'Just Moving Electrons': Fleuret's Chess Analogy Against LLM Reductionism

francoisfleuret · x · 2026-09-26

François Fleuret argues that saying a model 'just predicts the next token' is as informative as calling a chess program 'just moving electrons around' — true but the wrong level of analysis.

This is tweet 2 of 7; his point is that LLMs should be understood at the level of a complex distribution over sentences, not individual tokens.

Original post →

More from AGI Musings

AGI Musings channel →