These AI threads are all super interesting. Yesterday r00t even got it to answer questions without dealing with ethics. It's crazy how this thing can break down questions into components. I have to think there are layers of non-AI / general sentence structure algorithms somewhere in there, which directs the questions to different algorithm subsystems.

It's all guesswork trying to figure out how these work. When I was in undergrad we had the green AI book and in grad school the students I had to pretend to like had the blue AI book ... and in both cases, all the really cool shit was still closed and proprietary ... even the basic decision tree stuff back then.

ID3 was one of the big ones (confusing because that's also mp3 metadata) and now we have GPT3 (confusing because GPT is a partition table type ... they're always bad at naming things).