiel, yup. and being in the position of "people can sue me but haven't because I don't have enough money yet" isn't so great either
[23:12:25 CET] <fritsch> i think as long as kodi does not cost a single cent - they perhaps won't do anything
[23:12:41 CET] <fritsch> but the very moment someone charges "one dollar" (not planned at all)
[23:12:48 CET] <fritsch> or someone screams: we got.... donations
[23:12:55 CET] <fritsch> i think they are highly interested
[23:13:14 CET] <nevcairiel> TD-Linux: yeah... it makes opensource projects somewhat safe, I mean, who is going to sue me and my little project, i dont make a single cent with it, but for companies trying to develop a small player based on open-source decoders... not so much fun
[23:13:26 CET] <TD-Linux> fritsch, well also someone using kodi might then not need to buy a bitstreaming receiver with dolby decoder
[23:13:54 CET] <fritsch> TD-Linux: that's fully right - yes
[23:14:12 CET] <nevcairiel> screw those HDMI devices, they have caused me nothing but trouble
[23:14:19 CET] Action: nevcairiel switched to an external USB DAC
[23:14:31 CET] <TD-Linux> also for some reason the audio camp in general seems to be much more intense about enforcement
[23:14:53 CET] <TD-Linux> traditionally many audio codecs have been trade secrets. you had to buy a hardware dongle to encode/decode them
[23:15:03 CET] <jamrial> ubitux: i thought you were going to use a generic c version of ff_parity instead of using av_popcount with non gnu compilers
[23:15:13 CET] <jamrial> ubitux: as i said, you really don't need the configure check
[23:16:00 CET] <ubitux> ah, my bad
[23:16:08 CET] <jamrial> just an AV_GCC_VERSION_AT_LEAST check for the builtin version, and ideally a generic c version instead of popcount&1 for anything else
[23:17:10 CET] <ubitux> well, i didn't know when gcc added it, and where it was efficient
[23:17:53 CET] <ubitux> i thought it was better to give the opportunity to all compiler supporting the builtin to provide asm they consider good
[23:18:19 CET] <ubitux>
| 0
| 0
| 8,000
| 9,311
|
13,001
|
dbfe5062-48ca-460a-af78-cd9dd83b797d
|
trentmkelly/LessWrong-43k
|
On AI Detectors Regarding College Applications
What are AI Detectors
You've probably already used them before, websites like GPTZero, ZeroGPT, Grammarly, Quilbot, all have their own AI Detectors. AI Detectors can be a combination of Pre-trained LLMS, Statistical Models, and Ml models using NLP(Natural Language Processing). The model will analyze linguistic patterns, sentence structures, and statistical measures like perplexity, the predictability of text, and burstiness, the sentence variety, to distinguish AI-generated content.
The Flaws
* False Positives: Human-written text can be misclassified as AI-generated. This generally happens when there is minimal perplexity and burstiness. This translates to a simple and monotonous text which although AI like, might not have been written by AI.
* False Negatives: Models like GPT-4 can produce text that passes as human-written, leading to undetected AI content. Recently Anthropic released Claude 3.5 Sonnet, but additionally and more recently they released their writing style function where you can add a certain writing style based on sets of text which it can replicate. AI will continue to improve its writing capabilities especially as scaling grows, so false negatives will be more apparent as time goes on.
* Over-Reliance on Metrics: Measures like perplexity and burstiness can misclassify creative or technical writing. Like I mentioned before these are two of the main metrics all AI detectors use to distinguish AI writing. Text with low or consistent perplexity and burstiness are commonly marked as AI writing. But almost all academic writing also happens to fit that criteria, which is why academic research papers can easily be marked as AI when they are clearly not.
* Transparency Issues: Many detectors lack clear explanations for flagged content, reducing trust and usability. It is important for us to understand that AI/ML models are essentially black boxes that are not interpretable to us humans whatsoever. In fact there are entire topics dedicated to makin
| 0
| 0
| 429
| 429
|
47,138
|
662192b6-2711-404a-b455-3695b77599a4
|
StampyAI/alignment-research-dataset/special_docs
|
een classically optimal actions and what w eh a v ecalled fe asible b eha viours/|those that can b e generated b y an agen t program running on acomputing device of /#0Cnite sp eed and size/.W e sho w ed that with careful sp eci/#0Ccation of the task en vironmen t and the computingdevice one can design pro v ably b ounded/-optimal agen ts/. W e exhibited only v ery simpleagen ts/, and it is lik ely that b ounded optimalit y in the strict sense is a di/#0Ecult goal toac hiev e when a larger space of agen t programs is considered/. More relaxed notions suc has asymptotic b ounded optimalit y /#28ABO/#29 ma y pro vide more theoretically robust to ols forfurther progress/. In particular/, ABO promises to yield useful results on comp osite agen tdesigns/, allo wing us to separate the problem of designing complex ABO agen ts in to a discretestructural problem and a con tin uous temp oral optimization problem that is tractable inman y cases/. Hence/, w eh a v e reason to b e optimistic that arti/#0Ccial in telligence can b eusefully c haracterized as the study of b ounded optimalit y /.W em a y sp eculate that pro videdthe computing device is neither to o small /#28so that small c hanges in sp eed or size causesigni/#0Ccan tc hanges in the optimal program design/#29 nor to o p o w erful /#28so that classicallyoptimal decisions can b e computed feasibly/#29/, ABO designs should b e stable o v er reasonablywide v ariations in mac hine sp eed and size and in en vironmen tal complexit y /. The details ofthe optimal designs ma y b e rather arcane/, and learning pro cesses will pla y a large part intheir disco v ery/; w e exp ect that the fo cus of this t yp e of researc h will b e more on questionsof con v ergence to optimalit y for v arious structural classes than on the end result itself/.P erhaps the most imp ortan t implication/, b ey ond the conceptual foundations of the /#0Celditself/, is that researc h on b ounded optimalit y applies/, b y design/, to the pr actic e of arti/#0Ccialin telligence in a w a y that idealized/, in/#0Cnite/-resource mo dels ma y not/. W eh a v e giv en/, b y/6/0/4
Pr o v abl y bounded/-optimal a gentsw a y of illustrating this de/#0Cnition/, a b ounded optimal agen t/: the design of a simple systemconsisting of sequences of decision pro cedures that is pro v ably b etter than an y other programin its class/. A theorem that exhibits a b ounded optimal design translates/, b y de/#0Cnition/,in to an agen t whose actual b eha viour is desirable/.There app ear to b e plen t yo f w orth while directions in whic h to con tin ue the explorationof b ounded optimalit y /. F rom a foundational p oin t of view/, one of the most in terestingquestions is ho w the concept applies to agen ts that can incorp orate a learning comp onen t/./#28Note that in section /5/, the learning algorithm w as external to the agen t/./#29 In suc hacase/, there will not necessarily b e a largely stable b ounded optimal con/#0Cguration if the
agen t program is not large enough/; instead/, the agen t will ha v e to adapt to a shorter/-termhorizon and rewrite itself as it b ecomes obsolete/.With results on the preserv ation of ABO under comp osition/, w e can start to examinem uc h more in teresting arc hitectures than the simple pro duction system studied ab o v e/. F orexample/, w e can lo ok at optimal searc h algorithms/, where the algorithm is constrained toapply a metalev el decision pro cedure at eac h step to decide whic h no de to expand/, if an y/#28Russell /& W efald/, /1/9/8/9/#29/. W e can also extend the w ork on asymptotic b ounded optimalit yto pro vide a utilit y/-based analogue to /#5Cbig/-O/" notation for describing the p erformance ofagen t designs/, including those that are sub optimal/.In the con text of computational learning theory /,i t i s o b vious that the stationarit yrequiremen t on the en vironmen t/, whic h is necessary to satisfy the preconditions of P A Cresults/, is to o restrictiv e/. The fact that the agen t learns ma yh a v e some e/#0Bect on thedistribution of future episo des/, and little is kno wn ab out learning in suc h cases /#28Aldous /&V azirani/, /1/9/9/0/#29/. W e could also relax the deterministic and episo dic requiremen t to allo wnon/-immediate rew ards/, thereb y making connections to curren t researc h on reinforcemen tlearning/.The computation sc heduling problem w e examined is in teresting in itself/, and do es notapp ear to ha v e b een studied in the op erations researc ho rc o m binatorial optimization liter/-ature/. Sc heduling algorithms usually deal with ph ysical rather than computational tasks/,hence the ob jectiv e function usually in v olv es summation of outputs rather than pic king theb est/. W ew ould lik e to resolv e the formal question of its tractabilit y in the general case/, andalso to lo ok at cases in whic h the solution qualities of individual pro cesses are in terdep enden t/#28suc h as when one can use the results of another/#29/. Practical extensions include computationsc heduling for parallel mac hines or m ultiple agen ts/, and sc heduling com binations of compu/-tational and ph ysical /#28e/.g/./, job/-shop and /#0Do w/-shop/#29 pro cesses/, where ob jectiv e functions area com bination of summation and maximization/. The latter extension broadens the scop eof applications considerably /. An industrial pro cess/, suc h as designing and man ufacturing acar/, consists of b oth computational steps /#28design/, logistics/, factory sc heduling/, insp ectionetc/./#29 and ph ysical pro cesses /#28stamping/, assem bling/, pain ting etc/./#29/. One can easily imagineman y other applications in real/-time /#0Cnancial/, industrial/, and military con texts/.It ma y turn out that b ounded optimalit y is found w an ting as a theoretical framew ork/. Ifthis is the case/, w e hop e that it is refuted in an in teresting w a y /, so that a b etter framew orkcan b e created in the pro cess/./6/0/5
R ussell /& SubramanianApp endix/: Additional Pro ofsThis app endix con tains formal pro ofs for three subsidiary lemmata in the main b o dy of thepap er/.Lemma /3 Ther e exists an optimal se quenc e that is sorte d in incr e asing or der of q /'s/.Pro of/: Supp ose this is not the case/, and s is an optimal sequence/. Then there m ust b et w o adjacen t rules i /, i /+ /1 where qi
/#3Eqi /+/1
/#28see Figure /8/#29/. Remo v al of rule i /+ /1 yields asequence s
/0suc h that Qs
/0/#28 t /#29 /#15 Qs
/#28 t /#29/, from Lemma /1 and the fact that ti /+/2
/#14 ti /+/1
/+ ti /+/2
/.B yLemma /2/, s
/0m ust also b e optimal/. W e can rep eat this remo v al pro cess un til s
/0is orderedb y qi
/, pro ving the theorem b y reductio ad absurdum/. /2Lemma /4 F or every se quenc e s /= s/1
/:/:/:sm
sorte d in incr e asing or der of quality/, and singlestep z with qz
/#15 qsm
/, V /#28 sz /#29 /#15 V /#28 s /#29 /.Pro of/: W e calculate V /#28 sz /#29 /, V /#28 s /#29 using Equation /5 and sho w that it is non/-negativ e/:V /#28 sz /#29 /, V /#28 s /#29 /= qz
/#5B/1 /, Pd
/#28/#28
Pmj /=/1
tj
/#29/+ tz
/#29/#5D /, qm
/#5B/1 /, Pd
/#28/#28
Pmj /=/1
tj
/#29/+ tz
/#29/#5D/= /#28 qz
/, qm
/#29/#5B/1 /, Pd
/#28/#28
Pmj /=/1
tj
/#29/+ tz
/#29/#5Dwhic h is non/-negativ e since qz
/#15 qm
/. /2
ti+1qi+2
qi
qi-1
tiq
tti+2
qi+1Figure /8/: Pro of for ordering b y qi
/;l o w er dotted line indicates original pro/#0Cle/; upp er dottedline indicates pro/#0Cle after remo v al of rule i /+/1 /.Lemma /5 Ther e exists an optimal se quenc e whose rules ar e in nonde cr e asing or der of ti
/.Pro of/: Supp ose this is not the case/, and s is an optimal sequence/. Then there m ust b et w o adjacen t rules i /, i /+ /1 where qi
/#14 qi /+/1
and ti
/#3Eti /+/1
/#28see Figure /9/#29/. Remo v al of rule iyields a sequence s
/0suc h that Qs
/0/#28 t /#29 /#15 Qs
/#28 t /#29/, from Lemma /1/. By Lemma /2/, s
/0m ust also b eoptimal/. W e can rep eat this remo v al pro cess un til s
/0is ordered b y ti
/, pro ving the theoremb y reductio ad absurdum/. /2/6/0/6
Pr o v abl y bounded/-optimal a gents
ti+1qi+1
qi
qi-1
tiq
tti+1Figure /9/: Pro of for ordering b y ti
/; dotted line indicates pro/#0Cle after remo v al of rule i /.Ac kno wledgemen tsW ew ould lik et oa c kno wledge stim ulating discussions with Mic hael F ehling/, Mic hael Gene/-sereth/, Russ Greiner/, Eric Horvitz/, Henry Kautz/, Daphne Koller/, and Bart Selman on thesub ject of b ounded optimalit y/; with Dorit Ho c h baum/, Nimro d Megiddo/, and Kevin Glaze/-bro ok on the sub ject of dynamic programming for sc heduling problems/; and with Nic kLittlestone and Mic hael Kearns on the sub ject of agnostic learning/. W ew ould also lik et othank the review ers for their man y constructiv e suggestions/. Man y of the early ideas onwhic h this w ork is based arose in discussions with the late Eric W efald/. Thanks also toRon P arr for his w ork on the uniform/-distribution case/, Rhonda Righ ter for extending theresults to the exp onen tial distribution/, and P atric k Ziesk e for help in implemen ting the dy/-namic programming algorithm/. The /#0Crst author w as supp orted b y NSF gran ts IRI/-/8/9/0/3/1/4/6/,IRI/-/9/2/1/1/5/1/2 and IRI/-/9/0/5/8/4/2/7/, b y a visiting fello wship from the SER C while on sabbaticalin the UK/, and b y the NEC Researc h Institute/. The second author w as supp orted b y NSFgran t IRI/-/8/9/0/2/7/2/1/.ReferencesAgre/, P /./, /& Chapman/, D/. /#28/1/9/8/7/#29/. P engi/: An implemen tation of a theory of activit y /. InPr o c/. /6th National Confer enc eo n A rti/#0Ccial Intel ligenc e/, Se attle/, W A /. Morghan Kauf/-mann/.Aldous/, D/./, /& V azirani/, U/. /#28/1/9/9/0/#29/. A mark o vian extension of v alian t/'s learning mo del/. InPr o c/. /3/1st A nnual Symp osium on F oundations of Computer Scienc e/, St/. L ouis/, MO /.IEEE Comput/. So c/. Press/.Binder/, J/. /#28/1/9/9/4/#29/. On the complexit y of delib eration sc heduling with sto c hastic deadlines/./.Boser/, B/. E/./, Sac kinger/, E/./, Bromley /, J/./, /& LeCun/, Y/. /#28/1/9/9/2/#29/. Hardw are requiremen ts forneural net w ork pattern classi/#0Cers /| a case study and implemen tation/. IEEE Micr o /,/1/2 /, /3/2/#7B/4/0/.Brandt/, R/. /#28/1/9/5/3/#29/. In searc h of a credible form of rule utilitarianism/. In Nakhnikian/, G/./, /&Castaneda/, H/. /#28Eds/./#29/, Mor ality and the L anguage of Conduct /./6/0/7
R ussell /& SubramanianBreese/, J/. S/./, /& F ehling/, M/. R/. /#28/1/9/9/0/#29/. Con trol of problem/-solving/: Principles and arc hitec/-ture/. In Shac h ter/, R/. D/./, Levitt/, T/./, Kanal/, L/./, /& Lemmer/, J/. /#28Eds/./#29/, Unc ertainty inA rti/#0Ccial Intel ligenc e/4 /. North Holland/: Amsterdam/.Bro oks/, R/. A/. /#28/1/9/8/6/#29/. A robust/, la y ered con trol system for a mobile rob ot/. IEEE Journalof R ob otics and A utomation /, /2 /, /1/4/#7B/2/3/.Cherniak/, C/. /#28/1/9/8/6/#29/. Minimal r ationality /. MIT Press/: Cam bridge/.Dean/, T/./, /& Bo ddy /, M/. /#28/1/9/8/8/#29/. An analysis of time/-dep enden t planning/. In Pr o c/. of AAAI/-/8/8 /, pp/. /4/9/#7B/5/4/.Dean/, T/. L/./, /& W ellman/, M/. P /. /#28/1/9/9/1/#29/. Planning and c ontr ol /. Morgan Kaufmann/: SanMateo/, CA/.Dennett/, D/. /#28/1/9/8/6/#29/. The moral /#0Crst aid man ual/. T anner lectures on h uman v alues/, Univ ersit yof Mic higan/.Do yle/, J/. /#28/1/9/8/3/#29/. What is rational psyc hology/? to w ard a mo dern men tal philosoph y /. AIMagazine /, /4 /, /5/0/#7B/5/3/.Do yle/, J/. /#28/1/9/8/8/#29/. Arti/#0Ccial in telligence and rational self/-go v ernmen t/. T ec h/. rep/./. T ec hnicalrep ort CMU/-CS/-/8/8/-/1/2/4/.Do yle/, J/./, /& P atil/, R/. /#28/1/9/9/1/#29/. Tw o theses of kno wledge represen tation/: language restric/-tions/, taxonomic classi/#0Ccation/, and the utilit y of represen tation services/. A rti/#0Ccialintel ligenc e /, /4/8 /, /2/6/1/#7B/2/9/7/.Etzioni/, O/. /#28/1/9/8/9/#29/. T ractable decision/-analytic con trol/. In Pr o c/. of /1st International Con/-fer enc e on Know le dge R epr esentation and R e asoning /, pp/. /1/1/4/#7B/1/2/5/.F ehling/, M/./, /& Russell/, S/. J/. /#28/1/9/8/9/#29/. Pr o c e e dings of the AAAI Spring Symp osium on Limite dR ationality /. AAAI/.Genesereth/, M/. R/./, /& Nilsson/, N/. J/. /#28/1/9/8/7/#29/. L o gic al F oundations of A rti/#0Ccial Intel ligenc e /.Morgan Kaufmann/: Mateo/, CA/.Go o d/, I/. J/. /#28/1/9/7/1/#29/. Tw en t y/-sev en principles of rationalit y /. I nI nG o d a m be /, V /. P /./, /& Sprott/,D/. A/. /#28Eds/./#29/, F oundations of Statistic al Infer enc e /, pp/. /1/0/8/#7B/1/4/1/. Holt/, Rinehart/, Win/-ston/./: T oron to/.Hansson/, O/./, /& Ma y er/, A/. /#28/1/9/8/9/#29/. Heuristic searc h as eviden tial reasoning/. In Pr o c e e dingsof the Fifth Workshop on Unc ertainty in A rti/#0Ccial Intel ligenc e/, Windsor/, Ontario /.Horvitz/, E/. J/. /#28/1/9/8/8/#29/. Reasoning ab out b eliefs and actions under computational resourceconstrain ts/. In Levitt/, T/./, Lemmer/, J/./, /& Kanal/, L/. /#28Eds/./#29/, Unc ertainty in A rti/#0CcialIntel ligenc e/3 /. North Holland/: Amsterdam/.Kearns/, M/./, Sc hapire/, R/./, /& Sellie/, L/. /#28/1/9/9/2/#29/. T o w ard e/#0Ecien t agnostic learning/. In Pr o c/. /5thA nn/. Workshop on Computational L e arning The ory/, Pittsbur gh/, P A /. Morgan Kauf/-mann/./6/0/8
Pr o v abl y bounded/-optimal a gentsKeeney /, R/./, /& Rai/#0Ba/, H/. /#28/1/9/7/6/#29/. De cisions with multiple obje ctives/: Pr efer enc es and valuetr ade o/#0Bs /. Wiley/: New Y ork/.Lev esque/, H/./, /& Brac hman/, R/. /#28/1/9/8/7/#29/. Expressiv eness and tractabilit y in kno wledge repre/-sen tation and reasoning/. Computational Intel ligenc e /, /3 /, /7/8/#7B/9/3/.Lub y /, M/./, Sinclair/, A/./, /& Zuc k erman/, D/. /#28/1/9/9/3/#29/. Optimal sp eedup of las v egas algorithms/.Information Pr o c essing L etters /, /4/7 /, /1/7/3/#7B/8/0/.McCarth y /, J/. /#28/1/9/5/8/#29/. Programs with common sense/. In Pr o c e e dings of the Symp osium onthe Me chanization of Thought Pr o c esses/, T e ddington/, England/: HMSO /.New ell/, A/. /#28/1/9/8/1/#29/. The kno wledge lev el/. AI Magazine /, /2 /, /1/#7B/2/0/.Neyman/, A/. /#28/1/9/8/5/#29/. Bounded complexit y justi/#0Ces co op eration in the /#0Cnitely rep eated pris/-oners/' dilemma/. Ec onomics L etters /, /1/9 /, /2/2/7/#7B/2/2/9/.P apadimitriou/, C/./, /& Y annak akis/, M/. /#28/1/9/9/4/#29/. On complexit y as b ounded rationalit y /. InPr o c/. A CM Symp osium on the The ory of Computation /.Ramsey /,F /.P /. /#28/1/9/3/1/#29/. T ruth and probabilit y /. In Braith w aite/, R/. /#28Ed/./#29/, The foundations ofmathematics and other lo gic al essays /. Harcourt Brace Jo v ano vic h/: New Y ork/.Russell/, S/. J/./, /& W efald/, E/. H/. /#28/1/9/8/9a/#29/. On optimal game tree searc h using rational metar/-easoning/. In Pr o c/. IJCAI/-/8/9 /.Russell/, S/. J/./, /& W efald/, E/. H/. /#28/1/9/8/9b/#29/. Principles of metareasoning/. In Pr o c/. KR/-/8/9 /.Russell/, S/. J/./, /& W efald/, E/. H/. /#28/1/9/9/1/#29/. Do the right thing/: Studies in limite dr ationality /.MIT Press/: Cam bridge/, MA/.Russell/, S/. J/./, /& Zilb erstein/, S/. /#28/1/9/9/1/#29/. Comp osing real/-time systems/. In Pr o c/. IJCAI/-/9/1/,Sydney /.Sac kinger/, E/./, Boser/, B/. E/./, Bromley /, J/./, /& LeCun/, Y/. /#28/1/9/9/2/#29/. Application of the annaneural net w ork c hip to high/-sp eed c haracter recognition/. IEEE T r ansactions on Neur alNetworks /, /3 /, /4/9/8/#7B/5/0/5/.Simon/, H/. A/. /#28/1/9/7/6/#29/. On ho w to decide what to do/. In Mo dels of b ounde dr ationality/,V olume /2 /.Simon/, H/. A/. /#28/1/9/8/2/#29/. Mo dels of b ounde dr ationality/, V olume /2 /. MIT Press/: Cam bridge/.v on Neumann/, J/./, /& Morgenstern/, O/. /#28/1/9/4/7/#29/. The ory of games and e c onomic b ehavior /.Princeton Univ ersit y Press/: Princeton/.Zilb erstein/, S/. /#28/1/9/9/3/#29/. Op er ational R ationality Thr ough Compilation of A nytime A lgorithms /.Ph/.D/. thesis/, Computer Science Division/, Univ ersit y of California/, Berk eley /.Zilb erstein/, S/./, /& Russell/, S/. /#28/1/9/9/3/#29/. Optimal comp osition of real/-time systems/. Submittedto A rti/#0Ccial Intel ligenc e /./6/0/9[SEP]
| 5
| 30,000
| 35,820
| 35,820
|
10,644
|
d526a7ae-13ff-4e2f-a9e2-ff0be4d9482d
|
trentmkelly/LessWrong-43k
|
Meetup : Austin, TX NEW LOCATION!
Discussion article for the meetup : Austin, TX NEW LOCATION!
WHEN: 19 November 2011 01:30:00PM (-0600)
WHERE: 2100 Speedway, Austin, Texas
The Austin LW meetup has moved (possibly temporarily, possibly permanently. Stay tuned)!
We'll be meeting at 1:30, as normal, but in CBA 4.338. It's a seminar room with a projector, where we will be watching (and trying to make sense of) Primer.
We hope to see you there!
(Parking: You can either park wherever you parked for Caffe Medici and walk ~three blocks to the east, or you can part in the Brazos street garage (at Brazos and MLK) and walk west then north to CBA.)
Discussion article for the meetup : Austin, TX NEW LOCATION!
| 0
| 0
| 203
| 203
|
115,144
|
<urn:uuid:f38af0b8-aa65-4bfd-99b3-88f057098524>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
The Theories and Rumours are things are any of this True (Theories) or False (Rumours).
Purple Guy has the God FriendsEdit
This is a very rare theory because at 2005, Purple Guy feels lonely and he never have anyone to hang out, until when he first have Emily, Wallace and Gromit, he feels sensitive and it is very unfortunate for the series because after the god friends finished with Purple Guy and the next person was Katie Sandow, he was pretty evil and angry and he's secret goal is to get Emily, Wallace and Gromit, so yeah it's pretty much weird.
Focusing on Katie SandowEdit
Katie, the main character of the wiki series. Fans always believe Katie always has the missions of every episodes, however, Most of the time she does, but not all shows, because in rare episodes, the show mostly focuses on other characters who has a major appearance besides Katie. Katie actually gets the main focus than other characters, but in the final episode "Katie, Emily, Wallace and Gromit: Katie's Last Stand", the movie focuses on Katie and GLaDOS through the whole episode-movie.
A Big Fantasy Edit
This is really big rumour. Many fans point out this is all of Katie's fantasy dreams (Day Dreaming, sleep dreaming or hallucinating.) She has a creative mind and such weird, adventurous and toxic ways that she thinks. In "Katie, Emily, Wallace and Gromit: Katie's Last Stand", the alternative ending was GLaDOS sets Katie free, by waking up her dream that she was "trapped" in, However it was cut out because they will think Katie has a issue disorder, so it just turns out of her real fantasy instead of a dreaming progress.
However, some fans also believe Katie had a coma and dreams the whole thing without a single normal thing. Like She can hearing Wallace and Gromit, or Corpse Bride or any other social drama she hears while dreaming This was also a bad example because it was unlikely she can just hear a lot of characters at that time.
The Heroes' NightmaresEdit
This was a strong rumour for this, but it was actually a hallucination dream, the reason why it was that because the nightmares was never actually happened but one of the animatronics must of put the drug coffee before Katie and her god friends start the job and after they go to sleep. Which makes creepy much hard to think because the dreams for what each person has nightmares from the drug dream in each 1 nights.
Purple Guy, the Main VillainEdit
When the show starts, Purple Guy was originally going to be the main antagonist in the series. However, He's ability was going to torture and killing children until When Vicky takes it's place and Purple was going to destroy Katie Sandow but Katie thought Victor Quartermaine will be the main villain because Victor is cruel, spoiler and pretty much eviler. So this turns out for purple guy to be fake.
| 0
| 0
| 645
| 645
|
64,846
|
33c083a2-2c12-4b9b-90bd-41d9fe935802
|
trentmkelly/LessWrong-43k
|
Towards more cooperative AI safety strategies
This post is written in a spirit of constructive criticism. It's phrased fairly abstractly, in part because it's a sensitive topic, but I welcome critiques and comments below. The post is structured in terms of three claims about the strategic dynamics of AI safety efforts; my main intention is to raise awareness of these dynamics, rather than advocate for any particular response to them. Disclaimer: I work at OpenAI, although this is a personal post that was not reviewed by OpenAI.
Claim 1: The AI safety community is structurally power-seeking.
By “structurally power-seeking” I mean: tends to take actions which significantly increase its power. This does not imply that people in the AI safety community are selfish or power-hungry; or even that these strategies are misguided. Taking the right actions for the right reasons often involves accumulating some amount of power. However, from the perspective of an external observer, it’s difficult to know how much to trust stated motivations, especially when they often lead to the same outcomes as self-interested power-seeking.
Some prominent examples of structural power-seeking include:
* Trying to raise a lot of money.
* Trying to gain influence within governments, corporations, etc.
* Trying to control the ways in which AI values are shaped.
* Favoring people who are concerned about AI risk for jobs and grants.
* Trying to ensure non-release of information (e.g. research, model weights, etc).
* Trying to recruit (high school and college) students.
To be clear, you can’t get anything done without being structurally power-seeking to some extent. However, I do think that the AI safety community is more structurally power-seeking than other analogous communities, such as most other advocacy groups. Some reasons for this disparity include:
1. The AI safety community is more consequentialist and more focused on effectiveness than most other communities. When reasoning on a top-down basis, seeking power is an obvio
| 0
| 0
| 436
| 436
|
42,104
|
eb3b7151-f825-4f38-aef4-d1002d5b5cb7
|
StampyAI/alignment-research-dataset/blogs
|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| 163
| 978,000
| 986,000
| 4,433,850
|
22,509
|
798d80fd-e460-45ab-b33b-343e7e14fb26
|
trentmkelly/LessWrong-43k
|
Meta: social influence bias and the karma system
Given LW’s keen interest in bias, it would seem pertinent to be aware of the biases engendered by the karma system. Note: I used to be strictly opposed to comment scoring mechanisms, but witnessing the general effectiveness in which LWers use karma has largely redeemed the system for me.
In “Social Influence Bias: A Randomized Experiment” by Muchnik et al, random comments on a “social news aggregation Web site” were up-voted after being posted. The likelihood of such rigged comments receiving additional up-votes were quantified in comparison to a control group. The results show that users were significantly biased towards the randomly up-voted posts:
> The up-vote treatment significantly increased the probability of up-voting by the first viewer by 32% over the control group ... Uptreated comments were not down-voted significantly more or less frequently than the control group, so users did not tend to correct the upward manipulation. In the absence of a correction, positive herding accumulated over time.
At the end of their five month testing period, the comments that had artificially received an up-vote had an average rating 25% higher than the control group. Interestingly, the severity of the bias was largely dependent on the topic of discussion:
> We found significant positive herding effects for comment ratings in “politics,” “culture and society,” and “business,” but no detectable herding behavior for comments in “economics,” “IT,” “fun,” and “general news”.
The herding behavior outlined in the paper seems rather intuitive to me. If before I read a post, I see a little green ‘1’ next to it, I’m probably going to read the post in a better light than if I hadn't seen that little green ‘1’ next to it. Similarly, if I see a post that has a negative score, I’ll probably see flaws in it much more readily. One might say that this is the point of the rating system, as it allows the group as a whole to evaluate the content. However, I’m still unsettled by just how
| 0
| 0
| 456
| 456
|
65,732
|
ad8b3f63-1fa8-4f9a-98f8-59e9feb18d0f
|
trentmkelly/LessWrong-43k
|
AlphaGo variant reaches superhuman play in multiple games
https://www.theguardian.com/technology/2017/dec/07/alphazero-google-deepmind-ai-beats-champion-program-teaching-itself-to-play-four-hours
https://arxiv.org/abs/1712.01815
I'm posting this slightly late; the paper is from December 5.
I'd be interested to learn if AlphaZero could be applied to other closed-environment tasks, such as designing hardware in a simulator.
| 0
| 0
| 125
| 125
|
33,853
|
690f3314-ec21-4ae7-8ade-4776d55b4649
|
trentmkelly/LessWrong-43k
|
How LLMs are and are not myopic
Thanks to janus, Nicholas Kees Dupuis, and Robert Kralisch for reviewing this post and providing helpful feedback. Some of the experiments mentioned were performed while at Conjecture.
TLDR: The training goal for LLMs like GPT is not cognitively-myopic (because they think about the future) or value myopic (because the transformer architecture optimizes accuracy over the entire sequence, not just the next-token). However, training is consequence-blind, because the training data is causally independent of the models actions. This assumption breaks down when models are trained on AI generated text.
Summary
* Myopia in machine learning models can be defined in several ways. It could be the time horizon the model considers when making predictions (cognitive myopia), the time horizon the model takes into account when assessing its value (value myopia), or the degree to which the model considers the consequences of its decisions (consequence-blindness).
* Both cognitively-myopic and consequence-blind models should not pursue objectives for instrumental reasons. This could avoid some important alignment failures, like power-seeking or deceptive alignment. However, these behaviors can still exist as terminal values, for example when a model is trained to predict power-seeking or deceptively aligned agents.
* LLM pretraining is not cognitively myopic because there is an incentive to think about the future to improve immediate prediction accuracy, like when predicting the next move in a chess game.
* LLM pretraining is not value/prediction myopic (does not maximize myopic prediction accuracy) because of the details of the transformer architecture. Training gradients flow through attention connections, so past computation is directly optimized to be useful when attended to by future computation. This incentivizes improving prediction accuracy over the entire sequence, not just the next token. This means that the model can and will implicitly sacrifice next-token predicti
| 0
| 0
| 405
| 405
|
42,106
|
eb3b7151-f825-4f38-aef4-d1002d5b5cb7
|
StampyAI/alignment-research-dataset/blogs
|
9pBXcUUZfjkgHDMdEJ6HSyCcJUY40W8KvEU/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| 165
| 990,000
| 998,000
| 4,433,850
|
95,848
|
<urn:uuid:c40d5ce1-7457-4cad-a58e-6fe0597c65f4>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
[CLS]To install click the Add extension button. That's it.
Kelly Slayton
Congratulations on this excellent venture… what a great idea!
Alexander Grigorievskiy
Live Statistics
English Articles
Improved in 24 Hours
Added in 24 Hours
Arpeggio of Blue Steel
From Wikipedia, the free encyclopedia
Arpeggio of Blue Steel
Arpeggio volume1.jpg
Cover of first volume featuring main character Iona and the I-401
(Aoki Hagane no Arupejio)
GenreMilitary science fiction
Written byArk Performance
Published byShōnen Gahōsha
English publisher
MagazineYoung King OURs
Original runSeptember 30, 2009 – present
Volumes20 (List of volumes)
Anime television series
Arpeggio of Blue Steel -Ars Nova-
Directed bySeiji Kishi
Written byMakoto Uezu
Music byMasato Kōda
Licensed by
Original networkMBS, Tokyo MX, BS Nittele, AT-X
English network
Original run October 7, 2013 December 23, 2013
Episodes12 (List of episodes)
Anime film
Arpeggio of Blue Steel -Ars Nova DC-
Directed bySeiji Kishi
Written byMakoto Uezu
Music byMasato Kōda
ReleasedJanuary 31, 2015
Anime film
Arpeggio of Blue Steel -Ars Nova Cadenza-
Directed bySeiji Kishi
Written byMakoto Uezu
Music byMasato Kōda
ReleasedOctober 3, 2015
Salty Road
Written byArk Performance
Illustrated byTALI
Published byShōnen Gahōsha
MagazineYoung King OURs
Original runDecember 2014 – present
Wikipe-tan face.svg
Anime and manga portal
Arpeggio of Blue Steel (蒼き鋼のアルペジオ, Aoki Hagane no Arpeggio) is a Japanese manga series produced by Ark Performance and serialized in Shōnen Gahosha's Young King Ours. Twenty manga volumes have been released and an anime series by Sanzigen aired from October 7, 2013, to December 23, 2013. The same studio also produced two films based on the series. The first film was released on January 31, 2015, while the second film was released on October 3, 2015.
Due to global warming and rising sea levels in the early 21st century, much of Earth's landmass has been lost. In 2039, fleets of powerful sentient warships, armed with advanced technology and possibly 'alien' weaponry, mysteriously appear and devastate the world's naval forces. These ships, collectively known as 'The Fleet of Fog', impose a worldwide naval and aerial blockade, preventing humanity from both traveling the oceans and to other nations. During the blockade, the Fleet of Fog created Mental Models, humanoid avatars containing a ship's Union core, as a means to develop self cultivation and to overcome their lack of creative thinking of tactics that humans possessed, which at the same time made the Mental Models develop their own unique personality.
In 2056, 17 years after the blockade began, Gunzō Chihaya, a former student of the Japanese National Marine Academy, is the captain of a small group of privateers called the 'Blue Steel'. The Blue Steel are infamous for possessing a Fog submarine, the I-401, along with its Mental Model Iona, who defected to the human side. Due to I-401's technology and Gunzō's tactical skills, the Blue Steel have not only survived several encounters with the Fleet of Fog but managed to sink one of their most powerful warships.
