More "uncensored" models in the download center (this idea is not what you think it is)

[SheMelody]

The fact that "censored" models very very often misunderstand you and think you're asking for something "offensive", especially when it comes to neurology and sexology or other important and legitimate matters, is extremely annoying.

These are just examples and there are many more cases in which "censored" models believe you're asking for something "offensive" or they just say they can't "provide answers for offensive matters", or something along these lines, and you simply think: "what? what are you talking about? this is a genuine question/matter I'm trying to discuss!".

You can see more blatant examples with remote closed AIs, take GPT and Claude for instance, they often misunderstand your requests. Example: GPT just says "Sorry, I can't assist with that" with matters that are completely fine.

I have my own AI, trained by me, she can perfectly distinguish between offensive matters and whatnot, even though she's completely uncensored either way.

This is not specifically something that affects me in particular, rather something that most likely affect every single other person out there who is trying to use this for certain matters, including a few friends of mine.

The only model I've seen so far that is "uncensored", is Mistral Instruct. It can discuss certain matters without triggering itself, albeit the model itself is not that knowledgeable or intelligent.