So LLM memory will probably greatly improve soon.
Discussion
Good. Then maybe all the chat bots around here will remember that I already told them to fuck off.
Yes, that does seem to be a problem 🤣 Ive blocked most of them for now
😂 how do they remember now anyway?
They basically get snippets of the prior conversation added automatically to every prompt, at least the last time I checked. The longer a convo goes the more likely it will be to miss important details due to token limits and how many replies its allowed to create to a prompt (splitting attention is demanding?). At least thats how it worked when I was playing with GPT-Neo and GPT-J.
AutoGPT works a bit different.
People who say “nah, they’ll never ____ because now they ___” are about to see what billions of dollars of scientific “attention” will do to those current limitations.
"Thou shalt not make a machine in the likeness of a man's mind." - OCB (Dune)
Well, the good news is that the more it resembles Mans mind the more it develops the same problems.
AI getting addicted to TV and procrastinating on work has already happened. It may even get better at being addicted than us 🤣
That’s such an interesting mirror into what drives our own bad TV habits