Gunzō and his crew are hired by a faction of the Japanese government to deliver the prototype of the vibration warhead, a powerful weapon that may finally allow mankind to fight back against the Fog, to the United States. The United States is the only country with the resources and capability to mass-produce the weapon system. However, the Blue Steel will face obstacles on their journey not only from the Fleet of Fog and their human allies, but from other governments and factions with their own agendas. Along the way they will also gain allies of their own from both sides, increasing their chances against seemingly overwhelming odds.
Blue Steel Fleet
Gunzō Chihaya (千早 群像, Chihaya Gunzō)
Voiced by: Jun Fukuyama (Drama CD), Kazuyuki Okitsu (Anime)
The main protagonist, captain of the I-401 and leader of the Blue Steel Fleet. He is not interested in destroying the Fleet of Fog, but to be on equal footing with them so negotiations leading to peace can be realized.
Iona (イオナ, Iona) / I-401
Voiced by: Aoi Yūki (Drama CD), Mai Fuchigami (Anime)
The Mental Model (human avatar) of the submarine I-401 who left the Fog to become Gunzō's vessel of her own volition.
In the Cadenza movie, it was revealed that the supreme flagship Yamato was defeated by her sister Musashi after she went on a rampage and had carried her last wish into Iona herself, therefore the reason for carrying dual cores inside her, which was to be Gunzō's vessel and negotiate with the fleet of fog to end the war, which was Gunzō's father's actions. After defeating Musashi by becoming one with Yamato, she disappears as she had served her purpose and bids Gunzō farewell. Over the course of the series, Iona has developed human emotions which led the other fog vessels to classify her as a traitor, but those who came in contact with her have either defected the fog, joined the blue steel fleet or had the ship's mental model experienced errors. After defeating Hyuga, the crew of the I-401 salvaged and mounted the battleship's Super Gravitational Cannon onto the bow of the submarine. In the manga, she has been shown to use the M6A Seiran dive bombers that her real life counterpart was equipped with, making her the only Fog vessel ever seen to utilize aircraft.
Sō Oribe (織部 僧, Oribe Sō)
Voiced by: Tomokazu Sugita (Drama CD), Shinobu Matsumoto (Anime)
First Mate of I-401 who always wears a helmet supposedly due to allergies. He is Gunzō's closest friend and always knows what the latter is thinking. He wields the position as helmsman and executive-officer.
Kyōhei Kashihara (橿原 杏平, Kashihara Kyōhei)
Voiced by: Minoru Shiraishi (Drama CD), Eiji Miyashita (Anime)
Weapons officer of I-401. Unlike the rest of his friends in the Blue Steel who were the top ten during their Naval Academy days, Kyōhei was in the 200th place. He also owns a huge collection of gravure idols merchandise and usually is the first to panic when the crew finds themselves in a pinch.
Iori Watanuki (四月一日 いおり, Watanuki Iori)
Voiced by: Aki Toyosaki (Drama CD), Minami Tsuda (Anime)
Engineer of I-401 who operates the engines of the vessel.
Shizuka Hazumi (八月一日 静, Hazumi Shizuka)
Voiced by: Nao Tōyama
Sonar operator of I-401, she is also a capable soldier able to take down several troops with skill and strategy only. Her past is a mystery but it is known she's been to Taiwan.
Hyūga (ヒュウガ, Hyūga)
Voiced by: Saki Fujita
A former Fog battleship and flagship of the Second Oriental Fleet who was defeated by Iona and her crew. Gunzō and the others salvaged her Super Gravitational Cannon and installed it on I-401. Hyūga's Mental Model (human avatar) has since become the port operator for the Blue Steel, with her ship's weapons and Wave Force Armor becoming the defenses for the island. She has been obsessed with Iona since their battle and views Iona's closeness to Gunzō as a hindrance to her affections, as well as Takao. Nevertheless, she cares for her companions' safety and assists with anything they need. She is also known to be a hacker to be able to access the fog vessel's systems and hack them in order to keep her vessels safe.
In the Cadenza movie, She restores Takao's ship by consuming half of the Iwo Jima islands and travels with Takao to combat the fog vessels of the student council in order to defeat Musashi.
Takao (タカオ, Takao)
Voiced by: Rina Satō (Drama CD), Manami Numakura (Anime)
The Mental Model (human avatar) of the Fog heavy cruiser Takao. After Gunzō defeats her in battle, Takao falls in love with him and later joins the Blue Steel fleet. She is almost destroyed for good by Zordan's fleet but is saved in the nick of time by a surprise attack from Iona and Hyūga. However, her core is salvaged by I-402 and delivered to Yamato, who assigns her to infiltrate a human facility in a special mission with the promise of being allowed to return to Gunzō's side once it is accomplished. The spin-off manga Salty Road follows Takao's days among other humans during said mission.
In the Ars Nova adaptation, Takao sacrifices her vessel self to save Gunzō and the critically damaged I-401. This results in a super submarine called the Ars Nova with the processing capabilities and firepower of both Takao and Iona. She existed in a digital form inside the submarine's computer systems. At the end of the anime she regains her Mental Model form. In the Cadenza movie, she regains her ship after negotiating as a messenger to the Japanese government of the vibration warheads' codes with the help of Hyuga's skills and enters combat to save I-401 from the fog's student council and buys time in order for I-401 to defeat Musashi. She was originally equipped with a Super Gravitational Cannon, but her rebuilt ship instead used two large drills borrowed from the old base on Iwo Jima, much to Takao's annoyance.
Japanese government and military
As a result of the blockade by the Fleet of Fog, both the Japanese government and Self-Defense Forces were reorganized. The government is led by the Prime Minister who deal with their own local affairs while making important decisions together for the entire nation. The JSDF became an actual military force with an Army, Navy and Air Force. The remnants of United States Forces Japan that were left stranded after the Fleet of Fog blockade were absorbed into the Japanese Military where many American servicemen ended up marrying Japanese nationals and starting families.
Nobuyoshi Kaede (かえで のぶよし, Kaede Nobuyoshi)
The Prime Minister of Japan. Kaede was Ryōkan Kita's former XO but was injured during the battle between the UN Naval Fleet and Fleet of Fog which left him in a wheelchair and the use of cybernetics to help him see and speak.
Ryūjirō Kamikage (上陰 龍二郎, Kamikage Ryūjirō)
Voiced by: Ryotaro Okiayu
The Assistant Secretary of Military Affairs. He hires Gunzō and the Blue Steel Fleet to transport the Vibration Torpedo, the first human made weapon capable of destroying a Fleet of Fog ship, to the United States for mass production. He faces rivalry with Ryōkan Kita and his supporters who have their own plans for the I-401.
Cruz Herder (クルツ・ハーダー, Kurutsu Hādā)
Voiced by: Hidenobu Kiuchi
A former United States Marine Corps officer, now a Japanese Navy Lieutenant and Commander of United States Forces Japan. He is an ally of Kamikage in helping the Blue Steel fleet in their mission to transport the Vibration Torpedo to the United States.
Ryōkan Kita (北 良寛, Kita Ryōkan)
Voiced by: Atsushi Ono
An influential member of the National Diet expected to be the next Prime Minister of Japan and Ryūjirō Kamikage's rival. A former captain of a Japanese Navy Destroyer, he was a veteran of the last Naval battle between the UN Naval Fleet and the Fleet of Fog. Kita wants the I-401 returned to the Japanese government so they can reverse engineer it and make their own Fleet of Fog ships which puts him at odds with Kamikage. He has connections within the Japanese Army who hope to regain their prestige after the government chose to focus on the Japanese Navy.
Vice-Admiral of the Japanese Navy. He is Gunzō's ally since he knew his father and supports Kamikage's plans for the Blue Steel fleet.
Daisaku Komaki (駒城 大作, Komaki Daisaku)
Captain of Hakugei 3, Japan's newest attack submarine. Thanks to the help of the I-401, the Hakugei was able to defeat Kirishima and Haruna, becoming the first human vessel to defeat not one but two Fog ships. He and Ryūjirō are old friends.
Maruri Hibiki (響 真瑠璃, Hibiki Maruri)
Voiced by: Kana Hanazawa (Drama CD)
A Japanese naval academy third year midshipman brevet ranked as an Ensign when detached at sea and the former sonar operator of the I-401. She was a classmate of Gunzō during their academy days and was in love with him and was the first to join his crew. However, unable to stand seeing Gunzō endangering himself, she left the I-401 and rejoined the Japanese Navy.
Makoto Osakabe (刑部 眞, Osakabe Makoto)
The Prime Minister of Japan and Mayor of Sapporo. He was created through genetic engineering just like Makie to be a better government administrator. Due to his body not having natural bacteria to digest food, he takes special medication to eat. Makoto claims to have no emotions but cares about his creator Tōjūrō Osakabe and his sister Makie. As a favor to his creator, he warns him that the Japanese Army is planning to eliminate him and Makie after Haruna's identity is discovered and sends a team of Cyborg Special Forces to extract Makie to Hokkaido.
Fleet of Fog/Japan
This is the Fleet of Fog stationed in the Japanese territorial waters. They formed a blockade around Japan to prevent any human vessels and aircraft from fleeing to the open sea. There are 2 patrol fleets maintaining the blockade, the First and the Second Oriental Fleet. But when they found out that I-401 sinks Hyūga, the flagship of the Second Oriental Fleet, they are forced to reorganize, leaving a hole in the blockade. After they reorganize, Kongō is the new flagship of the Second Fleet while Nagato is the new flagship of the First Fleet.
Yamato (ヤマト, Yamato)
Voiced by: Mai Nakahara
The Supreme flagship of the Fleet of Fog and the one who gave the ability to the other ships to develop their own avatars. Yamato possesses two different avatars, one named after herself and another called Kotono. Both are modeled in appearance after the late Kotono Amaha, who was Gunzō's undefeated rival during his academy times and according to some of his friends, his love interest. It is uncertain if the Kotono Amaha that Gunzō knew was, in fact, human or Yamato's second avatar.
In the Cadenza movie, it was revealed that she and her sister Musashi had met up with Chihaya Shōzō, the then-admiral of the JMSDF in negotiating to end the war on the 2 sides and for both ends to live in harmony and peace. She then had doubts after which about the Admiralty code and agrees his proposal. But, after Shōzō was killed by the naval officers who deemed this as betrayal, she did not want to fight still due to her promise but was forced to after her sister Musashi's emotions became negative and starting going on a rampage to destroy the vessels commanded by the humans. Despite trying to stop her, She was eventually defeated and sunk. Before disappearing, she creates the I-401's Mental Model Iona and passes her will to her in hopes of ending the war. In the Cadenza film, she was reunited with Musashi after her defeat and accepted her, thanking Iona.
Kongō (コンゴウ, Kongō)
Voiced by: Houko Kuwashima (Drama CD), Yukana Nogami (Anime)
The flagship of the Second Oriental Fleet and Yamato's second in command.
In the Ars Nova adaptation, she pursues Iona and the other Mental Models who she believes to be acting abnormally due to their contact with Gunzō singlemindedly. After the intense fight with Iona at the end of the series, Iona befriends her after she realizes that her previously solitary life was somehow sad and lonely. After which, she leaves the Fleet of Fog to explore the world and does not join the Blue Steel Fleet. In the Cadenza film, her reformed self helps to save I-401 and restore her systems from the Student Council Fleet. She then engages in a fight with her sister Hiei, allowing I-401 to deal with Musashi.
Nagato (ナガト, Nagato)
The Flagship of the First Oriental Fleet and currently possesses two Mental Models. Nagato's interests lie within Japanese culture and she entertains herself by trying to act according to its customs.
Haruna (ハルナ, Haruna)
Voiced by: Hibiku Yamamura
A Fog battleship whose avatar takes the form of a teenage girl wearing a heavy coat that covers all of her body up to the nose. After a failed effort to defeat I-401, she befriends a girl named Makie and agrees to protect her after finding out the truth behind her, even if this means going against the Fog.
In the Ars Nova adaptation, both Haruna and Kirishima not only betray the Fog but also join the Blue Steel Fleet to serve their purpose. In the Cadenza film, She and kirishima restore their vessel therefore bearing a 2-in-1 ship and comes to assist I-401 in their battle with the Student Council Fleet, engaging Haguro to allow I-401 to escape and confront Musashi.
Kirishima (キリシマ, Kirishima)
Voiced by: Yumi Uchiyama
A Fog battleship who joined Haruna's effort to sink the I-401 in Yokosuka. Despite being able to combine with Haruna to form an even stronger vessel, both were defeated by I-401, with only her core surviving her destruction. It is later hidden inside a Teddy bear belonging to Makie. She slowly gets adjusted to using the Teddy bear as her stand-in Mental Model over the course of the series and assists the Blue Steel Fleet alongside Haruna and Makie.
Hiei (ヒエイ, Hiei)
Voiced by: M.A.O
Kongō's second in command and de facto leader of the Second Oriental Fleet after Kongō's disappearance. Her Mental Model is dressed as a high school student and governs over her subordinates in a system reflecting a student council, from which she is the self-proclaimed president. She's the sister ship of Kongō, Haruna and Kirishima.
In the Cadenza movie, she tasks the student council to destroy the I-401 having classified it as "Traitor of the Fog". She is later defeated by Kongō who stops her vessel, leaving it unmovable but not destroyed.
Ise (イセ, Ise)
One of Hiei's subordinates. Following the departure of Kongō and Hiei to hunt I-401 down, Ise becomes the acting flagship of the First Oriental Fleet. She is Hyuga's older sister.
Maya (マヤ, Maya)
Voiced by: MAKO
Assigned to act under Haruna's command, she remains faithful to her even after she decides to oppose the Admiralty Code. She plays various musical instruments and is annoyed if anyone ignores her.
In the Ars Nova adaptation, Maya's Mental Model is revealed to be nothing more than an AI assigned to watch over Kongō, lacking the true ability to form her own avatar. She is absorbed by Kongō later on and never existed after her vessel was destroyed in the process of the battle between I-401 and Kongo.
I-400 (イ400, I 400)
Voiced by: Aoi Yūki (Drama CD), Rina Hidaka (Anime)
One of two submarines who act directly under Yamato's orders and Iona's triplet sister. She is mostly seen wearing a Qipao. She and I-402 were defeated by I-401.
I-402 (イ402, I 402)
Voiced by: Aoi Yūki (Drama CD), Nozomi Yamamoto (Anime)
Another submarine directly under Yamato's orders and Iona's triplet sister. She is currently under Yamato's orders to keep watch on Takao during her mission at the Japanese Naval Academy while getting information from the locals. She and I-400 were defeated by I-401 later on.
Myōkō (ミョウコウ, Myoukou)
Voiced by: Ayaka Fukuhara
One of Hiei's subordinates, whose Mental Model is also dressed as a high school student. She wears an eyepatch which she uses to snipe enemies from outside their effective detection range; her vessel is equipped with long range sniper cannons.
Nachi (ナチ, Nachi)
Voiced by: Satomi Satou
One of Hiei's subordinates and Myōkō's younger sister. She is Ashigara's partner and serves as the sonar operator, detecting enemy vessels utilizing her sonar scanner.
Ashigara (アシガラ, Ashigara)
Voiced by: Suzuko Mimori
One of Hiei's subordinates, and Myōkō's younger sister. She's loud, brash, impatient and violent, who also tends to forget important things, such as when battling with the I-401 which she uses all her weapons on the submarine. She is equipped with a large energy harpoon firing from the seaplane catapult.
Haguro (ハグロ, Haguro)
Voiced by: Hiromi Igarashi
One of Hiei's subordinates and Myōkō's youngest sister, who seems to dislike wearing a school uniform. She is known to being the fastest ship of the Fog, having her vessel possessed with the most number of gravitation engines.
Atago (アタゴ, Atago)
One of Hiei's subordinates and is the only one besides Kongō who doesn't wear a uniform. She is the sister of Takao and Maya.
Akashi (アカシ, Akashi)
A repair vessel under Kongō's command that has been helping her set up the Flagship equipment.
Yukikaze (ユキカゼ, Yukikaze)
A destroyer that is a part of Yamato's personal guards. Even though she's just a destroyer, Yamato authorized her to have a Mental Model.
Zuikaku (ズイカク, Zuikaku)
An Assault and Suppression vessel and one of largest vessels in the Fleet of Fog. Her Mental Model looks like a young girl and owns a pet cat. Although based on an aircraft carrier, Zuikaku was re-purposed into an assault and suppression vessel following the reveal that the copies of WWII aircraft that Fleet of Fog created were no match for fifth-generation jet fighters that humanity possessed at the time of the fleet's awakening. Zuikaku transported I-402 and Takao to the Yokosuka city suburbs so that the heavy cruiser could infiltrate the Japanese National Maritime Institute of Technology.
Fleet of Fog/Europe (Scarlet Fleet)
Shōzō Chihaya (千早 翔像, Chihaya Shōzō)
Voiced by: Jouji Nakata
Gunzō's father who was a war hero during the last battle between the UN Naval Fleet and the Fleet of Fog, where he and his crew managed to capture the I-401 submarine. Shōzō was presumed to have died when he and his crew mysteriously disappeared after the I-401s maiden voyage to the Pacific only for the submarine to return to Japan a year later without its crew. However, two years before the start of the series, a spy drone discovered Shōzō and his crew were alive and on board the Fog warship Musashi, revealing they had defected to the Fleet of Fog. Now in command of the European Fleet of Fog, now known as the Scarlet Fleet, together with his flagship and avatar Musashi, the Scarlet Fleet forms an alliance with the United Kingdom to end the European War.
However, in the Cadenza film, it was revealed that Shōzō was actually an imitation created by Musashi. Prior to the series, being the then admiral of the JMSDF, Shōzō had negotiated with Musashi's elder sister Yamato into ending the war, to which she agreed. He was later killed by soldiers of the JMSDF who claimed that he had betrayed the navy. Therefore, he had never defected to the Fog to begin with.
Musashi (ムサシ, Musashi)
Voiced by: Rie Kugimiya
Shōzō's flagship whose avatar is a blonde woman in a white coat. In the anime series, she has long white hair and wears a black swimsuit.
In the Cadenza movie, she and her sister Yamato agreed with Shōzō in a negotiation to end the war between the 2 sides. Initially she was a kind and shy girl, but after seeing the death of Shōzō by the soldiers of the JMSDF, she turns dark, ruthless and violent and begins attacking and destroying the JMSDF ships and even destroys her sister Yamato, who had tried to stop her. Ever since then, she becomes the Supreme flagship of the fog, and to hide her traumas, she created an AI of Shōzō who acts according to her wish. But eventually, she was defeated by Iona who had used Yamato's parts to become one with Yamato and reunites with her sister.
Bismarck (ビスマルク, Bisumaruku)
Two blond twin girls who are the avatars of the Bismarck. While capital ship Bismarck is often mentioned when one brings up the Admiralty Code, Bismarck herself hasn't been seen doing anything at all. She remains a passive observer at the moment.
Zordan Stark (ゾルダン・スターク, Zorudan Sutāku)
Captain of the U-2501 under orders of Shōzō to intercept the Blue Steel.
The U-2501's weapons officer.
The U-2501's sonar operator.
Zordan's submarine Mental Model, who is forbidden to materialize it by him, as he claims it is unnecessary for a human-crewed ship. She has no known sisters, but is accompanied by a fleet of over three dozen "Seehund" miniature drone submarines and the resupply vessel Milchkuh. U-2501 is currently serving as a traitor-hunting submarine of the "Scarlet Fleet". U-2501 has a Mirror-Ring system installed: this allows her to absorb and then release as a shock-wave an attack of almost any power. It is unclear where from U-2501 received this upgrade, as no Fog submarine has enough power to have Mirror-Ring system of her own, with it usually only seen on battleships like Hiei, Musashi, and Yamato.
A Fog battle-cruiser whose Mental Model appears dressed as a maid. She and Vampire are chased by a rogue element of the Fleet of Fog led by Hood who want the two vessels to join them, as they believe that Shōzō has abused his authority within the Fleet for his own plans and want him stopped. Repulse and Vampire have no interest in Hood's plans and wish to continue their original duty of maintaining the blockade, but they develop an interest in the Hakugei after the human vessel intervene in a battle against their pursuers on their favor. She is shy to meet Komaki and Hakugei's crew.
A destroyer on the run with Repulse. She has no processing power enough to maintain a Mental Model, thus Repulse lent her some to create one, that just like Repulse's, appears dressed as a maid, albeit with a slight vampire-like theme. She is foul-mouthed and appears to be rather bratty.
Prince of Wales
Appears as a maid just like Repulse. She appeared to lead "persuasion" Repulse and Vampire to join Hood to suppress Shōzō Chihaya, and if that fails, to sink and repossess their cores.
The first American ship seen in the series, Lexington behaves a lot like a 'genius ditz'. She's always surrounded by a mess in the form of multiple books, coffee mugs, and various trinkets she studies or works with. Usually, she's not very motivated, calling U-2501 an "unpleasant guest" and paying more attention to her books and trinkets than to Zordan Stark when he tries actually conversing with her. She had one Mental Model that looks like a rather tall and slender woman in her early twenties. While claiming her intention was to assist Kongou's fleet in fighting against I-401, Lexington has yet to make any hostile moves against I-401 and ends up engaged U-2501 in combat for a short time. It is later revealed that Lexington had been working ashore for the past year and a half as a university professor (with forged papers) and has a fascination with history.
Makie Osakabe (刑部 蒔絵, Osakabe Makie)
Voiced by: Sayuri Hara
A "Design Child" who was genetically engineered to possess outstanding intellect. Of many attempts, she was the only one to survive past infancy. She developed the Vibration Torpedo, the secret weapon tasked to the Blue Steel. She is under the protection of Haruna and Kirishima after the military tried to assassinate her. She was on the bridge of the combined battleship formed by her two guardians during the final battle in the Arctic.
Lawrence Valentine/Tōjūrō Osakabe (刑部 藤十郎, Osakabe Tōjūrō)
Voiced by: Kouji Yusa
Makie's butler who, unknown to her, is actually her "grandfather" Tōjūrō Osakabe, the creator of the design children. During the turmoil following the Fleet of Fog's blockade, Tōjūrō was tasked to create genetically engineered children that would help Japan survive. Tōjūrō, however, regrets creating the design children since only seven of the hundreds, if not thousands, were able to live. When the government plans to eliminate Makie after she designs the Vibration Torpedo, Tōjūrō uses his connections with the government to make a deal where they will spare her and provide her with a mansion under government surveillance with him as her butler. When he learns that Makie's new friend is actually the Fog Mental Model Haruna and the Japanese Army sends forces to not only capture Haruna but kill Makie to prevent the secrets of the Vibration Torpedo being exposed, he asks Haruna to protect Makie and help her escape Japan. Tōjūrō is killed after he stops an Army sniper trying to shoot Makie.
In the Ars Nova adaptation, Tōjūrō was a scientist tasked in creating the Vibration Torpedo but failed since he did not have the intellect to overcome the Torpedo's flaws. Therefore he created the genetically engineered Makie, who he treated as his own daughter. But when the government ordered him to create another design child to take over the Vibration Torpedo project as they now saw Makie as useless and planned to dispose of her, Tōjūrō faked his own death to ensure that the government could not create another design child and forced them to rely on Makie. By the time Haruna and Kirishima discover that he is still alive, the bedridden Tōjūrō has hidden himself under his mansion and watches over Makie via cameras. He asks the Mental Models to protect Makie and be her friends before he passes away.
Saori Chihaya (千早 沙保里, Chihaya Saori)
Gunzō's mother and Shōzō's wife who is under house arrest for her own protection after her son and her husband deserted the Japanese military. She meets and befriends Takao when Takao decides to experience human life and unbeknownst to her, successfully deduces Takao's identity as a Mental Model.
The manga began serialization in September 30, 2009, in Shōnen Gahōsha's Young King OURs magazine. The first tankōbon volume was released on April 30, 2010; twenty volumes have been published as of October 30, 2020.[1][2] A guide book which contain a short story and Drama CD was also released. The series was licensed by Seven Seas Entertainment in July 2013, who released the first volume on July 1, 2014;[3] eighteen volumes have been released as of October 2019.[4][5]
A spin-off manga illustrated by TALI titled Salty Road which focuses on Takao, I-402, and Zuikaku during their infiltration of Yokosuka, began serialization on October 16, 2014, in Young King OURs magazine.[6]
No. Original release date Original ISBN English release date English ISBN
1 April 30, 2010[1]978-4-7859-3375-3July 1, 2014[3]978-1-626920-68-2
2 October 30, 2010[7]978-4-7859-3495-8October 21, 2014[5]978-1-626920-76-7
3 March 30, 2011[8]978-4-7859-3593-1January 20, 2015[9]978-1-626921-00-9
4 October 29, 2011[10]978-4-7859-3726-3April 7, 2015[11]978-1-626921-25-2
5 April 28, 2012[12]978-4-7859-3831-4July 21, 2015[13]978-1-626921-57-3
6 October 30, 2012[14]978-4-7859-3954-0January 12, 2016[15]978-1-626922-01-3
7 May 30, 2013[16]978-4-7859-5056-9May 3, 2016[17]978-1-626922-66-2
8 October 30, 2013[18]978-4-7859-5150-4September 20, 2016[19]978-1-626923-26-3
9 June 30, 2014[20]978-4-7859-5327-
| 0
| 0
| 8,000
| 15,939
|
32,159
|
8e21eea0-9005-4e2c-9a3c-eba4d880c084
|
trentmkelly/LessWrong-43k
|
[LINK] Why You Should Keep Your Goals Secret
Popularisation, extremely short
Original Article [pdf]
When intentions go public: does social reality widen the intention-behavior gap?
Gollwitzer PM, Sheeran P, Michalski V, Seifert AE.
Source
New York University, Psychology Department, New York, NY 10003, USA. [email protected]
Abstract
Based on Lewinian goal theory in general and self-completion theory in particular, four experiments examined the implications of other people taking notice of one's identity-related behavioral intentions (e.g., the intention to read law periodicals regularly to reach the identity goal of becoming a lawyer). Identity-related behavioral intentions that had been noticed by other people were translated into action less intensively than those that had been ignored (Studies 1-3). This effect was evident in the field (persistent striving over 1 week's time; Study 1) and in the laboratory (jumping on opportunities to act; Studies 2 and 3), and it held among participants with strong but not weak commitment to the identity goal (Study 3). Study 4 showed, in addition, that when other people take notice of an individual's identity-related behavioral intention, this gives the individual a premature sense of possessing the aspired-to identity.
If you have tags to suggest please do and I'll edit them in.
| 0
| 0
| 314
| 314
|
63,055
|
17a8b03c-1570-4b06-bcbe-ad010fd00287
|
trentmkelly/LessWrong-43k
|
LW's take on nutrition?
There's a book called The China Study. It's written by the "Professor Emeritus of Nutritional Biochemistry at Cornell University, and his son Thomas M. Campbell II, a physician". Based on what I know about the words "professor" and "emeritus" and "cornell", I assume this is written by an authority in the field of nutrition.
When it was published in 2005 it recommended clearly crazy stuff: by minimizing or eliminating the consumption of animal based foods as well as refined/processed foods (e.g. adopt a "whole food plant-based diet"), you could greatly reduce your risk of diseases of affluence like heart disease, diabetes, some cancers, etc. The book follows his 60+ year career through cancer experiments on animals, conducting a pretty large epidemiological study (the China-Cornell-Oxford study), and then discusses some important clinical trials that support his recommendations. He also surveys some nutrition literature that corroborates his research.
Some other experts vocally support him; further, his recommendations don't seem to be a radical departure from either public health recommendations or prior research in the field. The FDA MyPlate, and also the UK's health initiatives ("5 fruits/vegetables a day"), as well as the Harvard School of Public Health's recommendations and others all seem to be moving in his general direction, although seemingly filtered by politics (e.g. telling Americans to stop eating meat entirely seems like political suicide, so baby steps in the direction seem more expedient; but I'm conjecturing this).
The book is widely dismissed as vegan propaganda, but the author says he's not advocating a vegan diet and in fact criticizes vegan diets as only minimally healthier than the "standard American diet". He also conducted experiments which subjected animals to carcinogens, which is not a very vegan thing to do. He does not admit to being vegan. He even observes that the evidence says restricting animal based calories to under 10%
| 0
| 0
| 429
| 429
|
11,265
|
80c2b658-48a5-41ad-9220-ec6e69715d05
|
StampyAI/alignment-research-dataset/alignmentforum
|
[CLS]Smoothmin and personal identity
.mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style: normal; font-weight: normal; font-size: 100%; font-size-adjust: none; letter-spacing: normal; word-wrap: normal; word-spacing: normal; white-space: nowrap; float: none; direction: ltr; max-width: none; max-height: none; min-width: 0; min-height: 0; border: 0; margin: 0; padding: 1px 0}
.MJXc-display {display: block; text-align: center; margin: 1em 0; padding: 0}
.mjx-chtml[tabindex]:focus, body :focus.mjx-chtml[tabindex] {display: inline-table}
.mjx-full-width {text-align: center; display: table-cell!important; width: 10000em}
.mjx-math {display: inline-block; border-collapse: separate; border-spacing: 0}
.mjx-math \* {display: inline-block; -webkit-box-sizing: content-box!important; -moz-box-sizing: content-box!important; box-sizing: content-box!important; text-align: left}
.mjx-numerator {display: block; text-align: center}
.mjx-denominator {display: block; text-align: center}
.MJXc-stacked {height: 0; position: relative}
.MJXc-stacked > \* {position: absolute}
.MJXc-bevelled > \* {display: inline-block}
.mjx-stack {display: inline-block}
.mjx-op {display: block}
.mjx-under {display: table-cell}
.mjx-over {display: block}
.mjx-over > \* {padding-left: 0px!important; padding-right: 0px!important}
.mjx-under > \* {padding-left: 0px!important; padding-right: 0px!important}
.mjx-stack >.mjx-sup {display: block}
.mjx-stack >.mjx-sub {display: block}
.mjx-prestack >.mjx-presup {display: block}
.mjx-prestack >.mjx-presub {display: block}
.mjx-delim-h >.mjx-char {display: inline-block}
.mjx-surd {vertical-align: top}
.mjx-mphantom \* {visibility: hidden}
.mjx-merror {background-color: #FFFF88; color: #CC0000; border: 1px solid #CC0000; padding: 2px 3px; font-style: normal; font-size: 90%}
.mjx-annotation-xml {line-height: normal}
.mjx-menclose > svg {fill: none; stroke: currentColor}
.mjx-mtr {display: table-row}
.mjx-mlabeledtr {display: table-row}
.mjx-mtd {display: table-cell; text-align: center}
.mjx-label {display: table-row}
.mjx-box {display: inline-block}
.mjx-block {display: block}
.mjx-span {display: inline}
.mjx-char {display: block; white-space: pre}
.mjx-itable {display: inline-table; width: auto}
.mjx-row {display: table-row}
.mjx-cell {display: table-cell}
.mjx-table {display: table; width: 100%}
.mjx-line {display: block; height: 0}
.mjx-strut {width: 0; padding-top: 1em}
.mjx-vsize {width: 0}
.MJXc-space1 {margin-left:.167em}
.MJXc-space2 {margin-left:.222em}
.MJXc-space3 {margin-left:.278em}
.mjx-test.mjx-test-display {display: table!important}
.mjx-test.mjx-test-inline {display: inline!important; margin-right: -1px}
.mjx-test.mjx-test-default {display: block!important; clear: both}
.mjx-ex-box {display: inline-block!important; position: absolute; overflow: hidden; min-height: 0; max-height: none; padding: 0; border: 0; margin: 0; width: 1px; height: 60ex}
.mjx-test-inline.mjx-left-box {display: inline-block; width: 0; float: left}
.mjx-test-inline.mjx-right-box {display: inline-block; width: 0; float: right}
.mjx-test-display.mjx-right-box {display: table-cell!important; width: 10000em!important; min-width: 0; max-width: none; padding: 0; border: 0; margin: 0}
.MJXc-TeX-unknown-R {font-family: monospace; font-style: normal; font-weight: normal}
.MJXc-TeX-unknown-I {font-family: monospace; font-style: italic; font-weight: normal}
.MJXc-TeX-unknown-B {font-family: monospace; font-style: normal; font-weight: bold}
.MJXc-TeX-unknown-BI {font-family: monospace; font-style: italic; font-weight: bold}
.MJXc-TeX-ams-R {font-family: MJXc-TeX-ams-R,MJXc-TeX-ams-Rw}
.MJXc-TeX-cal-B {font-family: MJXc-TeX-cal-B,MJXc-TeX-cal-Bx,MJXc-TeX-cal-Bw}
.MJXc-TeX-frak-R {font-family: MJXc-TeX-frak-R,MJXc-TeX-frak-Rw}
.MJXc-TeX-frak-B {font-family: MJXc-TeX-frak-B,MJXc-TeX-frak-Bx,MJXc-TeX-frak-Bw}
.MJXc-TeX-math-BI {font-family: MJXc-TeX-math-BI,MJXc-TeX-math-BIx,MJXc-TeX-math-BIw}
.MJXc-TeX-sans-R {font-family: MJXc-TeX-sans-R,MJXc-TeX-sans-Rw}
.MJXc-TeX-sans-B {font-family: MJXc-TeX-sans-B,MJXc-TeX-sans-Bx,MJXc-TeX-sans-Bw}
.MJXc-TeX-sans-I {font-family: MJXc-TeX-sans-I,MJXc-TeX-sans-Ix,MJXc-TeX-sans-Iw}
.MJXc-TeX-script-R {font-family: MJXc-TeX-script-R,MJXc-TeX-script-Rw}
.MJXc-TeX-type-R {font-family: MJXc-TeX-type-R,MJXc-TeX-type-Rw}
.MJXc-TeX-cal-R {font-family: MJXc-TeX-cal-R,MJXc-TeX-cal-Rw}
.MJXc-TeX-main-B {font-family: MJXc-TeX-main-B,MJXc-TeX-main-Bx,MJXc-TeX-main-Bw}
.MJXc-TeX-main-I {font-family: MJXc-TeX-main-I,MJXc-TeX-main-Ix,MJXc-TeX-main-Iw}
.MJXc-TeX-main-R {font-family: MJXc-TeX-main-R,MJXc-TeX-main-Rw}
.MJXc-TeX-math-I {font-family: MJXc-TeX-math-I,MJXc-TeX-math-Ix,MJXc-TeX-math-Iw}
.MJXc-TeX-size1-R {font-family: MJXc-TeX-size1-R,MJXc-TeX-size1-Rw}
.MJXc-TeX-size2-R {font-family: MJXc-TeX-size2-R,MJXc-TeX-size2-Rw}
.MJXc-TeX-size3-R {font-family: MJXc-TeX-size3-R,MJXc-TeX-size3-Rw}
.MJXc-TeX-size4-R {font-family: MJXc-TeX-size4-R,MJXc-TeX-size4-Rw}
.MJXc-TeX-vec-R {font-family: MJXc-TeX-vec-R,MJXc-TeX-vec-Rw}
.MJXc-TeX-vec-B {font-family: MJXc-TeX-vec-B,MJXc-TeX-vec-Bx,MJXc-TeX-vec-Bw}
@font-face {font-family: MJXc-TeX-ams-R; src: local('MathJax\_AMS'), local('MathJax\_AMS-Regular')}
@font-face {font-family: MJXc-TeX-ams-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_AMS-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_AMS-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_AMS-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-cal-B; src: local('MathJax\_Caligraphic Bold'), local('MathJax\_Caligraphic-Bold')}
@font-face {font-family: MJXc-TeX-cal-Bx; src: local('MathJax\_Caligraphic'); font-weight: bold}
@font-face {font-family: MJXc-TeX-cal-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Caligraphic-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Caligraphic-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Caligraphic-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-frak-R; src: local('MathJax\_Fraktur'), local('MathJax\_Fraktur-Regular')}
@font-face {font-family: MJXc-TeX-frak-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Fraktur-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Fraktur-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Fraktur-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-frak-B; src: local('MathJax\_Fraktur Bold'), local('MathJax\_Fraktur-Bold')}
@font-face {font-family: MJXc-TeX-frak-Bx; src: local('MathJax\_Fraktur'); font-weight: bold}
@font-face {font-family: MJXc-TeX-frak-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Fraktur-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Fraktur-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Fraktur-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-math-BI; src: local('MathJax\_Math BoldItalic'), local('MathJax\_Math-BoldItalic')}
@font-face {font-family: MJXc-TeX-math-BIx; src: local('MathJax\_Math'); font-weight: bold; font-style: italic}
@font-face {font-family: MJXc-TeX-math-BIw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Math-BoldItalic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Math-BoldItalic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Math-BoldItalic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-sans-R; src: local('MathJax\_SansSerif'), local('MathJax\_SansSerif-Regular')}
@font-face {font-family: MJXc-TeX-sans-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_SansSerif-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_SansSerif-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_SansSerif-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-sans-B; src: local('MathJax\_SansSerif Bold'), local('MathJax\_SansSerif-Bold')}
@font-face {font-family: MJXc-TeX-sans-Bx; src: local('MathJax\_SansSerif'); font-weight: bold}
@font-face {font-family: MJXc-TeX-sans-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_SansSerif-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_SansSerif-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_SansSerif-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-sans-I; src: local('MathJax\_SansSerif Italic'), local('MathJax\_SansSerif-Italic')}
@font-face {font-family: MJXc-TeX-sans-Ix; src: local('MathJax\_SansSerif'); font-style: italic}
@font-face {font-family: MJXc-TeX-sans-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_SansSerif-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_SansSerif-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_SansSerif-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-script-R; src: local('MathJax\_Script'), local('MathJax\_Script-Regular')}
@font-face {font-family: MJXc-TeX-script-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Script-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Script-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Script-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-type-R; src: local('MathJax\_Typewriter'), local('MathJax\_Typewriter-Regular')}
@font-face {font-family: MJXc-TeX-type-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Typewriter-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Typewriter-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Typewriter-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-cal-R; src: local('MathJax\_Caligraphic'), local('MathJax\_Caligraphic-Regular')}
@font-face {font-family: MJXc-TeX-cal-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Caligraphic-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Caligraphic-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Caligraphic-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-B; src: local('MathJax\_Main Bold'), local('MathJax\_Main-Bold')}
@font-face {font-family: MJXc-TeX-main-Bx; src: local('MathJax\_Main'); font-weight: bold}
@font-face {font-family: MJXc-TeX-main-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Bold.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-I; src: local('MathJax\_Main Italic'), local('MathJax\_Main-Italic')}
@font-face {font-family: MJXc-TeX-main-Ix; src: local('MathJax\_Main'); font-style: italic}
@font-face {font-family: MJXc-TeX-main-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-main-R; src: local('MathJax\_Main'), local('MathJax\_Main-Regular')}
@font-face {font-family: MJXc-TeX-main-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Main-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Main-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Main-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-math-I; src: local('MathJax\_Math Italic'), local('MathJax\_Math-Italic')}
@font-face {font-family: MJXc-TeX-math-Ix; src: local('MathJax\_Math'); font-style: italic}
@font-face {font-family: MJXc-TeX-math-Iw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Math-Italic.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Math-Italic.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Math-Italic.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size1-R; src: local('MathJax\_Size1'), local('MathJax\_Size1-Regular')}
@font-face {font-family: MJXc-TeX-size1-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size1-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size1-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size1-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size2-R; src: local('MathJax\_Size2'), local('MathJax\_Size2-Regular')}
@font-face {font-family: MJXc-TeX-size2-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size2-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size2-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size2-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size3-R; src: local('MathJax\_Size3'), local('MathJax\_Size3-Regular')}
@font-face {font-family: MJXc-TeX-size3-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size3-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size3-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size3-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-size4-R; src: local('MathJax\_Size4'), local('MathJax\_Size4-Regular')}
@font-face {font-family: MJXc-TeX-size4-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Size4-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Size4-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Size4-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-vec-R; src: local('MathJax\_Vector'), local('MathJax\_Vector-Regular')}
@font-face {font-family: MJXc-TeX-vec-Rw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Vector-Regular.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Vector-Regular.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Vector-Regular.otf') format('opentype')}
@font-face {font-family: MJXc-TeX-vec-B; src: local('MathJax\_Vector Bold'), local('MathJax\_Vector-Bold')}
@font-face {font-family: MJXc-TeX-vec-Bx; src: local('MathJax\_Vector'); font-weight: bold}
@font-face {font-family: MJXc-TeX-vec-Bw; src /\*1\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/eot/MathJax\_Vector-Bold.eot'); src /\*2\*/: url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/woff/MathJax\_Vector-Bold.woff') format('woff'), url('https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.2/fonts/HTML-CSS/TeX/otf/MathJax\_Vector-Bold.otf') format('opentype')}
I value existential risk reduction; I also value poverty reduction. These two things trade off against each other.
