Ai and the human mind

This Correspondence questions the symmetry assumed by Chen et al. (Nature 650, 36–40; 2026), who treat human minds and large language models as comparable black boxes evaluated through their outputs. The asymmetry is structural. Human intelligence is a natural phenomenon whose generative mechanisms remain incompletely understood, whereas LLMs are designed systems whose operating principle (statistical optimization of token prediction) is known. For LLMs, the known mechanism already suffices to explain behaviour, without invoking understanding. Establishing artificial general intelligence would require evidence that a system's behaviour cannot be fully accounted for by its generative mechanism alone.