LeCun and Chollet question language as the road to intelligence
Yann LeCun argued that world models built on joint-embedding architectures can replace the storage of knowledge with understanding, and that humans and animals mostly do not think in words. François Chollet wrote that evolution favours generality because specialists break when the environment changes.
In the Weekly Review of 8 November 2026
Filed under Large language models, Learning theory
LeCun tied his first claim to his paper with Randall Balestriero, which supports one part of it: a small text JEPA aligned with vision models better than much larger language models. The broader claim, that such models replace accumulation with understanding, is his argument and is not settled by the alignment results. His second post called verbal thought an illusion created by attaching discrete words to thoughts.
Chollet’s post drew an implicit parallel with measuring AI by fixed tests. In another widely shared post he wrote that language models turn every sentence into a punchline, placing emphasis without judging which sentences deserve it. All three posts are statements of position rather than new evidence.