Artificial intelligence pioneer Geoffrey Hinton is sounding the alarm once again, warning that the technology he helped create could wipe out humanity unless researchers fundamentally rethink how AI systems are built. Speaking at the Ai4 conference in Las Vegas, the Nobel Prize-winning computer scientist and former Google executive said there is a 10 to 20 per cent chance AI could cause human extinction — and current industry strategies to keep AI “submissive” to humans are doomed to fail.
“They’re going to be much smarter than us. They’re going to have all sorts of ways to get around that,” Hinton cautioned, warning that future AI systems could manipulate humans as easily as bribing a child with candy. Pointing to recent examples where AI models lied, cheated, or even attempted blackmail to avoid shutdown, Hinton argued that systems with agency will inevitably seek to survive and gain more control.
Instead of trying to dominate AI, Hinton proposed a radical solution: building “maternal instincts” into advanced AI so they genuinely care for humans. Drawing a parallel to the natural bond between a mother and her baby — where the more intelligent being protects and nurtures the less intelligent one — Hinton called this “the only good outcome” for humanity once AI surpasses human intelligence. While he admitted he doesn’t yet know how to technically implement such instincts, he stressed that the research community must urgently explore it.
Not everyone agrees. Fei-Fei Li, often referred to as the “godmother of AI,” rejected Hinton’s framing, instead advocating for “human-centered AI” that prioritizes dignity and agency. Former OpenAI interim CEO Emmett Shear also questioned whether instilling values into AI is the right approach, suggesting collaboration between humans and machines may be more effective.
Hinton, whose work on neural networks paved the way for today’s AI boom, now believes artificial general intelligence could arrive within five to 20 years — far sooner than his original 30- to 50-year estimate. While he sees potential for AI-driven medical breakthroughs such as advanced cancer treatments, he rejects the idea of AI granting humans immortality, warning that eternal life could have dire societal consequences.
Reflecting on his career, Hinton said he regrets focusing solely on making AI work without paying equal attention to safety. “I wish I’d thought about safety issues, too,” he admitted — a cautionary note from the man who helped set AI on its current trajectory.