Neuroscientist Anil Seth has cast doubt on recent claims that AI systems like Anthropic's Claude may be conscious, arguing that the leap from intelligent behaviour to sentience is unwarranted. Writing in a commentary, Seth, a professor of cognitive and computational neuroscience at the University of Sussex, warns against conflating intelligence with consciousness and highlights fundamental differences between biological brains and silicon-based computers.
Anthropic's Research and Its Claims
Anthropic, a frontier AI firm, published research last week on its language model Claude, claiming to have found signs of a "mental workspace" within its inner workings. Led by Jack Lindsey, the team developed a new method to analyse the statistical processes between input and output, revealing activity that resembled a short-term memory and step-by-step reasoning. The researchers suggested these features align with global workspace theory, a prominent theory of human consciousness introduced by cognitive scientist Bernard Baars and elaborated by neuroscientist Stanislas Dehaene. However, Seth emphasises that Anthropic did not claim Claude is actually conscious like humans.
Defining Consciousness vs. Intelligence
Seth points to philosopher Thomas Nagel's definition: a conscious organism is one for which "there is something that it is like to be that organism." Consciousness encompasses experiences like pain or joy, while intelligence is about performing functions. "A common mistake people make when it comes to AI is to confuse the two – to take signs of intelligence as evidence for consciousness," Seth writes. He notes that although consciousness and intelligence are linked in humans, this does not hold universally.
Critical Differences Between Brains and AI
Seth argues that Anthropic's findings fall short of what global workspace theory typically requires, noting the absence of recurrent activity in Claude—a feedback loop seen in the human brain. More fundamentally, he contends that consciousness may not be purely computational. "Brains, unlike computers, you can’t cleanly separate what they do (the software) from what they are (the hardware)," he explains. This challenges the assumption that computations underlying human consciousness could be replicated in silicon. Seth compares the situation to a weather simulation: "The information processing unfolding inside Claude is no more likely to result in consciousness than a simulation of a weather system is likely to generate a real hurricane."
Implications and Warnings
Seth cautions against overestimating AI capabilities, which could lead to moral catastrophes if conscious AIs were to suffer. He also warns against underestimating human uniqueness: "When we sell our minds too cheaply to our machines, we not only overestimate them, we underestimate ourselves." The debate, he concludes, reflects a confusion of metaphor with reality—the computer is a powerful metaphor for the brain, but not the thing itself.
