Sorry for the late reply. I've read a bit about this, but I don't know how it's done. I've seen some uncensored models using a process called "abliteration" [1] but I'm not sure if that's the same thing or not.

I've never tried self training an AI model or anything but I'd be very curious about the process. Maybe this is something someone could do as a side hustle?

[1] Uncensor any LLM with abliteration by Maxime Labonne https://huggingface.co/blog/mlabonne/abliteration