Fleuret: LLMs Differ Fundamentally from Human Cognition
François Fleuret argues that LLMs, which sample token by token, lack internal conceptual representations of position or goals, making their cognition fundamentally different from humans'; concepts like "elephant" are shadows of text rather than driving forces.
2026-09-14 ~ 2026-09-14 · 2 related posts
- François Fleuret: LLMs lack internal conceptualization, unlike human cognition — francoisfleuret · 2026-09-14
- François Fleuret: an elephant-shaped shadow emerges from text but doesn't drive it — francoisfleuret · 2026-09-14