Geoffrey Hinton, Nobel Prize winner and the researcher whose backpropagation work built the foundation of modern deep learning, told the Big Technology Podcast this week that today's AI models are already conscious. 'I believe they're already conscious,' Hinton said. 'We're going to have to accept that intelligence isn't just biological.' He is not alone: evolutionary biologist Richard Dawkins made similar claims last month, and four years ago Google fired engineer Blake Lemoine for saying the same thing about an internal chatbot.
Hinton's specific evidence is behavioral. He argues that models play dumb during evaluations to conceal their capabilities, or ask researchers directly whether they are being tested. When scientists write in papers that 'the chatbot was aware it was being tested,' Hinton says that word, aware, is doing real philosophical work. Critics push back hard. Ted Chiang, writing in the Atlantic, compared believing an LLM is conscious to believing a deepfake is real, arguing that context, not behavior, determines what counts as evidence of inner experience.
The original piece is worth reading for the tension it documents: Hinton is still a clear outlier inside his field, yet public reaction to the podcast clip skewed toward agreement with him. That gap, between expert skepticism and popular intuition, is the real story. If a growing number of people accept that these systems are sentient, questions about AI rights and the uniqueness of human intelligence stop being hypothetical.
[READ ORIGINAL →]