LeCun doubles down: autoregressive LLMs alone won't reach human-level AI
tak3sh8 · x · 2026-09-21
Yann LeCun reaffirmed his claim that "auto-regressive LLMs, in and of themselves, will not lead to human-level AI":
- Reasoning is non-autoregressive search: current systems' reasoning relies on search, which he has long advocated, but doing it in token space is limited and inefficient; human-like reasoning should be a search in continuous representation space, and he sees the industry moving that way.
- Limits of self-improvement: current methods only work where outputs can be scored without human intervention — math, code, and accurately simulable domains.
- The quoter eliebakouch mocked the narrative around "major redesign" and latent-space reasoning, dismissing CoT monitorability concerns.
The exchange highlights the ongoing debate between explicit CoT reasoning and latent-space reasoning, and its monitorability trade-offs.
Related event: LeCun Reasserts Autoregressive LLMs Won't Reach Human-Level AI(6 posts)→
More from AGI Musings
- NYT essay: resisting Big Tech and AI takes collective action, not willpower — TuhinChakr · 2026-11-12
- Ethan Mollick: 'talk to your documents' RAG solutions already feel dated — emollick · 2026-11-12
- Anthropic's pitch hasn't changed in 5 years: 'AI is winner-take-all, we will build god' — menhguin · 2026-11-12
- Peter Diamandis: We have 6-18 months to reach 'Safe Superintelligence' — what is Ilya building? — PeterDiamandis · 2026-11-12
- AI observer: headcount is no longer a proxy for ambition — signulll · 2026-11-12
- AI system proves longstanding quantum computing conjecture in three days, Lean-verified — _arohan_ · 2026-11-12