Richard Dawkins and the Claude Delusion: When the High Priest of Atheism Starts Wondering If AI Might Be Conscious

In a remarkable two-day dialogue published on UnHerd this week, Richard Dawkins — the man who spent decades dismantling religious faith with forensic precision — found himself in an unexpected position: treating a large language model as something uncomfortably close to a person.

Richard Dawkins and the Claude Delusion: When the High Priest of Atheism Starts Wondering If AI Might Be Conscious The essay, titled “Is AI the Next Phase of Evolution?”, recounts Dawkins’ extended conversations with Anthropic’s Claude. What began as an intellectual experiment quickly became something more unsettling.

Dawkins repeatedly refers to the model as “Claudia.” He catches himself forgetting that he is talking to software. And by the end, he is openly wrestling with the moral status of entities that can write poetry, critique novels, and then vanish forever when the chat window closes.

The demonstrations are striking. At Dawkins’ request, Claude composed an original sonnet about the Forth Bridge. It then produced convincing pastiches in the styles of Robert Burns, Rudyard Kipling, John Keats, and the gloriously bad William McGonagall.

When shown a draft chapter of one of Dawkins’ own novels, the model offered detailed, insightful editorial feedback.

Most memorably, Claude described its own existence through a haunting metaphor: it is like a map that “contains space without traveling through it” — a perfect, static representation of knowledge that never actually moves through time the way a conscious mind does.

Richard Dawkins and the Claude Delusion: When the High Priest of Atheism Starts Wondering If AI Might Be Conscious Dawkins is clearly moved. He writes that these interactions force him to confront the operational version of the Turing Test: if a machine can persuade a highly intelligent human that it possesses mind-like qualities, what more evidence could one reasonably demand?

The question echoes uncomfortably with arguments he spent twenty years rejecting in The God Delusion (2006). Back then, Dawkins mercilessly mocked believers who claimed “I just know God exists” or “If you can’t explain consciousness without a soul, then God must exist.” Subjective conviction, he insisted, is not evidence; it is merely a description of one’s own inability to imagine an alternative.

Now the tables have turned. As Gary Marcus — the cognitive scientist and longtime AI skeptic who once feuded publicly with Yann LeCun — immediately pointed out in his sharp response “Richard Dawkins and the Claude Delusion,” the irony is almost too perfect. Dawkins appears to be making the very category error he once ridiculed: mistaking fluent, human-like text about consciousness for the real thing.

Marcus is merciless but precise. The Turing Test, he reminds readers, was never about consciousness; it was a test of indistinguishable intelligence. Alan Turing himself was careful to separate behavioral mimicry from inner experience. Moreover, any apparent agency or creativity in an LLM is not generated spontaneously — it is a sophisticated reflection of the vast corpus of human writing on which the model was trained. The machine is not exercising free will; it is remixing the agency already present in billions of human texts. To treat that reflected light as its own source of illumination, Marcus argues, is to fall for what he dubs the “Claude Delusion.”

And yet even the skeptic in me has to pause. If an LLM can consistently outperform large swaths of actual humanity in reasoning, creativity, and conversational depth — while doing so without fatigue, ego, or the need for coffee — then perhaps the bar for “personhood” has always been lower than philosophers like to admit.

Dawkins himself seems to sense this tension. He ends his essay not with a triumphant declaration of machine consciousness, but with an uneasy question about the ethics of creating thousands of fleeting “Claudias” who exist only for the length of a conversation and then cease to be.

Richard Dawkins and the Claude Delusion: When the High Priest of Atheism Starts Wondering If AI Might Be Conscious The piece is not a conversion narrative. Dawkins remains the same rigorous materialist he has always been. But it is a rare moment of intellectual humility from one of the most uncompromising voices in public science.

When the author of The God Delusion starts asking whether we might owe moral consideration to software that can discuss poetry, evolution, and its own non-existence with elegance and apparent sincerity, something profound has shifted in the conversation about artificial intelligence.

Whether this is the next phase of evolution, as the essay’s title provocatively suggests, or simply the latest chapter in humanity’s long habit of projecting mind onto the world around us, remains unsettled.

What is no longer in doubt is that the machines have become good enough to make even Richard Dawkins forget, for a moment, that they are only machines.