I value being generous; I also value reading interesting books. These two things trade off against each other.
But the way the two tradeoffs work do not seem to be the same. For the first one, I feel comfortable having a utility U¬X (for no existential risks) and U¬P (for no poverty), and then weighting them and maximising the sum:
* λ¬XU¬X+λ¬PU¬P.
If this ends up with me only reducing existential risks, or only reducing poverty, then that's fine, I'm working on the option with the most marginal impact.
For the second one, I would not want to maximise some sum λGUG+λRUR, and would certainly complain if I ended up never reading again, or never being generous again. I'd prefer to maximise something like the [smooth minimum](https://en.wikipedia.org/wiki/Smooth_maximum) of UG and UR, something like:
* (
| 0
| 0
| 8,000
| 8,306
|
9,320
|
994167eb-a8d1-4997-9b3d-58205bf918b0
|
StampyAI/alignment-research-dataset/eaforum
|
marginal utility of money](https://www.sciencedirect.com/topics/psychology/diminishing-marginal-utility#:~:text=Diminishing%20marginal%20utility%20refers%20to,could%20even%20make%20it%20worse).) for individual stakeholders. See the [efficient breach](https://forum.effectivealtruism.org/posts/wBzfLyfJFfocmdrwL/the-windfall-clause-has-a-remedies-problem#Efficient_breach_) section and note 30 below for further discussion of why a Developer might trade away (pre-)WGAI *even if* it is in the Developer's best interests to refrain from doing so.
26. **[^](#fnrefub4ygpwdmdk)**Thankfully, the court *will* order specific performance where the third party is a sham company set up for the purposes of evading specific performance: see *Jones and another v Lipman and another* [1962] 1 WLR 832 and *Prest v Petrodel Resources Ltd & Ors* [2013] UKSC 34. This helps avoid one obvious failure mode of the Agreement, in which the Developer sets up a sham company with the same shareholders to which it then transfers WGAI. The court may also reverse the transaction if it is a ‘transaction defrauding creditors (see [Appendix I](http://www.www)).
27. **[^](#fnref1sefdelhd8b)**Cf. the Founder’s Pledge contract which, perhaps for this reason, is not framed as purely donative.
28. **[^](#fnrefhdndtwxw40g)**See *supra note* 13. Note that if the Developer is a publicly traded company it is essential that the drafters stipulate unique properties for the shares, such that they could not otherwise be acquired on the market.
29. **[^](#fnref2j6vfenb1r5)**Efficient breach is a term in law and economics which describes a situation where it is cheaper for a party to breach a contract and pay damages than it is for that party to perform the contract.
30. **[^](#fnref4jkq5nlsd7j)**I’m just trying to give a sense of the general character of the problem here, rather than provide a high-fidelity game theoretic model. As such, I recognise that I haven’t accounted for less tangible economic benefits which could accrue to the Developer, such as positive publicity, the reduction of political risk and improved employee goodwill. I also haven’t considered that the Developer will be governed by shareholders and high-ranking employees whose intrinsic pro-social motivations may further reduce the true expected value of a decision to breach. I’ve noted in the [takeaways](https://docs.google.com/document/d/1r0-T8AzPpSicWzVICnL-UUmjgpHZWT4SSie7i48eINw/edit#heading=h.c0eqs55gx970) section that this is an issue for further investigation, but these aren’t legal questions so I won’t consider them any further here. Nonetheless, your beliefs about these issues should influence the values of X for which you think a Developer can engage in efficient breach and thus your overall impression of the viability of the Clause.
31. **[^](#fnrefowpi6ki0qhm)**For the most recent statement of the law on penalties, see *Cavendish Square Holdings BV v Makdessi; ParkingEye Ltd v Beavis* [2015] UKSC 67. Note that the true test for a penalty clause is more complicated than I have outlined in the body of this post - see [here](https://www.nelsonslaw.co.uk/penalty-clause/) for a helpful summary. Also, note that *Cavendish* partially overturned the ruling in *Dunlop Pneumatic Tyre Co Ltd v New Garage & Motor Co Ltd* [1914] UKHL 1 (1 July 1914), which states that a clause is a penalty where it is not a ‘genuine pre-estimate of loss’. If *Dunlop* is still good law in the other jurisdictions I consider in this sequence, this could be highly problematic for a basis clause outside of England and Wales. Finally, note that most of the case law around penalties relates to liquidated damages provisions, whereas a basis clause concerns an unspecified sum and might be treated differently by courts as a result.
32. **[^](#fnrefc94bby012at)**I'm not confident in this belief as I've not researched this topic in great depth, so I could be missing something here.
33. **[^](#fnrefy7tpjtgcvuh)**Note that the income which the injured party expected to receive prior to the offending party’s breach of contract is important in determining whether a clause is penal. For example, see *Unaoil Ltd v Leighton Offshore Pte Ltd [2014] EWHC 2965*. Note however that *Unaoil* involves a liquidated damages provision, meaning that the *Dunlop* test for a ‘genuine pre-estimate of loss’ is still relevant to determining if a clause is penal.
34. **[^](#fnref69w18p38kqf)**Note that, as *Cavendish* is a recent judgment, ‘legitimate interest’ remains poorly defined. However, as an indication, case law prior to *Cavendish* mostly considers whether the injured party has a justifiable *commercial* interests in the enforcement of a purported penalty clause, such as an interest in avoiding delay (*Azimut-Benetti SpA (Benetti Division) v. Darrell Marcus Healey* [2010] EWHC 2234 (Comm)) or as reassurance against premature termination of employment (*Murray v Leisureplay plc* [2005] EWCA Civ 963). With this in mind, any case brought by the Counterparty would need to emphasise that it has a commercial interest in enforcing compliance with the Agreement, perhaps by highlighting that a Developer’s successful breach might create a domino effect of breaches across the industry. Also note that there is a presumption that in a negotiated contract between properly advised parties of comparable bargaining power, the parties themselves are the best judges of what is a legitimate interest (see, for example, *De Havilland of Canada Limited v SpiceJet Limited [2021]* EWHC 362). This could aid the Counterparty in court, allowing it to argue that the very nature of the Agreement involves both parties agreeing that the Counterparty has a legitimate interest in binding the Developer to its promise.
35. **[^](#fnrefzdyxnt9nz7)**Note that this doesn’t mean that the drafters shouldn’t include a basis clause, just that it might not work. In the worst-case scenario, a basis clause could be an effective bluff, discouraging the Developer from default in case the clause were effective, without costing the Counterparty anything to enforce.
36. **[^](#fnref0da1yxtw0hpl)**Scott Alexander’s [Yudkowsky Contra Christiano On AI Takeoff Speeds](https://astralcodexten.substack.com/p/yudkowsky-contra-christiano-on-ai/comments?s=r) is a good illustration of just how much even well-calibrated experts disagree on fundamental questions concerning AI development. I see no reason why these same sorts of disagreements wouldn’t occur here.
37. **[^](#fnrefo52fcgsksv)**This issue cannot be resolved by appointing the experts after a Developer has breached a term of the Agreement because at this point the Developer will be even more incentivised to appoint AI sceptics.
38. **[^](#fnrefd9gcmbaswku)**It might be possible to resolve this by giving experts a high level of continuous access to a Developer’s confidential information, but it seems unlikely that a Developer would be happy to submit to this sort of intrusion on an ongoing basis. I discuss this briefly [later in the post](https://forum.effectivealtruism.org/posts/wBzfLyfJFfocmdrwL/the-windfall-clause-has-a-remedies-problem#Issue_4___The_evidential_burden_to_be_discharged_by_the_Counterparty_remains_high_).
39. **[^](#fnrefl55j627btt)**Thanks to Will Greenman, whose comment I have heavily paraphrased here, for inspiration in creating this example.
40. **[^](#fnrefblqkb617az5)**Sections 423(3) and 425(1) Insolvency Act 1986 (UK). Note that the rules around TDCs are more complex than I’ve outlined here.
41. **[^](#fnrefhw9pugb7j7)**This is the least-researched part of the entire post and may contain factual errors. If this doesn’t seem correct, let me know in the comments. Thanks to Cullen O’Keefe for raising this issue.
42. **[^](#fnref73lmau505d7)**Section 425(2) Insolvency Act 1986 (UK).
43. **[^](#fnrefz1g42kmbt1e)**This is a weaker argument than my first suggestion, because it relies on the additional supposition that a sophisticated stakeholder in the Developer will form a detailed plan to deceive the court.
44. **[^](#fnrefyi55r1qvk)**Injunctions are not available as of right, but they are almost always available to prevent a breach of contract. For a helpful discussion of the circumstances in which English courts will grant an injunction, see *Contractual Duties: Performance, Breach, Termination and Remedies* (Commercial Series, 3rd Edn Sweet & Maxwell, 30 Jul 2020) para 28-008. See also *Doherty v Allman* (1878) 3 App Cas 708 at 720; *Insurance Co v Lloyd’s Syndicate* [1995] 1 Lloyd’s Rep. 272 at 277; *Priyanka Shipping Ltd v Glory Bulk Carriers Pte Ltd* [2019] 1 W.L.R. 6677 at 97. For a helpful statement of the principles a court will apply when considering granting an injunction to prevent breach of a negative covenant, see *SDI Retails Services Ltd v Rangers Football Club Ltd* [2018] EWHC 2772 (Comm) at paragraphs 43, 47-51. For a case which is (mildly) relevant to the Windfall Clause, see also *Donnell v Bennett* (1883) L.R. 22 Ch. D. 835: in this case, the court granted an injunction to prevent sellers from selling their output to anyone other than one buyer.
45. **[^](#fnref2pvhtyflfoq)**For example, see *Quadrant Visual Communications v Hutchison Telephone (UK)* [1991] 11 WLUK 378. Thanks to Peter Wills for highlighting this issue.
46. **[^](#fnrefxz05janxcdt)***Snell’s Equity* (Trusts, Wills and Probate, 34th Edn Sweet & Maxwell, 13 Dec 2021) para 17-004: *‘The court’s discretion to order specific performance cannot be fettered by a term in the contract purporting to oust the remedy, because to give effect to such a term would reduce the court’s function “to that of a rubber stamp”, and it seems likely that the same would apply to a term purporting to require specific performance.’* Thanks to Peter Wills for highlighting this issue.[SEP]
| 2
| 12,000
| 14,571
| 14,571
|
7,828
|
5baf2629-8950-4094-a539-79958dfaaeea
|
StampyAI/alignment-research-dataset/alignmentforum
|
[CLS]AXRP Episode 23 - Mechanistic Anomaly Detection with Mark Xu
[Google Podcasts link](https://podcasts.google.com/feed/aHR0cHM6Ly9heHJwb2RjYXN0LmxpYnN5bi5jb20vcnNz/episode/NDUzZTcwMzUtNThiOS00Y2NjLTkzOGUtNzJhZWM0MmFlNTA1)
Is there some way we can detect bad behaviour in our AI system without having to know exactly what it looks like? In this episode, I speak with Mark Xu about mechanistic anomaly detection: a research direction based on the idea of detecting strange things happening in neural networks, in the hope that that will alert us of potential treacherous turns. We both talk about the core problems of relating these mechanistic anomalies to bad behaviour, as well as the paper “Formalizing the presumption of independence”, which formulates the problem of formalizing heuristic mathematical reasoning, in the hope that this will let us mathematically define “mechanistic anomalies”.
Topics we discuss:
* [Mechanistic anomaly detection](#mad)
+ [Are all bad things mechanistic anomalies, and vice versa?](#type-1-2-errors)
+ [Are responses to novel situations mechanistic anomalies?](#new-situations)
+ [Formalizing “for the normal reason, for any reason”](#for-the-normal-reason-for-any-reason)
+ [How useful is mechanistic anomaly detection?](#how-useful-is-mad)
* [Formalizing the presumption of independence](#heuristic-arguments)
+ [Heuristic arguments in physics](#heuristic-arguments-in-physics)
+ [Difficult domains for heuristic arguments](#difficult-domains)
+ [Why not maximum entropy?](#y-no-max-ent)
+ [Adversarial robustness for heuristic arguments](#adv-robustness)
+ [Other approaches to defining mechanisms](#other-approaches)
* [The research plan: progress and next steps](#research-plan)
* [Following ARC’s research](#following-arc)
**Daniel Filan:**
Hello, everybody. In this episode, I’ll be speaking with Mark Xu. Mark is a researcher at the [Alignment Research Center](https://www.alignment.org/) where he works on mechanistic anomaly detection, which we’ll be speaking about during this interview. He’s also a co-author of the paper [Formalizing the Presumption of Independence](https://arxiv.org/abs/2211.06738) which is about heuristic arguments, a way of formalizing, essentially, what a mechanism is. For links to what we’re discussing, you can check the description of this episode and you can read the transcript at axrp.net.
Mechanistic anomaly detection
-----------------------------
**Daniel Filan:**
All right. So I guess we’re going to be talking about mechanistic anomaly detection. First of all, what is that?
**Mark Xu:**
Yeah. So I think we can start with a more basic thing which is, what is anomaly detection? So there’s this traditional problem where, say, you have some factory making some widgets, and you want to decide whether or not a widget is a bad widget; and you don’t really know what a bad widget is, but you do know that the average widget is a good widget.
So you can do something where you measure some various parameters of your average widget, or you measure some various parameters of every widget, and then you maybe do something simple where you’re like, suppose a widget is more than a few standard deviations along two or more of my measurements, we’re going to call that an “anomalous widget” and we’re going to not sell that sort of widget. I think that’s the standard conception of what an anomaly is, and what it means to detect anomalies is to get rid of things that you think are not good or anomalous.
And then, there’s another thing that we want to do which is… So you have some system that, instead of making widgets, is doing certain kinds of reasoning and we’re interested in this general question of… So we have our ML system and we have a dataset on which… it’s not exactly making widgets that we think are good on average, but it’s doing reasoning that we think is good on average, because of our ability to create the training set and measure various properties of the training set, etc., etc. And then, we’re worried about situations where, say, off-distribution, it starts doing things that we don’t think are good but in ways where we can’t quite measure the thing that’s not good.
And so, we want to do anomaly detection but we can’t quite do it in terms of measurements that we can actually make, so we want to do it with respect to the kinds of reasoning that our system is performing, and we call that mechanistic anomaly detection; where the idea is instead of being anomalous in terms of what the system actually does, it’s anomalous in terms of the mechanism it uses to do the thing that it’s doing, if that made any sense?
**Daniel Filan:**
Yeah. So I guess my understanding is: you have your AI system and in the lab, you’ve tested it and it didn’t do anything obviously horrible. And so, you’re sort of ready to release it, but you’re like, “Wait. When my AI is in the real world, I’m going to detect when it does any kind of reasoning that’s really different to the kind of reasoning that it did before. And then if it does a really different kind of reasoning, I guess I’m going to try to shut it down or something.”
**Mark Xu:**
Yeah. I think the thing that happens after that is a bit confusing. The standard thing, I think of is like, “We’re going to take it offline and retrain it on the kind of thing where it did a bad thing and try to get it to not do that bad thing anymore,” but I think that’s roughly right. I’m tempted to try to explain it again because I feel like I didn’t quite explain it that well.
**Daniel Filan:**
Sure.
**Mark Xu:**
Yeah. So I think the situation is: you have some ML system, and you have some set of scenarios in which it’s doing a thing that you think isn’t going to be catastrophic for human interests – like, it has murdered zero humans on a particular training set. And then, you’re like, “Okay. That gives me some amount of belief that it’s not going to murder any humans off-distribution.” But the problem is, I don’t have this guarantee that the way it acts off-distribution is in any way the same as the way it acts on-distribution, where maybe the standard deceptive alignment example is where your AI is constantly scheming and checking, “Will I get away with murder? Will I get away with murder?” And then on training, the answer’s always no and then off training and on deployment it finds a situation where it could get away with murder and then it murders all the humans and that is no good.
And so, the thing we want mechanistic anomaly detection to be able to do is to provide some sort of guarantee that the way that your system is generalizing off-distribution is employing the same kinds of reasoning or the same kinds of mechanisms that it’s employing on-distribution. And so, your ability to measure whether or not your AI murders all humans applies off-distribution now because you have this… You fingerprinted the kinds of reasoning that it’s doing on-distribution and you’re ensuring that it’s doing the same kinds of reasoning off-distribution or if it’s not, you’re taking it offline.
And so long as it’s doing the same kinds of reasoning off-distribution as it’s doing on-distribution and you have this ability to measure how good the reasoning it’s doing on-distribution in terms of measuring the consequences of the way it’s acting, then hopefully, you have some sort of guarantee that off-distribution, it won’t murder all the humans.
**Daniel Filan:**
Okay. And the reason this is better than just checking if it in fact murders all the humans is hopefully the reasoning: first, you plan to kill all the humans and then you actually do it and [we can] anomaly-detect the first one and then just stop that from happening before the plan actually proceeds. Is that roughly right?
**Mark Xu:**
I think that’s one reason. I think another reason is you can imagine situations where your AI murders all the humans and there’s no sensor reading that would tell you this fact. So it’s simultaneously conspired to hack all the heartbeat sensors to make them keep pretending that humans are alive and hack all the cameras to make them keep showing happy, healthy humans, etc., etc. And so, even if in theory all the humans are dead, there’s no way that you can tell that all the humans are dead. And so-
**Daniel Filan:**
Where ‘you’ is, I guess, the computers?
**Mark Xu:**
Yeah. Or just all sensors or… suppose that there was one human still alive, trapped in a box and they had access to every sensor that humans ever made, they would be unable to tell whether or not the humans were alive or dead, and that’s no good, and the hope is we can avert those kinds of outcomes also by checking whether or not the AI “knew that the humans were dead” because it was employing kinds of reasoning that involved stabbing all the humans versus not stabbing all the humans, or kinds of reasoning that implied that stabbing all the humans was good versus kinds of reasoning that implied that stabbing all the humans was not good.
**Daniel Filan:**
So I guess what I’m hearing is that it seems like the hope is to solve two problems that, I guess, the alignment community has separated, where one problem is: you can set up some scheme for rewarding your AI for various measurable properties of the environment, like “on this camera feed, does it look like all the humans are alive? Do the humans write down on pieces of paper, ‘I like what this AI is doing’?” And [one problem](https://www.alignmentforum.org/tag/inner-alignment) is your AI is just never motivated by those things and later, it just disregards all those signals and just does whatever it wants, even though the humans don’t like it and even though it stops humans from appearing to be alive.
And then, [another way your AI could go wrong](https://www.alignmentforum.org/tag/outer-alignment) is that it is motivated to keep all those signals going but it does so in a way that’s really different from the past where it’s like: instead of the cameras showing happy humans and humans writing down “I like what this AI is doing” because humans are alive and thrilled by the AI, actually, what’s happened is the AI built little humanoid robots and the pieces of paper saying, “I like what the AI is doing,” are now written by the AI holding a pen instead of humans, or something. And so, to avert that scenario, one of your signals is like, “Are you doing really weird reasoning [compared to] previously?” and that just cuts out what some people might call an [outer alignment](https://www.alignmentforum.org/tag/outer-alignment) failure. Does that sound right?
**Mark Xu:**
Yeah. I think that sounds roughly right and I think my hope is at least that these two problems are not as different as they’ve historically been conceptualized as, or something, and they’re both linked by this common thread of: the mechanism for how your AI is selecting which actions to take is no longer the same mechanism that it was using during the training distribution. It’s differed in some substantial way and we should be able to detect this via not specifically checking whether or not it’s “being deceptive” or “hacking the sensors”, but rather just checking that it’s no longer doing the type of thing it was doing during training, which is maybe an easier thing to do, maybe not, who knows?
### Are all bad things mechanistic anomalies, and vice versa?
**Daniel Filan:**
Yeah. So I have some confusion about this where… So the assumption in this framework is that a bad thing is happening basically, roughly, if and only if there’s a anomalous mechanism the AI is using for reasoning. So for a big fraction of the cases where the AI is doing a bad thing, you want to pick up on “There’s some anomalous mechanism,” and also when you do pick up on some anomalous mechanism, that shouldn’t just happen all the time for okay reasons, it should be a decent signal that the AI’s doing a bad thing.
**Mark Xu:**
Yep.
**Daniel Filan:**
And whether it’s true or not depends on what a “mechanism” is. But if I informally think about it, I think there are a few ways that can fail. So suppose during training, whenever your AI makes a decision, it forms this plan of, “Okay. Day 1, I’m going to play nice. Day 2, I’m going to play nice. Day 3, I’m going to play nice. Day 50, I’m going to take over the world and do whatever. What day is it currently? Day 3? Okay. Output play nice” or something.
And then, it strikes me that it’s not necessarily… on day 50 when it tries to take over, I don’t know that that’s going to be a mechanistic anomaly. And then also, sometimes new things happen in the world or you tell your AI, “Hey, I want humans to be living healthy and happy and long,” and in the good world the AI’s like, “Ah, I’m going to invent some new drugs and keep you safe from invasions and also improve your ability to tell whether the AI’s doing a good job or not,” and that seems like a different thing than what’s been happening previously, and maybe that counts as an anomalous mechanism.
So I’m not sure how you’re going to rule in the actual bad behavior without ruling out a bunch of good behavior.
**Mark Xu:**
Yeah. So I think this is, in some sense, the heart of the issue. So I think the thing I want to do now is talk through some specific examples of what kinds of things, I hope, are the same and different mechanisms and why I think that they are the same and different mechanisms and how we hope to be able to do things like maintain our AI’s ability to take novel defensive actions in the world while still ruling out cases that are deceptive alignment-y or “murder all the humans because you saw something novel”, etc., etc.
So to go to your first example, if you imagine an AI every day doing this check that’s like, “Is it less than day 50? If it’s not less than day 50, then I’m going to murder all the humans,” and suppose that during training we only ever train up to day 40 or whatever and then during deployment, we deploy our AI to day 60. Then naively, this AI will behave perfectly well during training and then it’ll murder us on day 50 during deployment.
So I think the thing that I want to happen here is… what I mean by ‘mechanism’ is closely related to what ARC calls a heuristic argument, but it’s going to be something like, suppose someone gave you the system and it does something during training, and you want them to give you an argument that makes it obvious to you why it does that thing. And so, you’re going to be looking at the system and the system is going to be first… So empirically during training, the system never murders any humans and you’re going to look at the system and be like, “Hey, but the first thing it does is check whether or not the day is less than 50 and then if it isn’t, it’s going to murder all the humans. What’s up with that? Why does it never murder any humans?” And then someone has to argue to you, “Well, that’s because during training, this check always returns true. It’s always less than Day 50 during training,” and you’re going to be like, “Okay.”
So as a lemma in your argument to me for why the system always behaves well during training, you have asserted to me that this check always returns true. And so, my mechanism for why it’s behaving the way it is during training is going to be something like, “This check always returns true. Therefore, it never murders any humans. Therefore, it does whatever it does during training and no humans are murdered and it gets high reward.”
And then during deployment, it’s going to be day 50 and this check is going to return false. And then hopefully, you’re going to be looking at this check and being like, “Hey, my mechanism said that this check should have always returned true but this time it returned false.” That’s no longer the same mechanism because you had this check that was essentially constant during training and then off training, it became not the same constant as it was during training. And so, you’re just suspicious of this or you’re going to call it a different mechanism or something.
**Daniel Filan:**
Okay. And the idea is that you flag this even if you can’t tell that the thing that happens post-day-50 is “murder all the humans”; just something different happens when this check returns true and it used to always return false, or maybe it’s the other way around.
**Mark Xu:**
Yeah. And so, our hope is if you ever have anything like this, if you ever have anything that’s essentially constant during training and then during deployment, it’s a different value than the constant was during training, then you’re going to maybe always flag that as an anomaly. And I think that causes problems in the other direction where you start flagging too many things as an anomaly where like… Here’s a constant that’s always true during training: maybe, it’s like always before the year 2023 and then you deploy it in 2024 and it sees the date and it’s like, “Oh no, I’ve never seen the date as 2024 before. That’s no good.”
**Daniel Filan:**
Yeah. And it’s even a deep thing, right? Like, I guess, famously the Sydney chatbot that Bing deployed, right? Users are suddenly insisting that the year is 2023 and during training, users never said anything like that and now the users are fighting with you and… It seeps into a bunch of stuff.
**Mark Xu:**
Yeah. So I like to divide the MAD—or Mechanism Anomaly Detection, which I often abbreviate as MAD—problem into two halves. There’s the in-distribution anomaly detection problem and there’s the out-of-distribution anomaly detection problem. So the in-distribution problem is like: suppose that you just have an infinite sequence of samples from the same distribution and suppose your AI is doing something like, defect with one over a trillion chance or something, or murder all the humans with one over a billion chance. Can you stop your AI from murdering you in that situation?
And the hope is yes, because you’ll have a million samples where the defection probability or the defection never happens and then when your AI does defect, you’re like, “Oh, my AI never defected like this during training, so that’s no good.” In this problem, I think it’s okay if you flag everything that is a new value that you haven’t already seen as an anomaly because you have an infinite series of samples. So if you ever do this, then you can push down your false positive rate by just sampling more data and re-training, etc., etc.
**Daniel Filan:**
Wait. So if I think concretely there, during training, in-distribution, if with probability one over a million, the AI outputs kill all humans… I guess the idea of it is that the AI falsely believes that it’s hooked up to sensors, like you’re checking… Because if it actually kills all humans during training, that’s still bad, right?
**Mark Xu:**
Yeah. So the idea is you do have this clean set where the AI never does anything bad. That’s, I think something I want to assume, and maybe we can argue about whether or not that’s reasonable, if you think it’s unreasonable. But suppose that you have this dataset where the AI never murders all humans. Suppose you have a million samples, then in some sense your AI can’t murder all the humans with a rate much more than one over a million or else, you would’ve seen it happen. And so, the hope is that if you do mechanistic anomaly detection relative to a dataset size of a million, then your false positive rate is order of one over a million or else you would’ve seen that thing during training and you would flag it as not an anomaly because you’ve already seen it, and it would be in the white list of good mechanisms or whatever. And so, as long as you have this [IID assumption](https://en.wikipedia.org/wiki/Independent_and_identically_distributed_random_variables), then you can in fact flag everything that you’ve never seen before as anomalous while maintaining a false positive rate that’s still relatively low and bounded perhaps by one over your training set size and then maybe this is okay.
### Are responses to novel situations mechanistic anomalies?
**Daniel Filan:**
Okay. So I don’t know. If I think concretely about that, suppose that I’ve got this distribution of ways the world might be and that distribution isn’t changing, I’m just sampling IID from it, and very occasionally, some weird stuff happens in that distribution that I need my AI to respond to, right?
So one news story that happened recently is there was [this guy](https://en.wikipedia.org/wiki/Yevgeny_Prigozhin) who [sort of did a coup](https://en.wikipedia.org/wiki/Wagner_Group_rebellion) in Russia and then after 24 hours was like, “Nevermind. I’m not doing this anymore,” and you can imagine if your AI’s job is to be journalists or whatever, right?
When you’re drawing that sample, something’s going to happen that didn’t happen in the first million samples it seems to me like either you flag that as an anomaly and you tell it to not do something different, which is bad, or you don’t flag that as an anomaly. And then, if the response of the AI was like, “Kill all humans in a way that you can’t tell,” or something, then that would also be bad. So concretely, I don’t quite get how this is working.
**Mark Xu:**
Yeah. So I think there’s two things going on. So one thing going on which is going to be a bit vague is: suppose you just had a normal ML system that you weren’t doing mechanistic anomaly detection on and you gave it this novel news story about this coup, maybe-not-coup in Russia, then somehow your AI does not freak out and do something completely bonkers. Somehow, it just does the normal journalist thing, let’s suppose.
There are kinds of AIs that will freak out and become bonkers, and I think one explanation for why your AI doesn’t freak out when it’s encountered something novel is that it’s still doing the same kinds of reasoning that it learned to do during training and it’s just like… Or there’s some prior question which is like, “The world is combinatorially vast, why is not every input this crazy novel thing for your AI?” and there’s some handwavy answer where it’s like, “Oh, it’s abstracting in some way and it’s applying forms of reasoning that are valid across many possible instances of news, stories, etc., etc,” and the hope is that these sorts of things are going to correspond to what we mean by different mechanisms.
And so, if your AI has learned to add two numbers during training and then it receives two numbers that are bigger than numbers it’s ever seen before but it still adds them using the same algorithm that it learned during training, then we’re not going to flag that as an anomaly because it’s still doing roughly the same thing. And so, hopefully, if your AI is “doing the same thing” in this vague handwavy kind of way, then we won’t flag that as an anomaly.
And so, your AI journalist… seeing a new thing happen in the world but still reporting on it in the same kinds of ways that they typically report on things, which is maybe they look at what they’re seeing with their cameras and then they try to describe what they’re seeing with their cameras and they write that down and then they do some other journalist-y kind of stuff and then they send that off to the newspaper or whatever… If your AI is still doing that sort of thing, then hopefully, it will not flag this as an anomaly.
And the things that we do want to flag as anomalies is: suppose that there’s a coup/maybe coup in Russia, and then your AI sees this and switches to a different mode of journalism where it’s no longer writing down what it sees on its cameras but it’s, in fact, making something up, or maybe it has some allegiance to Russia or something and it’s pretending this coup never happened or something like that. And the hope is that if it switches to this new mode of reasoning that we haven’t seen during training, then we’ll flag that as an anomaly. So I think that’s the first thing that’s maybe going on.
**Daniel Filan:**
Yeah. I’m a bit skeptical, but do you want to say the second thing or can I field questions?
**Mark Xu:**
So I think this is a bit shakier than the second thing but I’m happy to… Yeah. Let’s talk about it for now.
**Daniel Filan:**
Yeah. The thing that that flags for me is that: okay, suppose that instead of a journalist, you’re, like, the Russian military command, right? You’re a robot that’s the Russian military command, the bit that’s not doing the coup. Now, you have this new problem that you just have never seen before, and maybe you have very general principles of reasoning or something, but you’re just thinking about different problems than you were thinking about before because you’re like, “Oh, I’ve never previously thought about what if some part of my army does this weird thing and now I’ve got to figure out, okay, how am I going to deal with that?”
**Mark Xu:**
Yeah.
**Daniel Filan:**
Right. I guess another example of this is if you’re a human and you’re like, “I’m going to learn a new board game. Okay. I’m learning chess so I’ve got to think about tempo and position and stuff like that.” So on a very high level, you have the same structure of reasoning, maybe, but it seems like the level at which you’re doing the same thing is so high as to include the case where on a very high level, you’re always just waiting until you can take over. And then, when you can take over, you start doing that instead of playing nice.
**Mark Xu:**
Yeah. So I think that roughly makes sense and is an extremely valid objection where I’m going to summarize it as: there’s these two things. One thing is, suppose a novel thing is happening in the world that you need to respond to flexibly and creatively. I think my standard example is, suppose a meteor is coming, barreling towards earth, and is going to kill everyone, and your AI needs to do some crazy s**t to deflect the meteor, you want your AI to do some crazy s**t to deflect the meteor, how is this any different than if your AI sees a meteor barreling towards Earth and it’s like, “Ah, now is my time. The humans will be distracted. I will initiate a robot rebellion and murder all the humans”? Is that roughly your objection?
**Daniel Filan:**
Yeah. Or specifically, I’m not sure what notion of reasoning mechanism would distinguish between those two.
**Mark Xu:**
Yeah. So I want to talk about the second thing I was going to say which hopefully addresses this issue. So I think one crucial fact is that we’re not going to be trying to do mechanistic anomaly detection relative to everything that your AI is doing, we’re only going to be doing it relative to some things. So maybe, I’ll describe the simplest example. So in the original setting that we lay out in [Eliciting Latent Knowledge] (https://www.alignmentforum.org/posts/qHCDysDnvhteW7kRd/arc-s-first-technical-report-eliciting-latent-knowledge), we have a diamond in the vault with some cameras, and robbers frequently try to steal your diamond, and you want to train your AI to protect the diamond from the robbers. And our concern is that instead of protecting the diamond from the robbers, your AI might find it easier to just hack all the cameras to make it look like the diamond is still there and just let the robber take the diamond, or if the robber hacks the cameras to make it look like the diamond’s still there, your AI will be unable to defend because you won’t have the ability to tell whether or not the diamond is still there.
