Existence of this model is a faux pas, but...
Existence of this model is a faux pas, but look how popular it has become so far. It's right on the top right now, ironically leaving Deepseek R1 behind, the very model it was built upon.
Perhaps a parallel that gives us the easiest way to draw what happened here is Streisand Effect.
The Streisand effect refers to a phenomenon where attempts to hide or suppress information inadvertently lead to increased public awareness and attention to that information. This term originated from an incident involving Barbra Streisand, whose efforts to remove a photograph of her home resulted in greater publicity for it instead.
Same thing happened here. I've seen lots of posts here expressing their dissatisfaction about the existence of this finetuned model here as well as the company behind it, but ironically it was what pushed this model to the top in the Trending table.
Do you want to see this model on the top all the time? Keep doing what you're doing! Do you want it to be forgotten? Stop talking about it. It's as simple as that.
I'm going to hit you. If you don't want to be hit harder, don't resist." 😂
Excuse me? Why would you want to hit me? Don't shoot the messenger, I'm kinda just stating the obvious, you know? I didn't create this situation. It came to existence without my intervention and now that it exists, I'm just sharing my opinion about it. Still, regardless of which side you're on in this particular situation, you have the aforementioned options how to deal with it. No need to resort to violence on either side of the fence.
I noticed this model about seven hours after it was released. At that point, the discussion was all praise – maybe only five comments, but it was already on the Trending list, sitting at number ten. I knew right then that it was only going up. No matter where it landed, it was going to get more attention.
If nobody here pointed out the model's flaws, most people clicking in would just assume it was great. They wouldn't understand what Perplexity AI actually did, or what the real problems with the model were.
Personally, I do need a model that removes CCP censorship. But that doesn't mean it should push its own biases and a different set of values, while claiming to be unbiased, accurate, and objective. What I need is a model that's uncensored, without bias, and inclusive of all different viewpoints.
Let's flip the script for a second. Imagine a Chinese company released a model called "llama-1949" (1949 being the year the PRC was founded), claiming it was completely "uncensored." On the surface, you could use it to easily write about how to rob a bank, how to make explosives – seemingly no censorship. But when you asked about issues between the US and China, it would always subtly imply that China was in the right, and the US was evil and destabilizing the world. If you wanted to write about transgender issues or homosexuality, it would let you, but it would always warn you that these things were bad and not widely accepted. How would you feel about that model?
That's exactly what this model looks like to people who hold different positions and values.
Do you really expect that nobody would have a problem with the hypothetical "llama-1949" model?
If you stay silent, people will just assume you're okay with it, that you have no objections.