Three reasons AI isn't conscious, Seth, Block, Tononi argue

Three senior consciousness researchers used a Nature comment to argue that consciousness in today's AI is 'extraordinarily unlikely, at least in the digital, silicon-based systems that underpin today's technologies.' Anil Seth of the University of Sussex, Ned Block of New York University and Giulio Tononi of the University of Wisconsin–Madison lay out three reasons, and conclude that a precautionary approach to AI welfare 'would be misguided.'

Their first move separates intelligence from feeling. 'Consciousness is not the same thing as intelligence,' they write, framing consciousness as what 'goes away with dreamless sleep or under general anaesthesia.' The cerebellum, which contains 'nearly four-fifths of the neurons in the human brain,' performs fast, adaptive motor computation yet, by their account, does not affect consciousness directly. Smart behaviour, in that reading, is not evidence of an inner life.

The second reason targets computational functionalism head-on. 'Brains are not — or at least not just — computers that are made of meat,' the authors write, rejecting the idea that running the right algorithm on silicon is sufficient for experience.

The third reason is about language. 'LLMs are mirrors that reflect an image of our collective digitized past,' they say. Fluent output invites humans to project a mind where none exists. The piece notes that protein-structure systems like AlphaFold draw none of the same intuitions, pointing to linguistic mimicry rather than AI generally as the source of the illusion.

Seven of the researchers we track shared the piece on the day of publication, a sign the line-up is doing real work. The byline is itself part of the news: Tononi, associated with integrated information theory, and Block, associated with biological naturalism, start from different theories of consciousness and converge on the same verdict about silicon.