LeCun: Superintelligence isn't coming in 2 years; AI still lags a cat at physical understanding

ylecun · x · 2026-11-19

Yann LeCun pushes back on near-term AGI timelines, calling the idea of superintelligent systems within one or two years a misconception. He argues today's AI excels at symbols like math and code but merely regurgitates accumulated declarative knowledge rather than reasoning. Pointing to the absence of domestic robots, he says current models understand the physical world worse than a cat, and true "Physical AI" remains far off.

Related event: Hinton says AI beats most mathematicians; LeCun retorts it understands less than a cat(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →