Seth, Block and Tononi argue in Nature that AI isn't conscious

Three of the field's most senior consciousness researchers used a Nature comment on 19 October to argue that today's AI systems are not candidates for subjective experience, and that current trajectories of AI development are unlikely to produce it. Anil Seth of Sussex, the philosopher Ned Block of NYU and Giulio Tononi of the University of Wisconsin–Madison lay out their three reasons in Nature, responding in part to Anthropic's research suggesting computational signatures of consciousness in Claude, and to prominent voices like Geoffrey Hinton and Richard Dawkins who have gestured at machine sentience.

Their first move is to pry intelligence apart from consciousness. Doing is not feeling. The authors point to the human cerebellum, which holds nearly four-fifths of the brain's neurons and performs, in their words, "precise, adaptive, highly intelligent motor computations entirely outside conscious awareness." Language is where the illusion of mind takes hold: an LLM mirrors human text and someone seems to appear on the other end. "LLMs are mirrors that reflect an image of our collective digitized past," they write; that sense of a mind is largely projection. AlphaFold does the same kind of hard work without provoking the intuition.

The second reason goes after the quiet assumption behind most AI-consciousness talk: computational functionalism, the idea that the right algorithm on any substrate is enough. "Brains are not… computers that are made of meat," they argue. The material details of biology, the chemistry of synapses, metabolism, may not be features a functionalist can abstract away.

The third is that the leading scientific theories of consciousness do not fit today's models. Tononi's integrated information theory requires tight interconnection patterns absent in feed-forward deep learning. Biological naturalism ties experience to properties of living systems, electrochemical synapses and metabolism, that silicon systems lack. On both accounts, present AI falls outside the circle.

Their bottom line is terse: today's systems are "smart tools, not smart people," and precautionary welfare protections for AI risk pulling resources from beings whose suffering is not in doubt. Two researchers on our radar had shared the piece within hours of publication.