François Chollet argues robotics needs causal reasoning beyond autoregressive sampling

François Chollet says autoregressive sampling alone cannot solve real-world physical tasks, because the physical world punishes hallucination and approximate pattern matching.

In a post on X, he argues that autonomous robotics requires engineering causality and discrete reasoning back into the loop. He also praises Aether AI's Biwei and team for work in this direction. The post is Chollet's opinion; the report does not describe a specific system, result, or benchmark.