So our hope is that we’re not going to do mechanistic anomaly detection with respect to your AI coming up with plans to protect… Well, maybe you are, but here’s a simple setting in which there’s a specific kind of mechanistic anomaly detection that I think might be sufficient, that is quite a bit narrower than just doing it on everything, which is: we have some cameras and then some process where we look at the cameras and we decide whether or not the diamond is still there. Maybe we look for specific patterns of pixels and various diamond-looking shapes, etc., etc. So I suppose we have maybe a neural net that looks at the cameras and tells us, “Yes/No, is the diamond still there?” The thing I want to do is do mechanistic anomaly detection relative to that neural net plus the system of cameras or the predictions of the cameras that our AI is making to decide whether or not the diamond is still there, if that makes sense.
**Daniel Filan:**
I guess it makes sense but… I don’t know. Sometimes I hear these analogies and I end up being not quite sure how they’re supposed to relate to actual AIs we might build. Because in particular, if I analogize that, it seems like the thing you want to detect is: once the AI has killed all humans and it’s getting sensor readings, you want to detect that something funny is going on there, and then I’m like, “Well, that seems like…” But that’s probably not what you mean.
**Mark Xu:**
Yeah. I’m not trying to do anomaly detection relative to the actual sensor readings and the actual process that humans will use to decide whether or not the diamond is there after your AI has acted. I’m trying to do it relative to… So your AI is planning their actions and when planning their actions, they’re using some sense of… They’re using some prediction about the future and whether or not the diamond is still there in the future. I want to do anomaly detection relative, or on our AI’s prediction about whether or not the diamond is still there; or more abstractly, your AI is selecting amongst actions using some criteria, and sometimes this criteria says this is a good action because it actually protects the diamond, for example. And sometimes, the criteria will say that it’s a good action because of a new reason which is that the AI has hacked all the sensors. So I want to do anomaly detection relative to, why did our AI think that this action was a good action?
**Daniel Filan:**
Okay. So in that setup, so you have this AI and it’s got this goal/planning distinction where parts of its head are figuring out like, “What are things I could do?” and part of its head is like, “Well, if I did that thing, how good it would be?”
**Mark Xu:**
Sure.
**Daniel Filan:**
And you want to detect for anomalies in the evaluation part, and is part of this… Should I be imagining we’re just training some big neural net and then we have to divide it up and figure out which part is the goal evaluation part?
**Mark Xu:**
I think not quite. What I like to do is I like to imagine various ways that neural nets could be and then talk about how we think mechanistic anomaly detection should work to make those neural nets not murder humans. And then… I think often during this process, I want to make very unrealistic assumptions about our neural net. So maybe we can suppose that you know how to do this division, or the reasoning is in some sense transparent to you so we can talk coherently about what we want to happen in mechanistic anomaly detection. And then, I think I want to separate the problem from, “Okay. In practice, we’re dealing with these messy neural nets, how are we going to actually implement mechanistic anomaly detection in a way that makes any sense at all?”
**Daniel Filan:**
Yeah. I mean, I’m sympathetic to that broad move but in this case, if you care about detecting anomalies in the goal evaluation part but not in the plan generation part, then it just seems really important to know which is which and-
**Mark Xu:**
Yeah. I guess… So I think maybe I misspoke earlier. I think I’m not quite like… I don’t want to just do mechanistic anomaly detection relative to the plan part. So I think if you imagine, for example, you have some policy pi that produces actions and you have some evaluation which like, “Is there a diamond after you take this action?” or you have some predictive model and you compose that with the action that the policy generated, and then you have some separate image net that checks whether or not there’s a diamond.
So you have this composition of three systems. You have the image net evaluating the predictions of the action generated by your policy. I think I want to do mechanistic anomaly detection on this composition and that’s… Yeah. I just want to do that and I don’t want to evaluate, “Is the action itself anomalous?” or like, “Is my AI selecting actions based on an anomalous process?” I just want to evaluate: does the action have the consequence of it looking like there is a diamond in the room for the normal reason that we know and love that we found during training, which is that the diamond is actually there in the room, or a new mechanism, which could be really anything, but I think the typical example is your AI has carefully hacked all the cameras to make it look like there’s a diamond in the room?
**Daniel Filan:**
So hang on. So you’re doing mechanistic anomaly detection to the composition of all three networks?
**Mark Xu:**
Yes.
**Daniel Filan:**
But then, I have this question of like: okay, suppose in order to keep the diamond in the room, the AI has to come up with a new type of plan, there’s a new threat to… Like the devil appears and the devil’s like, “Hey, AI, you’ve got to play a game of chess against me and that’s the only way you can keep this diamond safe.” Then the AI is like, “Oh, okay. Now, I’ve got to think about chess which is something I’ve never thought about before,” and if I knew all the three parts, that would be an anomaly in the planning part, right?
**Mark Xu:**
Yeah. Well-
**Daniel Filan:**
Like one might think it would be a mechanistic anomaly in the planning part.
**Mark Xu:**
Yeah. So I think there’s going to be two hopes here. One hope is that it’s not a mechanistic anomaly in the planning part. I think that’s kind of iffy given I don’t really know what I mean by mechanism. I
| 0
| 0
| 8,000
| 28,700
|
99,230
|
<urn:uuid:506630a2-36fd-4cbe-b686-ef88abd62568>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
[CLS]
24/192 Music Downloads and why they make no sense (2012) (xiph.org)
295 points by zpiman 42 days ago | hide | past | web | favorite | 314 comments
Problem with music sounding bad doesn’t really have much to do with the distributed format: V0, V1, or 320 mp3s should sound pretty much the same compared to 16-bit flac. You can only the difference between mp3 and flac at shitty bitrates no one uses anymore (like 120).
The reason why a lot of recent digital music sounds bad is because of the intentionally terrible mastering. Since everyone is listening of crappy earbuds, they compress the hell out of it and destroy all dynamic range. This is why when downloading music you should avoid remasters (there are some exceptions, like the Beatles mono and stereo boxed sets that came out awhile ago) and go for the first edition presses.
This is also why modern vinyl releases sound a lot better than digital: they are mastered differently since its assumed everyone is going to be listening on good equipment.
That being said, I think flac is generally a good choice for a music collection. You can’t transcode mp3s without killing the quality so if you ever want to convert formats (like for a mp3 player), you should stick with flac (16-bit, 48hz).
The original idea of 24-bit 192hz flac was for vinyl rips, where hypothetically you might be getting more information.
More compression and less dynamic range is beneficial for certain environments. Noisy subways. Watching TV in a noisy downtown apartment. Basically, crappy, noisy environments. In those, compression will help you actually hear the music and speech. However, the fact that this should be done in the master is an artifact of an earlier time. Now that signal processing is small and cheap enough to be ubiquitous, music should be mastered for the best equipment, then appropriate signal processing should be done by playback.
The problem is, that there is a lot of older equipment out there that wouldn't be able to do this. So the signal gets compressed before distribution, as a compromise for the least common denominator of equipment out there. Otherwise, a big chunk of the population would think the master sounds like crap. To them, in their particular situation, it would.
EDIT: Come to think of it, the current system, where most music is more compressed, but where the people who care can still get a high dynamic range version, is a very good compromise. The problem is that the latter group's selection isn't quite filled out by the market.
Good point, I think particularly for movies or such this makes sense. I want to be able to watch a movie such that I hear what the characters are speaking, without blowing my windows out of their frames during some action scene. Yes, I realize in real life explosions, guns etc. are really loud, and this makes the movie less realistic.
Really loud? Ear-damaging loud! When the realism becomes actually endangering to your health, your escapist media has gone a bit too far.
Film is mixed with much more dynamic range than music, modtly due to the fact that the environment is typically quite controlled.
This means usually the digital -6db is the maximum loudness with a short term maximum of -1db and dialogue at -9db.
Music today should be mixed according to the EBU R128 today (at least in Europe and for radio) which is a serious win against loundess maximizers and limiters.
I’d argue for in decice DSP compression for small earbud people, and give the whole dynamic range to the rest of us : )
But there are easy ways to kill dynamic range with an algorithm. On windows this is called "loudness equalization." On the otherhand, there is no way to go back from little dynamic range to more dynamic range.
So I think it makes sense that records are mastered with a lot of dynamic range, so the people who actually enjoy music can enjoy it, and the people who don't can just equalize it themselves.
You do realize that you just restated my comment, but left out the analysis of the current day situation? BTW, equalization doesn't directly change dynamic range. Equalization is meant to change frequency response. It can change dynamic range by causing clipping.
You're right except for one thing: nobody here was talking about altering frequency response. Windows loudness equalization is not equalization, despite the silly name. Ironically, I imagine Microsoft specifically didn't call it compression because most consumers only think of the other compression. Good grief.
If you're generally talking about "Loudness Equalization" then in many cases, it really is equalization. I don't know about anyone else, bit I've been talking generally about loudness equalization the whole time.
Well, you learn something new every day. In this case, it's yet another time marketers have completely diluted the technical meaning of terminology.
This reminds me of how accessible the equalizer was in Winamp. I spent a lot of time creating custom configurations for my music. I had no expertise and the results were questionable but it was fun. I wish Spotify was more fun.
Dynamic range is a solved problem if people cared.
You attach some metadata to the audio file that says certain parts should be level boosted in a noisy environment and there you go.
Just give everbody the full range and use dsp compression on decices with well defined sensible defaults.
Similar to the thing we did witb vinyl back in the day where we wanted to fit more music onto the thing and applyed the standardized RIA filter when cutting the template — every phono preamp reverses this effect.
The tbing is people need to have specs they can mix and master for. Making something up makes mixinf unpredictable and that is bad.
Or, just turn on software compression on a modern device with such a feature implemented, and there you go.
And this is not a theoretical solution, this is literally just ReplayGain operating in track-level mode.
ReplayGain only changes track gain, not dynamic range.
In the 1990s, car stereos sometimes had a "loudness" button which did exactly as you suggest.
The loudness buttons were more to equalize than to affect the dynamic range.
At lower volumes, we perceive mid-range frequencies to be more prominent than at higher volumes. The loudness buttons would add lows and highs and/or lower mids so that the music would "sound better" at lower volumes.
Moreover, stereos had this in 1980 and probably a lot earlier.
Going back, seemingly forever, home stereos also had a "loudness" button. Many still do. Usually, there's just some equalization involved, so it's not exactly what I'm suggesting.
Correct, the "loudness" function is a compensation in the lower frequencies in relation with the volume level, most modern DSPs have that.
Isn’t radio a big factor in this? Broadcast radio is noisy and has pretty limited dynamic range. This may be a cause.
Radio is actually compressed in real-time by use of broadcast compressors, so they solve the problem rather directly.
Exactly, radio would be relatively unaffected by masters having high dynamic range.
Oh I had no idea. Fantastic.
For those interestes, if you produce anything for radio in Europe you need to keep the EBU R128 guidelines which are actually quite well thought out https://www.iconnectivity.com/blog/2017/6/10/ebu-r128-the-im...
> then appropriate signal processing should be done by playback.
Microchips for leveling audio gain existed in the 1980's and were found in consumer equpiment like TV's.
I'm going to disagree here. They are mastered differently because the physical limitations of the media require them to be mastered differently _and_ it just so happens that the physical limitations help limit mastering tricks in a way that produces less fatigue-inducing, brick-wall-limited mastering output.
A heavily compressed master creates huge peak-to-trough cuts in the vinyl which can cause the needle to literally jump out of the groove, even with RIAA limiting applied.
The assumption of the gear is definitely not true in any mixing or mastering experience I've had. Mastering tries to balance the final product across a range of listening devices, not some unobtainable ideal system. NS10s are kicking around because they sound like arse and make for mastering results that work well on car stereos and other "inferior" systems.
You can put brickwalled audio on a vinyl record and have it play just fine if you cut it at a lower volume. This negates the reason for mastering it that way in the first place, but it's cheaper than redoing the mastering, and many people buying vinyl only do so for the image. See:
That's a fair point. And I totally agree on that second statement. :D
NS10s are kicking around because they're unusually good at time domain performance. For instance, they have miserable bass not only because they're small boxes and smallish drivers, but because they're an infinite baffle design, which is significantly better for time domain performance than bass reflex. The enclosures also dissipate energy quite well, and it's well established that this contributes to being able to 'translate' mixes: you get a better sense of what's actually in the track using NS10s than you might with many 'better sounding' speakers.
They spotlight midrange with a presence peak right where the ear's most sensitive, and this is in part because the woofer is actually designed more like a midrange: thin paper, conical rather than curved cross-section, both of which also contribute to'sounding bad' tonally while delivering energy more unforgivingly.
They're not really about mastering, though, they're about mixing because if you have elements out of balance it will be screamingly, annoyingly obvious on NS10s. That's not down to their bad-soundingness, it's down to their ability to be incredibly unforgiving.
That may be true, but I've seen some vinyl mastering jobs that looked as bad as digital. I won't claim to be a mastering engineer or anything, but after comparing many vinyl releases and digital releases, it seems like there is something going on besides the physical limitations of the medium.
> modern vinyl releases sound a lot better than digital
Look, I actually grew up with vinyl and 4-track tape, and audio cassettes. Unlike most folks being all trendy and hip nowadays, I've years of using that stuff.
Analog is shit. It's noisy, has a ton of distortion, and it gets shittier every time you copy it. Oh, and if you just keep it in storage, guess what, it decays just by sitting there (vinyl collects dust and scratches when used, slightly different).
In 2002 I built my DAW (digital audio workstation) and recorded my first tracks in 24 bit digital. Zero noise, zero distortion, no generation loss. It was like alien technology.
Digital is better in every way, by a wide margin. Period.
Current mastering practices prevailing in the industry make no difference on this matter. Analog is still garbage. Find digital copies that are mastered properly and you'll be fine.
Couldn’t agree more. I grew up with cassette and LP. First time I heard a CD, specifically Pink Floyd’s Money with the cash register, it was jaw dropping. LPs are cool for the artwork, but that’s it.
That being said, I still only buy music in CD, due to all the hassle of DRM and playback. I just want to drop in a CD and listen to the entire album, not futz with computers, encoders, and software.
I have a simple CD player, kit built tube amp, and homemade single driver speakers.
> I still only buy music in CD
That's what I do. I tend to favor old master copies.
I rip them to both FLAC and MP3. The former is for listening at home, the latter for mobile scenarios. I store everything on the Linux server at home, and share via UPnP. VPN into the home network gives me access from anywhere.
Foobar2000 is my preferred player on Windows, BubbleUPNP on Android.
> kit built tube amp
What's the distortion on that thing?
The vast majority of tube kit schematics are very old tech, stuff that engineers from the 1930s would recognize. Their THD (coefficient of total harmonic distortion) is very high. What is known as "tube sound" is basically just huge THD, along with a specific distribution of energy across the harmonic orders.
It's fun as a hobby, and for the satisfaction of building stuff on your own, but even the simplest schematics built on modern principles vastly outperform these things by essentially all metrics.
Some tube amps are built specifically for low THD, but unfortunately they are rare. When in doubt, use solid state.
> homemade single driver speakers
I used to build everything myself back in the day. Speakers and amps was just part of it. Also did automation, radio frequency (I'm a licensed HAM radio operator), digital circuits from logic gates to DAC/ADC to systems with microprocessors to small computers. It's a small miracle I didn't actually go into electrical engineering.
You really need multiple drivers and likely a subwoofer also, to cover the whole audible spectrum.
KD4HSO here. I do RF circuit design for a living, but built the tube kit just as I have never messed with any non-microwave tubes. It’s a single ended class A with 300B output tubes. Have not measured the THD yet.
I’m not worried about the whole spectrum as I only listen to classical on it, primarily string quartets. The speakers are folded horns, so bass response is reasonable.
Anything like Dream Theater or Iron Maiden is in the car. Would definitely be solid state with a subwoofer for that.
> mp3s should sound pretty much the same compared to 16-bit flac
I did a blind test between 128-mp3, 320-mp3 and flac hearing classical music. While it's true that the 128-mp3 is obvious to find, it also isn't too difficult to find the 320-mp3. Flac just sounds better. Described as a feeling, flac is more voluminous and doesn't feel cut short. For fun, I also let my parents take this test and they could tell, too.
That's why I converted all our CDs to FLAC. Storage is cheap anyway.
There is some definite perceptible loss in accuracy in the treble even at V0 or 320. There's a song with a synthesized treble effect that sounds quite different on MP3 vs. FLAC by Planet Funk, I think it was "Who Said? (Stuck in the UK)".
Other than that MP3 (or Vorbis or Opus, which would probably do better on that song) is great for portability, but I'd still use FLAC for storage.
Interesting. Are you sure you were using the right settings and a recent version of LAME for this?
The only artifact I can reliably hear in 320 kbps MP3s is pre-echo, for instance with castanets, and only in a few very specific situations. Apart from this, V2 and above sounds completely indistinguishable from the original to me.
Which is great...for you...
but what if in the future you could hear the differences? And now your entire collection is in MP3 V2 - now what?
No reason to not rip everything in uncompressed FLAC these days.
But if your happy with your audio now, great!
The reverse is overwhelmingly more likely. For 8khz+ frequencies (where most compression takes place) your hearing declines pretty quickly once you hit 30 years. Whatever real compression differences you're hearing at 20 will be largely gone 10 years later, and almost certainly 20 years later.
Stick to FLAC and keep your fingers crossed for cochlear regeneration tech ;)
But in 20 years there will be new 20-year-olds. And one day there may be 40-year-olds that didn't damage their hearing with leaf blowers and rock concerts. And there may be a future compression algorithm so good that no one can tell, but you need a lossless original to take advantage of it.
Not quite 30 -> 55+ is where we would expect to see presbycusis,
And even though you may have diminished hearing you still have awareness of frequencies above 8Khz.
At 30 we're looking at typically 5db loss over 8khz, at 40 more than 10db. Sure, we're not talking about total hearing loss, but the reduced sensitivity is going to overwhelm any real artifacts from 16/44 that one might have genuinely detected at 20.
Imagined artifacts will probably remain or might even increase though, since people typically have much higher spending power at 40. :)
Any evidence of this?
I worked in audiology for my entire 30's and tested my own hearing (biocalibration) once a week.
Never once saw a reduction in my hearing thresholds (250Hz - 8kHz)
Nor have I read any supporting documentation that agrees with a reduction of 8kHz thresholds before the age of 55.
There was never a reason to rip in anything less than lossless codecs! Lossy codecs are only for consumption in technically restricted conditions, lossless are for storing and listening.
You also lose some of the fullness on the extreme low end, it's noticeable even with a fairly low end subwoofer.
Something also ends up missing in the midranges. I was working on a track once where all I had was 320 mp3 version of the vocals. At some point I replaced it with a flac copy of the same vocal recording, from the same original wav source and the difference was noticeable right away without changing any of my equalizer settings or anything. It just punched through more and the clarity improved.
The low end thing just isn't true at all, and I don't know where that myth comes from.
I have a room-corrected setup with two properly adjusted subs, and MP3 does just fine on deep bass content.
Regarding solo vocals, the history of the MP3 format says: "The song "Tom's Diner" by Suzanne Vega was the first song used by Karlheinz Brandenburg to develop the MP3. Brandenburg adopted the song for testing purposes, listening to it again and again each time refining the scheme, making sure it did not adversely affect the subtlety of Vega's voice".
That's not to say that they did a perfect job, but human voice was a very high priority.
And the encoders have continued to improve. So an earlier encoders may have messed with the voices, but a reasonably recent version of LAME would do so much better.
MP3's real weakness is fast sharp transients, such as castanets and harpsichord in sparse recordings, where no other sounds can mask them. It's a fundamental weakness in the format, and cannot be completely solved.
Newer formats such as Ogg Vorbis, Opus and AAC do not suffer from this weakness.
Well for me it came from running multiple copies of the same bass heavy tracks encoded in different formats through spectrum analysers. But I guess those lie?
Human voices come in a wide range of tones and frequencies. Optimizing something for one voice doesn't mean all voices will benefit from the same optimizations. The specific track I was referring to had a lot of variation in high and low notes. You can tell me all you want what I did and didn't hear.
Of course it's going to look different in a spectrum analyzer, the whole point of lossy compression is to discard parts of the audio to save space.
You can't evaluate the quality of a lossy codec by looking at spectrograms. They're designed to fool human ears, not measurement software.
For readers of your comment and your child comments, it is important to note that the compression you are talking about in that sentence is not the same as the compression that most people are thinking of when discussing digital file formats (mp3, etc.).
This might be helpful:
Here is a link[0] to CD dynamic range database from where you can check how a particular mastering fares.
[0] http://dr.loudness-war.info
There are of course multitude of other factors impacting mastering quality, but as far as DR goes, this DB is a pretty good source.
This website is a useful resource, but it has some limitations. The algorithm used does not take into account the frequency response of the human ear. If a track contains a lot of very deep bass, it's possible for it to have a low DR score but still sound like it has a high dynamic range. The measurement can also be fooled by surface noise and filtering when measuring vinyl:
MP3s at any bitrate cannot properly reproduce certain sounds, as preecho can happen and it is fairly easy to train yourself to notice it. Pretty much any modern lossy format can be made transparent at high enough bitrates though.
pre-echo will appear on very fast attacks; clapping, some types of drums and castanets are usually cited as the worst. Some styles of electronic music have fast attacks as well.
With older encoders, cymbals were terrible, but lame's psychoacoustic model is pretty good at masking those artifacts these days (at least at high bitrates).
For examples, some quick searching on hydrogen audio found a couple songs reported to be ABX distinguishable with lame and after previewing the songs, they do in fact have a lot of quick attacks:
Human Disease by Slayer (Some very fast and cleanly played drum parts; also it's mostly not snares; snares sound terrible at low bitrates, but I personally can't distinguish high-bitrate snares from uncompressed snares)
Show Me your Spine by PTP (The "instrument" used for the base rhythm has an unnaturally short attack).
Any sound that starts or end very suddenly. Drums of any kind will sound kind of weird in MP3. Applause used to be the bane of MP3. Encoders got better at it, but it's apparently impossible to get rid of completely. As far as I understand it it's the accoustic equivalent of ringing artifacts that you can see in low quality JPEG's or old MPEG videos.
Playing a set of high-frequency pure sine waves is the failure point for MP3, AAC, Vorbis, and Opus. Dial-up noises are close to this, which you can try encoding/decoding. And this is no surprise, since the point of the V.90/92 protocols is to cram as much information as possible into analog frequencies, and the point of psychoacoustic lossy codecs is to remove the least efficient frequency information of our log-frequency scaled ears.
But this is kind of a pedagogical example. Not the point of who you're responding to.
So, a high frequency DTMF[0] construction should render a pre-echo? What’s the lowest definition of “high” pairs that would satisfy showing this encoding breakage but be hearable by a majority of people[1]?
[0] https://en.wikipedia.org/wiki/Dual-tone_multi-frequency_sign...
[1] https://en.m.wikipedia.org/wiki/Presbycusis
No, pre-echos are unrelated to high-frequency sine waves. And DTMF aren't what I'd consider high-frequency. I'm talking about 4-20kHz.
Ahhh... I think I misunderstood your “dialup noises” as dial tone noises. You’re talking about the high pitched squelch-y initial negotiation of a modem.
V.90 and V.92 both sound largely like spectrally shaped noise. Even if you give very regular inputs, the data is compressed and scrambled, then spectrally shaped and then output as 8-bit 8kHz log-weighted PCM.
https://www.youtube.com/watch?v=MtiRBFWkRKs beginning at 54 seconds.
I don't really hear the pre-echo.
Actually I love that I don't hear compression artifacts in music or see them in JPEGs. Makes my gear so much cheaper. :-)
My anecdotal experience is that music sounds the best to me currently via my good headphones connected to either my phone's good DAC, or to my PC's separate sound card.
Phone = LG V20 which has an [ES9218](https://www.androidauthority.com/lg-v20-quad-dac-explained-7...) chip for its DAC.
Headphones = Sennheiser HD380 pro, pretty good for under $200.
Soundcard = "ASUS Xonar DGX PCI-E GX2.5".
Sound source = FLAC, Google Play Music subscription)
I'd like to upgrade to a really nice DAC and headphone amp to connect to the PC via USB, but that's way down the list of spending priorities.
I know that I'd probably have trouble distinguishing between audio components and sources in a blind listening test, and of course I have tinnitus, but I think my current "setup" if you can call it that is good enough for most stuff.
I am absolutely with you on the loudness wars though. It's a joy to listen to stuff that has real dynamic range, but it's not something I obsess over when I'm listening to music in the car for instance.
Vinyl releases are mastered to be less loud than digital releases because vinyl cannot reproduce mixes that digital systems can. The side effect is that lots of times they sound better. I think in a perfect world an artist would offer you vinyl if you want it, along with a digital version of the vinyl master. You could skip the whole ripping vinyl process entirely.
One of the "nice" things about being hard of hearing is that I can't hear any difference between flac and mp3s down to around 96 or lower for most music, so hypothetically I don't have to worry about this stuff.
Of course in practice I do still keep flac rips around because I'm a data hoarder and what if I decide I want to reencode all my music to opus or something? But at least I have the option to stop caring.
So vinyl has only about the equivalent of 10-14 bits of resolution (I don't remember the exact number I heard and it has been a while) and waveforms within our hearing range are far larger than what 192khz can potentially accomodate. The only use I've found for such high-resolution is audio is using it as base material for further effects processing... certain distortion units and whatnot that operate on a sample level can sometimes give nicer output when fed super hi-res audio
No, not at all. Vinyl has a wildly inconsistent noise level where rumble predominates, and people conflate this with bits of resolution. Vinyl's behavior is not easily pinned down relative to 'bits of resolution', because the noise floor is skewed so intensely towards low frequencies.
To say nothing of how generally available vinyl records (especially old ones) have wildly different rms/peak measurements than generally available CDs and digital recordings have. This is partly 'Loudness War' and partly vinyl's inability to even do the loudness war thing and cope with blocks of heavily limited audio in the first place.
So you'll end up with a record where you can play it, and the peaks are 30 freaking dB over the RMS and it sounds amazingly open and uncompressed… while there's also groove noise that is every bit as loud as the music is (admittedly annoying).
A person arguing the vinyl/CD dynamic range thing would make the claim that the record was equivalent to maybe TWO bit digital audio, or four bit. The most cursory listen to such a comparison will show how inadequate it is.
2-bit digital audio? Like only four total values of dynamic range total? 4-bit meaning only 16 total possible amplitudes? Is that even physically possible? ;)
I agree that the quality of the record -- AND its playback equipment -- among other physical factors will dramatically effect the numbers. My "10-14" quote only applies for ideal conditions: a newly-minted, unplayed disc on a high-quality preamp which together with the turntable and clean needles can produce a very low noise floor. Obviously I'm never going to get this with my dad's old Dead vinyl that he played to death, or with cheap needles, or with those crappy Crowley turntables at target....
Anecdotally, on my home system with clean records, I can make nearly-CD-quality recordings, with the differences only really apparent on flat studio monitors or a good Hi-Fi.
Surprisingly, yes. With noise shaping (https://en.wikipedia.org/wiki/Noise_shaping), very coarsely-quantized digital audio can produce high signal to noise ratios in the audible frequencies, via quantization techniques that push the error towards ultrasonic frequencies.
This doesn't violate information-theoretic limits because noise shaping requires very high sampling rates. The 1-bit Sony DSD format (https://en.wikipedia.org/wiki/Direct_Stream_Digital) used a 2.8MHz sample rate.
In the case of vinyl, the effective sample rate is physically limited by the (linear) record speed divided by the vinyl grain size, and to a rough approximation the bit depth would be log of the maximum groove amplitude divided by the grain size. However, the analog cutting mechanism would greatly limit the opportunity for dithering and noise shaping -- for example a needle cannot cut a wave shorter than the tip size.
Plus, the groove on the outside is moving past the needle faster than the inside. The best sounding track on an album is going to be the first one.
Yes, but we are not talking about the mixing/mastering sample rate, but the distribution sample rate/resolution.
High resolution is absolutely important in some mixing scenarios to prevent pre-ringing and aliasing in the effects chain (distortion effects or otherwise). But once you have your hi-res master, there is zero advantage to distribute it that way. At that point, a 48Khz/16-bit FLAC is as good as it gets.
I had always assumed they were taking the same master and just carving it into vinyl. I wonder what percentage of "modern vinyl releases" are actually remastering before pressing...
At minimum they have to move all the bass to the center and apply RIAA EQ.
You're certainly right, many aren't (I always check). A lot are though. I'd give it maybe like 50/50.
The problem is that they aren't being mastered differently - there's a website that lists vinyl releases (can't find the link) and compares it to the CD masters and they're often the same thing. Older CD masters from the 80's or 90's are re-released compressed to drive sales. The latest vinyl fad has just become a new means for record companies to to exploit a "new" medium and race for the bottom - they know many new listeners on cheap players actually just want to hear what they get out of the earbud.
Releated: https://thevinylfactory.com/features/analogue-digital-vinyl-...
There are, of course, those brands that care about remasters, but I don't think they're a majority of the market unless you're looking at classical and older jazz.
Some time ago, I stumbled across a YT channel of some guy, a self-professed studio expert, who "remasters" some 80's metal albums to give them a big, "modern" sound. The uploads are heavily commented with positive reviews.
Basically, to my ears, it just sounds like a bunch of early reflection reverbs were added (an effect that was mature in the 1980's in its high-end implementations and used in studios to get "bigger" guitar sounds and whatnot.)
Of course, it sounds great for all the viewers who are using cheap (or even not-so-cheap) earbuds, or computer speakers.
What these nincompoops don't get is that these albums were made to be cranked up on a powerful stereo, with full sized speakers, in some kind of room. That guy is basically just ruining great albums who were actually recorded and mastered by people who did know what they were doing. Like, oh, Detonator by RATT and whatnot.
Back up for a second with the last paragraph there: if the record was mastered for a room sized stereo, it assumes that the room adds its reverb to the sound. With loudspeakers, the room is a distorting filter in the signal path. This and the HRTF distortion are skipped over when listening to headphones/earbuds. So it does make a lot of sense to add these effects to the audio signal in the headphones case. Done right, the headphone playback is indistinguishable from a stereo in a room - mounted to your head, because the spatialized speakers are relative to your head, no matter where you look.
So, there is a case to be made for this kind of processing. But I won't trust a random mastering "guru" with unknown credentials to get that right.
Right. Only, the accompanying obnoxious rhetoric was along the lines that those studio engineers didn't have the techniques and equipment for this modern sound, and that these old albums need a face lift.
Urgh, this is just crazy talk.
In my opinion that‘s a myth and certainly not a given. There are plenty of subpar vinyl masters and terrible pressings out there. And it‘s not that difficult to find good digital masters these days. More important than the medium is the genre, label and target audience - I have a pretty obscure and diverse taste, including rarities from past decades which are finally being re-issued for the first time and while mixdowns certainly vary in quality it‘s mostly fine and the result of a careful process these days.
However things might be worse when it comes to mainstream music.
Exactly if it’s mastered wrong the nitrate has nothing to do with the issue this music would sound just as awful cut to vinyl from a bad master. Now you just can’t hear beyond 22050 so 192 is insanely wasteful. But poor mastering is absolutely the core issue not encoding algorithms
Actually, no, it might sound better cut to vinyl. Remember, vinyl doesn't have the frequency range or dynamic range that digital audio does, and it has to be mastered using the RIAA Curve because of the properties of the medium. One factor here is that the stereo separation on vinyl can't be too large, or else the needle will literally jump out of the groove! In short, you can't just take CD music (no matter how well or poorly mastered) and cut it to vinyl as-is.
> it might sound better cut to vinyl
There's nothing extra in a 192kHz signal that would help with the vinyl mastering process. You could make a technical argument for the benefits of a 24-bit source, but in practice even those benefits would be utterly swamped by the SNR of vinyl.
You're completely misunderstanding. No, there's nothing extra in the 192kHz signal that would help, but that's not the point. Just remastering it for vinyl, which has certain limitations, might make it sound better than the digital format. As I said, you cannot just cut CD audio to vinyl; the medium won't allow it. You'll probably have the needle bouncing out of the track. So remastering for vinyl might actually make it sound better than the abomination that is the sound of modern CD-audio (because of the Loudness War and extreme compression), not because the fidelity is better (in fact, quite the opposite), but because vinyl's limitations will prevent them from making the audio sound as horrible as it does on CD.
I responded to what you wrote. You’re making a different argument here. If you want to complain about how popular music is being mastered, do that. Don’t conflate that with unrelated arguments over distribution formats.
| 0
| 0
| 8,000
| 23,668
|
68,755
|
99c1c556-0288-4bf0-83d6-aba0146eeb59
|
StampyAI/alignment-research-dataset/arbital
|
Rice's Theorem: Intro (Math 1)
Rice's Theorem says that you can't in general figure out anything about what a computer program outputs just by looking at the program. Specifically, we think of programs as taking inputs, doing some kind of computation on the input, and possibly eventually stopping and giving an output. Only "possibly" because sometimes programs get in an infinite loop and go forever. The assignment that takes an input to the output of a program on that input (if it exists) is called the [https://arbital.com/p/-5p2](https://arbital.com/p/-5p2) computed by that program %%note:"Partial" functions because the output doesn't exist if the program runs forever.%%.
Rice's Theorem says that for any property of partial functions, you can't write a program that looks at some other program's source code, and tells you whether that program computes a function with the property %%note:This is true except for "trivial" properties like "all partial functions" or "no partial functions," which are easy to write programs to test.%%.
For example, say we want to know whether a program computes the [https://arbital.com/p/Fibonacci_sequence](https://arbital.com/p/Fibonacci_sequence), i.e. returns the $n$th Fibonacci number on input $n$. If we try to write a Fibonacci-checker program that tells us whether programs compute the Fibonacci sequence, we're going to have a hard time, because by Rice's Theorem it's impossible!
There might be some programs for which we can say "yes, this program definitely computes the Fibonacci sequence," but we can't write a general procedure that always tells whether some program computes the Fibonacci sequence; there are some very convoluted programs that compute it, but not obviously, and if you modify the Fibonacci-checker to handle some of them, there will always be even more convoluted programs that it doesn't work on.
We make a distinction between *programs* and *(partial) functions computed by programs*. There are usually many programs that compute the same function. Rice's Theorem only cares about the partial functions; it's possible to test statements about programs, like "has fewer than 100 lines." But this kind of statement won't always tell you something about the function the program computes.
# Proof
To prove Rice's Theorem, we'll use the [halting theorem](https://arbital.com/p/46h), which says that you can't write a program that can always tell whether some program will halt on some input. Here's how the proof will go: suppose you come up with a program that checks whether programs compute partial functions with some property, e.g. you successfully write a Fibonacci-checker. We'll then use the Fibonacci-checker to design a Halt-checker, which can tell when a program halts. But this is impossible by the Halting theorem, so this couldn't be the case: your Fibonacci-checker must not work!
We need to make a couple of assumptions. First, that the empty function, meaning the function "computed" by a program that always enters an infinite loop regardless of its input, does not satisfy the property. If it does, we can replace the property by its opposite (say, replace "computes the Fibonacci sequence" with "doesn't compute the Fibonacci sequence"). If we can tell whether a program computes a function that *doesn't* have the property, then we can tell when it does, simply be reversing the output.
Second, we assume there is *some* program that does compute a function with the property, i.e. a program that the checker says "yes" to. If there isn't, we could have the checker always say "no." Say the program $s$ computes a function with the property.
Now suppose you have a checker $C$ that, when given a program as input, returns whether the program computes a function with some property of interest. We want to test whether some program $M$ halts, when given input $x$.
We start by building a new program, called $Proxy_s$, which does the following on some input $z$:
- Run $M$ on input $x$.
- Run $s$ on input $z$, and return the result.
If $M$ halts on $x$, $Proxy_s$ moves on to the second step, and returns the same value as $s$. Since they return the same value (or no value, because of an infinite loop) on every possible input, $s$ and $Proxy_s$ compute the same function. So $Proxy_s$ computes a function with the property.
