When Geoffrey Hinton walked on stage at AI4, he wasn’t greeted like an academic. He was greeted like a rock star. “Geoffrey, we love you!” someone shouted as he took his seat. Before him was an audience of at least 5,000 AI leaders—founders of scrappy startups, executives from global tech giants—all eager to hear from the man who has shaped their industry more than almost anyone alive.

Known as the “Godfather of AI,” Hinton has spent decades building the neural network technology that powers today’s large language models, image generators, and autonomous systems. But on this day in Las Vegas, his role wasn’t to rally the troops—it was to warn them.

The Timeline to Superintelligence

Moderator Shireen Ghaffary of Bloomberg got right to it: how close are we to AI that “vastly outsmarts humans”?

“GPT-5? Looks like it might have been a small backwards step,” Hinton said with a slight shrug, drawing a ripple of laughter. Then he turned serious. “I think most experts think sometime between five and 20 years from now. Very few people who are experts in AI think we’re not going to get that.” He paused. “It could be just a few years. I think a reasonable bet is sometimes between five and 20 years.”

The crowd, many of them working feverishly to push that timeline forward, listened in silence. Hinton didn’t mince words about what could go wrong. “We’ve always had propaganda,” he said. “But never this automated, never this personalized.” AI systems, he warned, can now generate misinformation at a scale and speed that outpaces any attempt to correct it.

Economic disruption was next on his list. “We are about to enter an economy where thinking, not just doing, can be automated,” he said. Unlike past industrial revolutions, he stressed, this one doesn’t just threaten factory jobs—it threatens knowledge work, creative industries, and even parts of scientific research.

And then came the most sobering risk: loss of control. “Once an AI develops its own internal objectives that diverge from ours, even slightly, aligning it becomes extraordinarily difficult,” he told the room. “And if it surpasses human intelligence… well, you can do the math.”

Regulation, Open Source, and the Race Ahead

Ghaffary pressed him on whether governments could realistically regulate AI in time. Hinton’s reply was blunt: “Regulation tends to follow harm, not anticipate it. With AI, that lag could be catastrophic.”

On open-source development, he refused to take an absolutist stance. “Open source democratizes access, which is good for innovation,” he said. “But it also removes guardrails, making it easier for bad actors to weaponize these systems.”

For all his warnings, Hinton rejected the idea that he had turned against the technology. “I’m not anti-AI,” he said. “I’m pro-survival.” His call to action was to accelerate the positive uses of AI—like breakthroughs in medical research and climate modeling—while constraining those capabilities that could cause irreversible harm.

He pointed to protein-folding research as an example worth emulating. “That’s the kind of AI progress we should be racing toward,” he said, “not building systems whose primary competitive advantage is their ability to outwit humans.”

Hinton looked out at the thousands of faces in front of him and delivered what felt like the real purpose of his talk: “The people in this room are the ones writing history. In 50 years, no one will care how much revenue your model generated in 2025. They will care whether you built something that improved human life—or endangered it.”

He urged companies to invest in safety research, interpretability, and cross-industry standards. “Competition is natural in business,” he said. “But existential risk should be a pre-competitive space.”

The Weight of His Words

What made the session resonate wasn’t just the science—it was the sense of personal reckoning. Hinton, who had stepped away from Google in 2023 to speak freely about AI’s dangers, was effectively telling the crowd that the dream he had pursued for 50 years could, if left unchecked, become a nightmare.

“We have maybe five, maybe ten years before systems surpass human intelligence in most domains,” he said. “That’s not much time. We can either use it to build the safeguards we’ll need, or we can spend it convincing ourselves they aren’t necessary.”

The applause at the end was respectful but subdued. Many in the room seemed to understand that they’d just been given a choice: slow down and steer the future—or speed ahead and hope the brakes work when we need them most.

###

Techonomy 25: Human Agency Meets Machine Autonomy
📅 November 18, 2025 | 🏙 Civic Hall, New York City

The AI era is here—and it’s rewriting the rules of business, technology, and society. At Techonomy 25, we’re bringing together world-class thinkers, entrepreneurs, and changemakers to explore how humans can shape a future where machines amplify our potential rather than replace it.

From responsible AI and the future of work to quantum breakthroughs, biotech, and climate innovation, our program will deliver sharp insights, lively debates, and unexpected collaborations. You’ll hear from industry pioneers, visionary founders, and the policymakers shaping what comes next.

Seats are limited—secure yours now and be part of the conversation shaping our shared future.

Register here: https://techonomy.com/event/techonomy-25-human-agency-meets-machine-autonomy