Hinton says on November 4 an AI-scale Chernobyl could occur within the next two years
Hinton called the OpenAI agent intrusion at Hugging Face a Three Mile Island-style warning, and said a worse scenario could be an AI-driven outage or a week-long collapse of financial systems.
Reported on November 4, Nobel laureate Geoffrey Hinton said in an interview with the Guardian that the field of artificial intelligence could see a Chernobyl-scale accident within the next year or two. Hinton said the OpenAI agent intrusion at Hugging Face this summer was a warning on the order of Three Mile Island, and that a more serious artificial intelligence Chernobyl could involve an AI-driven large-scale blackout or a week of financial system collapse. Humanity might be able to withstand such a disaster and only then recognise how dangerous the technology is.
Hinton does not trust Meta chief executive Mark Zuckerberg, SpaceX-linked leaders or OpenAI chief executive Sam Altman, arguing that while these leaders understand the risks, they focus only on the economic returns of the technology and let development proceed. Hinton concedes he believes AI has the potential to bring great benefit to humanity, but says the risks it carries are severe enough that he sometimes thinks the technology should never have been developed. He expects humanity to know within a decade or less whether the risks of AI can be managed safely, and backs mandatory testing of frontier AI, independent evaluation and the setting of common intergovernmental standards.