LeCun called out as 'not very smart': critics say JEPA can't deliver superintelligence at 1000 TPS

teortaxesTex · x · 2026-09-14

AI commentator teortaxesTex published a blunt critique of Yann LeCun, calling him "not very smart, on account of being extremely stubborn and a normie" — arguing that LeCun wants AI to be a scientific field moving at a comfortable pace, so he imagines it that way.

In the quoted exchange, jmbollenbacher contended that embedding prediction (the JEPA approach) will likely become a dominant tool in parts of AI, particularly as efficient scaling for realtime video — meaning robots will use it, and AI for robots will set the broader trend. teortaxesTex pushed back: even if JEPA wins at realtime video, what does it offer for the race toward "SUPERINTELLIGENCE AT 1000 TPS"?

Original post →

More from AGI Musings

AGI Musings channel →