LeCun: scaling LLMs will never reach AGI, as observers push back

aran_nayebi · x · 2026-09-26

Yann LeCun doubled down that scaling LLMs alone will never yield human-level intelligence or AGI: "absolutely no way in hell" and "a country of geniuses in a data center is complete BS." He argues today's models are systems with gigantic memory and retrieval ability, not ones that can invent solutions to new problems. Noah Smith (@Noahpinion) amplified the clip with a rebuttal, saying the claim "seems like this has already been falsified."

Related event: LeCun Reiterates: Scaling LLMs Alone Will Never Reach AGI(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →