In brief
- Geoffrey Hinton warned that artificial intelligence could surpass human intelligence and become uncontrollable.
- He outlined threats from both human misuse and autonomous AI, including cyberattacks, misinformation, and biological weapons.
- Hinton called profit motives one of the key reasons AI development won’t slow down.
Geoffrey Hinton, widely known as the “Godfather of AI,” issued his starkest warning yet in a new interview, cautioning that artificial intelligence poses not only a threat to jobs, but also an existential risk to humanity as a whole as the world races toward superintelligent machines.
Speaking on the “Diary of a CEO” podcast, Hinton outlined a grim vision of the future, suggesting that AI could eventually decide that humanity itself is obsolete.
“There’s no way we’re going to prevent it getting rid of us if it wants to,” Hinton said. “We’re not used to thinking about things smarter than us. If you want to know what life’s like when you’re not the apex intelligence, ask a chicken.”
Hinton said the threat will take two distinct forms: those stemming from human misuse, such as cyberattacks, the spread of misinformation, and the creation of autonomous weapons; and those arising from AI systems that become fully autonomous and uncontrollable.
“They can make lethal autonomous weapons now, and I think all the big defense departments are busy making them,” he said. “Even if they’re not smarter than people, they’re still very nasty, scary things.”
In May 2023, Hinton, a pioneer in neural networks, left Google and the University of Toronto after more than a decade working on artificial intelligence, so that he could speak freely about the technology’s dangers.
Hinton’s warning comes amid a surge in military applications of AI. Recent developments have highlighted the rapid integration of technology into defense operations, with the United States leading a rise in funding and partnerships.
In November, in its bid to augment the military with AI and autonomous weaponry, the U.S. Department of Defense requested $143 billion for research and development in its 2025 budget proposal to Congress, with $1.8 billion allocated specifically to AI. Earlier that year, software developer Palantir was awarded a $175 million contract with the U.S. Army to develop AI-powered targeting systems. In March, the Pentagon teamed with Scale AI to launch a battlefield simulator for AI agents called Thunderforge.
Hinton likened the current moment to the advent of nuclear weapons, except that AI is harder to control and is useful in far more domains.
“The atomic bomb was only good for one thing, and it was very obvious how it worked,” he said. “With AI, it’s good for many, many things.”
This combination of corporate profit motives and international competition is why AI development won’t slow down, Hinton explained.
“The profit motive is saying: Show them whatever will make them click, and what will make them click is things that are more and more extreme, confirming their existing biases,” he said. “So you’re getting your biases confirmed all the time.”
How would AI kill off humans? Hinton said a superintelligent AI could design new biological threats to kill off the human race.
“The obvious way would be to create a nasty virus—very contagious, very lethal, and very slow—so everyone would have it before realizing,” he said. “If superintelligence wanted to get rid of us, it would likely choose something biological that wouldn’t affect it.”
Despite the bleak outlook, Hinton isn’t entirely without hope.
“We simply don’t know whether we can make them not want to take over and not want to hurt us. I don’t think it’s clear that we can, so I think it might be hopeless,” Hinton said. “But I also think we might be able to, and it’d be sort of crazy if people went extinct because we couldn’t be bothered to try.”
Generally Intelligent Newsletter
A weekly AI journey narrated by Gen, a generative AI model.