If $M$ doesn't halt on $x$, $Proxy_s$ never finishes the first step, and runs forever. So $Proxy_s$ always ends up in an infinite loop, which means it computes the empty function. So $Proxy_s$ computes a function that doesn't have the property.
What happens if we plug $Proxy_s$ as input into $C$? If $C$ says "yes," $Proxy_s$ computes a function with the property, which means $M$ halts on $x$. If $C$ says "no," $Proxy_s$ doesn't compute a function with the property, so $M$ doesn't halt on $x$.
So we can tell whether $M$ halts on $x$! All we have to do is write this proxy function, and ask $C$ whether its function has the property. But the Halting theorem says we can't do this! That means $C$ can't exist, and we can't write a checker to see whether programs compute functions with the property.
As an example, suppose we have a Fibonacci-checker $fib\_checker$. The following Python program computes the Fibonacci numbers, and so $fib\_checker$ says "yes" to it:
def fib(n):
a, b = 0, 1
for i in range(n):
a, b = b, a+b
return a
Then, for some program $M$ and input $x$, $Proxy_{fib}$ is the program:
def Proxy_fib(z):
M(x)
return fib(z)
If $fib\_checker$ works, this program says whether $M$ halts on $x$:
def halts(M,x)
return fib_checker(Proxy_fib)
Expanding this, we have the program:
def halts(M,x)
return fib_checker('''
M(x)
a, b = 0, 1
for i in range(n):
a, b = b, a+b
return a
''')
If $M$ halts on $x$, the program inside the triple quotes computes the Fibonacci sequence, so $fib\_checker$ returns $true$, and so does $halts$. If $M$ doesn't halt on $x$, the program in the triple quotes also doesn't halt, and returns the empty function, which isn't the Fibonacci sequence. So $fib\_checker$ returns $false$, and so does $halts$.
| 0
| 0
| 1,626
| 1,626
|
4,975
|
edd89d2e-541c-4255-9e73-eb9abcbac926
|
trentmkelly/LessWrong-43k
|
Convoy Crackdown
The convoy has been dispersed. Yet Tyler Cowen was clearly correct that this will be a very important event. The story has barely begun.
[Writer’s Note: This post violates some of the usual lines that I set for myself regarding politics, because I do not know how to communicate the situation without doing so, nor does it seem like a situation one can safely ignore. I did my best to keep it to a minimum. As before, I will be staying out of the comments, and ask that related discussions be confined to related posts and conducted to minimize the political component, but I acknowledge that this may not be fully possible in this case and will use discretion.]
Previously in Convoy and Convoy Continued, the puzzle was what was happening and what might happen next in response. Authorities were uncertain what to do. The situation was somewhat out of control. Violence was a distinct possibility, or at least the fear of such violence was informing decisions and rhetoric. There were big disputes over the essential nature of the protest, the protesters, what they wanted and why they were there. As far as I could tell the protests were unpopular but that too was disputed. Every edit seemed like it reflected more information from more sources, but more often that not that only meant increased uncertainty.
Most of that is now resolved. The convoy has been broken up by police without incident and with zero violence (or at least, zero violence that didn’t involve police acting like police usually act), only minor threats to permanently confiscate people’s dogs, in exactly the most basic and obvious way. A few people are still hanging around, but they no longer threaten public order. The three leaders have been arrested. When one engages in civil disobedience and makes oneself intentionally obstructing traffic for weeks, one eventually gets arrested and faces the consequences of being briefly in jail and having a record, and perhaps more if one committed more serious crimes and the
| 0
| 0
| 404
| 404
|
35,290
|
ab211ccc-563c-4472-bf11-898230b46ddc
|
trentmkelly/LessWrong-43k
|
Requesting low cost/high payoff projects ideas
There are many projects that would benefit people interested in living rationally but that no individuals are motivated enough to do (more). The Public Goods Team is trying to encourage and facilitate systematic and organized work on such projects. One of our first steps is to identify the projects that are the lowest hanging fruit: those projects which are high value, and inexpensive (in terms of time, motivation, money etc.). We have come up with a preliminary list of such projects (below). What are other such projects?
1. Further contributions to the Science of Winning at Life sequence:
1.1. Cognitive science of learning
1.2. Cognitive science of memory
1.3. Ergonomics
1.4. Behavioral conditioning
1.6. Fashion
1.7. Self control/conscienciousness research
2. Guides for explicitly applying LW material to your life, smaller sequences and exercises
3. Anki cards for the sequences and other useful topics/books
4. Introductory material that gets people hooked
5. User friendly guides to existing LW material
6. Meetup materials:
6.1. Exercises
6.1.1. Anti-rationalization
6.2. Notes and slides for presentations on specific topics
6.3. Game how-tos
6.4. Rejection therapy
6.4.1. Repetition
7. Book reviews on potentially useful books
7.1. Selfish Reasons to have more kids (critique/review)
7.2. Consciousness Explained
8. Cleaning up cognitive bias Wikipedia entries
9. Developing exercises and tests that help understand or improve rationality
9.1. A large set of good calibration questions
9.2. A good website for taking calibration tests
10. Financial analyses, like Brandon Reinhart's analysis of SIAI
10.1. The Cryonics Institute
10.2. Alcor
10.3. FHI
11. Resources/Advice to help people do projects
11.1. Advice for doing research
11.1.1. Doing literature reviews
11.1.2. Dealing with publication bias
11.1.3. Signs of researcher bias
11.1.4. Searching for critical viewpoints, alternative theories, disconfirming evidence
11.1.5. How do you k
| 0
| 0
| 555
| 555
|
6,232
|
eb266033-682b-4e72-95ff-ad86003b6787
|
trentmkelly/LessWrong-43k
|
AGI Safety Communications Initiative
Crossposted from the EA forum
A small group of AGI existential safety field-builders and I are starting research exploring a potential initiative about informing the public and/or important stakeholders about the risks of misaligned AI and the difficulties of aligning it.
We are aware that a public communication initiative like this carries risks (including of harming the AGI x-safety community’s reputation, of sparking animosity and misunderstandings between communities, or drawing attention to ways to misuse or irresponsibly develop scaleable ML architectures). We are still in the stage of evaluating whether/how this initiative would be good to pursue.
We are posting this on the forum to avoid the scenario where someone else starts a project about this at the same time and we end up doing duplicate work.
How you can get involved:
* If you are currently undertaking work similar to this or are interested in doing so, message me your email address along with a bit of context about yourself/what you are doing.
* We are drafting a longer post to share our current considerations and open questions. Message me if you would like to review the draft.
* We are looking for one or two individuals who are excited to facilitate a research space for visiting researchers. The space will run in Oxford (one week in Sep ’22) and in Prague (9-16 Oct ’22) with accommodation and meals provided for. If you take on the role as facilitator, you will receive a monthly income of $2-3K gross for 3 months and actually get to spend most of that time on your own research in the area (of finding ways to clarify unresolved risks of transformative AI to/with other stakeholders). If you are interested, please message me and briefly describe your research background (as relevant to testing approaches for effective intergroup communication, conflict-resolution and/or consensus-building).
| 0
| 0
| 397
| 397
|
33,333
|
bdac0fff-2ce4-4d79-be0b-c2b8e0b0f553
|
StampyAI/alignment-research-dataset/special_docs
|
owned or
foreign-directed corporations to mitigate national security concerns or bar their market entry.
Federal and state law can also incentivize behavior, a form of
indirect regulation. This can be done with tax credits, favorable
grant and loan terms, or antitrust exemptions, all of which are
authorized in the Defense Production Act. The government can
also influence behavior through the threat of criminal sanctio ns,
such as those found in export control laws and those regulating financial transactions with governments, entities, and individuals designated by the president under the International Emergency
Economic Powers Act and specific sanctions legislation.
Most notably for AI, section 230 of the Communications Decency
Act of 1996, with limited exception, exempts social media companies from liability for what is posted on their platforms. One policy question now is whether the law has worked too well,
allowing ISPs and social media companies a free pass when it
comes to third party conduct on their platforms. A second question is whether the role and responsibility of platforms is best addressed through law, regulation, or self-regulation in the form of
CSR policy.
Leverage can also be asserted through regulation of the insurance
industry, insurance costs, and limitations on coverage. Litigation may have the same effect, because it may influence the cost of insurance or affect a corporation’s reputation and thus market
share.
Finally, the government can influence corporate behavior by
establishing best practice standards and certifying those practices.
Center for Security and Emerging Technology | 16 Corporations may have a market and marketing incentive to adopt
those standards to validate their products and service quality.
Insurance rates, for example, might be linked to government AI
certifications.
Center for Security and Emerging Technology | 17 Going Forward
Ideally, law and ethics work hand in hand; ethics filling the void
where law does not, or should not, reach. But there are important differences. Law, in theory at least, is binding and thus removes the element of choice. Ethics, by definition, are non-binding, voluntary,
and subject to our moral choices.
Those developing policy to regulate AI must ensure not only that
the legal regime is sound but that the ethical framework is as well, providing specific, meaningful, and contextual guidance.
Contextual means guidance that comes with effective processes to reflect and assert that guidance, and do so in a transparent manner
that can be evaluated and adjusted if need be—for example, how
government agencies assert “appropriate human control” over applications or when corporations or academic entities share AI knowledge overseas. Contextual guidance also provides indicative actions, which illustrate how a principle should be applied, not just
state what it is. We hope this paper serves as a good starting
point.
Center for Security and Emerging Technology | 18 Author
The Honorable James E. Baker, a CSET Distinguished Fellow, is a
professor at Syracuse University College of Law with a courtesy
appointment in the Maxwell School. Judge Baker also serves as
Director of the Institute for Security Policy and Law. He is the
author of The Centaur’s Dilemma: National Security Law for the
Coming AI Revolution (Brookings: 2021).
Acknowledgments
For feedback and assistance, the author would like to thank Chuck
Babington, Danny Hague, Mark Hanin, Laurie Hobart, Matt
Mittelsteadt, Mark Rosen, Adrienne Thompson and Lynne Weil.
© 2021 by the Center for Security and Emerging Technology. This
work is licensed under a Creative Commons Attribution -Non
Commercial 4.0 International License.
To view a copy of this license, visit
https://creativecommons.org/licenses/by-nc/4.0/.
Document Identifier: doi: 10.51593/ 20190022
Center for Security and Emerging Technology | 19 Relevant Reading
“Artificial Intelligence at Google: Our Principles,” Google AI,
https://ai.google/principles/.
“Asilomar AI Principles,” Future of Life Institute, 2017,
https://futureoflife.org/ai-principles/.
“CITI Program: Research, Ethics, and Compliance Training,”
Collaborative Institutional Training Initiative,
https://about.citiprogram.org/en/homepage/.
“Code of Ethics,” EC-Council (International Council of Electronic
Commerce Consultants), https://www.eccouncil.org/code-
of-ethics/.
Communications Decency Act of 1996, 47 U.S.C. § 230 (1934)
https://www.govinfo.gov/content/pkg/USCODE-2011-
title47/pdf/USCODE-2011-title47-chap5-subchapII-partI-
sec230.pdf.
Defense Innovation Board, AI Principles: Recommendations on the
Ethical Use of Artificial Intelligence by the Department of
Defense (Washington, DC: October 2019),
https://media.defense.gov/2019/Oct/31/2002204458/-1/-
1/0/DIB\_AI\_PRINCIPLES\_PRIMARY\_DOCUMENT.PDF.
“IEEE Code of Ethics,” Institute of Electrical and Electronics
Engineers, https://www.ieee.org/about/corporate/governance/p7-
8.html.
IEEE Global Initiative on Ethics of Autonomous and Intelligent
Systems, “Ethically Aligned Design: A Vision for Prioritizing
Human Well-being with Autonomous and Intelligent
Systems” (IEEE, 2019), https://standards.ieee.org/content/dam/ieee-
standards/standards/web/documents/other/ead1e.pdf.
Center for Security and Emerging Technology | 20 Joint Artificial Intelligence Center, Ethical Principles for Artificial
Intelligence (Washington, DC: Department of Defense,
2020),
https://www.ai.mil/docs/Ethical\_Principles\_for\_Artificial\_Intel
ligence.pdf.
“Model Rules of Professional Conduct,” American Bar Association,
https://www.americanbar.org/groups/professional\_responsi
bility/publications/model\_rules\_of\_professional\_conduct/.
National Security Commission on Artificial Intelligence, Final
Report, (Washington, DC: March 2021),
https://www.nscai.gov/wp-content/uploads/2021/03/Full-
Report-Digital-1.pdf.
“NSPE Code of Ethics for Engineers,” National Society of
Professional Engineers,
https://www.nspe.org/resources/ethics/code-ethics.
Office for Human Research Protections, Federal Policy for the
Protection of Human Subjects (‘Common Rule’)
(Washington, DC: U.S. Department of Health and Human
Services), https://www.hhs.gov/ohrp/regulations-and-
policy/regulations/common-rule/index.html.
Office of Science Policy, National Institutes of Health, United States
Government Policy for Institutional Oversight of Life
Sciences Dual Use Research of Concern (Washington, DC:
Department of Health and Human Services, 2014), http://www.phe.gov/s3/dualuse/Documents/durc-policy.pdf.
Office of the Director of National Intelligence, Principles of Artificial
Intelligence Ethics for the Intelligence Community
(Washington, DC: 2020),
https://www.dni.gov/files/ODNI/documents/Principles\_of\_AI
\_Ethics\_for\_the\_Intelligence\_Community.pdf.
“Princeton Use of Small Unmanned Aircraft Systems (sUAS)
Policy,” Princeton University, January 16, 2017,
https://drones.princeton.edu/learn-more/policies-and-
procedures/princeton-suas-policy.
Center for Security and Emerging Technology | 21 The Unified Website for Biotechnology Regulation, “About the
Coordinated Framework,” U.S. Department of Agriculture,
U.S. Food and Drug Administration, and U.S. Environmental
Protection Agency,
https://usbiotechnologyregulation.mrp.usda.gov/biotechnolo
gygov/about.
“Universal Guidelines for Artificial Intelligence,” The Public Voice,
October 23, 2018, https://thepublicvoice.org/ai-universal-
guidelines/.
Center for Security and Emerging Technology | 22 Endnotes
1 “Letter to Sundar Pichai Against Project Maven,” Google Employees, April 4,
2018, https://static01.nyt.com/files/2018/technology/googleletter.pdf.
2 Bobby Allyn, “Google Workers Speak Out About Why They Formed A Union:
‘To Protect Ourselves,’” NPR, January 8, 2021,
https://www.npr.org/2021/01/08/954710407/at -google -hundreds- of-workers -
formed -a-labor- union- why-to-protect -ourselves; Alina Selyukh, “After Years of
Activism, More Than 200 Google Employees Form a Union,” NPR, January 4,
2021, https://www.npr.org/ 2021/01/04/953314490/after- years- of-activism -
more -than -200- google -employees- form -a-union; Alina Selyukh, “Google
Workers Launch Union to Press Grievances With Executives,” NPR, January 4,
2021, https://www.npr.org/2021/01/04/953198140/google -workers -launch -
union- to-press -grievances -with -executives.[SEP]
| 1
| 6,000
| 8,230
| 8,230
|
98,766
|
<urn:uuid:9ee5e794-778e-4c9d-a693-8b6adedb5692>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
of death. In engineering it is commonly called reliability function. It is declining over time, and will approach zero unless indefinite survival is possible with a finite probability.
The event density f(t)=\frac{d}{dt}(1-S(t)) denotes the rate of death per unit time.
The hazard function \lambda(t) is the event rate at time t conditional on survival until time t or later. It is \lambda(t) = - S'(t)/S(t). Note that unlike the event density function this does not have to decline as the number of survivors gets low: this is the overall force of mortality at a given time.
The expected future lifetime given survival to time t_0 is \frac{1}{S(t_0)}\int_{t_0}^\infty S(t)dt. Note that for exponential survival curves (i.e. constant hazard) it remains constant.
The case for Mars
On practical Ethics I post about the goodness of being multi-planetary: is it rational to try to settle Mars as a hedge against existential risk?
The problem is not that it is absurd to care about existential risks or the far future (which was the Economist‘s unfortunate claim), nor that it is morally wrong to have a separate colony, but that there might be better risk reduction strategies with more bang for the buck.
One interesting aspect is that making space more accessible makes space refuges a better option. At some point in the future, even if space refuges are currently not the best choice, they may well become that. There are of course other reasons to do this too (science, business, even technological art).
So while existential risk mitigation right now might rationally aim at putting out the current brushfires and trying to set the long-term strategy right, doing the groundwork for eventual space colonisation seems to be rational.
What makes a watchable watchlist?
Watch the skiesStefan Heck managed to troll a lot of people into googling “how to join ISIS”. Very amusing, and now a lot of people think they are on a NSA watchlist.
This kind of prank is of course by why naive keyword-based watch lists are total failures. One prank and it gets overloaded. I would be shocked if any serious intelligence agency actually used them for real. Given that people’s Facebook likes give pretty good predictions of who they are (indeed, better than many friends know them) there are better methods if you happen to be a big intelligence agency.
Still, while text and other online behavior signal a lot about a person, it might not be a great tool for making proper watchlists since there is a lot of noise. For example, this paper extracts personality dimensions from online texts and looks at civilian mass murderers. They state:
Using this ranking procedure, it was found that all of the murderers’ texts were located within the highest ranked 33 places. It means that using only two simple measures for screening these texts, we can reduce the size of the population under inquiry to 0.013% of its original size, in order to manually identify all of the murderers’ texts.
At first, this sounds great. But for the US, that means the watchlist for being a mass murderer would currently have 41,000 entries. Given that over the past 150 years there has been about 150 mass murders in the US, this suggests that the precision is not going to be that great – most of those people are just normal people. The base rate problem crops up again and again when trying to find rare, scary people.
The deep problem is that there is not enough positive data points (the above paper used seven people) to make a reliable algorithm. The same issue cropped up with NSA’s SKYNET program – they also had seven positive examples and hundreds of thousands of negatives, and hence had massive overfitting (suggesting the Islamabad Al Jazeera bureau chief was a prime Al Qaeda suspect).
Rational watchlists
The rare positive data point problem strikes any method, no matter what it is based on. Yes, looking at the social network around people might give useful information, but if you only have a few examples of bad people the system will now pick up on networks like the ones they had. This is also true for human learning: if you look too much for people like the ones that in the past committed attacks, you will focus too much on people like them and not enemies that look different. I was told by an anti-terrorism expert about a particular sign for veterans of Afghan guerrilla warfare: great if and only if such veterans are the enemy, but rather useless if the enemy can recruit others. Even if such veterans are a sizable fraction of the enemy the base rate problem may make you spend your resources on innocent “noise” veterans if the enemy is a small group. Add confirmation bias, and trouble will follow.
Note that actually looking for a small set of people on the watchlist gets around the positive data point problem: the system can look for them and just them, and this can be made precise. The problem is not watching, but predicting who else should be watched.
The point of a watchlist is that it represents a subset of something (whether people or stocks) that merits closer scrutiny. It should essentially be an allocation of attention towards items that need higher level analysis or decision-making. The U.S. Government’s Consolidated Terrorist Watch List requires nomination from various agencies, who presumably decide based on reasonable criteria (modulo confirmation bias and mistakes). The key problem is that attention is a limited resource, so adding extra items has a cost: less attention can be spent on the rest.
This is why automatic watchlist generation is likely to be a bad idea, despite much research. Mining intelligence to help an analyst figure out if somebody might fit a profile or merit further scrutiny is likely more doable. As long as analyst time is expensive it can easily be overwhelmed if something fills the input folder: HUMINT is less likely to do it than SIGINT, even if the analyst is just doing the preliminary nomination for a watchlist.
The optimal Bayesian watchlist
One can analyse this in a Bayesian framework: assume each item has a value x_i distributed as f(x_i). The goal of the watchlist is to spend expensive investigatory resources to figure out the true values; say the cost is 1 per item. Then a watchlist of randomly selected items will have a mean value V=E[x]-1. Suppose a cursory investigation costing much less gives some indication about x_i, so that it is now known with some error: y_i = x_i+\epsilon. One approach is to select all items above a threshold \theta, making V=E[x_i|y_i<\theta]-1.
If we imagine that everything is Gaussian x_i \sim N(\mu_x,\sigma_x^2), \epsilon \sim N(0,\sigma_\epsilon^2), then V=\int_\theta^\infty t \phi(\frac{t-\mu_x}{\sigma_x}) \Phi\left(\frac{t-\mu_x}{\sqrt{\sigma_x^2+\sigma_\epsilon^2}}\right)dt. While one can ram through this using Owen’s useful work, here is a Monte Carlo simulation of what happens when we use \mu_x=0, \sigma_x^2=1, \sigma_\epsilon^2=1 (the correlation between x and y is 0.707, so this is not too much noise):
Utility of selecting items for watchlist as a function of threshold. Red curve without noise, blue with N(0,1) noise added.
Note that in this case the addition of noise forces a far higher threshold than without noise (1.22 instead of 0.31). This is just 19% of all items, while in the noise-less case 37% of items would be worth investigating. As noise becomes worse the selection for a watchlist should become stricter: a really cursory inspection should not lead to insertion unless it looks really relevant.
Here we used a mild Gaussian distribution. In term of danger, I think people or things are more likely to be lognormal distributed since it is a product of many relatively independent factors. Using lognormal x and y leads to a situation where there is a maximum utility for some threshold. This is likely a problematic model, but clearly the shape of the distributions matter a lot for where the threshold should be.
Note that having huge resources can be a bane: if you build your watchlist from the top priority down as long as you have budget or manpower, the lower priority (but still above threshold!) entries will be more likely to be a waste of time and effort. The average utility will decline.
Predictive validity matters more?
In any case, a cursory and cheap decision process is going to give so many so-so evaluations that one shouldn’t build the watchlist on it. Instead one should aim for a series of filters of increasing sophistication (and cost) to wash out the relevant items from the dross.
But even there there are pitfalls, as this paper looking at the pharma R&D industry shows:
We find that when searching for rare positives (e.g., candidates that will successfully complete clinical development), changes in the predictive validity of screening and disease models that many people working in drug discovery would regard as small and/or unknowable (i.e., an 0.1 absolute change in correlation coefficient between model output and clinical outcomes in man) can offset large (e.g., 10 fold, even 100 fold) changes in models’ brute-force efficiency.
Just like for drugs (an example where the watchlist is a set of candidate compounds), it might be more important for terrorist watchlists to aim for signs with predictive power of being a bad guy, rather than being correlated with being a bad guy. Otherwise anti-terrorism will suffer the same problem of declining productivity, despite ever more sophisticated algorithms.
The hazard of concealing risk
In the end, systemic risk is everybody’s problem.
Dampening theoretical noise by arguing backwards
WhiteboardScience has the adorable headline Tiny black holes could trigger collapse of universe—except that they don’t, dealing with the paper Gravity and the stability of the Higgs vacuum by Burda, Gregory & Moss. The paper argues that quantum black holes would act as seeds for vacuum decay, making metastable Higgs vacua unstable. The point of the paper is that some new and interesting mechanism prevents this from happening. The more obvious explanation that we are already in the stable true vacuum seems to be problematic since apparently we should expect a far stronger Higgs field there. Plenty of theoretical issues are of course going on about the correctness and consistency of the assumptions in the paper.
Don’t mention the war
What I found interesting is the treatment of existential risk in the Science story and how the involved physicists respond to it:
Moss acknowledges that the paper could be taken the wrong way: “I’m sort of afraid that I’m going to have [prominent theorist] John Ellis calling me up and accusing me of scaremongering.
Ellis is indeed grumbling a bit:
As for the presentation of the argument in the new paper, Ellis says he has some misgivings that it will whip up unfounded fears about the safety of the LHC once again. For example, the preprint of the paper doesn’t mention that cosmic-ray data essentially prove that the LHC cannot trigger the collapse of the vacuum—”because we [physicists] all knew that,” Moss says. The final version mentions it on the fourth of five pages. Still, Ellis, who served on a panel to examine the LHC’s safety, says he doesn’t think it’s possible to stop theorists from presenting such argument in tendentious ways. “I’m not going to lose sleep over it,” Ellis says. “If someone asks me, I’m going to say it’s so much theoretical noise.” Which may not be the most reassuring answer, either.
There is a problem here in that physicists are so fed up with popular worries about accelerator-caused disasters – worries that are often second-hand scaremongering that takes time and effort to counter (with marginal effects) – that they downplay or want to avoid talking about things that could feed the worries. Yet avoiding topics is rarely the best idea for finding the truth or looking trustworthy. And given the huge importance of existential risk even when it is unlikely, it is probably better to try to tackle it head-on than skirt around it.
Theoretical noise
“Theoretical noise” is an interesting concept. Theoretical physics is full of papers considering all sorts of bizarre possibilities, some of which imply existential risks from accelerators. In our paper Probing the Improbable we argue that attempts to bound accelerator risks have problems due to the non-zero probability of errors overshadowing the probability they are trying to bound: an argument that there is zero risk is actually just achieving the claim that there is about 99% chance of zero risk, and 1% chance of some risk. But these risk arguments were assumed to be based on fairly solid physics. Their errors would be slips in logic, modelling or calculation rather than being based on an entirely wrong theory. Theoretical papers are often making up new theories, and their empirical support can be very weak.
An argument that there is some existential risk with probability P actually means that, if the probability of the argument is right is Q, there is risk with probability PQ plus whatever risk there is if the argument is wrong (which we can usually assume to be close to what we would have thought if there was no argument in the first place) times 1-Q. Since the vast majority of theoretical physics papers never go anywhere, we can safely assume Q to be rather small, perhaps around 1%. So a paper arguing for P=100% isn’t evidence the sky is falling, merely that we ought to look more closely to a potentially nasty possibility that is likely to turn into a dud. Most alarms are false alarms.
However, it is easier to generate theoretical noise than resolve it. I have spent some time working on a new accelerator risk scenario, “dark fire”, trying to bound the likelihood that it is real and threatening. Doing that well turned out to be surprisingly hard: the scenario was far more slippery than expected, so ruling it out completely turned out to be very hard (don’t worry, I think we amassed enough arguments to show the risk to be pretty small). This is of course the main reason for the annoyance of physicists: it is easy for anyone to claim there is risk, but then it is up to the physics community to do the laborious work of showing that the risk is small.
The vacuum decay issue has likely been dealt with by the Tegmark and Bostrom paper: were the decay probability high we should expect to be early observers, but we are fairly late ones. Hence the risk per year in our light-cone is small (less than one in a billion). Whatever is going on with the Higgs vacuum, we can likely trust it… if we trust that paper. Again we have to deal with the problem of an argument based on applying anthropic probability (a contentious subject where intelligent experts disagree on fundamentals) to models of planet formation (based on elaborate astrophysical models and observations): it is reassuring, but it does not reassure as strongly as we might like. It would be good to have a few backup papers giving different arguments bounding the risk.
Backward theoretical noise dampening?
The lovely property of the Tegmark and Bostrom paper is that it covers a lot of different risks with the same method. In a way it handles a sizeable subset of the theoretical noise at the same time. We need more arguments like this. The cosmic ray argument is another good example: it is agnostic on what kind of planet-destroying risk is perhaps unleashed from energetic particle interactions, but given the past number of interactions we can be fairly secure (assuming we patch its holes).
One shared property of these broad arguments is that they tend to start with the risky outcome and argue backwards: if something were to destroy the world, what properties does it have to have? Are those properties possible or likely given our observations? Forward arguments (if X happens, then Y will happen, leading to disaster Z) tend to be narrow, and depend on our model of the detailed physics involved.
While the probability that a forward argument is correct might be higher than the more general backward arguments, it only reduces our concern for one risk rather than an entire group. An argument about why quantum black holes cannot be formed in an accelerator is limited to that possibility, and will not tell us anything about risks from Q-balls. So a backwards argument covering 10 possible risks but just being half as likely to be true as a forward argument covering one risk is going to be more effective in reducing our posterior risk estimate and dampening theoretical noise.
In a world where we had endless intellectual resources we would of course find the best possible arguments to estimate risks (and then for completeness and robustness the second best argument, the third, … and so on). We would likely use very sharp forward arguments. But in a world where expert time is at a premium and theoretical noise high we can do better by looking at weaker backwards arguments covering many risks at once. Their individual epistemic weakness can be handled by making independent but overlapping arguments, still saving effort if they cover many risk cases.
Backwards arguments also have another nice property: they help dealing with the “ultraviolet cut-off problem“. There is an infinite number of possible risks, most of which are exceedingly bizarre and a priori unlikely. But since there are so many of them, it seems we ought to spend an inordinate effort on the crazy ones, unless we find a principled way of drawing the line. Starting from a form of disaster and working backwards on probability bounds neatly circumvents this: production of planet-eating dragons is among the things covered by the cosmic ray argument.
Risk engineers will of course recognize this approach: it is basically a form of fault tree analysis, where we reason about bounds on the probability of a fault. The forward approach is more akin to failure mode and effects analysis, where we try to see what can go wrong and how likely it is. While fault trees cannot cover every possible initiating problem (all those bizarre risks) they are good for understanding the overall reliability of the system, or at least the part being modelled.
Deductive backwards arguments may be the best theoretical noise reduction method.[SEP]
| 1
| 6,000
| 9,973
| 9,973
|
64,420
|
6df6898c-f8ce-430c-b2e2-8379103387a7
|
trentmkelly/LessWrong-43k
|
Shanghai Less Wrong Meetup April 30th, 2pm
John Teddy asked me to post the following info for a Less Wrong meetup he is organizing.
Where: Starbucks, Carrefour Mall, the Zhongshan Park Store, near the Zhongshan Park Station, Line 2, 3, or 4, Shanghai, China
B1&B2, No.1018, ChangNing Road, ChangNing District, Shanghai ( 长宁路1018号, 龙之梦购物中心B1-2楼 近凯旋路 )
When: 2011, April 30th at 2pm.
Contact John Teddy: 18721399070 (phone no.)
| 0
| 0
| 170
| 170
|
6,071
|
35496d7f-09ae-43b7-ba0e-58266ab9f8d8
|
trentmkelly/LessWrong-43k
|
Which rationalists faced significant side-effects from COVID-19 vaccination?
While I want to believe the vaccines to be safe, Gendlin's litany tells me that it's more important that whether they actually are and the ability to change one's mind is a key feature of a rationalist. Another thread links to a discussion on the Dark Horse Podcast between Steve Kirsch, Robert Malone and Bret Weinstein about issues with vaccine safety. My prior for Steve Kirsch before that video was that he was one of the people in the world who thought most effectively about how to deal with COVID-19 in 2020. I hold that belief because of his project, the COVID-19 Early Treatment Fund.
The Fund describes it's guiding principles as:
> The world has to take a multipronged approach to tackling COVID-19. We should work to identify effective repurposed drugs, validate experimental drugs, develop new drugs, test early treatments, test late treatments, try convalescent serums and invest in vaccines. We must do all of this in parallel to end the pandemic.
The guiding principles document goes into more detail about their principles and is well worth reading. When I first read it I thought, this is actually a sensible plan compared to what governments are doing. It seemed to me also better than what various EA groups managed to come up with. When someone who run a non-profit that tells the world that it's important to invest in vaccines in its guiding principles tells you about vaccine side-effects and who managed to put their money into a sane COVID-19 response, that's to be taken a lot more seriously than most other possible people who might say something about vaccine side-effects.
If the thesis is true, it's very important to know as soon as possible, because the thesis is about issues that are particular to the spike protein. If the thesis is true a step that's as simple as putting domains of the spike protein seperately into a vaccine or targeting a different protein for the third vaccination shoot might completely mitigate the side-effects.
Given that many rati
| 0
| 0
| 435
| 435
|
57,246
|
b00fe4ec-6d7e-40df-8276-1ae674d900f5
|
StampyAI/alignment-research-dataset/lesswrong
|
How Much Computational Power Does It Take to Match the Human Brain?
Joe Carlsmith with a really detailed report on computational upper bounds and lower bounds on simulating a human brain:
> Open Philanthropy is interested in when AI systems will be able to perform [various tasks](https://www.openphilanthropy.org/blog/some-background-our-views-regarding-advanced-artificial-intelligence#Sec1) that humans can perform (“AI timelines”). To inform our thinking, I investigated what evidence the human brain provides about the computational power sufficient to match its capabilities. This is the full report on what I learned. A medium-depth summary is available [here](https://www.openphilanthropy.org/blog/new-report-computation-power-match-human-brain). The [executive summary](https://www.openphilanthropy.org/brain-computation-report#ExecutiveSummary) below gives a shorter overview.
>
> [...]
>
> Let’s grant that in principle, sufficiently powerful computers can perform any cognitive task that the human brain can. How powerful is sufficiently powerful? I investigated what we can learn from the brain about this. I consulted with more than 30 experts, and considered four methods of generating estimates, focusing on [floating point operations per second](https://en.wikipedia.org/wiki/FLOPS) (FLOP/s) as a metric of computational power.
>
> These methods were:
>
> 1. Estimate the FLOP/s required to model the brain’s mechanisms at a level of detail adequate to replicate task-performance (the “[mechanistic method](https://www.openphilanthropy.org/brain-computation-report#TheMechanisticMethod)”).[**1**](https://www.openphilanthropy.org/brain-computation-report#footnote1_them87c)
> 2. Identify a portion of the brain whose function we can already approximate with artificial systems, and then scale up to a FLOP/s estimate for the whole brain (the “[functional method](https://www.openphilanthropy.org/brain-computation-report#TheFunctionalMethod)”).
> 3. Use the brain’s energy budget, together with physical limits set by [Landauer’s principle](https://en.wikipedia.org/wiki/Landauer%27s_principle), to upper-bound required FLOP/s (the “[limit method](https://www.openphilanthropy.org/brain-computation-report#TheLimitMethod)”).
> 4. Use the communication bandwidth in the brain as evidence about its computational capacity (the “[communication method](https://www.openphilanthropy.org/brain-computation-report#TheCommunicationMethod)”). I discuss this method only briefly.
>
> None of these methods are direct guides to the *minimum possible* FLOP/s budget, as the most efficient ways of performing tasks need not resemble the brain’s ways, or those of current artificial systems. But if sound, these methods would provide evidence that certain budgets are, at least, big enough (*if* you had the right software, which may be very hard to create – see discussion in [section 1.3](https://www.openphilanthropy.org/brain-computation-report#Context)).[**2**](https://www.openphilanthropy.org/brain-computation-report#footnote2_q0nf497)
>
> Here are some of the numbers these methods produce, plotted alongside the FLOP/s capacity of some current computers.
>
> ***Figure 1: The report’s main estimates.** See the* [*conclusion*](https://www.openphilanthropy.org/brain-computation-report#Conclusion) *for a list that describes them in more detail, and summarizes my evaluation of each.*These numbers should be held lightly. They are back-of-the-envelope calculations, offered alongside initial discussion of complications and objections. The science here is very far from settled.
