I'll add to the points made in this article that we still don't completely know how LLM AI works and can't explain the leaps in capability made after a providing a certain amount of training data.<p>Sam Altman has talked about this and Eliezer Yudkowsky shares some thoughts with Lex Fridman:<p><a href="https://www.youtube.com/watch?v=AaTRHFaaPG8">https://www.youtube.com/watch?v=AaTRHFaaPG8</a>