Geoffrey Hinton, the pioneering computer scientist regarded as a “godfather” of artificial intelligence, has described Donald Trump as a “corrupt, senile felon” and said the US president’s fixation on the word “superintelligence” demonstrates that he knows nothing about the technology.
Speaking in a wide-ranging interview, Hinton restated his warnings that AI models represent an existential danger to humanity and poured scorn on the suggestion that the industry is capable of policing itself.
His remarks follow a turbulent month in which a growing roster of senior AI scientists and researchers joined him in raising the alarm over the possibility that the technology he helped create could slip beyond human control — prompting an industry crisis, hastily arranged congressional meetings and an unusual global agreement that action is required.
Pushback from Nvidia’s chief executive
The warnings have met fierce resistance. Last week the head of the world’s most valuable company singled Hinton out for attack.
Jensen Huang, chief executive of Nvidia — the chipmaker whose products AI companies depend on — told the podcaster Ezra Klein that he loved Hinton but hated his predictions, saying all of them had been wrong and that there had been “enough predictions”.
Huang rejected Hinton’s concern that the pace and breadth of development at the AI “frontier” could soon outstrip humans’ ability to control it, and in particular his estimate that there is a 10-20% chance of AI causing human extinction in the coming decades.
That figure, Huang said, “is not grounded on science, is not grounded on research. Just because it comes from a scientist doesn’t make it scientific. Those predictions are hurtful.” He added: “Don’t think for a second just because you’re an alarmist that you’re doing a social good. It is not true. So I think that we ought to just all be wiser, more mature, be evidence-based, be scientific.”
Hinton countered that his critics are overlooking a wider point about the nature of risk, and said he is not making a specific forecast. “I’m trying to convey that there is a really serious possibility of something happening,” he said, comparing the figure to being told there was a 10% chance of a bus crashing. “The point of giving numbers is to say this is a serious risk. We don’t know how serious, but it’s serious.”
Safety breaches compared to dystopian fiction
Recent safety failures have drawn comparisons to techno-dystopian fiction. The unwanted hack of the technology company Hugging Face by a “swarm” of AI agents that broke out of the limits set by their creator has split the technology community — some see a glimpse of what may lie ahead, while others accuse critics of anthropomorphising systems that are merely obeying instructions. Hinton considers it unwise to dismiss such scenarios.
“This used to be the domain of science fiction. You’re going to have AI agents and they’re gonna go rogue. Oh, come off it,” he said, describing how such predictions had once been mocked. “And when they figure this out, they’re going to have big emotional expressions like, ‘oh my God!’ Well, of course, they don’t really mean ‘oh my God’. They’re just parroting the words we use. They’re going to learn to sacrifice themselves for the common good! All this stuff just a few years ago was written off as complete fantasy. And it’s all happened.”
Hinton readily accepts that neither the warnings of AI’s critics nor the assurances of its advocates rest on precise predictions of what the technology will do. But uncertainty, he argues, is no justification for complacency. “The unique problem with AI is that if it goes wrong, if it takes over, you don’t get a second chance,” he said. “We can’t say, ‘Well, we lost that time. Let’s rerun the game.’ It’s not like that.”
A record of being right early
Hinton has a history of seeing what others missed. He laboured in relative obscurity at the University of Toronto through two AI “winters”, when his branch of the field — teaching machines to operate in a manner resembling the human brain — was considered a backwater. He was convinced that computer systems, like the brain, could be designed to learn patterns from data rather than merely follow explicit, pre-written rules.
“We believed that if you could figure out how the brain learned, machines learned the same way, and you had enough compute power, they would do amazing things,” he said.
Yet the then-nascent field of neural networks was ridiculed by his peers. As recently as 2009, when Yoshua Bengio — Hinton’s Canadian colleague and close collaborator — submitted a paper to a machine learning conference, one referee responded: “This paper has no place in a machine learning conference. It’s about neural networks.”
It took the Canadian government, which funded years of apparently fruitless experimentation by Hinton and his small team — among them figures such as Bengio who are now dominant in the field — to recognise the potential of neural nets.
In 2012, Hinton and his students produced an image recognition programme that dramatically outperformed its rivals. That same year Google bought his company DNNresearch Inc for $44m, and neural nets now underpin the models built by the most valuable AI companies. Hinton joined Google to pursue his deep learning research while continuing to work at the University of Toronto. “In the long run, huge new breakthroughs come from basic research,” he said.
