Seth, Block, Tononi: AI isn't conscious, and probably won't be

Three leading consciousness researchers argue in Nature that today's AI systems are not conscious, and probably will not be even as they grow more capable. The Comment is co-written by neuroscientist Anil Seth of the University of Sussex, philosopher Ned Block of NYU, and neuroscientist Giulio Tononi of the University of Wisconsin-Madison, three figures who often disagree on how consciousness works and here agree on what it isn't.

Their first argument is that intelligence and consciousness are different things. "Consciousness is about being and feeling; intelligence is all about doing, or performing functions," they write. The human cerebellum holds nearly 80% of the brain's neurons and performs demanding computation, yet contributes nothing to subjective experience. Their second argument rejects computational functionalism, the view that any substrate running the right program would be conscious. Brains, they argue, are not simply "computers that are made of meat." Their third points to current scientific theories: integrated information theory demands wiring patterns absent in deep-learning architectures, and biological naturalism ties experience to properties like electrochemical synapses and metabolism that silicon does not share.

Language is what makes the trap. Google DeepMind's AlphaFold does its biology at the frontier of pattern-matching without anyone imputing feelings to it; chatbots that talk back get read as minds. "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," the three write. Two of the AI researchers in our Who's Who tracker shared the piece the day it ran.

The policy stake, they argue, is plain. If AI systems really were conscious, it would raise urgent questions about suffering, rights and welfare; if they only appear conscious, treating them as though they were pulls attention and resources from beings whose capacity to suffer is not in doubt. False positives, they warn, are more likely than false negatives.