LeCun tells Fleuret: auto-regressive prediction isn't search, and neither is chain-of-thought
ylecun · x · 2026-10-10
Turing Award winner Yann LeCun pushed back on François Fleuret's suggestion that chains of thought are a form of stochastic search.
LeCun's core argument: search/optimization means examining multiple configurations and evaluating them to pick a good one. Auto-regressive prediction — stochastic or not — involves no such process, so CoT fundamentally does not constitute search. A representative debate on the nature of LLM reasoning.
Related event: LeCun and Lample Push Back on Fleuret's AlphaZero Challenge to LLM Skeptics(9 posts)→
More from AGI Musings
- Museum visit laments: talks about industrial history now fixate only on job losses and inequality — Afinetheorem · 2026-10-10
- "AI is sophisticated search": dev argues model outputs trace back to uncredited authors — gerardsans · 2026-10-10
- AI isn't killing creative jobs — it's changing what clients pay for — kevinsurace · 2026-10-10
- AI recipes aren't AI creations: the original author was simply erased — gerardsans · 2026-10-10
- KKT Points in Imperfect-Recall Games Correspond to CDT+GT Optimal Policies — jessi_cata · 2026-10-10
- AI consciousness meme: classicalists throwing darts at the blank edge of the 'space of minds' chart — eigenhector · 2026-10-10