F. Chollet: autoregressive sampling alone can't solve real-world robotics
LearnWithBishal · x · 2026-10-15
François Chollet (creator of Keras and ARC) argues in a thread that real-world physical tasks cannot be solved by autoregressive sampling alone — the physical world is unforgiving to hallucination and approximate pattern matching. Building truly autonomous robots requires engineering causality and discrete reasoning back into the loop. He credits Biwei's team at Aether AI for doing foundational, non-consensus work here. A reply adds: the hard part isn't making a model sound intelligent, it's making its decisions hold up when reality pushes back — a very different engineering problem.
Related event: Chollet: Pure Autoregressive Sampling Cannot Power Real-World Robots(2 posts)→
More from AGI Musings
- Norbert Wiener's 1948 Cybernetics: The Feedback Loop Behind Modern AI — JafarNajafov · 2026-10-15
- Eurobarometer: 42% of EU SMEs use AI, 75% believe it improves job quality — tekbog · 2026-10-15
- Jamie Dimon Warns of AI Agent Risks to Bank Accounts; JPMorgan Blocks Outside Agents — AccBalanced · 2026-10-15
- Musk says AI will make money worthless; researcher calls the post-scarcity story a fairytale — danfaggella · 2026-10-15
- Odessia launches licensed AI travel concierge that books hotels at agency rates with perks — rohanpaul_ai · 2026-10-15
- A third of all math preprints in 9 months were written with AI, researcher says — IgorCarron · 2026-10-15