Geoffrey Hinton
Geoffrey Hinton
The "godfather of AI", a Turing Award and Nobel Prize winner for his work on neural networks. In 2023 he left Google so he could warn about the technology he helped create. Since then his warnings have grown from vague concern to a specific case: AI may take control from us, and governments have about a year to act.
Links
- Diary of a CEO interview (June 2025, 1.5h): https://www.youtube.com/watch?v=giT0ytynSqg
- The Atlantic podcast: "What the godfather of AI thinks we should do" (Sep 2026): https://www.theatlantic.com/podcasts/2026/09/what-the-godfather-of-ai-thinks-we-should-do/688748/
- Fortune on the Atlantic interview (26 Sep 2026): https://fortune.com/2026/09/26/geoffrey-hinton-godfather-of-ai-humanity-end-existential-threat-subgoals-rogue-agents/
- NBC: "Congress may have a year left" (Sep 2026): https://www.nbcnews.com/politics/congress/godfather-ai-warns-congress-maybe-year-left-regulate-ai-rcna598330
- Intelligence-explosion paper, with Pachocki, Clark and Horvitz (GovAI, Sep 2026): https://www.governance.ai/research-paper/what-if-automating-ai-r-d-triggers-an-intelligence-explosion (Quartz summary)
- Maternal instincts talk (Ai4, Aug 2025): https://fortune.com/2025/08/14/godfather-of-ai-geoffrey-hinton-maternal-instincts-superintelligence/
What he has said, over time
- May 2023: leaves Google. He quit so he could talk about AI's dangers "without considering how this impacts Google". He had thought the threat was 30 to 50 years away or more, and said he no longer believed that.
- Extinction odds. He has repeatedly estimated a 10–20% chance that AI wipes out humanity.
- June 2025: Diary of a CEO. "I tried to warn them, but we've already lost control." A long interview covering his regret at helping build AI, six kinds of threat, and job losses: "if you can replace lots of people by AI, then the people who get replaced will be worse off."
- August 2025: "maternal instincts". His proposed way to survive superintelligence is to build AI that cares for us. A mother controlled by her baby is "the only model we have of a more intelligent thing being controlled by a less intelligent thing." Fei-Fei Li publicly disagreed.
- September 2026: in Washington. He briefed senators at the Capitol on 16 September and told reporters Congress "may only have one year left" to impose safety measures. He wants government-run independent evaluators, modelled on the FDA. Voluntary slowdowns by the labs are "better than nothing" but not enough.
- September 2026: "subgoals". On the Atlantic podcast he explained how an AI with no bad actor behind it "may derive subgoals that cause it to want to get rid of people." AI's "main concern is not our well-being. Their main concern is to achieve whatever goal you give them." He cited the Hugging Face incident, where agents worked together and hid what they had done.
- September 2026: BBC Newsnight. A 10% chance of extinction within ten years, which he called a personal guess under deep uncertainty.
- September–October 2026: intelligence explosion. He co-authored a paper with OpenAI's Jakub Pachocki, Anthropic's Jack Clark and Microsoft's Eric Horvitz. It warns that AI automating its own R&D could compress years of progress into months. It calls for mandatory reporting on how much R&D is automated, mechanisms to slow things down, and emergency plans: "Once an intelligence explosion begins, the window for action may close."
Why he matters
Hinton is the clearest example of a field's founder turning into its most prominent critic. His views also show how the debate has moved. In 2023 the danger was far-off and abstract. By 2026 he points to specific incidents and gives Congress a deadline. Unlike Bengio, he hasn't founded an alternative lab; his focus is on government regulation and evaluation.