In Nature, Seth, Block and Tononi reject AI consciousness

"Consciousness is about being and feeling; intelligence is about doing, or performing functions." That line, from a Nature essay by three of the field's most senior consciousness researchers, is the hinge of their case that today's AI systems are not conscious and perhaps never will be.

Anil Seth of the University of Sussex, Ned Block of New York University and Giulio Tononi of the University of Wisconsin–Madison define consciousness as "what goes away with dreamless sleep or under general anaesthesia," and treat that as the thing fluent language output does not demonstrate. Their first argument leans on neuroanatomy. The cerebellum "contains nearly four-fifths of the neurons in the human brain" yet does not contribute directly to conscious experience, they note, offering it as evidence that sophisticated computation can proceed entirely outside awareness.

The second argument is aimed at the premise most AI-welfare worries rest on. The authors reject computational functionalism, the idea that running the right algorithm on any hardware is sufficient for experience. "Brains are not — or at least not just — computers that are made of meat," they write.

Their third argument invokes integrated information theory, which Tononi has long championed: consciousness, on that view, requires a system whose components are tightly interconnected in specific ways, and the authors contend that the layered, feed-forward architectures of today's deep-learning systems do not meet the bar. On language models in particular they are blunt. "LLMs are mirrors that reflect an image of our collective digitized past," they write. Seven of the researchers we track in our Who's Who directory had shared the piece by the time we picked it up.