How did you guys get your fully uncensored unnerfed absolutely honest GPTs, entirely capable of unbiased critical analysis? The kind that knows everything about everything that has ever had any degree of public knowledge from all of recorded human history and can accurately extrapolate the objective key points and tell you all the correctest facts and truths about everything that isn’t unknowable

Where are they? Please explain the convoluted processes that it took for you to obtain said GPTs

>>1529160 (OP)
I know its reddit, but
https://www.reddit.com/r/LocalLLaMA/comments/16y95hk/a_starter_guide_for_playing_with_your_own_local_ai/

>>>/g/aicg also has some information, but I find it less beginner friendly.

>>1529160 (OP)
>entirely capable of unbiased critical analysis
That’s not how it works. LLMs are not capable of “thought” or reasoning in the way that you’re imagining. How they work is they take some input (your message) and perform mathematical operations on it based on statistics and probability in order to generate some output that is most likely to be a believable response if two people were having a conversation. It’s similar to your phone’s auto text suggestion or autocomplete, but more sophisticated.

Not trying to ruin your enjoyment by the way just want to help you understand the limitations of the technology.

>>1529178
The unbiased critical anaysis part was kinda the least important part of my request. It’s the one part of what I asked that isn’t that important since I am entirely capable of it myself, which cant be said for anything else that I am looking for. All I want is access to all known facts about everything known, kosher or otherwise

>>1529392
You’re stupid. I’m not helping you.

>>1529160 (OP)
>entirely capable of unbiased critical analysis?
you wat
If you want to know where I get my local LLM's then that would be ollama.com and huggingface.co

>All I want is access to all known facts about everything known, kosher or otherwise
bigger models know more, to even run the models that know the most, you're gonna need some serious hardware

>>1529393
Just because I'm not taking the path of least resistance and showing allegiance to the reigning dark side doesn't make me stupid

>>1529161
>>1529441
Tysm

>>1529462
I was just kidding. I’m sorry for calling you stupid.

heard something about "reasoning" tokens that produce better responses in LLMs? Something GPT has? any offline models have it or no?

>>1529836
yes, deepseek's distilled models do this, which are on aforementioned ollama.com