Seth, Block, Tononi argue AI isn't conscious, may never be

Current AI is not conscious, and present trajectories of development are unlikely to make it so. That is the argument Anil Seth, Ned Block and Giulio Tononi put forward in a 19 October Nature comment; three of the researchers we track had the link up within a day of publication.

Their first argument pries intelligence apart from consciousness. Consciousness, they write, is "what goes away with dreamless sleep or under general anaesthesia, and what returns after waking up or while dreaming." Intelligence is "about doing, or performing functions." The human cerebellum, which holds nearly four-fifths of the brain's neurons, "runs precise, adaptive, highly intelligent motor computations entirely outside conscious awareness." Being smart does not entail being sentient.

The second argument takes aim at computational functionalism, the assumption that running the right algorithm on any substrate suffices. "Brains are not… computers that are made of meat," they write. The piece also invokes integrated information theory, under which consciousness depends on components being wired "in very specific, tightly interconnected ways" that current deep-learning architectures do not have. Biological naturalism theories cited alongside it point to properties like electrochemical synapses and metabolism that silicon systems lack.

The authors close with a warning about mistaking fluency for mind. "LLMs are mirrors that reflect an image of our collective digitized past. We gaze into them and all too readily confuse the reflections with our own conscious minds," they write. Treating appearances of consciousness as the real thing, they argue, risks pulling welfare attention away from beings whose capacity to suffer is already established. Today's AI, in their phrasing, are "smart tools, not smart people."