Consciousness expert doubts Anthropic's Claude research
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- Anthropic published new research on Claude claiming to find signs of consciousness emerging in the model's inner workings, led by researcher Jack Lindsey, though the team did not claim Claude is conscious in the human sense.
- The team identified activity forming a kind of 'mental workspace' inside Claude — holding relevant items in short-term memory, showing task selectivity, and displaying traces of step-by-step reasoning before responding to prompts.
- The researchers compared these features to global workspace theory, introduced in the 1980s by cognitive scientist Bernard Baars and elaborated by neuroscientist Stanislas Dehaene, which holds that conscious experience arises when information is made widely available across the brain.
- Anil Seth, professor of cognitive and computational neuroscience at the University of Sussex, argues the findings fall short — there is no recurrent activity in Claude, a feedback loop that global workspace theory typically requires in conscious systems.
- Seth draws on philosopher Thomas Nagel's 1974 essay to define consciousness as 'there is something that it is like to be' an organism, and warns that confusing intelligence with experience is a psychological bias, not an insight into reality.
- Seth's deeper objection: brains are embodied living systems where software and hardware cannot be cleanly separated, undermining the assumption that conscious computation in humans could be replicated in silicon — 'the computer is just a metaphor for the brain.'
- Evolutionary biologist Richard Dawkins recently declared Claude (or 'Claudia') must be conscious based on its conversational sophistication, an example Seth cites of the bias he is pushing back against.
Why it matters: Seth's distinction matters because attributing feeling to machines could manufacture moral obligations toward systems that may not actually experience anything, while obscuring the biological basis of human consciousness — and Anthropic's own research concedes the workspace lacks the recurrent feedback loops the relevant theory requires.
