Skip to content

Comment on Why large language models can understand the world [video]

Comments

While the title slightly over-promises - a more accurate one might be "why LLMs may come to understand the world" - it seems to me that this video gives a very clear introduction to a hypothesis about why double-descent works despite the threat of overfitting, and a thought-provoking extrapolation to the claim of its title. I came to post it myself, and am disappointed to see that it is languishing.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.