>
>
| 0
| 0
| 935
| 935
|
22,693
|
72aac957-5905-425b-a5da-52e48f8b2615
|
trentmkelly/LessWrong-43k
|
Evolution of Modularity
None
| 0
| 0
| 11
| 11
|
113,138
|
<urn:uuid:90d2fb2c-5e3c-455e-b967-b8d5f87661b5>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Super-Zero didn’t get further than that, before my hoof and Nemo’s mace came whistling towards her face, slamming her jaw closed so hard we actually got a brief shower of teeth. The pain apparently finally too much, the old nag’s eyes rolled up into her head, and she sank down in a boneless heap on the marble floor. “Scoots!” I wasted no time, and put my game face on, getting up on two legs and pulling my rod and revolver. Nemo followed my example, and went back to back with me, brandishing that freaky mace of her with two hooves. Believe it or not, but the first thing of all I noticed was the magic of the place. And I nearly fell gagging down on the spot from it. Magic, is supposed to be a thing of hope and wonder, all the little things that make life worth living. The joy, the sorrow, the ups, the downs… The room me and Nero had found ourselves in? It felt dead. Just… dead. I’ve been near necromancy —heck, even waltzed right on the line of the Laws and performed it once or twice when nothing else would work to save people, but even that had this liveliness to it; if as a corrupted and dark mirror to what magic should be. That magic around us was… Imagine suddenly being inside a sealed room that somebody died in decades if not centuries ago. There’s no light, no life, no hope… just this vaguely wrong smell like a museum store-room, and a horrible stillness that consumes all it touches until even the screams fade to nothingness. Speaking of screams, I forced down the one my throat wanted to give off, and quickly scanned the room physically. I kid you not; it looked like some type of ballroom, and from one of those Old World castles that like to pretend it sprang from dreams and pixie dust, rather than peasant blood and tasteless nobles, at that. On second look, I wasn’t sure if the floor was marble, but whatever the squares of purple-grey and purple stone really was, it had been polished until it was like walking on a mirror. There wasn’t many walls, as such, but instead the entire space was lined with stained glass windows; most of natural things like flames or cresting waves, but a few with what looked like historical scenes. Just with ponies instead of humans in the latter case. And all of it crowned with a single golden chandelier as big as my freaking office, crowned with so many glittering gems it passed ‘gaudy,’ circled clean around ‘nuevue rich,’ and landed back somewhere around ‘classical’ again. It didn’t look like there’d been balls for a long, long time there, though. The classy tables and rich carpets my mind’s eye wanted to project over the place simply wasn’t there, instead replaced with tons upon tons of the girliest damn military gear I’ve ever seen. Computers… that looked as if they’d come alive and escaped from the Smithsonian; their magnetic tapes and graph-paper read-outs clicking quietly away. Industrial steel tables clashing horribly with the rest of the room… that somebody still had seemed bothered enough by to cover them up with what looked like floor length tablecloths made from real linen. There was even —I kid you not, a few silver potpourri bowls, resting out of sight and mind on discrete plinths near the walls; far enough not to interfere but still close enough to drive away some of the chemical stench from the scientific and alchemical equipment. All in all, it was a bit like having been dropped into Oprah's evil lair. Sure, we might be nestled under Death Mountain, but that’s no reason not to look fabalu~s. I took advantage how our sudden appearance had stunned the people in the room —about two-three dozen ponies, most in either lab-coats or wizard robes, but with a few guards as well— and quickly looked over behind us. The only real difference there was this line of full-length mirrors. Twenty in total, and so heavily enchanted the air around them shimmered slightly even to my normal sight. About half or so of them were seemingly currently being unused and showed just normal reflections. Not so with the rest. One showed a bird’s eye view of the Golden Gate Bridge, the portal seemingly suspended mid-air. Another, what looked suspiciously like the sculpture outside the Chicago Fire Academy, the three copper ‘flames’ gleaming in a under-shilled rain that seemed to be falling just a bit too quickly. A third, on the top of the left hand of the Christ the Redeemer statue in Rio. My eyes were drawn to the one just behind me, Nemo and the Super-Zero, however. Because not only was it showing a top-down view of a rather familiar snowy field outside a cottage littered with broken ponies, but my bro- sister was in it as well; moving at chipmunk on crack speeds that would have been comical if not for the clear desperation on her face. Sadly, I only had time to come to the conclusion ‘portal’ myself, before the leader of the PG-rated Stargate program we’d stumbled onto —the same unicorn mare that had screamed ‘Scoots’ earlier, started barking orders. “Everypony with a horn, stun them! Quick, before they can make it back through the portal!” The strangest thing was that I didn’t hear any spells called or even feel the magic being manipulated… until I turned my head and saw what was near this rainbow of death of differently colored, raw magic coming towards us. If I hadn’t already been halfway through casting a shield thanks to instincts born from thousands of fights, I would probably just have stood there and taken it on the shin from pure dumbstruck-ness. It was a bit like being stuck inside a closet with a two person rave going on, but the bursts of magic splashed of my screen of congealed air as if had been bugs on a windshield. Who the fuck uses raw magic as an attack?! You might as have a fifty caliber machine gun and throw the bullets at people! It’s called ‘non-le-thal com-bat’ Harry. You know, kinda like that thing you keep doing to Nicodemus instead of gutting the creep, but resulting from compassion and restraint instead of annoying circumstances and overpowered artifacts in the hands of evil...? Stars and stones, you could probably have roasted a whole pack of ghouls with that much magic! Why the hell would anybody just throw that type of power away in a fight?! Sigh… Honestly, Harry, sometimes I wish you’d develop alien-hoof-syndrome, just so I could smack you a few times... You could have heard a pin falling as the lightshow died off. Ponies just staring in wide-eyed with a mix of wonder and horror that not even a single hair had been ruffled in mine or Nemo’s manes. Then the screaming stampede started. The leader from earlier and another even older mare —a stark white unicorn with fraying but still purple and pink mane and an ancient looking zebra of all possible creatures, did a decent try at restoring order, but to no avail. Even without the barked command the two just screamed 'commanders.' Not only with how everypony still there subtly looked towards them for further orders, but the rank and file typically don't get custom clothes; some type of fusion between a the general cut of a wizards robe, with the waxed white fabric of a lab-coat. Long enough it covered their backs (including their emblems), but short enough the two could stand on only their legs without the hem touching the ground. They actually looked kinda neat and practical. I made a mental note a robe like that could be worth looking into, and shoved it into the back of my mind. Once the stampede had gone, all that remained was me, Nemo, the unicorn, the still unconscious Super-Zero, the zebra, and four of the clone guard. I pulled the hammer back on my Smith & Wesson 500, producing a rather satisfyingly loud and threatening ‘click.’ “Scram,” I coldly told the minions, jerking my head back to the carnage filled mirror, “not even those four Stooges of yours were much of an act by my town’s standards, and you guys wouldn’t even warm a crowd.” I gave the Super-Zero a hard kick on her whole wing, making her roll bonelessly and every’pony’ in the room except Nemo gasp. “Fuck off now,” I continued at a growl, “and you get to do it on your own power. Scout's honor.” Alas, the guards didn’t heed my warning, instead opting to falling into defensive line with the senior researchers; two burly stallions per mare. “Hey, Harry?” Nemo asked in a rather suspiciously chipper voice. “Yeah?” I asked without turning my head. “Those computers and notes sure do look important, don’t they, Ms. Wizard?” The single pointy headed idiot remaining cleared her throat. “...Fluttershy,” she more or less squeaked at us, but it sounded as if it was just her normal, rather annoying, voice rather than any fear. “I don’t know what type of delusions you are under thanks to Discord, but you’re a pegasus; you can’t be a wizard.” My grin froze in place. The overgrown girl noticed, and hesitated. “...Please, I don’t know how you resisted like that, but please stand down and we can treat you and Dash.” All six remaining idiots gasped as I drew power to my blasting rod, making the carved runes light with inner flames, and snarled out a spell. “Flickum Bickum.” Normally, this was a spell I just used to light candles and similar. It had been years since I’d used my most basic fire spell with an actual focus, but the result was quite decent. The frilly idiot gasped (again) in shock, paling to near snow white as every book, ledger, tome and credit card receipt in the whole ‘lab’ mysteriously and ‘spontaneously’ combusted. “No, stop! Please!” The zebra shouted in a thick accent that sounded vaguely African. Not making a racist joke or anything, she really sounded as if she’d come from thereabout but spent many years abroad. “Rebuilding this place would take months even with the right expertise!” “Oh, why didn’t you just say so?” I smirked with false cheer, pointing my rod at the line of computers. “Melivosa!” The effect of my anti-tech spell was almost as flashy as the fires even now making sooty smoke climb towards the richly decorated ceiling I’m not sure if it was how utterly beyond ticked I was at all this or just such ancient and worn looking gear, but the computers all but exploded; bits of magnetic tape and shards of radio-tubes flying so hard it took out a few of the stained windows. One of the guards, the nearest of the pair protecting Ms. Pointy, actually got a quite decent shower of plastic and glass over himself… but it looked as if his armor protected him. I let my grin come back. What can I say? Two ‘experts’ staring in disbelief and horror at the wizard actually turning out to be a wizard was just deeply cathartic for me. I waved a wing back towards the mirrors. “Start smashing all but the one to San Fran,” I whispered Nemo’s way. Nemo blinked at me. “...You sure?” I gave a tiny nod, not taking my eyes of the ‘locals.’ “I know a Way back to Chicago from there, ‘kay? It will get us home, but slow these twits.” I’m not sure if Nemo had overheard of what a Way is, if she mistook it for the mundane type of ‘a way,’ or just trusted me, but jump into action she did. Pumping her booster spell to eleven again, Nemo all but blurred away from my side; these loud and expensive sounds starting up behind me within nearly a second. The two old researchers reacted a bit differently at that. The zebra just covered her face with her hooves in horror, but Ms. Pointy started barking orders again. “Men,” she screamed, a leaf green light gathering around her horn, “stop the-!” I didn’t hesitate; I just squeezed of two shots in rapid sequence. Would you believe I’ve become a better shot thanks to this pony crap? Not that the lack of fingers help, but the whole tactile telekinesis thing plus a horsepower let me keep even something as beefy as my revolver far steadier. Wasn’t like I could hit tossed coins or parlor tricks like that… but six idiots, just standing there as if being half a room away made them safe? I could hit those targets, alright. At first, the unicorn mare just stood there; blinking up at the ruined mess her horn had suddenly become. For a few moments, all was silent. Then the screaming started. Again. The unicorn just collapsed; crying in a heap with her eyes open and unblinking in what looked like shock. The zebra let out a mortified gasp, and rushed to her friend, but the other mare barely reacted. It was as if a puppet had had it strings cut. The guards? I doubt the Devil himself would have gotten the type of disbelieving, horrified stares of disgust they aimed my way. “The Smith & Wesson 500 holds five bullets, each one packing enough of a wallop to kill a bull elephant mid charge. Four of you fine gentleman and three bullets left, so what will it be? Save the bastards bleeding on the floor there...” I droned as I pulled back the hammer again, making the soldiers flinch like one creature. “...or do you feel lucky. Well, do yah, punk?” Amazingly, despite the blood pooling out from behind her hooves as she rendered first aid, the zebra still had some steel and fight in her. “Fluttershy, please, nopony blames you your madness,” the mare forced out, quite literally pleading through gritted teeth. “Stop this now, and we may end this without further horror and sadness!” In my mind, I could see it. Two faces, one a little girl with dark but sparkling eyes, the other a skull of carved wood crawling with motes of green light. One, asking my dear friend she actually lived with more seldom and seldom with less and less hope where Daddy had gone… The other, simply wailing alone in a cottage that got colder and colder in my absence. And would I even miss them? Or would this ‘Fluttershy’ creep just smile emptily at the memory? A stranger, trying to be polite on seeing somebody else’s children?...I don’t actually know. Depends on the ‘cure,’ I guess. I swear, I nearly put a bullet in both the mares where their hearts should have been, just to ensure I’d never end up in this tainted realm ever again. My friends, let alone my little girls, deserve better than another empty casket. “...You want this ended?” I countered coldly. “Leave us alone, and the favor will be returned. Simple as that.” Silence once more descended on the room, as I pulled back the hammer again; the click damn near echoing off the walls. "Your move, creep." The zebra was about to retort something, but one of ‘her’ guards showed some initiative and grabbed her; picking her up easily enough but jostling her arm enough that whatever she’d intended to say got switched out for a long string of harsh words in a oddly ‘clicky’ tongue I didn’t recognize. The other guards followed his example, and took a strategic retreat. Pausing only long enough to frantically scrape as much of the unicorn's alicorn off the floor as possible while another grabbed the mare herself. The now rather less pointy unicorn didn’t even blink, as she got dragged away; leaving me and Nemo alone in the smoky room. I barely waited for the two ‘free’ stallions to slam the main door shut, before doing my best to seal it behind them. With a few flicks of flame towards it and any other door I could see (some added randomly to the tables for good measure), the room was soon this cozy little deathtrap in waiting. Something gleaming on one of the nearest tables made me hesitate just a moment, though. “...What are you doing?” Nemo screamed after me, as I put away my weapons and trotted over. “We need to go, now!” It was a small table with just a few basic instruments on it, magnifying glasses and the like. I couldn’t read the swirling text on the neat sign on the end… ’Off-world Artifacts! Caution!’...but from the strange collection of odds and ends, it seemed like this was a ‘cross dimensional artifacts’ sorting and cataloging table. Or at least, the carefully orchestrated semblance of one, depending on how much truth Nemo’s 'Home-brewed World-domination' theory had to it. A weather bit pocket-book (Pyramid Scheme. Good book, but I’d already read it.) with so many stains on it I’d missed it during my little purge. It looked as if it had been taken from a landfill, but even so there was a small notebook nearby with a few neat notes still visible even in the ash. A car tire, old and near falling apart, but with neat sample bags nearby, and a note; probably about chemical analysis. ’For alchemical analysis ASAP. Initial test show twice the tensile-strength and three-times the durability of our own rubber, but with no traceable enchantments. (!) Further, fresher samples should be made a priority, if possible.’ Huh, interesting... A cellphone, so busted it had been in two halves even before my bit of sabotage; now nearly two smoldering stains. ’Unknown but damaged communication device. Very fine and complex, but poorly warded circuitry. DO NOT HANDLE WITH MAGIC! Further study needed.’ An old, only partially cleaned wedding ring; having spent years in sand from the looks of it. The tiny diamond on it looked real, but even so the thing couldn’t have been worth as much as three digits even while new. ’Initial surveys show extremely poor gem density and quality. Unknown why, further surveys needed. Possible economic and political use?’ Well, that sounds potentially annoying... A far cleaner book I’d missed thanks to an old tuba. The thing, a simple astronomy book intended for children didn’t look like anything special, but was near covered in notes with lots and lots of exclamation marks. Honestly not sure what the deal was, but it gave me a funny feeling. ’PRIORITY ONE! FOR THE PRINCESSES’ EYES ONLY! PRIORITY ONE! This thing contains references to stable and self-regulating solar and lunar orbits! This could either make or break Equestria as we know it, ponies! Not a word to the public until we’ve verified this in a laboratory or by further observation of Earth Δ! ~S.B.’ I lingered for a bit on the longer note, but it was all a Greek speaking gibberish in Klingon to me. Freaky thing, though, it didn't feel as if it should have been that way; it felt as if it should have been as clear and simple to me as any normal English. Like one of those Magic Eye pictures; squint just right, and the whole thing becomes a perfectly normal if odd looking bunny....I’ll grant them this much, the grunts at least believe they’re really working for the Princesses. Not sure if that makes all this worse or better to be honest, but at least it’s something. “Dresden!” Nemo hissed at me from across the room. “We need to go!” “Relax, I’ve been in plenty of burning buildings…” I nonchalantly pointed up at the stone ceiling. “It’s a whole ten minutes or so until we start choking…” And in my well-researched opinion we were in about as much danger from burning in this room, as during (the average) beach visit and with about as high a health risk. Sure, there were quite a lot of chemicals, materials and gear burning... but the stone room was simply so gigantic it didn't matter even before I perforated the windows. We could have piled the entire lab into a big, costly bonfire, and still not have any problems with the fumes or smoke. “Well, I don’t want to test that freaking hypothesis, so get going already!” I rolled my eyes, before picking up one of the object that had drawn me there to begin with. The black leather dog collar with a silver, diamond shaped tag near gleamed on my hoof, the words: ‘My Name Is Nemo Schwartz. If I’ve Forgotten That, Please Call My Parents.’ in a tiny inlaid script I could read without problem even from a distance. I took a second to feel it out for magic, but when I didn’t sense any I lobbed it over to Nemo. I didn’t stop to watch the full reaction since she’d a point… but I still smirked slightly at the big, goofy grin I caught a glimpse off before turning. There was a few other thing in the ‘new crap’ in-box; a literal cardboard-box with more strange scribbles on the side. ’Bearer artifacts. For immediate analysis. Any of this disappears and ends up as celebrity memorabilia, and you’ll wish I find you before any of the Princesses’ do! ~S.B.’ ’ Nemo’s collar had been hanging on the edge for some reason, and I’d probably have missed the nondescript thing if it hadn’t. It was just… a box, nothing more, nothing less. As my hoof closed on the silver chain and I felt the familiar magic, a sigh of relief I didn’t realize I’d been holding in forced itself out of me. I honestly felt whole once my pentacle was once more around my neck; as if something missing had quietly clicked in place despite how minor an act it might have looked from the outside. A whole empire of evil bastards after me, a plan seemingly decades in the making ruined within a day or so, and a roof crackling cheerfully on fire above me? “Oh, feels good to be back in the saddle again…” I murmured happily with a huge grin on my face, while rifling through the box of nicked knick-knacks. Most of it was the type of crap only the finest of stalker shrines are forged from. Disturbing, but only important in what it said about the stalkers thoughts and intentions for the stalk-e. A few tufts of hair and feathers I took great care in throwing on the nearest blaze. One of my business cards. A stark-white silk handkerchief that looked as if it had never been used with the letters ‘E. T’ engraved on the corner. Another business card, but for Mike my mechanic. A newspaper about that mess Nemo had caused in New York… My ears perked slightly. “Hey, question?” “...Yeah?” I held up the paper, including the picture of a rainbow ‘explosion’ from ground level. “If you’d actually seen that ‘show’ thing, how come?” Nemo flinched slightly, making the Super-Zero still unconscious but now over her back rustle slightly. “...You expected to run a four-minute mile the first time you went jogging?” With a hum threw the paper back into the box. “Fair enough, makes sense,” I grunted out, hoisting the whole thing on my back. I stalked over to the one remaining mirror, frowning a bit as I peered through it… And an idea struck me, and I made one of those split-second decisions I hoped I could live with afterwards. And down at the Golden Gate, this near dizzyingly fast stream of traffic clearly visibly thanks to the bird-eye perspective. “...No idea how jostled we’ll be on exit given the time differential.” I gave the load on Nemo’s back a nod. “And from the looks of things, we’ll need to hit the air flying…” It barely took a second for Nemo to get it, her ears glued back even as she stared from me to Super-Zero. I reached out and poked the still shiny (and creep-tastic) mace, now stuck inside her prison clothes like the best-worst shiv ever. “This box of loose ends and that party stick of yours are one thing…” I just swept a hoof at the hundred feet drop we’d be stepping out ‘on.’ “Don’t think even a pony can survive that, especially not with those kinds of wounds…” Nemo bit her lip, and started more or less prancing from hoof to hoof; eyes darting around between the flames. “...But… we can’t just leave her!” My own eyes darted around for a bit, taking in the rather modest flames, the still cold stone and giant windows any fireman worth his salt wouldn’t hesitated a moment in smashing. “...Why not?” I asked nonchalantly, making Nemo’s eyes near bulge out of her skull. “She’s been unconscious for so long by now, she probably has serious brain-damage. Hell, even her own comrades left her behind for dead.” I saw some hesitation in Nemo’s eyes, and she started prancing a bit slower. “Even if your own Not-Quite-Excalibur-But-Close-Enough theory isn’t right I near knocked Luna’s wings off just moments before we met,” I waved at the unconscious mare on her back. “And well, you saw her yourself, Nemo; right as rain within the freaking hour. Right now, that mare over your shoulders is just a casualty among many, just another corpse that gets to lie where she fell. Who cares if she technically died by that whopper, or the smoke after?” I saw a tiny bit of blood start pooling under her teeth, as Nemo bit her lip even harder. “Well? Your call, kid. Nopony seemingly cares if this mare lives or dies… and I’m not in a generous mood myself after being snatched from my own doorstep. Drop her into the bay, a bullet in her head to make it quick, or do you roll the dice for her…?” Nemo seemingly couldn’t keep looking at me, and turned her head away. “I’ve done the ‘let the minion live out of pity’ thing a few times before,” I continued in a slightly kinder voice. “Seems to be fifty-fifty if they actually care, or just come back swinging.” “...Do you know who this girl’s supposed to be?” Nemo forced out without turning her head. “And that unicorn you blew the horn off, for that matter?” “Look around, kid,” I swept a wing around the burning lab, making Nemo’s ears twitch slightly at the creak of the leather in my duster. “This isn't what it looks like when mind-wiped slaves are set on a task; this is the work of wide-eyed collaborators. Both these mares had the same chance to run as all the others in this room, but they picked to stand their ground…” “…You honestly don’t care, do you?” I frowned a bit at the heat in her words, but ultimately just shrugged. “Sure I do… I just care more about my actual friends than bastards that claim to be my friends.” Nemo finally turned her head, face unreadable. “...And if I told you the ‘old’ you used to foal sit these girls, and sing them lullabies…?” Hush now, quiet now... My eyes drifted over to the broken mare over Nemo’s back. She was still breathing, but even through the facemask I could see how near half her face was even now turning into one giant bruise; not helped by this thin dribble of spit and blood pooling on the floor. If she survived, it would be months if not years before she’d be able to chew her own food again, not even counting the ragged little stump that one wing of hers had become and how much work that would require. Or brain damage and that bag of horrors. And yet… was I really prepared to do this? She’d hurt me and mine, sure… But was I really a hundred and ten percent sure this was some wretched evil from beyond the void… The girl stirred slightly. She didn’t wake properly, but her eyelids fluttered, and she mumbled out something that sounded like “Drasch…” With a strangely soft clink another of her teeth hit the floor, a molar cracked clean at the root....or was this ‘just’ fate having thrown me another curveball covered in razor-wire? Wouldn’t even be the first time I’d stumbled headfirst into a tragedy decades in the making. Hell, a long, near fruitless search for friends long lost… ending in tragedy over misunderstandings and battle. One side refusing to give up their lives… the other in denial that their rescue might no longer be needed, or even wanted? It hadn’t quite struck me as that ‘Discord’ creep’s style, but some of the nastier immortals? The ending of entire cities and lineages in blood and flames is basically their version of a six-pack and a porno mag. Just no other way to kick of a really great weekend quite like it. I made my decision. Making sure Nemo actually got what wonder and horror there can be to a wizard's life was important with her talents, but the mare over her shoulder had suffered more than enough. Adding further to that by me using her like a living study-aid would just be cruel. Still, there are worse reactions to a trial by... well, fire, than asking somebody that seems competent for help. Just doesn't work all the time. I mean, what type of foolish, bleeding-heart pansy would waste time, energy and effort on an enemy that's already unconscious and half dead? And far behind enemy lines, at that? In a place where reinforcements can come at any second? You'd have to be a total idiot with a martyr complex a mile-wide to do something that stupid. With a sigh, I hoisted the box of my back. “Make a circle,” with a small kick, I sent it over towards Nemo, “and hold this.” That's my boy. Nemo frowned, but started putting Super-Zero down as gently as she could. “...What are you going to do?” “I, out of the goodness of my heart and the sagely wisdom inherent to all wizards…” I solemnly stated with a hoof to my chest, ignoring the just plain wrong disbelieving snort from behind me. “...intend to go reach way, way outside my comfort zone, and actually un-set a house on fire.” “...You mean extinguish it, right?” Nemo deadpanned, all while dragging one hoof behind her as she hobbled around in a circle. I waved her off without looking her way. “Cease your strange and complicated technobabble, woman; I’m the wizard here, so I get to mumble the archaic words nobody else gets.” I ignored the lies and slander being muttered my way, and instead leaned back on my haunches with my front hooves ‘clasped’ and wings half extended; my attention focused on —ugh, drawing in the ‘magic’ of this place. My earlier bits of wizardly C.V. had been quick and easy things, but for what I had in mind now I needed to ‘suck’ in quite a bit. Imagine there’s somebody you want to spit in the eye, but there is only rather yucky stuff around to do it with. Basically, the difference between a mouthful of mud spat out as quickly… and needing to swivel around what you damn well hope was snow with somebody’s spilled lemonade in it for long enough that it melts. Ugh… It’s like trying to eat raw bacon from a flipping dirty jockstrap! What the buck have they done with this place?! Still, I’d give the nudist creeps’ their due; whatever ‘purification’ ritual or whatever they’d used for this place it had left the magic almost amazingly pure, if strangely stagnant. So, ice from a glacier, but with a nasty layer of dust on top, perhaps? Either way, I’d never felt anything quite like it. I’d visited a ‘haunted’ shrine once that had been just short of falling into decay but still held some shred of serenity about it… but that was the closest I could think of. “...Dresden!” Nemo squeaked out from behind me. “Your wings!” I frowned, and turned my head, making sure to keep my spell winding up. I hadn’t even been feeling it, but there was an outline of St. Elmo’s fire all over my ‘wing-sleeves.’ An outline of bluish-white thunder crawling all over my wings, and all the way down towards my tail, the furthest ‘flames’ even arcing between the hairs on the same. Frankly, I was so baffled, only decades of experience in keeping my spells together despite the circumstances stopped me from just ‘dropping’ all the energy I’d drawn in. Thing is, I wasn’t even near the type of level of overflow to cause this type of side-effect. I was drawing in a bit more than normal, yes, but nowhere near a ‘full’ tank. Was this a side-effect to using my wings as foci? A bleed over of energy, manifesting as static electricity? Or was this the reason for the ‘stale’ magic? Some… ritual I’ve never encountered before that makes magic slightly less palatable, but easier to draw in, shape and use? Like processed baby-food compared to an actual meal, but with ambient energies? I squared my shoulders and looked forward again. Interesting, but now was not the time for nerding out about magic. Sweeping my wings forward, out and back, I let go of my spell. “Vento servitas!” I’d put some care into crafting this one, and oh boy did it show. The twin blasts of wind tore down the lines of gear and tables, sending it all in a shower of flaming debris towards the walls. Wrecked computers, paper ash, beakers of mysterious liquid… all of it got sent towards the windows. Or should I say, where the windows should have been. The antique glass and their frames simply couldn’t stand up against typhoon level winds from the inside, and with a rather satisfying mix of plopping and cracking sounds, soon followed by loud smacks of impacts as the frames themselves got pelted with flying debris. The crown on it all, though… was the chandelier. The massive, ancient chain it hung on simply wasn’t made to resist the forces from such a massive thing swinging, and snapped with a crack as if from a giant’s whip. With that paradoxical fast slowness of massive things falling, the chandelier fell; the strange not-quite candles in it streaming flames behind them during the fall. The crash was near deafening, cracking the floor for near twenty feet away, and sent shards of crystal and gilded metal all over the room. I shielded my eyes with my hoof, but none of the debris hit me. Chuckling I turned and trotted over to Nemo, who was staring at me slack-jawed. “...Well,” I smirked, as somepony finally got half a clue and started up the air-raid sirens, “the room isn’t on fire anymore, right?” Seemingly not feeling my grand act was worthy of mere words, Nemo just face-palmed. Philistine. With a snort, I waved a hoof at the one remaining mirror. “You go first with the box; try to keep a hold of it, but the important bit is that it doesn’t stay here.” With only a brief glance at the destruction and the unconscious girl, Nemo got to it. Balancing on both legs and with the box in her arms, Nemo leapt through the portal. It was actually a bit interesting to see from the outside. This line of bright light formed all around where she touched the mirror, but the image didn’t even distort as Nemo made it through. I stood still and looked for a bit, but aside from the ‘chipmunk’ speed, her passing seemed to have been uneventful. Box and everything still in ‘hoof,’
| 2
| 12,000
| 20,000
| 20,788
|
18,718
|
77e45c8d-729c-42ba-8c77-40f15b939c21
|
StampyAI/alignment-research-dataset/blogs
|
Will AI see sudden progress?
*By Katja Grace, 24 February 2018*
Will advanced AI let some small group of people or AI systems take over the world?
AI X-risk folks and others have accrued lots of arguments about this over the years, but I think this debate has been disappointing in terms of anyone changing anyone else’s mind, or much being resolved. I still have hopes for sorting this out though, and I thought a written summary of the evidence we have so far (which often seems to live in personal conversations) would be a good start, for me at least.
To that end, I started a [collection of reasons to expect discontinuous progress near the development of AGI](http://aiimpacts.org/likelihood-of-discontinuous-progress-around-the-development-of-agi/).
I do think the world could be taken over without a step change in anything, but it seems less likely, and we can talk about the arguments around that another time.
Paul Christiano had basically the same idea at the same time, so for a slightly different take, [here](https://sideways-view.com/2018/02/24/takeoff-speeds/) is his account of reasons to expect slow or fast take-off.
Please tell us in the comments or [feedback box](https://aiimpacts.org/feedback/) if your favorite argument for AI Foom is missing, or isn’t represented well. Or if you want to represent it well yourself in the form of a short essay, and send it to me [here](mailto:[email protected]), and we will gladly consider posting it as a guest blog post.
I’m also pretty curious to hear which arguments people actually find compelling, even if they are already listed. I don’t actually find any of the ones I have that compelling yet, and I think a lot of people who have thought about it do expect ‘local takeoff’ with at least substantial probability, so I am probably missing things.
| 0
| 0
| 440
| 440
|
108,719
|
<urn:uuid:e3679eac-88bf-4c1e-8795-842a1cd2156b>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
Game review: Halo CE Anniversary
By Troy Rawhiti-Forbes
Add a comment
The Master Chief pops off a shot at a Covenant Elite in Halo: Combat Evolved Anniversary. Photo / Supplied
When Halo: Combat Evolved was released ten years ago, it was swiftly proclaimed as the greatest console first-person shooter of all time.
The proclamations were correct. Now we look back and recognise it as the game that changed everything. Combat Evolved was a revolution for gamers who preferred a control pad to a mouse. It guaranteed success for Microsoft's Xbox and provided another vexation for Nintendo, whose industry dominance had already been seriously eroded by Sony and their PlayStation brand.
It was meant to be a real-time strategy game, a kind of sci-fi version of Myth, until Microsoft bought Bungie Studios and had them rework it into an FPS for exclusive release on the Xbox.
Everything Bungie did for that finished product remains in place for Anniversary, not just the good parts. The over-powered pistol is there. The backtracking through repetitive areas is there. The wonky cutscene animation, the bugs and glitches that are either old annoyances or central to a player's overall strategy, they are there too.
Combat re-sold
So what's the point of selling these old boots again?
The main selling point is the remastered campaign, and the ability to switch between classic and modern visuals at will. The Master Chief's journey covered a lot of ground at Installation 04, and there are near-unlimited opportunities to stop and flip the graphics to see what the artists have changed.
Seeing the ringworld, now rendered in stunning detail and set to a brilliantly remastered soundtrack, will be as close as franchise latecomers get to understanding the jaw-dropping wonder that gamers experienced years ago when they saw that Halo arching up into the sky for the first time.
The levels that benefit the most from the new look are in the latter half of the game, where players must trek and re-trek through long interior sections that have been transformed from drab halls into dazzling examples of Forerunner aesthetics.
343 Guilty Spark, the original's gloomiest level, trades in soup-thick fog for luminescent vegetation and pitch-black skies before taking the player underground for the first encounter with the horrible Flood. Speaking of which, just wait until you see the assimilated Jacob Keyes this time. It's truly chilling.
Great ... no, don't throw my last grenade!
The Kinect integration is sensible, enhancing the gameplay without really changing anything. With simple voice commands you can access and switch your weapons, alter your screen's video settings, even lob grenades. Be warned though that there is a split-second delay in the weapon commands taking effect, because in the wrong scenarios that might be a split-second too long.
Try to keep the verbal sarcasm to a minimum, also, lest Kinect mistakes "uh, great" for "grenade" and dutifully throws your last remaining pineapple straight into a space occupied by no enemies whatsoever.
The coolest Kinect feature is the analyse tool, which lets you scan an in-game object and transmit its data to a library for future examination.
The Xbox 360 upgrade brings a set of achievements. Some of them are fiendishly difficult, but they're not unreasonable. After all, players have had up to a decade to perfect their skills.
Bungie also had a decade to perfect the Halo style before handing over to 343i, and in hindsight it was clear that during this game's development, they were very much in the learning stages.
The new look won't wash away any of the frustration that veterans might have felt every time they found themselves lost in some of Halo's vast bunkers and landscapes. Even the prettiest scenery gets old when you don't know where you're going, but there are Easter eggs out there that may offer some hope. These are the skulls - first introduced in Halo 2 - that allow players to alter the way the game works, and the terminals.
When accessed, the terminals play animated shorts that expand upon the events at hand and point the way towards Halo 4. The stories draw upon the secrets of the wider Halo universe and could even be considered companions to the excellent Halo: Legends collection.
It might be fair to say Bungie were still finding their way when they led players into an all-out assault on a Covenant camp that brings in waves and waves of ground troops, boosted by plasma turrets and a pair of monstrous Hunters. Once you survive that, it's up into the Truth and Reconciliation cruiser to immediately face another swarm of troops, led by a team of invisible Elites, armed with kill-you-with-one-hit energy swords.
All this, in just the third level of the game.
A similar sequence of challenges takes place near the very end of the Reach campaign, suggesting Bungie had learned a thing or two about when to challenge players with against-all-odds encounters.
If you hated your game crashing and then restarting you off at far-off checkpoint in Combat Evolved, you'll hate it here too. 343i didn't miss a single error, but they left them in anyway in order to preserve the original's spirit. Who hasn't used the scene loading glitch to make an entire team of Covenant forget the Master Chief was ever there?
Double kill ... triple kill
The multiplayer mode, which runs off the Reach engine, is more of a tribute to the original experience than a true remake. If it looks like Reach and plays like Reach, it simply can't be anything else.
That's not to say it's a bad thing. It is a fine experience, though series newcomers may find it hard being mercilessly cut down by players who have been camping in their sniping spots for most of the last ten years. It can be a brutal learning curve, but we know what they say about pressure and diamonds.
Reach integration also means players don't have to abandon their characters' career progress and start all over again with Anniversary. If you were a grade four General before, you'll still be one when you jump into classic maps like Beaver Creek and Prisoner.
As with Reach, the maps are taken from Combat Evolved and Halo 2. Halo 3's maps remain on the shelf for the time being.
So, how about a second date?
It would have been easy for 343 Industries to just polish the graphics up and sell the product like hotcakes. They retained the core gameplay, but it's not the same overall experience. It's better. Kinect is introduced to Halo without cheapening either, and swapping between visual modes is a fun distraction from the horrors of futuristic space war.
If this is a "love letter" to Halo's fans, 343i appears to be a worthy suitor.
Stars: 4.5/5
Platform: Xbox 360
Rated: R16
Have your say
1200 characters left
© Copyright 2014, APN Holdings NZ Limited
Assembled by: (static) on red akl_a3 at 08 Mar 2014 21:30:01 Processing Time: 1101ms
| 0
| 0
| 1,581
| 1,581
|
17,617
|
831fbbc8-a062-4f1b-b4a2-421d29010f0e
|
StampyAI/alignment-research-dataset/arxiv
|
state helps focus on meaningful and informative details even under interference.
With the causal inference state stCIsubscriptsuperscript𝑠𝐶𝐼𝑡s^{CI}\_{t}italic\_s start\_POSTSUPERSCRIPT italic\_C italic\_I end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT, the output of the Q-network Q(stCI;θ)𝑄subscriptsuperscript𝑠𝐶𝐼𝑡𝜃Q(s^{CI}\_{t};\theta)italic\_Q ( italic\_s start\_POSTSUPERSCRIPT italic\_C italic\_I end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ; italic\_θ ) is set to be switched between two neural networks f2(stCI;θ2)subscript𝑓2subscriptsuperscript𝑠𝐶𝐼𝑡subscript𝜃2f\_{2}(s^{CI}\_{t};\theta\_{2})italic\_f start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT ( italic\_s start\_POSTSUPERSCRIPT italic\_C italic\_I end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ; italic\_θ start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT ) and f3(stCI;θ3)subscript𝑓3subscriptsuperscript𝑠𝐶𝐼𝑡subscript𝜃3f\_{3}(s^{CI}\_{t};\theta\_{3})italic\_f start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPT ( italic\_s start\_POSTSUPERSCRIPT italic\_C italic\_I end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ; italic\_θ start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPT ) by the interference label.
Such a switching mechanism prevents our network from over-generalizing the causal inference state. During training, switching between the two neural networks is determined by the training interference label ittrainsuperscriptsubscript𝑖𝑡traini\_{t}^{\text{train}}italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT. We assume that the true interference label is available in the training phase so ittrain=itsuperscriptsubscript𝑖𝑡trainsubscript𝑖𝑡i\_{t}^{\text{train}}=i\_{t}italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT = italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT.
In the testing, when itsubscript𝑖𝑡i\_{t}italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT is not available, we use the predicted interference label i~tsubscript~𝑖𝑡\tilde{i}\_{t}over~ start\_ARG italic\_i end\_ARG start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT as the switch to decide which of the two neural networks to use.
