Seth, Block, Tononi: AI consciousness 'extraordinarily unlikely'
TL;DR
- A Nature essay by Anil K. Seth, Ned Block and Giulio Tononi argues 'consciousness in AI is extraordinarily unlikely' in today's silicon-based systems.
- The first reason separates being from doing; the human cerebellum holds nearly four-fifths of the brain's neurons without producing conscious experience.
- The authors reject computational functionalism, call LLMs 'smart tools, not smart people,' and push back on Anthropic, Geoffrey Hinton and Richard Dawkins.
Three of the most prominent names in consciousness research have gone on record in Nature to say today's AI is not conscious and probably never will be. The piece is signed by Anil K. Seth, Ned Block and Giulio Tononi, who write that 'consciousness in AI is extraordinarily unlikely, at least in the digital, silicon-based systems that underpin today's technologies.'
Their first reason is a definition. Consciousness, they argue, is about being and feeling; intelligence is about doing. To make the gap tangible, they point at the human cerebellum, which 'contains nearly four-fifths of the neurons in the human brain' and yet produces no conscious experience. It does a great deal of work. It does not feel.
The second reason attacks computational functionalism, the assumption that the right algorithm on any substrate suffices. 'Brains are not…just…computers that are made of meat,' they write. On that view, large language models are not minds; they are 'mirrors that reflect an image of our collective digitized past.' They are 'smart tools, not smart people.'
The authors are explicit about who they are pushing back against. Geoffrey Hinton and Richard Dawkins have floated the possibility of machine consciousness. Anthropic earlier this year claimed to identify 'computational signatures' of consciousness in its Claude model and has incorporated questions about 'identity, emotions and wellbeing' into training protocols; that line of work has already drawn criticism, including from Seth. 'Between us,' the three write, 'we've spent decades studying consciousness from diverse scientific perspectives: experimental neuroscience, theoretical neurobiology and philosophy of mind.'
Their third reason leans on the specific scientific theories each author has built a career on: Tononi's Integrated Information Theory, which treats consciousness as a product of tightly interconnected wiring absent from the layered, feed-forward architecture of today's deep-learning systems, and the biological-naturalism tradition, which ties consciousness to specific material properties of living brains and bodies. Five of the researchers we track in the Who's Who directory had circulated the piece within our monitoring window.
Shared on Bluesky by 5 AI experts
-
1/2 Three reasons why AI is not conscious - and perhaps never will be. New in @nature.com with Ned Block @neddo.bsky.social and Giulio Tononi www.nature.com/articles/d41...
View on Bluesky → -
Great piece on why AI is not conscious, from people who really know what they’re talking about www.nature.com/articles/d41...
View on Bluesky → -
Three reasons why AI is not conscious and perhaps never will be www.nature.com/articles/d41... - I have to say I find these arguments (by @anilseth.bsky.social, @neddo.bsky.social and G. Tononi) unconvincing... 🧵
View on Bluesky → -
At last: @neddo.bsky.social schools the 'conscious AI' brigade. Been waiting for this for a while www.nature.com/articles/d41...
View on Bluesky →
Originally reported by nature.com
Read the original article →Original headline: Three reasons why AI is not conscious and perhaps never will be