Nature comment: three reasons AI isn't conscious, maybe never

Three of the field's most-cited consciousness researchers used a comment in Nature on October 19, 2026 to argue that the scientific default should be that today's AI systems are not conscious.

The authors are Anil Seth of the University of Sussex, Ned Block of New York University's philosophy department, and Giulio Tononi of the University of Wisconsin–Madison. They define consciousness as subjective experience, "what goes away with dreamless sleep or under general anaesthesia." Intelligence, in their account, is something else. The cerebellum holds nearly four-fifths of the brain's neurons and runs sophisticated computations entirely outside awareness, which they take as evidence that doing complex things does not entail feeling anything.

Their second reason rejects computational functionalism, the view that any substrate running the right program would be conscious. "Brains are not… computers that are made of meat," they write. The leap from a program producing intelligent outputs to a program producing felt experience, they argue, has no evidence behind it.

The third reason draws on Tononi's integrated information theory. Deep-learning systems fall short, they write, "because of the way in which they are wired": consciousness requires a system's components "to be tightly interconnected in specific ways" that today's layered architectures do not satisfy. Current models, in their phrase, are "smart tools, not smart people."

The thesis is explicit: "The default position of the research community should therefore be that AI is not conscious." Seven researchers in our Who's Who tracker circulated the comment the week it ran.