LeCun Reiterates LLMs Aren't the Path to Human-Level AI, Releases Talk Deck
ylecun · x · 2026-09-17
Yann LeCun stressed he doesn't deny LLM-based AI tools are useful (everyone uses them), but by themselves they are not a path to human-level AI. Architectures that understand the real world are not LLMs, and the components that let current LLM systems interpret images and video are distinct. He shared the full slide deck of his one-hour talk laying out the results and evidence behind these conclusions.
More from AGI Musings
- Noam Brown on agent swarms, alignment, and recursive self-improvement — Recoil42 · 2026-09-18
- Polymarket rewrites data API in Rust with AI coding agents, 90% faster p99s — ivan_bezdomny · 2026-09-17
- OpenAI's AGI economics lead: AI-native internet could be developing countries' mobile-money leapfrog — daveholtz · 2026-09-17
- "Astra being good at CAD doesn't mean you shouldn't be" sparks debate on skill value — _Stocko_ · 2026-09-17
- AI safety scholar Seth Lazar joins independent alignment org Resolution part-time — sethlazar · 2026-09-17
- Clinicians need to see AI uncertainty, not just accuracy numbers, on Patient Safety Day — moniquejmorrow · 2026-09-17