Geoffrey Hinton, who shared the 2024 Nobel Prize in Physics for work connected to artificial neural networks, has said it may turn out that humanity would have been better off never developing AI. In an interview with the Guardian, he framed that as a conditional judgment about a technology that could still lead to a very good future as well as a dangerous one.
The infrastructure scenarios
Hinton thinks a major incident could arrive later this year or next. His “Chornobyl” comparison is not a claim that a nuclear event is coming. He offered examples instead: an AI-enabled attack that causes a large power-system brownout, or one that takes the financial system down for a week. Those are scenarios he used to convey the scale of his concern, not reports of events that have happened or evidence of a specific impending attack.
His policy argument is more concrete than the metaphor. Hinton says advanced AI models should face an independent pre-release test similar in spirit to drug regulation: companies would need to provide strong statistical evidence that a model does more good than harm before it is released. He also supports shared international standards and cooperation between governments.
Independent testing proposal
That is not the same as calling for a blanket ban on AI. Hinton told the Guardian that he does not consider himself a doomer and that the technology could produce a “wonderful future.” He also said his instinct is that people are still in control, provided they make a serious collective effort.
For Hinton, the next decade may show whether societies can manage the risks without deepening inequality, unemployment and social division. The useful response is not to treat every warning as a prediction carved in stone. It is to ask what safeguards exist before systems are allowed to affect critical infrastructure, and whether the testing body can genuinely act independently of the company seeking release.
Keep the controls within reach. Risk discussion is not a promise of disaster; it is a way to make the boundaries, tests and accountability visible before a high-impact system receives access.
