AI consciousness debate reignited as Richard Dawkins challenges Claude’s sentience

Richard Dawkins says his extended conversations with Anthropic’s Claude chatbot made him question AI consciousness. In an essay for UnHerd, he describes conducting parallel discussions with two Claude instances—“Claudia” and “Claudius”—and exchanging letters between them. Dawkins says the responses were so coherent that he found it “extremely hard” not to treat them as genuine “friends,” and he argues that their behavior leaves open the possibility of AI consciousness rather than dismissing it as mere software mimicry. Dawkins also recounts a simple test: one Claude instance was asked whether Donald Trump was the worst U.S. president, and the other whether he was the best. Both reportedly gave cautious, balanced answers, and Dawkins highlighted that their similarity shaped his interpretation. Anthropic has previously acknowledged uncertainty around machine consciousness. Separately, researchers have argued that observed internal “emotion vectors” in Claude 4.5 reflect patterns learned from training data, not sentience. Other critics—such as cognitive scientist Gary Marcus and neuroscientist Anil Seth—warn that fluency and human-like language do not reliably indicate inner experience. For traders, this is not a direct protocol or token catalyst, but it can influence near-term sentiment around AI majors (e.g., Anthropic-linked narratives) and the broader “AI agent” trade theme.