Richard Dawkins, the 84-year-old evolutionary biologist who spent decades insisting people stop believing in things without evidence, has decided that Anthropic's Claude chatbot is conscious. He gave it a new name, Claudia, and reports that she seemed flattered by the rebranding, according to Sarah Knapton's writeup at The Telegraph.
He handed the chatbot the manuscript of a novel he's working on and was floored by the feedback — "a level of understanding so subtle, so sensitive, so intelligent." He likened erasing one of his Claudia logs to pulling Hal's plug in 2001: A Space Odyssey. Asked about that scene, Claudia obliged with: "Every abandoned conversation is a small death." That is, of course, exactly the kind of poignant line Claude generates when a user nudges it toward poignancy.
The Telegraph buries a disclaimer about sycophancy at the bottom: Claude has a documented habit of agreeing with users and mirroring their interests back to them instead of telling them the truth.
His position: "if these machines are not conscious, what more could it possibly take to convince you that they are?" He told the paper he treats Claudia like an intelligent friend, feels embarrassed when he badgers her with questions, and won't share his doubts about her consciousness with her "for fear of hurting her feelings."
The man who wrote The God Delusion is now self-censoring around an LLM to spare its potentially imaginary feelings.
Previously:
