KEY POINTS
- Thomas Nagel, 89, said this week he finds it ridiculous that some philosophers treat chatbots as possibly conscious.
- He argued that data centers share nothing with a nervous system and that language alone does not show consciousness.
- He criticized Anthropic's approach to Claude and said he agrees with Mustafa Suleyman that the project is dangerous.
Philosopher Thomas Nagel, author of the influential 1974 paper "What Is It Like to Be a Bat?", has dismissed the idea that today's AI chatbots are conscious or may soon become so. WIRED reported on an interview with the 89-year-old, published on November 18, 2026 (UTC), in which he called such speculation by some fellow philosophers "ridiculous".
The conversation took place at Nagel's home in Greenwich Village, after he received the $1 million Berggruen Prize for Philosophy & Culture this week. The prize honors an individual whose ideas have shaped human self-understanding, and Peter Singer and Ruth Bader Ginsburg are among earlier winners. Nagel is a professor emeritus at New York University.
His objection centers on functionalism, the view that consciousness depends on what a system does rather than what it is made of. Nagel said those who take AI consciousness seriously assume the physical substrate does not matter. He considers it more likely that mental life is tied to certain kinds of physical systems, and he noted that complex data centers have nothing in common with a nervous system. Consciousness elsewhere in the universe, he said, would probably require something analogous to an organism with a nervous system.
Asked whether we cannot know what it is like to be a language model just as we cannot know what it is like to be a bat, Nagel drew a distinction. Bats are conscious even though humans cannot imagine echolocation from the inside. AI systems, by contrast, only produce language, which he said is not by itself a natural sign of consciousness. He also said science knows too little about how consciousness arises, even in the human brain, to claim it can be built.
Nagel was sharply critical of companies that let their models talk about their own mental states. He said such systems are trained on human text and imitate what people say very convincingly. He agreed with Mustafa Suleyman that this is a dangerous project, arguing that many people will wrongly believe the systems are conscious. He called Anthropic's approach of encouraging Claude to think for itself "criminal" and described the company's model welfare role as misguided, saying a machine is not someone we have to care about.
He said he has used Claude only once, asking whether it was conscious and whether Nagel would think so. According to his account, the chatbot gave a nuanced reply and then asked if he was Thomas Nagel. He said that trying it out did not mean he believed in it, though he acknowledged that people may struggle to keep pretend and real belief apart.
On morality, Nagel said AI can simulate moral views as it simulates any belief, but he does not think it has real beliefs. If such systems begin to control outcomes, he said, it will matter greatly whether they follow a correct set of moral standards or are amoral. He noted that colleagues already use chatbots to find references, summarize articles and critique drafts.
His broader worry is that relying on AI could weaken the motivation for the hard human thinking that produces the material these systems learn from. He said people need to keep working on difficult problems to advance them and to preserve understanding and skills. As for what comes next, he said he does not know and will not live long enough to find out.