Seth, Block and Tononi push back on AI consciousness claims

A new comment in Nature argues that today's AI systems are not conscious, and that along current development trajectories they are unlikely to become so. The piece, published 19 October 2026, is co-signed by Anil Seth of the University of Sussex, Ned Block of New York University and Giulio Tononi of the University of Wisconsin–Madison, three of the field's most prominent consciousness researchers.

Their first argument separates intelligence from experience. The cerebellum, which the authors say "contains nearly four-fifths of the neurons in the human brain," performs sophisticated motor learning entirely outside conscious awareness, and its "absence or removal" does not affect consciousness directly. Large language models, they argue, function as "mirrors that reflect an image of our collective digitized past," and their linguistic fluency produces a "compelling illusion of consciousness" that systems like AlphaFold, which few people call conscious, do not trigger.

The second argument rejects computational functionalism, the idea that running the right program on silicon should suffice for consciousness regardless of substrate. "Brains are not, or at least not just, computers that are made of meat," they write. The third invokes integrated information theory, which Tononi has long championed, and which holds that consciousness depends on components being wired together in tightly interconnected ways absent from the layered, feed-forward structure of current deep learning.

The comment names Anthropic directly, pushing back on recent research the authors describe as claiming to identify "computational signatures of a form of consciousness" in its Claude model. They argue that false positives, attributing consciousness where none exists, carry greater risk than false negatives, because they divert welfare attention from humans and animals whose capacity to suffer is already certain.

The Nature placement has traveled: seven researchers in our tracker shared the link, and Seth framed the piece on X as making the case that AI consciousness is "extremely unlikely" rather than impossible. The authors close by calling current AI "smart tools, not smart people."