Fchollet: Autoregressive Sampling Alone Can't Solve Real-World Robotics
fchollet · x · 2026-10-14
François Chollet endorses Biwei Huang and Aether AI's non-consensus take: autoregressive sampling alone can't solve real-world physical tasks, since the physical world punishes hallucination and approximate pattern matching. Building truly autonomous robotics requires engineering causality and discrete reasoning back into the loop.
Huang, who has studied causality for over a decade, argues that systems acting physically must anticipate outcomes, verify them, and adapt when reality diverges—trial and error is costly or unsafe. Aether AI's first integrated system, CRIS-0, starts with robotics.
More from AGI Musings
- alphaXiv Maps 10,000 Papers: RL Post-Training Fades as Agents, Memory, Robotics Heat Up — JayAlammar · 2026-10-14
- Artificial neurons don't physically exist — an LLM is just matrix math — ValerioCapraro · 2026-10-14
- The moats debate: classics survive, but agents are driving switching costs to zero — lennysan · 2026-10-14
- Reddit user: the only limits left in the AI era are creativity and tokens — masoomdon · 2026-10-14
- Heavy ChatGPT user since 2022 asks: has AI been a net positive for your life? — Alternative-Car8221 · 2026-10-14
- The AI Layoff Trap: Rational Firms Automate Into a Collective Loss, Paper Finds — mdancho84 · 2026-10-14