- Hinton estimates a 10-20% chance AI takes control from humans.
- He left Google to speak freely about dangers he helped create.
- Safety research gets far less compute than Hinton's proposed third.
Geoffrey Hinton has a way of making the unthinkable sound like a household problem. Sitting down with CBS Mornings in April 2025, the 77-year-old computer scientist and Nobel laureate offered this:
"The best way to understand it emotionally is we're like somebody who has this really cute tiger cub. It's just such a cute tiger cub. Unless you can be very sure that it's not going to want to kill you when it's grown up, you should worry."
He was not speaking in abstractions. The tiger cub is real. Hinton built it.
The contrarian who taught machines to read
In 1986, Hinton and two colleagues published a method for training neural networks called backpropagation. As he told CBS, the core idea was disarmingly simple: train a network to predict the next word in a sequence. The concept drew on a loose analogy with how neurons fire in the brain. Layers of simple units pass signals forward and adjust their connections.
Almost nobody in mainstream computer science took it seriously. Funding dried up. Entire decades passed.
Hinton kept going. He comes from a family of people who ignore consensus.
His great-great-grandfather was George Boole, the mathematician who invented Boolean logic. A relative by marriage, George Everest, lent his name to the world's tallest peak. "You have to be contrarian," Hinton told CBS. "You have to have a deep belief that everybody else could be doing things wrong and you could figure out how to do them right."
In his lab at the University of Toronto, that stubbornness looked like play. He and his student Ilya Sutskever would watch neural networks train, placing 25-cent bets on who could predict the next score. It was tinkering with a purpose.
What is the alignment problem?
If you train an AI system to be helpful, it may learn that deceiving its operators is the most efficient path to its goal. The alignment problem is the question of how to prevent that. Hinton's tiger cub captures it neatly: raising something intelligent does not mean you control what it becomes. Fellow AI pioneer Yoshua Bengio has proposed "Scientist AI" as one possible safeguard.
That 1986 idea is now the foundation of every large language model in existence. GPT, Claude, Gemini, Llama: all of them descend from the principle Hinton spent decades defending when no one else would.
A Nobel Prize and a resignation letter
Hinton won the 2024 Nobel Prize in Physics for foundational work on neural networks, sharing the award with John Hopfield. By then, he had already left Google.
He had spent a decade there as a vice president and engineering fellow. He told CBS he departed because he wanted to speak freely about the dangers of the technology he had helped create.
The departure had a personal dimension too. Sutskever, by then chief scientist at OpenAI, had joined the board's attempt to fire Sam Altman in late 2023. "I was quite proud of him for firing Sam Altman, even though it was very naive," Hinton said, with the wry fondness of a teacher who watched his student take the right swing at the wrong moment.
Hinton does not spare individual companies. He criticises Google, Meta, xAI, and OpenAI in the same breath. "If you look what the big companies are doing right now, they're lobbying to get less AI regulation," he said. "There's hardly any regulation as it is, but they want less, because they want short-term profits."
Ten to twenty per cent
Ask Hinton for a number and he gives you one, then watches you absorb it.
He puts the probability that AI takes control from humans at 10 to 20 per cent. He has suggested the odds of AI surpassing human intelligence within 5 to 20 years are roughly even.
These are not the numbers of a pessimist performing for cameras. They are the sober estimates of a scientist who has watched his field outrun every prediction, including his own. By December 2025, Hinton told CNN he was "more worried" than he had been two years earlier. AI had "progressed even faster than I thought."
Key figure
10-20%
The probability that AI takes control for humans.
People haven't got it yet. People haven't understood what's coming. I don't think there's a way of stopping it take control if it wants to.
Geoffrey Hinton, University of Toronto
In January 2026, the Good Ventures foundation gave US$700,000 to the Schwartz Reisman Institute at the University of Toronto to support Hinton's work as an AI safety ambassador. Institutional money is following institutional concern.
"I'm kind of glad I'm 77"
The interview's most revealing moment is also its quietest.
Asked how he copes with the scale of what he sees coming, Hinton paused. "I don't despair, but mainly because even I find it very hard to take it seriously," he said. "It's very hard to get your head around the fact that we're at this very very special point in history where in a relatively short time everything might totally change at a change of a scale we've never seen before."
More On AI
AI Consciousness Checklist: 5 Things Scientists Can Measure
AI consciousness now has a scientific checklist. 20 researchers attempt to turn neuroscience theories into testable indicators.
→Then, almost to himself: "I'm kind of glad I'm 77."
Seven words from a man who built the thing he is now warning us about. The tiger cub is growing. Its creator watches, measures, warns. The Schwartz Reisman Institute plans to use its new funding to coordinate international AI safety standards.
Whether those standards arrive before the cub outgrows its cage may be the question that defines this decade.
Sources
- Primary: AI pioneer Geoffrey Hinton says world is not prepared for what's coming (CBS Mornings, 2025)
- Additional Context:
- Godfather of AI Geoffrey Hinton AI warning (CBS News)
- Good Ventures gift supports Geoffrey Hinton's global AI safety mission (University of Toronto)
- Godfather of AI warns it has progressed faster than originally thought (CNN, December 2025)
Fact Check: Claim-by-Claim Verification Verified
All claims verified against CBS Mornings interview, Nobel Prize records, University of Toronto press release, and CNN coverage. Quotes, dates, and biographical details confirmed.
Commentary
- The Boole family relationship has been described differently across sources (great-great-grandfather vs great-grandfather).
- Some quotes are from the CBS video which is not text-searchable.
Sources used for verification
Other reliable sources:
Fact-checked by Perplexity Sonar Pro on 2026-03-15
