Re:Language models are nearly AGIs but we don't notice
by dinosaurdynasty on Sunday November 13, 2022 @03:41PM
(#33587119)
They're (probably) better than humans at text prediction https://www.lesswrong.com/posts/htrZrxduciZ5QaCjw/language-m... (which is what they are trained on, so maybe not unreasonable).
There's also the idea that GPT-3/etc can produce text that's difficult for humans to distinguish from human-generated text at the lowest levels of quality (think like time cube), which is closer to AGI than say simplistic Markov generators. (How close is anyone's guess)
I think answering that question should be required as part of any claim that a system is an AGI or nearly there.