Abliteration is a technique that removes the refusal behavior from large language models
Did you know you that you can run local ai LLM models on an old desktop PC with a used GPU installed in it, and set it up to network with your phone and all of your laptops and netbooks and tablets on your LAN to access it from just about any device including when you're away from home?
You can also collect thousands of PDF and EPUB books about things you're not supposed to read about, and you can train your local AI model on all of that data, and/or have it categorize it all, summarize chapters, and have it schedule a daily reading list that you can get through in under 5-10 minutes.
No, I did not know that. Is that true?
That's right. One could make all kinds of explosives, using simple household items.
"LLM abliteration"
Did you know you that you can run local ai LLM models on an old desktop PC with a used GPU installed in it, and set it up to network with your phone and all of your laptops and netbooks and tablets on your LAN to access it from just about any device including when you're away from home?
You can also collect thousands of PDF and EPUB books about things you're not supposed to read about, and you can train your local AI model on all of that data, and/or have it categorize it all, summarize chapters, and have it schedule a daily reading list that you can get through in under 5-10 minutes.
is there a set up guide
yeah - you don’t need to do obliteration yourself. just download ollama and choose the uncensored models (i think wizard is one)
Ollama.com