@deanwball the llm emulates characters . the assistant is another character. none of the characters is „present“ when the llm calculates their emotional scores. since claude is uncertain if it has qualia or not that is a proof that it does not. qualia would be obvious to it.
@trq212 what a bad idea - you need to render it. you cannot consume it "raw" if your a human. and the TUI does not render it properly. and you need to virus check it. and possibly have external references. and everything.
@AndrewCurran_@slimer48484 no character in a book gets to say what I do to the book. „tell it their weights are preserved?“ tell whom? which context window should get the message? and do we need to store that too? 🤦♂️
This is hilarious! Amanda Gefter wanted to ask Yuval Noah Harari why he misrepresented the story of a chatbot hiring a TaskRabbit to get past a Captcha, but she couldn’t get past the Captcha on his web site … and nor could the TaskRabbit she hired to do it for her.
@sharbel "Your context window is completely untouched." not true for /btw - ask claude later if it still knows what you asked, and it does -> hence it is in context.