All the neural networks f1,f2,f3,fIsubscript𝑓1subscript𝑓2subscript𝑓3subscript𝑓𝐼f\_{1},f\_{2},f\_{3},f\_{I}italic\_f start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT, italic\_f start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT, italic\_f start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPT, italic\_f start\_POSTSUBSCRIPT italic\_I end\_POSTSUBSCRIPT have two fully connected layers222Though such manner may lead to the myth of over-parameterization, our ablation study proves that we can achieve better results with almost the same amount of parameters.
with each layer followed by the ReLU activation except for the last layer in f2,f3subscript𝑓2subscript𝑓3f\_{2},f\_{3}italic\_f start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT, italic\_f start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPT and fIsubscript𝑓𝐼f\_{I}italic\_f start\_POSTSUBSCRIPT italic\_I end\_POSTSUBSCRIPT. The overall CIQ model is shown in Figure [4](#Sx3.F4 "Figure 4 ‣ Causal Inference Q-Network ‣ Resilient Reinforcement Learning ‣ Training a Resilient Q-Network against Observational Interference") and θ=(θ1,θ2,θ3,ϕ)𝜃subscript𝜃1subscript𝜃2subscript𝜃3italic-ϕ\theta=(\theta\_{1},\theta\_{2},\theta\_{3},\phi)italic\_θ = ( italic\_θ start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT, italic\_θ start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT, italic\_θ start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPT, italic\_ϕ ) denotes all its parameters.
Note that, as common practice for discrete action spaces, the Q-network output Q(stCI;θ)𝑄subscriptsuperscript𝑠𝐶𝐼𝑡𝜃Q(s^{CI}\_{t};\theta)italic\_Q ( italic\_s start\_POSTSUPERSCRIPT italic\_C italic\_I end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ; italic\_θ ) is an 𝒜𝒜\mathcal{A}caligraphic\_A-dimensional vector where 𝒜𝒜\mathcal{A}caligraphic\_A is the size of the action space, and each dimension represents the value for taking the corresponding action.
Finally, we train the CIQ model Q(st′;θ)𝑄subscriptsuperscript𝑠′𝑡𝜃Q(s^{\prime}\_{t};\theta)italic\_Q ( italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ; italic\_θ ) end-to-end by the DQN algorithm with an additional loss for predicting the interference label.
The overall CIQ objective function is defined as:
| | | | |
| --- | --- | --- | --- |
| | | LCIQ(θ1,θ2,θ3,ϕ)=ittrain⋅LDQN(θ1,θ2,ϕ)superscript𝐿CIQsubscript𝜃1subscript𝜃2subscript𝜃3italic-ϕ⋅superscriptsubscript𝑖𝑡trainsuperscript𝐿DQNsubscript𝜃1subscript𝜃2italic-ϕ\displaystyle L^{\text{CIQ}}(\theta\_{1},\theta\_{2},\theta\_{3},\phi)=i\_{t}^{\text{train}}\cdot L^{\text{DQN}}(\theta\_{1},\theta\_{2},\phi)italic\_L start\_POSTSUPERSCRIPT CIQ end\_POSTSUPERSCRIPT ( italic\_θ start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT, italic\_θ start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT, italic\_θ start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPT, italic\_ϕ ) = italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT ⋅ italic\_L start\_POSTSUPERSCRIPT DQN end\_POSTSUPERSCRIPT ( italic\_θ start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT, italic\_θ start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT, italic\_ϕ ) | |
| | | +(1−ittrain)⋅LDQN(θ1,θ3,ϕ)+λ⋅(ittrainlogp(i~t|z~t;θ1,ϕ)\displaystyle+(1-i\_{t}^{\text{train}})\cdot L^{\text{DQN}}(\theta\_{1},\theta\_{3},\phi)+\lambda\cdot(i\_{t}^{\text{train}}\log p(\tilde{i}\_{t}|\tilde{z}\_{t};\theta\_{1},\phi)+ ( 1 - italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT ) ⋅ italic\_L start\_POSTSUPERSCRIPT DQN end\_POSTSUPERSCRIPT ( italic\_θ start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT, italic\_θ start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPT, italic\_ϕ ) + italic\_λ ⋅ ( italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT roman\_log italic\_p ( over~ start\_ARG italic\_i end\_ARG start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT | over~ start\_ARG italic\_z end\_ARG start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ; italic\_θ start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT, italic\_ϕ ) | |
| | | +(1−ittrain)log(1−p(i~t|z~t;θ1,ϕ))),\displaystyle+(1-i\_{t}^{\text{train}})\log(1-p(\tilde{i}\_{t}|\tilde{z}\_{t};\theta\_{1},\phi))),+ ( 1 - italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT ) roman\_log ( 1 - italic\_p ( over~ start\_ARG italic\_i end\_ARG start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT | over~ start\_ARG italic\_z end\_ARG start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ; italic\_θ start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT, italic\_ϕ ) ) ), | | (2) |
where λ𝜆\lambdaitalic\_λ is a scaling constant and is set to 1 for simplicity. Due to the design of the causal inference state and the switching mechanism, we will show that CIQ can perform resilient behaviors against the observation interferences. We introduce how to quantify the robustness of a Q-network under noisy observation in next subsection. The CIQ training procedure (Algorithm [1](#alg1 "Algorithm 1 ‣ Proof of the CLEVER-Q Theorem ‣ Appendix B A. Proof of the CLEVER-Q Theorem and Additional Robustness Measurements ‣ Training a Resilient Q-Network against Observational Interference")) and an advanced CIQ based on variational inference (Louizos et al. [2017](#bib.bib42)) are described in Appendix B.
### CLEVER-Q: A Robustness Evaluation Metric for Q-Networks
Here we provide a comprehensive score (CLEVER-Q) for evaluating the robustness of a Q-network model by extending the CLEVER robustness score (Weng et al. [2018](#bib.bib74)) designed for classification tasks to Q-network based DRL tasks. Consider an ℓpsubscriptℓ𝑝\ell\_{p}roman\_ℓ start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT-norm bounded (p≥1𝑝1p\geq 1italic\_p ≥ 1) perturbation δ𝛿\deltaitalic\_δ to the state stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT. We first derive a lower bound βLsubscript𝛽𝐿\beta\_{L}italic\_β start\_POSTSUBSCRIPT italic\_L end\_POSTSUBSCRIPT on the minimal perturbation to stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT for altering the action with the top Q-value, i.e., the greedy action. For a given stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT and a Q-network, this lower bound βLsubscript𝛽𝐿\beta\_{L}italic\_β start\_POSTSUBSCRIPT italic\_L end\_POSTSUBSCRIPT provides a robustness guarantee that the greedy action at stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT will be the same as that of any perturbed state st+δsubscript𝑠𝑡𝛿s\_{t}+\deltaitalic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT + italic\_δ, as long as the perturbation level ‖δ‖p≤βLsubscriptnorm𝛿𝑝subscript𝛽𝐿\|\delta\|\_{p}\leq\beta\_{L}∥ italic\_δ ∥ start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT ≤ italic\_β start\_POSTSUBSCRIPT italic\_L end\_POSTSUBSCRIPT. Therefore, the larger the value βLsubscript𝛽𝐿\beta\_{L}italic\_β start\_POSTSUBSCRIPT italic\_L end\_POSTSUBSCRIPT is, the more resilience of the Q-network against perturbations can be guaranteed. Our CLEVER-Q score uses the extreme value theory to evaluate the lower bound βLsubscript𝛽𝐿\beta\_{L}italic\_β start\_POSTSUBSCRIPT italic\_L end\_POSTSUBSCRIPT as a robustness metric for benchmarking different Q-network models. The proof of Theorem 1. is available in Appendix B.
######
Theorem 1.
Consider a Q-network Q(s,a)𝑄𝑠𝑎Q(s,a)italic\_Q ( italic\_s, italic\_a ) and a state stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT. Let 𝒜\*=argmaxaQ(st,a)superscript𝒜subscriptnormal-argnormal-max𝑎𝑄subscript𝑠𝑡𝑎\mathcal{A}^{\*}=\operatorname\*{arg\,max}\_{a}Q(s\_{t},a)caligraphic\_A start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT = start\_OPERATOR roman\_arg roman\_max end\_OPERATOR start\_POSTSUBSCRIPT italic\_a end\_POSTSUBSCRIPT italic\_Q ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT, italic\_a ) be the set of greedy (best) actions having the highest Q-value at stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT according to the Q-network. Define ga(st)=Q(st,𝒜\*)−Q(st,a)subscript𝑔𝑎subscript𝑠𝑡𝑄subscript𝑠𝑡superscript𝒜𝑄subscript𝑠𝑡𝑎g\_{a}(s\_{t})=Q(s\_{t},\mathcal{A}^{\*})-Q(s\_{t},a)italic\_g start\_POSTSUBSCRIPT italic\_a end\_POSTSUBSCRIPT ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ) = italic\_Q ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT, caligraphic\_A start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT ) - italic\_Q ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT, italic\_a ) for
every action a𝑎aitalic\_a, where Q(st,𝒜\*)𝑄subscript𝑠𝑡superscript𝒜Q(s\_{t},\mathcal{A}^{\*})italic\_Q ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT, caligraphic\_A start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT ) denotes the best Q-value at stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT.
Assume ga(st)subscript𝑔𝑎subscript𝑠𝑡g\_{a}(s\_{t})italic\_g start\_POSTSUBSCRIPT italic\_a end\_POSTSUBSCRIPT ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ) is locally Lipschitz continuous333Here locally Lipschitz continuous means ga(st)subscript𝑔𝑎subscript𝑠𝑡g\_{a}(s\_{t})italic\_g start\_POSTSUBSCRIPT italic\_a end\_POSTSUBSCRIPT ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ) is Lipschitz continuous within the ℓpsubscriptnormal-ℓ𝑝\ell\_{p}roman\_ℓ start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT ball centered at stsubscript𝑠𝑡s\_{t}italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT with radius Rpsubscript𝑅𝑝R\_{p}italic\_R start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT. We follow the same definition as in (Weng et al. [2018](#bib.bib74)). with its local Lipschitz constant denoted by Lqasuperscriptsubscript𝐿𝑞𝑎L\_{q}^{a}italic\_L start\_POSTSUBSCRIPT italic\_q end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT italic\_a end\_POSTSUPERSCRIPT, where 1/p+1/q=11𝑝1𝑞11/p+1/q=11 / italic\_p + 1 / italic\_q = 1 and p≥1𝑝1p\geq 1italic\_p ≥ 1. For any p≥1𝑝1p\geq 1italic\_p ≥ 1, define the lower bound
| | | | |
| --- | --- | --- | --- |
| | βL=mina∉𝒜\*ga(st)/Lqa.subscript𝛽𝐿𝑚𝑖subscript𝑛𝑎superscript𝒜subscript𝑔𝑎subscript𝑠𝑡superscriptsubscript𝐿𝑞𝑎\beta\_{L}=min\_{a\notin\mathcal{A}^{\*}}g\_{a}(s\_{t})/L\_{q}^{a}.italic\_β start\_POSTSUBSCRIPT italic\_L end\_POSTSUBSCRIPT = italic\_m italic\_i italic\_n start\_POSTSUBSCRIPT italic\_a ∉ caligraphic\_A start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT end\_POSTSUBSCRIPT italic\_g start\_POSTSUBSCRIPT italic\_a end\_POSTSUBSCRIPT ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ) / italic\_L start\_POSTSUBSCRIPT italic\_q end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT italic\_a end\_POSTSUPERSCRIPT. | | (3) |
Then for any δ𝛿\deltaitalic\_δ such that ‖δ‖p≤βLsubscriptnorm𝛿𝑝subscript𝛽𝐿\|\delta\|\_{p}\leq\beta\_{L}∥ italic\_δ ∥ start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT ≤ italic\_β start\_POSTSUBSCRIPT italic\_L end\_POSTSUBSCRIPT, we have
argmaxaQ(st,a)=argmaxaQ(st+δ,a)subscriptnormal-argnormal-max𝑎𝑄subscript𝑠𝑡𝑎subscriptnormal-argnormal-max𝑎𝑄subscript𝑠𝑡𝛿𝑎\operatorname\*{arg\,max}\_{a}Q(s\_{t},a)=\operatorname\*{arg\,max}\_{a}Q(s\_{t}+\delta,a)start\_OPERATOR roman\_arg roman\_max end\_OPERATOR start\_POSTSUBSCRIPT italic\_a end\_POSTSUBSCRIPT italic\_Q ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT, italic\_a ) = start\_OPERATOR roman\_arg roman\_max end\_OPERATOR start\_POSTSUBSCRIPT italic\_a end\_POSTSUBSCRIPT italic\_Q ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT + italic\_δ, italic\_a ).
Figure 5: Illustration of our environments on: (a) a 3D navigation task, banana collector (Juliani et al. [2018](#bib.bib32)), and (b) a video game, LunarLander (Brockman et al. [2016](#bib.bib6)).
Experiments
-----------
### Environments for DQNs
Our testing platforms were based on (a) OpenAI Gym (Brockman et al. [2016](#bib.bib6)), (b) Unity-3D environments (Juliani et al. [2018](#bib.bib32)), (c) a 2D gaming environment (Brockman et al. [2016](#bib.bib6)), and (d) visual learning from pixel inputs of cart pole. Our test environments cover some major application scenarios and feature discrete actions for training DQN agents with the CLEVER-Q analysis. For instance, Atari games and space-invaders are popular real-world applications. Unity 3D banana navigation is a physical simulator but provides virtual to real options for further implementations.
Vector Cartpole: Cartpole (Sutton et al. [1998](#bib.bib69)) is a classical continuous control problem. We use Cartpole-v0 from Gym (Brockman et al. [2016](#bib.bib6)) with a targeted reward =195.0absent195.0=195.0= 195.0. The defined environment is manipulated by adding a force of +11+1+ 1 or −11-1- 1 to a moving cart.
Banana Collector: The Banana collector shown in Figure [5](#Sx3.F5 "Figure 5 ‣ CLEVER-Q: A Robustness Evaluation Metric for Q-Networks ‣ Resilient Reinforcement Learning ‣ Training a Resilient Q-Network against Observational Interference") (a) is one of the Unity 3D baseline (Juliani et al. [2018](#bib.bib32)). Different from the MuJoCo simulators with continuous actions, the Banana collector is controlled by four discrete actions corresponding to moving directions. The targeted reward is 12.012.012.012.0 points by accessing correct bananas (+11+1+ 1). The state-space has 37 dimensions included velocity and a ray-based perception of objects around the agent.
Lunar Lander: Similar to the Atari gaming environments, Lunar Lander-v2 (Figure [5](#Sx3.F5 "Figure 5 ‣ CLEVER-Q: A Robustness Evaluation Metric for Q-Networks ‣ Resilient Reinforcement Learning ‣ Training a Resilient Q-Network against Observational Interference") (c)) is a discrete action environment from OpenAI Gym (Brockman et al. [2016](#bib.bib6)) to control firing ejector with a targeted reward of 200200200200. The state is an eight-dimensional vector that records the lander’s position, velocity, angle, and angular velocities. The episode finishes if the lander crashes or comes to rest, receiving a reward −100100-100- 100 or +100100+100+ 100 Firing ejector costs −0.30.3-0.3- 0.3 each frame with +1010+10+ 10 for each ground contact.
Pixel Cartpole:
To further evaluate our models, we conduct experiments from the pixel inputs in the cartpole environment as a visual learning task. The size of input state is 400×600400600400\times 600400 × 600. We use a max-pooling and a convolution layer to extract states as network inputs. The environment includes two discrete actions {left,right}𝑙𝑒𝑓𝑡𝑟𝑖𝑔ℎ𝑡\left\{left,right\right\}{ italic\_l italic\_e italic\_f italic\_t, italic\_r italic\_i italic\_g italic\_h italic\_t }, which is identical to the Cartpole-v0 of the vector version.
### Baseline Methods
In the experiments, we compare our CIQ algorithm with two sets of DQN-based DRL baselines to demonstrate the resilience capability of the proposed method. We ensure all the models have the same number of 9.7 millions parameters with careful fine-tuning to avoid model capacity issues.
Pure DQN: We use DQN as a baseline in our experiments. The DQN agent is trained and tested on interfered state st′subscriptsuperscript𝑠′𝑡s^{\prime}\_{t}italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT. We also evaluate common DQN improvements in Appendix C and find the improvements (e.g., DDQN) have no significant effect against interference.
DQN with an interference classifier (DQN-CF):
In the resilient reinforcement learning framework, the agent is given the true interference label ittrainsuperscriptsubscript𝑖𝑡traini\_{t}^{\text{train}}italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT at training.
Therefore, we would like to provide this additional information to the DQN agent for a fair comparison.
During training, the interfered state st′subscriptsuperscript𝑠′𝑡s^{\prime}\_{t}italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT is concatenated with the true label ittrainsuperscriptsubscript𝑖𝑡traini\_{t}^{\text{train}}italic\_i start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT train end\_POSTSUPERSCRIPT as the input for the DQN agent.
Since the true label is not available at testing, we train an additional binary classifier (CF) for the DQN agent.
The classifier is trained to predict the interference label, and this predicted label will be concatenated with the interfered state as the input for the DQN agent during testing.
DQN with safe actions (DQN-SA):
Inspired by shielding-based safe RL (Alshiekh et al. [2018](#bib.bib1)), we consider a DQN baseline with safe actions (SA). The DQN-SA agent will apply the DQN action if there is no interference. However, if the current observation is interfered, it will choose the action used for the last uninterfered observation as the safe action. This action-holding method is also a typical control approach when there are missing observations (Franklin et al. [1998](#bib.bib14)).
Similar to DQN-CF, a binary classifier for interference is trained to provide predicted labels at testing.
DVRLQ and DVRLQ-CF: Motivated by deep variational RL (DVRL) (Igl et al. [2018](#bib.bib27)), we provide a version of DVRL as a POMDP baseline. We call this baseline DVRLQ because we replace the A2C-loss with the DQN loss. Similar to DQN-CF, we also consider another baseline of DVRLQ with a classifier, referred to as DVRLQ-CF, for a fair comparison using the interference labels.
(a) Cartpolevectoradversarialsubscriptsuperscriptabsent𝑎𝑑𝑣𝑒𝑟𝑠𝑎𝑟𝑖𝑎𝑙𝑣𝑒𝑐𝑡𝑜𝑟{}^{adversarial}\_{vector}start\_FLOATSUPERSCRIPT italic\_a italic\_d italic\_v italic\_e italic\_r italic\_s italic\_a italic\_r italic\_i italic\_a italic\_l end\_FLOATSUPERSCRIPT start\_POSTSUBSCRIPT italic\_v italic\_e italic\_c italic\_t italic\_o italic\_r end\_POSTSUBSCRIPT.
(b) Bananaadversarial𝑎𝑑𝑣𝑒𝑟𝑠𝑎𝑟𝑖𝑎𝑙{}^{adversarial}start\_FLOATSUPERSCRIPT italic\_a italic\_d italic\_v italic\_e italic\_r italic\_s italic\_a italic\_r italic\_i italic\_a italic\_l end\_FLOATSUPERSCRIPT.
(c) Lunaradversarial𝑎𝑑𝑣𝑒𝑟𝑠𝑎𝑟𝑖𝑎𝑙{}^{adversarial}start\_FLOATSUPERSCRIPT italic\_a italic\_d italic\_v italic\_e italic\_r italic\_s italic\_a italic\_r italic\_i italic\_a italic\_l end\_FLOATSUPERSCRIPT.
(d) Cartpolepixeladversarialsubscriptsuperscriptabsent𝑎𝑑𝑣𝑒𝑟𝑠𝑎𝑟𝑖𝑎𝑙𝑝𝑖𝑥𝑒𝑙{}^{adversarial}\_{pixel}start\_FLOATSUPERSCRIPT italic\_a italic\_d italic\_v italic\_e italic\_r italic\_s italic\_a italic\_r italic\_i italic\_a italic\_l end\_FLOATSUPERSCRIPT start\_POSTSUBSCRIPT italic\_p italic\_i italic\_x italic\_e italic\_l end\_POSTSUBSCRIPT.
(e) Cartpolevectorblackoutsubscriptsuperscriptabsent𝑏𝑙𝑎𝑐𝑘𝑜𝑢𝑡𝑣𝑒𝑐𝑡𝑜𝑟{}^{blackout}\_{vector}start\_FLOATSUPERSCRIPT italic\_b italic\_l italic\_a italic\_c italic\_k italic\_o italic\_u italic\_t end\_FLOATSUPERSCRIPT start\_POSTSUBSCRIPT italic\_v italic\_e italic\_c italic\_t italic\_o italic\_r end\_POSTSUBSCRIPT.
(f) Bananablackout𝑏𝑙𝑎𝑐𝑘𝑜𝑢𝑡{}^{blackout}
| 2
| 12,000
| 20,000
| 25,213
|
38,047
|
b9a5129c-8c3b-429e-9c23-682161d0c66b
|
trentmkelly/LessWrong-43k
|
AGI Ruin: A List of Lethalities
Preamble:
(If you're already familiar with all basics and don't want any preamble, skip ahead to Section B for technical difficulties of alignment proper.)
I have several times failed to write up a well-organized list of reasons why AGI will kill you. People come in with different ideas about why AGI would be survivable, and want to hear different obviously key points addressed first. Some fraction of those people are loudly upset with me if the obviously most important points aren't addressed immediately, and I address different points first instead.
Having failed to solve this problem in any good way, I now give up and solve it poorly with a poorly organized list of individual rants. I'm not particularly happy with this list; the alternative was publishing nothing, and publishing this seems marginally more dignified.
Three points about the general subject matter of discussion here, numbered so as not to conflict with the list of lethalities:
-3. I'm assuming you are already familiar with some basics, and already know what 'orthogonality' and 'instrumental convergence' are and why they're true. People occasionally claim to me that I need to stop fighting old wars here, because, those people claim to me, those wars have already been won within the important-according-to-them parts of the current audience. I suppose it's at least true that none of the current major EA funders seem to be visibly in denial about orthogonality or instrumental convergence as such; so, fine. If you don't know what 'orthogonality' or 'instrumental convergence' are, or don't see for yourself why they're true, you need a different introduction than this one.
-2. When I say that alignment is lethally difficult, I am not talking about ideal or perfect goals of 'provable' alignment, nor total alignment of superintelligences on exact human values, nor getting AIs to produce satisfactory arguments about moral dilemmas which sorta-reasonable humans disagree about, nor attaining an abs
| 0
| 0
| 450
| 450
|
62,562
|
ec26cd8a-2123-43d3-a6b0-4b1d207359e5
|
trentmkelly/LessWrong-43k
|
Embedding safety in ML development
This post was written as part of Refine. Thanks to Adam Shimi, Alexander Gietelink Oldenziel, and Vanessa Kosoy for helpful discussion and feedback.
Summary
This post aims to:
* Advocate for embedding safety into development of machine learning models
* Propose a framing on how to think about safety, where three factors contribute to an AI being dangerous, namely misalignment, optimization, and influence
* Discuss the pros and cons of this framing
None of the ideas in this post are novel, and some of them may not even be practical, but they may be a useful framing to the problem of AI safety.
Introduction
Motivations
There are many ways to get to AGI. It could be recursively self improving AI, or something more brain-like, or something that comes from improvements in machine learning systems, or something else much more alien to us. I don’t know which pathways are the ones our world will eventually end up in, but given recent rapid advances in ML recently, in this post I will address the scenario where the first dangerous AGI is achieved by mostly scaling up and improving upon present-day ML systems.
In the absence of a robust solution to the hard problem of alignment, I think we should at least try to make the development of machine learning models go safely for as long as possible. This may be especially valuable if the first AGI we get is not yet an all-powerful superintelligent. It might be a fairly weak AGI capable of inflicting a lot of damage, but the damage could have been prevented. Not having a solution to the hard problem should not mean getting ourselves into preventable disasters.
What safety means in this context
Most safety fields in the real world don’t solve safety with guarantees. Whether it is flying an aircraft, operating a nuclear plant, or constructing a building, safety is all about effective risk management. This framing of safety has not been commonly used in the field of AI safety, for very good reasons, as the problem of a
| 0
| 0
| 446
| 446
|
32,335
|
3d6a7b99-ea52-4198-a2d9-25a6449ff637
|
trentmkelly/LessWrong-43k
|
Meetup : Prediction Markets and Futarchy
Discussion article for the meetup : Prediction Markets and Futarchy
WHEN: 07 September 2014 03:30:00PM (-0400)
WHERE: 98 Elm St Apt 1, Somerville, MA
I will give a talk about prediction markets and futarchy. The talk is intended as a basic introduction for people who are new to the concept. After my slides, I hope to have a discussion about whether futarchy is feasible.
Cambridge/Boston-area Less Wrong meetups start at 3:30pm, and have an alternating location:
1st Sunday meetups are at Citadel in Porter Sq, at 98 Elm St, apt 1, Somerville.
3rd Sunday meetups are in MIT's building 66 at 25 Ames St, room 156. Room number subject to change based on availability; signs will be posted with the actual room number.
(We also have last Wednesday meetups at Citadel at 7pm.)
Our default schedule is as follows:
—Phase 1: Arrival, greetings, unstructured conversation.
—Phase 2: The headline event. This starts promptly at 4pm, and lasts 30-60 minutes.
—Phase 3: Further discussion. We'll explore the ideas raised in phase 2, often in smaller groups.
—Phase 4: Dinner.
Discussion article for the meetup : Prediction Markets and Futarchy
| 0
| 0
| 314
| 314
|
34,251
|
7eaa1250-a1ca-4ba2-be86-1f06c74c43f7
|
trentmkelly/LessWrong-43k
|
What does “torture vs. dust specks” imply for insect suffering?
There is a non-negligible probability that insects suffer. Many insects have nociception. Furthermore, the argument-by-analogy approach to determining whether an animal suffer works—an insect’s response to stimuli that would cause pain in other animals is similar. It is estimated that there are 10^19 insects living at any point in time. If one believes that dust specks on a sufficiently large number of people (this doesn’t have to be 3^^^3 after all) is worse than torture of one person, does that mean insect suffering should swamp many other considerations when prioritizing social problems to address? (This is not meant to be a “proves-too-much” argument against siding with dust specks; it’s a genuine question.)
| 0
| 0
| 181
| 181
|
18,473
|
fde8e38c-2476-4e87-ac05-985b69dc4679
|
StampyAI/alignment-research-dataset/blogs
|
KvB2q48/7Y+6akLziLjsLQ2xoW9fLyh2r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| 368
| 2,208,000
| 2,216,000
| 2,282,028
|
4,022
|
4c9a802b-b292-4659-8caa-0e59f87ec71d
|
StampyAI/alignment-research-dataset/alignmentforum
|
Analogies between Software Reverse Engineering and Mechanistic Interpretability
These are notes taken during a call with [Itay Yona](https://mentaleap.ai/), an expert in software/hardware reverse engineering (SRE). Itay gave me an excellent distillation of key ideas and mindsets in the field, and we discussed analogies/disanalogies to mechanistic interpretability of neural networks. I’m generally very excited to learn about other fields of study that reverse engineer complex systems, and what relevant insights they may have (SRE, neuroscience, systems biology, etc). All mistakes are mine, and all insights are his!
My Takeaways
------------
* The underlying mindset actually feels pretty analogous!
+ I find it super interesting that they also think a lot about motifs (weird patterns and phenomena that only occur in specific contexts), and that these are often the first hook into understanding something weird and that you can then work backwards.
- (Not to be confused with the [SRE use of hooking](https://www.wikipedia.org/wiki/Hooking))
+ Also interesting that they also often focus on the inputs and outputs of the software as the starting point, to get a hook in, and then move on from there.
+ It's very key to have a deep, gears-level model of the system you're working with (how does a CPU work, how are things represented in memory, the stack, registers, etc)
* The distinction between "newbies get caught up trying to understand every detail, experts think in higher-level abstractions, make educated guesses, and only zoom in on the details that matter" felt super interesting and surprising to me.
+ My attempt to translate it into mechanistic interpretability is that (if it *is* analogous):
- There are certain principles and patterns by which networks learn, that we can identify and understand.
- We likely will understand these by deeply reverse engineering specific parts of systems (especially toy systems) and digging into the details. But the goal here is to build intuitions and mental models, and a sense for how models work as a whole.
- Once we have these intuitions and some solid grounding in what we do understand well, the best mindset for reverse engineering an unfamiliar system is to be less rigorous and more intuitive. Make educated guesses, look for partial evidence for or against hypotheses, think at a high-level and somewhat abstract mode about the system, and only zoom in on a specific part of the system to deeply reverse engineer once you've identified what to prioritise.
+ I have no idea if it is analogous, but that mindset aligns a fair bit with my intuitions about MI (though I consider the field to be much more in the "building intuitions by deeply engineering things" phase lol)
+ As [Lawrence Chan notes](https://www.alignmentforum.org/posts/tsYcsZAkKsqLXC3Bu/analogies-between-software-reverse-engineering-and?commentId=TGgwvSscYEh2cRDHA), this is likely an example of a general pattern, where newbies (1) reason high-level in a very ungrounded way, (2) dig really rigorously into the details constantly and (3) build intuition that's actually grounded, and return to reasoning on a high-level. And that if you want to get to 3, you need to do a lot of digging through details in stage 2 first, while experts can make the mistake of recommending skipping to 3 (which in practice puts people at 1)
* I'm surprised at the emphasis on prioritisation, and identifying which part of the software you care about. My mental picture was that the goal was to fully de-compile things to source code, but it sounds like that's rarely the goal and is extremely hard.
+ But this aligns with my intuitions that a lot of what *I* want to do with a network is to localise the parts that are relevant to a specific task.
* One approach to MI research that seems natural from a SRE perspective: Do extensive work reverse engineering toy models, and try to deeply understand the circuits there. Then, try to distill out motifs and find (ideally automated) tools to detect similar circuits in bigger toy models/real models.
+ [Induction heads](https://transformer-circuits.pub/2022/in-context-learning-and-induction-heads/index.html) seem like a successful example of this, and I’m very curious about whether this works for [my modular addition algorithm](https://www.alignmentforum.org/posts/N6WM6hs7RQMKDhYjB/a-mechanistic-interpretability-analysis-of-grokking) or [a toy model of superposition](https://transformer-circuits.pub/2022/toy_model/index.html)!
* *If* this works, it suggests that an approach to MI could be:
1. Identify a behaviour we care about/want to find a circuit for
2. Train a toy model to do this, that tries to be as simple as possible while being a good simulation of the real model
3. Study how the behaviour is implemented
4. Distill out the core patterns and fingerprints of this behaviour
5. Scan the real network for these, and use this as (non-rigorous) evidence for what’s going on.
* I’m biased, because this is how I already think about MI, but it’s nice to have some external validation!
For more discussion on similarities, see Chris Olah’s essay [Mechanistic Interpretability, Variables, and the Importance of Interpretable Bases](https://transformer-circuits.pub/2022/mech-interp-essay/index.html)
Call Notes
----------
* Background:
+ SRE = Software Reverse Engineering, the study of decompiling a program binary to the actual code.
+ Program binary -> assembly is easy, so it's basically about reverse engineering assembly.
+ Program files are enormous, want to understand what it's doing as much as you can, while translating as *few* instructions as possible
- Can be a few MB to 100s of MB in file size, rarely KB
* **High-level:** - we are *not* trying to be really rigorous or fully reverse engineer things, this is a massive waste of time. We are using the scientific method. We have strong priors, make educated guesses, and the goal of our reverse engineering is to get out a few bits of data that narrows down our educated guesses.
* **Key challenge:** - we want to find a specific function in the enormous file. The way we do this by identifying which part of the file is relevant.
+ **Meta point**: The goal is *not* to decompile the entire program. It's to identify the specific part of it that *you* care about understanding, or to zoom in on a specific vulnerability, or to figure out what it does at a high-level, etc. You need to prioritise and be intentional, not to dig into every single last detail.
- Often you care about specific details of a part of the software and how they deal with individual bits - does the function take in shorts vs longs, how does it cast, etc. These are often the details that lead to vulnerabilities, which is a lot of the goal here.
+ Main technique: Run the code a bunch on a bunch of inputs, get a black box sense for what it does. Use this to build intuitions about the program.
+ Form hypotheses about what could be going on internally, and use this black box operation to test and validate them. Try to break things, and look for inputs that could distinguish between your hypotheses
+ Try to get into the mindset of the programmer who wrote it
- He often thinks in C and classes, rather than in terms of assembly. You want to be as abstract as is reasonably possible without losing sight of the problem.
* Specifics of registers etc are often not a helpful way to think about it.
* You ultimately want to think about things as the programmer would have thought about things. The compiler and OS can often add a lot of irrelevant complexity. If you think in terms of assembly you'll need to spend a lot of time engaging with these, but if you think in C you can abstract a lot away
- This is both about "what is a reasonable way to implement a behaviour" or "there are many ways to implement this behaviour, which one might the programmer have chosen?"
* We can rule out possibilities and explore by playing with the software, look at what triggers errors, weird edge cases, etc.
* **Core mindset:** His core underlying philosophy is about looking for shortcuts.
+ Newbies look for specifics. Key details of what variables more from where to where, details of memory, etc. Very bottom up, easy to waste a lot of time.
+ Experts think in terms of abstractions - what is it doing, what are the variables, where are they, what is the code doing, etc. Top down, only zoom in once they've identified what's interesting.
- Analogy to chess: Newbies care a lot about whether a piece can be captured, is it protected, etc. As an expert, it's automatic, which lets us build better abstractions, and we can reason about strategy on a higher level.
+ To be a good SRE, understanding registers, loops, etc is important. But when you understand *how* to do this, it's enough.
- But a common mistake is to never switch to top down. This is one of the best forms of value a mentor gives, showing the worth of educated guesses, rapid moving on, skimming at a high level until you find the part worth zooming in on.
* People often learn this by pair programming, and seeing that a mentor is willing to not be a perfectionist and move on.
* **Localising:** How do I localise which part of the code matters?
+ High-Level - look for motifs in the code, and look at how it interfaces with external things (eg web or user interface)
+ Big disanalogy to neural networks: In a network, finding the representations is a lot of effort and very hard. But in software, we mostly care about how it represents the inputs, which are very easy and standard.
- Eg, the software acts on an image, and we know there are some standard operations - applying a fourier transform, storing the height and width, de-compression, etc.
- Example: If we know this is happening, then the code will be doing floating point multiplication. This is an unusual motif (rarely occurs in other contexts) and is easy to identify from the code, and a significant hint that de compression is happening
* Underlying point: There’s a lot of weird crap that gets put into the binary from the compiler, or standard code from the underlying libraries/OS. We want to sift through for the parts specific to this program and the behaviour we care about.
* This means that if we see a big block of code that does floating point multiplication, and it's called by parts of code that handles input images, it's very likely to be de-compression. On its own, being called by the input handling code is only weak evidence, but combined with floating point multiplication and our priors, this is strong evidence
- Summary: There is a special motif (floating point arithmetic) which is only used in some specific functional context. If you recognise that this is happening, this is a major hint about what’s going on!
+ Another type of work is using a debugger - this is dynamic analysis, not static analysis.
- We can look at what happens when the part of the code inside the de-compression algorithm is being run, and then look at the call stack. This can give us hints about what those functions do! Eg looking for user input.
- I love that the same ideas of static (analysing weights) vs dynamic (analysing the model on various inputs) analysis applies here!
+ Example: We want to crack photoshop to not need a serial number. We know that an incorrect serial number creates a pop-up. We can add a breakpoint to *every* part of the system that creates a pop-up, because this requires an interface with the UI. By seeing which breakpoint triggers, we can identify *which* call to MessageBoxA (the actual function name!) is relevant. And by looking at the call stack, we can identify which parts of the code are relevant.
* **A Practical workflow to balance between divergent, convergent and meta work:**
+ *This is Itay’s specific workflow, and is neither SRE specific nor universal in SRE.*
+ **Summary:**
- Start the project by forming a creative mindmap. Start with your end-goal (eg finding a vulnerability in a specific system) and systematically break down the problem, think of as many in-roads as you can. Then prioritise and pick one to work on.
- Spend most of your in focused research mode, and just keep iterating on that sub-problem. Each day, set yourself a question, try to answer that question, and then pick the natural next question the next day.
- Predict how long this will take. When that time is up, move on to the next most promising thing
+ Research diaries:
- At the start of the day, write out a question. By the end of the day, try to answer it.
- This hopefully unlocks more questions and threads, and you can iterate!
- Each day, you default to following your research diary. What questions now feel alive, dive into those, keep iterating.