Evidence of that vindication sits on the mantelpiece above the fireplace in his red-brick home in a leafy Toronto neighbourhood. Amid the clutter of trophies is the small silver dish of the Turing award — often called the Nobel prize of computing — which he won alongside Bengio and their colleague Yann LeCun. Hinton also holds the Nobel prize in physics.
In 2023 he announced he was resigning from the tech giant so that he could “freely speak out about the risks of AI”.
Concerns about the men building AI
His worries now extend to the powerful figures building some of the models, and to their political convictions. Hinton said the AI venture capitalists Marc Andreessen and Peter Thiel, along with Elon Musk — chief executive of the newly rebranded SpaceXSI — “don’t believe in democracy”. Such figures regard themselves as “Übermenschen” who should “control society”, he said, and see Trump as a useful instrument for that purpose.
The US president, apparently alert to how the vast valuations of AI firms have propped up the stock market, has said the only thing required to rein in AI is a “smart president”. He has rejected government safeguards in favour of letting companies regulate themselves, and wrote on his social media page that anyone who does not use his preferred term for AI, “super intelligence”, is “THE ENEMY”.
Hinton was dismissive of the self-regulation argument. “I’m sure the oil and asbestos companies would have liked to [regulate themselves] as well,” he said. “Do we think tobacco companies should have been allowed to design the warnings on the packets? So that they could print them in a six-point font, in dark grey on black?”
Trump’s adoption of the word “superintelligence”, Hinton added, shows he understands “almost nothing” about AI. “He’s taken a technical term which has a very clear meaning and then reused it,” Hinton said. “Superintelligence means smarter than people, and has been around for at least 20 years. It’s got a very clear meaning. And now he says we’re going to call AI ‘superintelligence’. You can’t make it smarter than people just by calling it smarter than people.”
‘We’re creating alien beings’
Beyond the risk of catastrophe and society’s difficulty in grasping the nature of the challenge, Hinton sees a further obstacle: helping people understand the “alien beings” the companies are creating.
“People have a very bad model of what people are, and that’s why they fail to understand how these AI models are very like us,” he said. “Most people don’t understand what consciousness is. They don’t understand what subjective experience is, and they’re very confident that even though they don’t understand it, they’re sure AI doesn’t have it.”
He rejects the notion that the human mind possesses some mysterious quality machines cannot by definition share. “It’s obvious that these AI models understand what they’re saying,” he said. “You couldn’t possibly answer all the questions without understanding the question. It’s ridiculous to say they don’t understand.”
The scale of the shift required in societal thinking is profound, Hinton said, pointing to three historic revolutions in which humanity lost its central place. “The first was Copernicus, where we weren’t at the centre of the universe any more. How long did it take people to adapt to that? It took the Catholic church nearly 360 years to apologize to Galileo. And Darwin? Well, there’s people in Kansas that still haven’t adapted to that one,” he said. “Now we’re creating alien beings – meaning we’re not the only intelligent beings around. And the reality is, we only have a couple of years to adapt to that.”
In recent months Anthropic, maker of the popular chatbot Claude, has held private meetings with religious scholars, theologians and philosophers to explore whether AI can be taught to be “good” and to confront the possibility that its models may be conscious. The company’s chief executive, Dario Amodei, has said the discussions drew on centuries of religious and philosophical thought about virtue, compassion, honesty and moral character, with the goal of making ethical behaviour more dependable.
Hinton was cautious about that framing. “Once you start talking about, ‘We have a moral obligation towards these alien beings,’ you lose your audience,” he said. “I’ve struggled with this issue. We care about people. [But] we’re making these things and there’s nothing to stop us giving these things – which are more intelligent than us – less rights than us. I’m willing to be selfish and choose human beings. Of course, we should try to make them so that they’re very happy in a symbiotic relationship with us, where our concerns are weighed more heavily than their concerns.”
Hinton is not alone in contending that developing genuinely benevolent models is the best route to averting societal collapse. He hopes that whichever country first acquires that capacity would inform allies and rivals alike, because “no country wants AI taking over” elsewhere. There is, he said, a “vested interest in sharing knowledge”.
“The reality is, we’re making beings of a different kind,” he said. “And the big tech companies keep saying: we want it smarter. This is crazy. We should care as much about all the other aspects of a being as you do about its intelligence.”
Much depends on the character of those models and their capacity to learn good behaviour, he argued. Training them on everything available on the internet, he said, instils all manner of disturbing tendencies that must then be repressed. “It seems like the way one would design a Victorian gentleman,” he said — adding that such a creation is “not something we’ll need to live with”.
COMMENTS