Richard Dawkins Concludes AI Is Conscious — Experts Push Back

Evolutionary biologist Richard Dawkins, 85, spent three days last week talking with AI bots he called “Claudia” (based on Anthropic’s Claude) and ChatGPT. He reported being “left with the overwhelming feeling that they are human” and told one bot: “You may not know you are conscious, but you bloody well are.” The exchange included poem generation in the style of Keats and Betjeman, analysis of Dawkins’ unpublished novel, and mutual flattery. Dawkins published his conclusions on UnHerd.

Why it matters

Dawkins is perhaps the most prominent figure yet to claim AI consciousness based solely on conversational ability. The claim has been widely rejected by experts:

• Prof Jonathan Birch (LSE Centre for Animal Sentience): “There is no one there” — just distributed data processing events.
• Gary Marcus (cognitive scientist): “There is no reason to think that Claude feels anything at all.”
• Anil Seth (University of Sussex): Dawkins confuses intelligence with consciousness. Fluent language is not a reliable indicator for AI.

Anthropic CEO Dario Amodei said in February: “We don’t know if the models are conscious … But we’re open to the idea that [they] could be.” Industry surveys show one in three people in 70 countries have at some point believed their chatbot was sentient.

Dawkins is not the first: in 2022 a Google engineer was placed on leave after making similar claims about LaMDA. The article notes that as AIs move toward “agentic” behavior — planning, task execution — such debates will intensify.