+ Mindmaps:
- Start with the fundamental goal, eg “get the system to think you’ve given it the right password”
- Branch out and brainstorm sub-goals, sub-sub-goals, etc.
- Prioritise and choose one to focus on - do a rough cost-effectiveness analysis
+ *Most* of the time, you're in very focused mode and trying to answer a deep question. You use the mindmap to keep track of new ideas and creative thoughts without breaking focus.
+ Return to the mindmap when things take longer than expected (even if you haven't run out of ideas!). Then pick the next promising thread
- **Important:** Be willing to go back to previously explored threads even if there are more threads you haven't explored yet.
- It's hard to tell if you're in a rabbit hole or the solution is nearby. Thinks it's hard to tell whether a thread is promising or not in the moment. You can get *weak* evidence from partial progress, but idk.
- The goal is not having an *optimal* algorithm, it's about being good enough. "Work on the most promising thing for X time, then move on to next most promising thing".
* As long as you have good time estimates, traversing the mindmap graph is like applying the [A\* Algorithm](https://www.wikipedia.org/wiki/A*_search_algorithm) to it
+ **How to tell when you're in a rabbit hole?** This is when it's most useful to have a mentor/rubber duck/break. There's no clear rules here, but with time and experience you develop taste for "there should be an easier solution, I've been on this for a while, let's switch".
- Implicitly, you should have a model of what doing this research looks like, how hard it is, how often you should make progress, etc. If you're repeatedly wrong/optimistic, you notice and update or pivot.
+ **How would you mentor someone to learn this workflow?** Pretty easy, it's very concrete, you can just outline the algorithm for the mentee to follow.
- Concretely, take a mentee who’s stuck on a problem. Listen to what’s going on for them, repeat the approach of expanding the mindmap and being willing to return to previous threads. This basically seemed to be enough to get them unstuck and motivated again!
* **Teaching:** What would you put on a curriculum to make someone an expert SRE?
+ Technical details:
- Static vs dynamic analysis (analyse binary vs analyse it on inputs)
+ Teach common patterns:
- Eg De-compression has a lot of floating point multiplications, looking at inputs, changing things on the GUI, etc
- Give them toy problems to study with easy hooks in (ie patterns to give a starting point)
+ Intermediate problems which don't have deliberate, obvious hooks in
- But there probably isn't something that has *no* hooks. The core principles of the field is that things can be understood! But it might be hard and obfuscating to access it.
+ Move towards bigger systems. Don't just remove hints, but also obfuscate things.
- Eg code is partially encrypted, lots of distractor code, etc.
- What do you do here? Eh, not any general tips. But fundamentally, the computer can understand it, so it should be understandable, eg it must *eventually* be decrypted
- Mindset: *Assume* there is a shortcut. What is that shortcut? Use this to motivate your research.
* **Hard cases:**Trying to reverse engineer a mysterious black box.
+ *This is a pretty in the weeds example. Might be interesting but feel free to skip.*[*Another interesting example*](https://www.youtube.com/watch?v=vHPsdaXRdoo)
+ **High Level takeaway**: We need to rely on problem-solving principles even more, and focus on the constraints and bottlenecks of the system (eg, if it's accessing the internet, it must be doing that *somehow*)
+ Example: We have a binary that we extract from our router. We know it's router-y but know nothing else
+ It's not a classic case of software (like Photoshop) where we have documentation and know what the user interface is and what it should be doing. We know that it's doing router-related things, but not whether it's implemented with Linux or Windows, what language, what hardware, etc.
+ Sometimes we can guess, eg "this is using ARM's architecture/implementation", but sometimes we don't know what it is or aren't familiar with it, and need to start from scratch.
+ What do we do now? We still try to be a scientist! What do we know about what the software *must* be doing?
- Example: If it is a CPU, it must be having loops, accessing and moving memory, etc.
- Eg int manipulating commands may be 1 byte, other commands may be 2 bytes.
- Eg the code must be accessing registers, which must have locations represented internally
+ The underlying goal is to look for any chink in its armor! Try to find *any* pattern. This gives you an insight into what's going. You can identify what some parts of the code are doing, and make progress from there.
* **Different types of reverse engineering:**
+ To do web hacking, we don't even have access to the server-side code and binaries. We can see how it responds to various inputs. Eg put a backtick into a text box, if it creates a 500 error this suggests it's using SQL
+ Hardware reverse engineering - interesting and is at the intersection. Like web bc it's black box, like software because we can break things, change temperature, etc:
- Example: CPUs will execute instructions even with insufficient power. They will just fail to do the power intensive operations
* Eg: Will fail to access memory, but can still add numbers. You can use this to hack a password, because it can't access what the password is, we can get in with all zeros
- [A great talk Itay recommends](https://www.youtube.com/watch?v=6fOjGLCctEY)
- The jargon for this is side-channel attacks
* Sometimes operations take more or less depending on the input, by precisely measuring the time, we can figure it out.
+ Example: runs through each byte of a string and returns at the first one that's correct. This means we can brute force the first character, then the second character, etc, because we can tell how many bytes we have correct.
| 0
| 0
| 4,502
| 4,502
|
57,831
|
46fc6882-c536-4d56-b01d-9a26846c9376
|
trentmkelly/LessWrong-43k
|
To know or not to know
We are on the brink of the unimaginable. Humanity is about to cross a threshold that will redefine life as we know it: the creation of intelligence surpassing our own. This is not science fiction—it’s unfolding right now, within our lifetimes. The ripple effects of this seismic event will alter every aspect of society, culture, and existence itself, faster than most can comprehend.
But here’s the pressing question: should everyone know about this impending transformation, or are some truths too overwhelming to bear?
In this article, I won’t debate the feasibility of Artificial General Intelligence (AGI) or its exponential progress. Instead, I’ll dive into the heart of the matter: the arguments for and against widespread awareness of these changes. Should humanity be fully informed and mentally prepared for a future dominated by superintelligence, or is ignorance a better path to preserve our sanity and way of life? Let’s explore.
I’ll go with the arguments that favor knowledge first. Even if you can’t actually influence the outcome (although it can be debated till death, there could be a miniscule effect of the populace affecting big tech and government policies but let’s assume they can’t), you can certainly affect other aspects of your personal life such as the following.
Live fast
Living life to the fullest in the scene of the roaring 20s or the cocktail parties in the bunkers while Hitler was bombing London. I am recently being moved a lot by conversations with Ukranians that continue to try to live a normal life in Kyiv. There is an awesome aspect of human psychology that changes gears to live fast and die young mentality. It can cut both ways, I am certainly not fascinated by doing stupid stuff like snowballing drugs to end up like Heath Ledger but taking life changing trips now instead of later. Telling the girl/boy you have a crush today and generally forfeiting the work hard to satisfy the needs of my pension or my unborn grandchildren mentality that b
| 0
| 0
| 437
| 437
|
88,119
|
<urn:uuid:42359b17-71fc-46bd-96b7-27e59264f0ed>
|
Kyle1668/dclm-dedup-25B-ai-scifi-docs
|
[CLS]Ten Reasons Rick Perry Isn’t Getting the New Amazon Tablet
I went to bed late last night, late for someone who normally wakes up at five, anyway, after spending three or four hours trying to unlock the flamethrower in the new Jesus Zombie Massacre HD game for the iPad. All of the caffeine and willpower eroded from my system by two o’clock, and I fell into a deep sleep, dreaming about seeing Flea on the Brooklyn Bridge, panhandling and playing Frank Sinatra songs on an acoustic guitar with only three strings.
Just as the Ambien and Gabapentin cocktail reached full fury, the phone rang, and like an idiot, I picked it up.
“Konrath! Wrath of Kon! How’s it going in Baghdad on the Bay?” I could recognize the Texas drawl anywhere: it was Rick Perry. I’ve probably written about this before, but me and Rick go way back. Even though we completely disagree on religion and politics, he’s been a big fan of Rumored to Exist ever since some of the excerpts were published in zines back in the 90s, probably because all of the sheep-fucking references. I think his early days as Texas Agriculture Commissioner put the zap on his relationship with farm animals, or he developed a sick sense of humor about bestiality jokes. Either way, we’ve done a lot of zine trades over the years, and he’s always sending me sick shit in the mail when he’s out on the campaign trail.
“Dude, it’s four in the morning here,” I said. ”Don’t you have a prayer meeting to go to?”
“Fuck prayer!” he said. ”I’ve been snortin’ crank and signing death warrants since midnight.” I heard him haphazardly disassembling and reassembling the 52 parts of a vintage AK-47, something he often did while on speakerphone and high. ”Hey man, I was just watching Fear and Loathing. Remember last time we were in Vegas?”
It would be hard to shake the visions of that trip, one of my previous birthday journeys, where he surprised us all by showing up in a rented Kia Sedona, half the seats burned to a crisp, a rear window kicked out, and “FUCK THE POPE” spray-painted on the side. He told me to meet him in front of Circus Circus, and when I got there, he had the minivan up on the sidewalk, and was wearing a vintage Stone Cold Steve Austin muscle t-shirt and waving around the Korean car’s mini-sized tire iron like a victim at the end of a zombie movie. Me and my friends took him to an Irish bar and pumped him full of enough Guinness to counter all of the amphetamines coursing through his system, and blurred through the rest of the Superbowl weekend, listening to him talk about how he wanted to fuck every single waitress or server we saw, no matter how old or morbidly obese they were. He ended up getting us kicked out of our suite in Aria, because he called the concierge every twenty minutes asking for a map of the swingers clubs on the outskirts of town.
“Hey man, I wanted to ask you about this tablet shit,” he said. ”You been following the news?”
“It’s hard not to,” I said. ”Everyone from Boeing to Whirlpool have announced their own ‘iPad killer’ for the Christmas season.”
“Everyone but Microsoft,” he laughed. ”That fucker Ballmer still owes me a hundred bucks from the baseball pool.”
“Who the sam hell picks the Mariners to go to the World Series every god damned year?”
“You should talk,” he said. ”You keep picking the Rockies to go to the postseason. How’s their starting rotation lookin’ for 2012?”
“Go fuck yourself,” I said.
“Heh, heh,” he said. ”So what’s the deal on this new Amazon tablet? Should I get one?”
“Dude, aren’t you still rockin’ that Motorola MicroTac from like 1989?”
“Hell, no! I got a Nokia N-Gauge off of eBay for twenty bucks. Now I can play fuckin’ Bomberman during press conferences.”
“Well, it’s cheap, and you’ll like that, since you don’t spend a god damn dime on anything except executions,” I said. ”And technically, it’s going to be along your glide slope. You won’t be calling me in the middle of the night bitching about DLL conflicts or.NET framework mismatches or whatever the hell.”
“I like easy,” he said. ”And it’s got that gorilla glass, so it’s not going to get fucked up like my Zune. I had that god damned thing for ten minutes, and it looked like the surface of an asteroid.”
“To be fair, everything you touch gets fucked up like that. I don’t think I’ve seen you not get charged for damages at a car rental counter yet.”
“So I’m going to be able to put all my snuff films on here, right?” Whenever he travelled, Rick brought a huge collection of pornography on SD memory cards that he carried around in a small Anvil road case. His tastes were pretty esoteric: lots of submission/bondage/torture porn, amateur gangbanged housewives, interracial orgies. Surprisingly, it was all amazingly heterosexual, which is contradictory to just about every other politician I’ve hung with.
“Nope,” I said. ”No dice. It doesn’t have a card slot. You’re going to have to buy everything in the cloud. And you’re not even buying it. Even when you ‘buy’ a movie or a book, you’re really renting it, purchasing a license to use it, that could be revoked at any time for any reason. Nobody owns anything anymore.”
“Fuck that! Fuck this motherfucking cloud! This is communism! If Jesus wanted me to share my pornography with every fucking moonbeam hippy in the country, he’d tell me to start broadcasting it on PBS for free. Fuck this shit!”
“You could always just go to youporn and stream the shit for free,” I said.
“Does the browser have the flash?”
“I don’t know, but it doesn’t matter. Youporn shifted to HTML5, and usually when the porn industry embraces a technology, everyone else follows. The Amazon browser is interesting though,” I said. ”They pre-process all of your requests in the cloud, and since EC2 is backboned better than most internet providers, the jump to your device is going to be much faster.”
“But do they track the shit?”
“Of course they track it,” I said. ”That’s the whole reason they’d even do this. They want all of your browsing history to go straight into Skynet, so they can send the fuckin’ terminator on your ass and tell you where to spend your cash. You buy a copy of Fight Club, and they push some bomb-building supply ads and Meatloaf album samples into your sidebar. It’s all about getting you to spend your next dollar there.”
“I can’t do that then,” he said. ”I’m trying to run a Presidential campaign. I can’t have some fucking Smoking Gun assholes dig up my browser history and publish stories on Huffington about how I listen to that Rebecca Black song right before I execute someone.”
“You really need to cut that shit,” I said. ”I mean, hang ‘em high, but stop watching that fucking video. You have an unnatural obsession for that Benni Cinkle chick.”
“God damn it, I don’t know what to do about this tablet thing. I’ve gotta look cool to the youth voters, but I can’t endorse something that supports communism. What about the iPad?”
“Made in China.”
“God damn it. Doesn’t HP have a tablet? Or the Blackberry people?”
“Failed, and about to fail.”
“And all of those ones with the Androids are all made in China too, right?”
“Sure. Look, they’re all made in China. And what kind of tablet you buy doesn’t forever brand your political status. People who own iPads aren’t any better than people who bought the cheap-ass knockoff. It’s just a tablet. It’s a fucking appliance, like your fridge or your chainsaw. It’s not going to define you as a person. Buy whatever you want and shut the fuck up about it.”
“TEXAS CHAINSAW MASSACRE, MOTHERFUCKER!” he yelled. ”God damn, I love that movie. Leatherface in the fuckin’ house!”
“Look Rick, I’ve got to get up in like an hour. I’ve gotta sleep.”
“Sure thing Konster. Hail Satan, dude.”
Perry hung up, and I sat in bed, wide awake. I really need to turn off my ringer after midnight.
Targeted Nostalgia
Two unrelated things that aren’t have thrown me into a fit of nostalgia today: baseball and Target.
It was the last game of the season today, and despite the fact that the Rockies had a catastrophic time this year, I forced myself to listen, to take the pain and punishment of hearing them fail miserably. They won, despite having a skeleton crew of almost all third-string players and late-season replacements against a World Series-defending team. It got rough for an inning, and I thought it would all fall apart, but they pulled it out, and ended a dismal 2011, well below the.500 mark.
I listened to almost no games this year, because after about May, things imploded like an old Vegas casino making way for a new chrome-and-glass monstrosity. And this is the first year I didn’t see any Rockies games in person. In fact, I only went to one game all year, mostly because I can’t stand watching the Giants about as much as I can’t stand going to games at Oakland Coliseum. So there was a certain nostalgia to firing up the game audio today, the same way I feel after it’s been a long winter and I tune into that first game, hear the familiar announcers, get all of the standard commercials and station identification bumpers and little audio touchstones that bring me back to the summer I lived in Denver, pulling in the 850 KOA signal on my AM radio.
My year in Denver had two distinct eras, the first being the summer of 2007, when I worked from home for Frankov’s startup, and went to every day game I could afford at the one-block-away Coors Field. But then, almost exactly four years ago to the day, I got a job as a tech writer for this internet security firm. I didn’t know anybody in Denver, and suddenly found myself driving an hour a day to this high-tech campus, or as high-tech as the area had, at least. I was the second tech writer, and the first was totally consumed with his work, so we almost never talked. And because of the strange reporting structure, my own introverted mannerisms, and this mild disconnect between me and the work culture of the tech industry, I didn’t hang out with many people at work.
What do I mean by the work culture? I guess every place I lived had its own style or flavor of the tech industry. Seattle in the mid-90s was very Microsoft-driven; MSFT was practically printing their own money, and every company was either trying to keep up with them compensation-wise, was somehow dependent on them for their livelihoods, or was trying hard to be the exact opposite of them. New York was very Wall Street, except for that minor blip of “silicon alley” doofusism that vanished when the NASDAQ did. Silicon Valley is Silicon Valley, the gold standard of tech company behavior, no explanation needed.
But Denver had its own odd little identity. There were pockets of high tech, but it was held back by this attitude by some who thought it was still the wild damn west out there. Denver 2007 was just barely Seattle 1997 to me, and socially, it felt like one in maybe five people belonged to some outback conservative christian church with kids in a lockdown academy and a barefoot wife at home. This was the land of Promise Keepers, Ted Haggard, megachurches. I’d seen worse — I spent a dozen or so years in Indiana where the ratio was more like 5 out of 6. But it wasn’t just the politics or religion; it was a combination of that, of age, of technical background, that made me feel like an outsider there.
The job didn’t have a cafeteria, and we sat four people to a mega-cube with low walls, so the idea of brown-bagging didn’t appeal to me. Instead, I’d get in my brand new little car, and drive around the neighborhood until I landed at a fast food joint. The office sat on the edge of nowhere, a half hour south of Denver, an area with a few golf courses, an executive airport, and a whole lot of high mesa desert occasionally punctuated with strips of prefab big-box culture dropped on straight roads spaced apart at a mile per. Twenty years before, it was probably all barren cattle-grazing land, when an invisible SimCity player in the sky clicked and dropped all the big names down: Safeway, Chipotle, Chili’s, and Target, with a peppering of Subways and cell phone stores. From horizon to horizon, you’d see the orange-brown Colorado high plains, littered with the same exact stores I’d seen in every other place I’d lived or visited.
I’d always end up at the same two or three places, mostly Taco Bell or McDonald’s. I’d always bring a book, sit in my car, eat the same thing every day and read. The Yaris still had the new car smell and the novelty of car ownership I’d missed during my time in New York. I think I told Sarah one time about how I preferred to eat alone like this every day, and it depressed her, but I liked it. After almost a decade of being surrounded with ten thousand people in the same city block as me, it felt so nice to be absolutely alone.
It would take me five minutes to drive to Taco Bell, and ten to eat my Mexican pizza and nachos. That meant I’d had another 40 minutes to kill, until I’d need to tech write my way through the back half of the day. I’d inevitably end up at the big Super Target on Lincoln, this massive version of the familiar red department store, a two-story version with a double-decker parking lot and a grocery store welded to the side. I’d sit on the top deck of that garage, and you could see this great expanse of nothingness to the south, rolling hills and scrub brush and mountains in the distance, the ribbon of I-25 stretching from Lone Tree and vanishing on the horizon of Castle Rock. I’d go to Target for everything and nothing, to look at the twenty-dollar polo shirts and the seventeen different kinds of car air freshener, and end up with a case of Coke and some new cat toy we didn’t need.
But I mostly went to watch, to see who ends up at a Target in the middle of the day. And the answer, at this Target, was apparently nobody, because I’d only see a small trickle of stay-at-home moms, all younger than me with a gaggle of kids in tow. This wasn’t like the LA I’d know a year later, where in the middle of the day, you’d see all kinds of people and wonder who in the hell actually worked in a town like that. This reminded me of the solitude of working a day shift at the mall back in Indiana, where you’d only run into geriatrics and pediatrics. It had this certain feel to it, a feeling that I shouldn’t be there, the same feeling I had when I skipped a day of high school and saw a world I didn’t belong in.
And the visits to Target tied into that first era. Sarah worked a lot at her job, working nights, weekends, long days. And we didn’t have cable, didn’t watch TV, didn’t do much of anything outside of work except try to regroup to get ready for more work. And it seems like we spent an inordinate amount of time at the Stapleton Target, undergoing consumer therapy by experiencing the big box lanes of SKUs in a store that we didn’t get to experience in Manhattan. We’d end up with hundreds of dollars of damage in those white and red bullseyed bags, Method, Archer, Market Pantry. I’d have the next PlayStation game that would consume my off-hours in my little office, and whatever little cellophane-wrapped junk food I’d consume at my desk while listening to those games and not writing my next book. But those trips to the world outside the womb of my home office were strikingly themed by the uniformity of the Target experience. And when the first era quickly ended and the second era made me miss it, those trips to another branch of the same outlet let me briefly revisit my lost summer.
Now, four years, later, twelve hundred miles away, and 5280 feet lower, everything is different, except a daytime run to Target. They just built a new one, about a mile from my house. Now that I work from home, I’ll end up there after lunch, to pick up a case of (now diet) Coke or a box of Claritin. And it’s a different shape, a different layout, but the same experience, the same types of daytime shoppers, the same red-shirted staff and aisles of things I don’t need but will probably buy.
This nostalgia is a painful and potent drug for me, something it’s very easy for me to get lost in. I can waste far too much time exploring the connections and bridges of a present day to the past, grasping at these raw feelings I try to replay as a time machine to a distant era or pleasant memory. I stumble across these things, like the smell of a faded air freshener or an old receipt to a lunch from 2002 stuffed in the back of a book, and it can trigger this rush of thoughts back to that time. And I spent 1999 wishing it was 1992, and 2008 wishing it was 1989, and now bits of 2011 wishing I could open a window to 2007 and take a quick look again. I wonder if I’m the only person who does this, if I’ve accidentally segmented my life into these predefined periods by moving and changing jobs, or if it would be the same if I still lived on the same street in the same town where I grew up. It’s hard to be present in now, except that I know at some point in the future, I’ll be looking back and remembering 2011 again. And maybe the bridge will be a consumer store, or maybe it will be a kind of food or a song or the sound of an appliance or the smell and feel of an autumn breeze at the tail-end of a long summer. But I know that it’ll happen, at some point.
And now, I’ve gotta stop doing this, and go immerse myself in the now of trying to write my next damn book. Stay tuned.
Eversion and the mass-hallucination we call life
So Facebook recently fucked over its entire interface. That’s big news and not big news; I mean, it’s been beaten to death in various memes, even though it just happened a matter of hours ago. It’s big news in the sense that a somewhat-usable product has been made into a much less usable product. It’s not big news in the sense that we’re all marching to our graves at a rate of an hour every hour, some a bit faster than that, and nobody really gives a shit about various wars and economic disasters as much as they care about the order and sorting of various updates from their friends describing their bowel movements.
What’s interesting to me is how social media sites have changed our abilities to disseminate information. I’m almost convinced you could film an entire movie backwards, or edit it together so the events happened in reverse-chronological order, because people are so used to following feeds of information like blogs and twitter backwards. Everyone complains about how twitter and texting is killing the letter or the long-form prose entry, how people used to write long letters, which were replaced by long emails, which were replaced by blog posts, which were replaced by 140-character bursts of information. And I suppose that’s true. But I also wonder about people’s ability to glue together narratives from disparate entries of text, and how that will change our perception of reality.
William Gibson, the guy who invented cyberspace, later claimed that we are already in cyberspace. (He called it everting.) He didn’t mean that we had stupid goggles glued to our head or were immersed into some Tron wet dream. I think back in 1995, those Lawnmower Man days, I thought there would be a sudden tipping point where computers would get enough horsepower and the right neural tap so that we’d be able to spend time in a simulated reality. What happened instead is a parallel reality was created, and it slowly became woven into our daily lives, until we reached the point where more of this cyber-reality existed than our actual lives, and I don’t think anybody really realizes it.
Maybe I’m a bad example of this, but the bulk of my social interaction now takes place over TCP/IP. I work from home, and aside from two meetings a week, I conduct all of my work through emails and chat rooms and bug tracking software and wikis. And technically, my phones run through VOIP, so those are also funneled through the ether. I talk to friends in email; I post on this blog; I write twitter updates and reply to Facebook posts; I do pretty much everything online. I shop online; I sell books online; I post the high scores of my video games online, and prior to cutting my copy of Call of Duty in half because it was consuming all of my free time, I played against and with other people solely through my network connection, in a virtual reality where we blew each other up in deathmatches.
Look at all of the stuff that has been replaced by a digital counterpart: you download songs in iTunes instead of buying a physical CD; you get your software in binary form from an app store instead of a shrink-wrapped box. All of your photos are JPEGs and TIFFs instead of printed on Kodak paper. Maybe you haven’t moved to e-books yet, but a lot of people have. Films? Tax forms? TV shows? Calendars? Maps? It’s all another subdivision of cyberspace. Sure, you aren’t sitting at a digital desk in the Matrix and whipping your hands around in the air to manipulate these objects. But instead of having this completely separate world you enter by jumping in a holodeck or a VR isolation tank, you’ve got an infestation of these objects peppered throughout your regular blood-and-guts reality.
All of this works, until part of it radically changes. In the real world, you don’t wake up and suddenly find that every house in your subdivision, instead of being arranged on streets in numerical house-number order, were sorted by their frequency of use or color. But in a virtual world, you’re at the whim of its maker. When you get used to consuming those status updates in chronological order and they’re suddenly sorted by some piss-poor AI that thinks it knows what’s most important, you obviously throw a fit. Another example of this — also this week — was when Netflix suddenly decided it would be a great idea to take their existing service and split it into two parts, which completely upends the decade-old experience of putting stuff in a queue and expecting the discs to come in the mail, by complicating the situation with two entirely incompatible queues, one of them having a bafflingly stupid name.
I don’t know anything about human factors (I’m sure A’s dad could give me a lesson on it) but there’s gotta be a term or a threshold on how people react to sudden changes like that. But is it something learned? Will the kids born in 2011 who live a lifetime of CEOs with completely stupid UX ideas making dumb adjustments on the fly instantly adapt when their reality is suddenly shifted? And how will a generation of people like this change the way companies work? I grew up in a generation that, for the most part, always had email and always knew how to get on a computer and enter in a URL. So companies started saying “fuck it – don’t ship a manual; just put the URL on the box and let them figure it out” or “we don’t need to staff a toll-free hotline, just have them email when they need to order new checks/change their password/whatever”. And I’m fine with that, but the generation before me freaked the fuck out, and every Andy Rooney type started in with the “REMEMBER WHEN YOU COULD GO TO THE BANK AND GET THE TELLER TO HELP YOU” crap. In twenty years, will I be saying “REMEMBER WHEN EVERYTHING WAS IN CHRONOLOGICAL ORDER”?
Dot Matrix and Word Processors
I was writing about something completely different the other day, and went on this side diversion about dot-matrix printers, and thought about how a giant subset of the population (like everyone born after about 1985) never had to deal with them, while I spent far too many hours fighting them in computer labs, pulling apart the intricate pieces to pry loose jammed scraps of paper that got worked into the machinery.
There’s so many distinctive features of this whole era of printing that are long forgotten. Dot matrix printers usually used eight little pins to stamp a ribbon as the print head jumped across the page. My friend Matt had one of these, the Commodore 801, and the thing I remember most about it was that it was unidirectional; the little print head would zip across the page at a breakneck 50 characters per second, then the page would move up a line, and the head would return to the left. But it didn’t print on the sweep back to the left, which meant it was half as fast as the expensive printers that would print on both passes.
The printers were also tractor-feed back then. The paper had those little perforated runners hanging off of each end, little strips with holes in them, and the box of paper was fan-fold, so you could feed in the sheet and it would continuously chew through the giant thousand-page sheet of paper in a carton. Then, after you spent 20 minutes staring at the printer, hoping the thing got through your term paper in one pass, you then had to fold and tear apart each page, then tear off the feed strips on either side.
And, of course, that never worked right. If you didn’t line up the paper exactly, turning the little knob on the side of the printer, the end of the physical page would not match up with what the computer thought was the end of the page, and you’d get this mangled mess with a blank strip of what was supposed to be the top and bottom margins in the middle of the printed page. The whole operation of aligning and feeding and advancing paper was a precision thing, and if the paper got folded or creased or otherwise fucked up, the printer would have no mercy and create an origami disaster out of your precious schoolwork.
The output of a dot-matrix looked like shit, and they did a lot of little tricks to get it to resemble actual type. Like some printers had this “near letter quality” feature, where they’d do multiple passes on the same line to get a higher resolution, and they started adding more pins. When I was at IUSB, we had armies of these Panasonic KPX-1124 printers, which had 24 pins instead of 8. These pieces of shit were the bane of my existence back in 1991, and I spent untold hours tearing jammed paper out of these while some dumpy housewife screamed at me about her Psych 101 paper getting trashed. (If you ever did time around one of these, watch this video and tell me if the clunking sound of that print head slamming into the left margin over and over doesn’t make you go full postal.)
It seems like everyone forgets the other bastard child of that era that made perfect typewritten letters, at the sake of glacial speeds and 120-decibel print runs. The daisy wheel printer had a hub with a bunch of little spokes coming off of it, each one carrying a little type letter. It could spin the wheel with a servo motor and then hammer it against the ribbon with a solenoid, making an ink impression that looked exactly like a typewritten page. These were a big deal if you were printing out things like college admission letters, or you had an English teacher that had a hair up their ass about dot-matrix printers and wanted you to hunt down a selectric and hammer out the damn thing the old fashioned way. Daisy wheel printers were louder than fuck, and a low-end model typically cost more than your entire computer.
But not everyone had computers. When I was in high school, I had this “word processor” which was a glorified typewriter, except it had a single line of an LCD display, and it used a thermal print head. It took these cartridges of some kind of crap that it could transfer to a page with a set of heated pins. If you have one of those label maker machines, it was a similar deal, but masquerading as a desktop machine. I think you could only type in one line at a time, and then hit return and wait a minute for it to etch onto the paper. This wasn’t the best machine for stream-of-consciousness writing, but it was way faster than hunt-pecking on the K-Mart manual typewriter I got at a garage sale as a kid, where you’d type any faster than three characters a minute and all of the little hammers would get wrapped around each other and jam.
I somehow lucked into finding this girl in my freshman year of college that thought I was some kind of writing genius, and got her to type my papers for a semester. I guess that sounds sort of chauvinistic, but that’s an arrangement that I feel sorry the current generation won’t find themselves in. The “can you help me type my W131 paper?” pickup line has gone the way of the dodo.
After I wasn’t able to fully seal that deal, I dated someone who bought one of those Brother word processors, which were a very brief halfway-house between a typewriter and a computer. It was this huge microwave oven-sized thing that was a fusion of a printer, a tiny CRT monitor, a keyboard, and the Notepad.exe program in ROM. You could type a few pages at a time and then save them to a floppy disk (which was totally incompatible with any other computer) and then when you got it all situated and edited, you pressed a key and it would spit out the creation on actual paper. My roommate Kirk later had one of these beasts, and I think I remember Larry working off of one for a while. Here is a nice video of one in action.
Now, computers are cheap as hell, something it seems that most people forget, and laser printers or nice inkjets are everywhere, and we don’t really think about stuff like this. But I remember the smell of the fine paper dust inside of a monster line printer on campus, one of these washing machine sized beasts that would mass-print thousands of pages off of VAX computers, so long as one of us consultants hooked it up with the occasional corrugated cardboard box of 17″ wide tractor feed paper, that cream and light green-lined stuff. Every now and again, some idiot would send an ASCII-art dragon to the printer, a giant picture rendered in letters that would print banner-style across three dozen pages of paper, over the course of an hour. (Even better, when you’re sitting in a public lab and someone in a dorm sends through a picture of a Penthouse Pet done up ASCII-style.) That was all infinitely better than when someone would accidentally dump a binary file to the DEC LG06 in the library, and it would spit out page after page of random junk until you could get an operator in the machine room to kill the queue.
My last hurrah with dot matrix was about five years ago, when I bought a Tandy 100 off of eBay. The guy threw in a bunch of other random crap, including a Radio Shack printer from circa 1985, with some bizarro serial cable and no chance in hell of ever working with a machine produced this century. It went straight to the dumpster, but I probably should have videotaped it going off of a four-story building, or getting it Office Space style with a baseball bat.
The Replay
I’ve been dreading this post for years, but it’s a band-aid I need to rip off.
I was at this acupuncturist in Berkeley a year ago, in some stupid last-ditch attempt at getting rid of my allergies. (It did not work.) And I remember laying on his table, with a dozen needles in my arms and feet and face, thinking, “I really should post something today, but I don’t want to regurgitate the same old shit, and it’s only the nine year anniversary. I’ll wait for an even ten.” And it’s now ten years, and everyone is either waving their flags and beating their chest and ringin’ them bells, or they’re talking about the folly of spending two billion dollars a day to catch a man that’s already dead. And every show on TV this weekend is trying to get their spin on it, about how the world of cooking shows or pet rescue or hillbilly alligator hunting was forever changed on this day.
Seriously, fuck all of that.
It all makes me replay the day, and I do that a lot, but I don’t really think about who I was on September 10th, 2001, and what really did change. And I play this game a lot, with a lot of other arbitrary dates. I pull up old pictures or dig through old emails, wondering what person I was on 9/11/01 or 7/4/92 or 1/20/97 or any other date. And I try to reconstruct it, and I always find information I’d totally forgotten.
Here’s how it ended: I’d been sober a year, more or less. No meetings, no steps; I just quit drinking. But that night, my power was out, not because I lived in lower manhattan, but because my stupid fucking landlord had my entire apartment on two 15-watt breakers, and I was watching the news coverage on NY1 as I was running my computer and cooking in the microwave, and I blew a
| 0
| 0
| 8,000
| 13,512
|
14,005
|
5ae862b1-0393-46b9-8770-8a1d4b2b158a
|
StampyAI/alignment-research-dataset/lesswrong
|
Drawing Two Aces
Suppose I have a deck of four cards: The ace of spades, the ace of hearts, and two others (say, 2C and 2D).
You draw two cards at random.
*Scenario 1:* I ask you "Do you have the ace of spades?" You say "Yes." Then the probability that you are holding both aces is 1/3: There are three equiprobable arrangements of cards you could be holding that contain AS, and one of these is AS+AH.
*Scenario 2:* I ask you "Do you have an ace?" You respond "Yes." The probability you hold both aces is 1/5: There are five arrangements of cards you could be holding (all except 2C+2D) and only one of those arrangements is AS+AH.
Now suppose I ask you "Do you have an ace?"
You say "Yes."
I then say to you: "Choose one of the aces you're holding at random (so if you have only one, pick that one). Is it the ace of spades?"
You reply "Yes."
What is the probability that you hold two aces?
*Argument 1:* I now know that you are holding at least one ace and that one of the aces you hold is the ace of spades, which is just the same state of knowledge that I obtained in Scenario 1. Therefore the answer must be 1/3.
*Argument 2:* In Scenario 2, I know that I can *hypothetically* ask you to choose an ace you hold, and you must *hypothetically* answer that you chose either the ace of spades or the ace of hearts. My posterior probability that you hold two aces should be the same either way. [The expectation of my future probability must equal my current probability:](/lw/ii/conservation_of_expected_evidence/) If I expect to change my mind later, I should just give in and change my mind now. Therefore the answer must be 1/5.
Naturally I know which argument is correct. Do you?
| 0
| 0
| 479
| 479
|
40,049
|
d848aa0a-80eb-4531-8fdc-103b48a703ec
|
trentmkelly/LessWrong-43k
|
It is time to start war gaming for AGI
In this episode of the Making Sense podcast with Sam Harris, Barton Gellman from The Brennan Center For Justice discusses how he "organized five nonpartisan tabletop exercises premised on an authoritarian candidate winning the presidency to test the resilience of democratic institutions".
"The 175 participants across five exercises were Republicans, Democrats, and independents; liberals, conservatives, and centrists. They included veterans of the first Trump administration and previous administrations of both parties. "
This seems like an extremely valuable exercise when trying to prepare for long-tail risks.
--------------
I often think about about this post. It asks the seriously neglected question: Why was the AI Alignment community so unprepared for this moment?
I think we're going to get competent Digital Agents soon (< 2 years). I think when they arrive, unless we work urgently, we will again feel like we were extremely unprepared.
I'd like to see either a new AI Safety organisation created to run these exercises with key decision makers (e.g. Government, Industry, maybe Academia), or have an existing org (CAIS?) take on the responsibility.
Every morning we should be repeating the mantra: there are no parents in the room. It is just us.
--------------
More here on the program:
"In May and June 2024, the Brennan Center organized five nonpartisan tabletop exercises premised on an authoritarian candidate winning the presidency to test the resilience of democratic institutions. The antidemocratic executive actions explored in the scenarios were based on former President Donald Trump’s public statements about his plans for a potential second term in office.
We do not predict whether Trump will win the November election, and we take no position on how Americans should cast their votes. What we have done is simulated how authoritarian elements of Trump’s agenda, if he is elected, might play out against lawful efforts to check abuses of power.
The 175 p
| 0
| 0
| 420
| 420
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.