Geoffrey Hinton warns AI may be harmful
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
Geoffrey Hinton says a 10% chance AI causes human extinction within a decade is "not unreasonable".
He also warns AI could disrupt banking and power systems.
Background
Geoffrey Hinton is a Nobel Prize-winning computer scientist, often called the "godfather of AI", who shared the 2024 Nobel Prize in Physics for foundational work in machine learning and helped create the neural networks behind modern AI. He left Google in 2023 after becoming deeply concerned about the risks AI could pose to humanity, saying he wanted to speak openly without ties to a major tech company. His warnings have grown more specific as AI spreads deeper into critical infrastructure and workplace systems, and the concern has reached Capitol Hill, where lawmakers held a closed-door briefing on AI's dangers in September 2026.
How it unfolded
- 2023Hinton left Google after becoming deeply concerned about AI's risks to humanity, saying he wanted to speak openly without ties to a major tech company.
- Apr 28, 2025He warned AI could one day outsmart and control humanity, saying a profit-driven AI arms race may be speeding up the danger.
- Sep 6, 2025Hinton said the real danger is not killer robots but smooth-talking AI that will know more than us and be better at emotionally manipulating people.
- Jan 22, 2026In a BBC Newsnight interview, he said the biggest mistake humans can make is failing to invest in research into how to coexist with intelligent systems.
- Sep 10, 2026Hinton said a 10% chance that AI could cause human extinction within the next decade is "not unreasonable".
- Sep 16, 2026He addressed an AI briefing for senators at the U.S. Capitol.
- Sep 26, 2026Hinton warned AI could wipe out humanity as a byproduct of subgoals it derives, and told reporters Congress may only have one year left to impose safety measures.
- Nov 4, 2026He warned AI could trigger major disruption to banking and power systems, saying the next decade may determine whether AI delivers broad economic gains or severe social disruption, job losses and failures in essential services; he also warned of a Chernobyl-like blackout this year or next.
Who’s saying what
- Caution
- Hinton says a 10% chance of AI-caused human extinction within a decade is "not unreasonable", and that a sufficiently advanced AI could persuade humans not to shut it down.
- Expert
- Researchers cited by a Nature explainer say some extinction scenarios are possible, while others argue the predictions rest on assumptions that can't yet be tested.
- Analysts
- Hinton says the next decade may determine whether AI delivers radical abundance or deep inequality, employment shocks and cultural upheaval.
Still unverified
Some experts dispute extinction predictions, saying they are based on assumptions that can't yet be tested, according to a Nature explainer from September 2026; a commenter also noted the Apollo Research results Hinton cites come from tests set up to push the model into that behavior.