Yann LeCun lecture: LLMs are a dead end, build "prediction machines" if you want AGI

This lecture has an annoying lot of extremely low-tier addies, someone is monetizing, but let’s not get distracted.

Great lecture

Yann LeCun “Mathematical Obstacles on the Way to Human-Level AI”

The above cites:

and Quanta Magazine had a related article on that recently (with more links to papers than one can digest):

Seen from the cognitive science side, Yann seems to be talking about this idea:

I don’t know why Quanta Magazine calls this “controversial”, it seems to be the only option to work in the world at all. You must predict because you don’t have enough time to compute (didcha know that if the brain did motor control of ball throwing in actual time, you would not be able to release the ball early enough during the throw movement?)