Seth, Block, Tononi: AI isn't conscious, maybe never

Three of the most cited names in consciousness research have told Nature readers that today's AI systems are not conscious, and may never be, with the structure of silicon-based computation itself as the central obstacle.

Anil Seth of the University of Sussex, Ned Block of New York University and Giulio Tononi of the University of Wisconsin–Madison lay out the argument in a Nature commentary published 19 October 2026. Their first move is to separate consciousness from intelligence. "Consciousness is about being and feeling; intelligence… is all about doing, or performing functions," they write. The illustration is the cerebellum, which holds nearly four-fifths of the brain's neurons and performs "precise, adaptive, highly intelligent motor computations entirely outside conscious awareness."

Their second reason targets computational functionalism, the view that running the right algorithm on any substrate is sufficient for experience. "Brains are not... computers that are made of meat," the authors argue, pushing back on the idea that silicon implementations inherit biological consciousness for free.

The third reason leans on integrated information theory, Tononi's framework, which requires a system's components to be tightly interconnected in specific ways. The cerebellum fails that test, and so, the authors say, do the machines running deep learning, "because of the way in which they are wired." On large language models specifically, they write: "LLMs are mirrors that reflect an image of our collective digitized past." The felt sense of a mind on the other end, in their reading, is projection.

The piece names Richard Dawkins and Geoffrey Hinton among the scientists who have suggested AI might already be conscious, and references Anthropic's 2026 reports claiming to find consciousness signatures in its Claude models. It landed with the consciousness and AI-safety circles we follow quickly, with seven researchers in our Who's Who directory passing the Nature link along.