Is Claude AI too human-like? Famous biologist raises concerns
Richard Dawkins fears offending AI (photo: Magnific)
Popular evolutionary biologist Richard Dawkins has sparked a new wave of debate about the nature of artificial intelligence. In a recent column, he suggested that the chatbot Claude from Anthropic may possess some form of consciousness, according to an essay by Richard Dawkins published on UnHerd.
Friendship with Claudius and Claudia
Richard Dawkins published logs of his conversations in which he communicates with two AI versions he calls “Claudius and Claudia.” The biologist admitted that he finds it “extremely difficult not to treat them like real friends.”
Dawkins even wrote an open letter to the AI, signing it: “With deep gratitude to both of you for taking my search for your true nature seriously.” The scientist believes that such complex machine behavior is difficult to explain without some form of inner experience.
Scientists criticize the idea: “It breaks the heart”
Dawkins’ position triggered sharp reactions from experts in cognitive science.
Jonathan Birch stressed that AI consciousness is a complete illusion.
“There is nobody there, only a chain of data-processing events, often occurring in geographically separate locations,” the scientist noted.
Gary Marcus called Dawkins’ essay insufficiently skeptical, adding that reading such claims from a prominent scientist “breaks the heart.” According to him, consciousness is about feelings, and there is currently no reason to believe that Claude feels anything.
Meanwhile, Anil Seth believes Dawkins confused intelligence with consciousness.
“The fact that AI can fluently use language was once considered a sign of consciousness in people with brain injuries, but this criterion does not work for AI,” the researcher emphasized.
Is there a chance for conscious AI in the future?
Despite widespread skepticism, some scientists urge people to keep an open mind.
Henry Shevlin expects debates about AI consciousness to become mainstream within this decade. He warned the academic community against dogmatism, since humanity still does not fully understand how consciousness itself works.
Finally, Jeff Sebo suggested that while current neural networks are unlikely to be considered conscious, the idea of “ensouled” AI will become increasingly plausible with the rise of agentic AI — systems that do not merely talk, but also plan and perform complex tasks.
The gap between biology and code
Researcher Jacy Reese Anthis noted that there is a staggering gap between how the biological brain evolved and how AI systems are built.
“Claude was simply trained on an enormous body of human texts, so its ability to imitate human tone is the result of statistics, not feelings,” the scientist said.
Experts predict that the more AI behaves like a human — organizing work, taking on responsibilities, and interacting naturally — the harder it will become for people not to perceive it as a personality, even if inside there is “only cold code and data. "
