The "godfather" of AI: an AI "Chernobyl" could come this year or next

When the man who laid the foundations of artificial intelligence says it might have been better if we had never created it, that's not just another column for technophobes. Geoffrey Hinton, one of its "godfathers" and a Nobel laureate in physics, has warned that humanity could face a serious crisis caused by artificial intelligence as soon as this year or next.

Hinton doesn't mean science-fiction robots. He's talking about something far more down to earth: major power outages or a banking system that stops working for days. In an interview with a British outlet, the 78-year-old scientist, whose work on artificial neural networks laid the groundwork for today's systems, said the next ten years or so will show whether humanity manages to rein in the risks.

Chernobyl without the radiation

Hinton compares recent safety incidents involving autonomous AI systems to the 1979 accident at the Three Mile Island nuclear plant in the US. A warning, but not a catastrophe. The catastrophe, he believes, is still to come - a kind of AI "Chernobyl". Not a nuclear accident, but a cyberattack that could cut off the power supply or temporarily paralyse the financial system.

The cynical part is that, in his view, exactly that kind of crisis might finally push world leaders to take regulation seriously. In other words, the lights have to go out first for anyone to react. How many times have we watched institutions wake up only after the damage was already done?

Who knows the risks and rushes anyway

Hinton doesn't spare the bosses of the biggest tech companies either. He voiced deep distrust of OpenAI chief Sam Altman, and accused him, Mark Zuckerberg and Elon Musk of knowing the risks but putting profit and the race against competitors first. That is his assessment of their motives, not an established fact - but it comes from a man who knows these systems from the inside.

His proposal is simple: the most advanced systems should be tested mandatorily before they reach the public, following the model of the US Food and Drug Administration, which demands proof of safety for every drug before approving it. Developers would have to prove to an independent regulator that the benefits outweigh the harm. That puts him in direct opposition to Donald Trump's approach of fewer rules so American companies can compete with China.

Hinton left Google in 2023 so he could speak more freely about the dangers. He hasn't lost hope - he believes people can still steer development and prevent the worst outcomes, but that it will take far greater effort from governments, scientists and industry. The only question is whether anyone listens to him before or after the first "Chernobyl".