After roughly 44 years working on artificial intelligence, Geoffrey Hinton has reached a conclusion that few investors in the sector want to hear: humanity "might have been better off never developing it". The 78-year-old Nobel laureate, who first sketched the idea of the neural network in the 1970s, told the Guardian that a large-scale AI-related incident could occur this year or next. His example was a temporary blackout of financial or power systems, the kind of failure that would be felt in trading rooms and hospitals alike. Hinton insists he is not a "doomer".

He still believes AI could deliver what he calls "radical abundance". But he argues that the answer will arrive within a decade, perhaps sooner, and that the two outcomes are starkly different. Success means shared prosperity. Failure means a huge gap between rich and poor, massive unemployment and, in his words, a society "torn apart".

He made the remarks before a screening of a new documentary, AI: Probably Nothing to Worry About, a title that was itself suggested by a chatbot. The sharpest criticism was aimed at the industry's leaders. Hinton said he does not trust Sam Altman, the chief executive of OpenAI, "at all", and called it "extremely worrying" that Altman helps steer one of the handful of American frontier companies racing China for the most powerful systems. Of Altman, Mark Zuckerberg and Elon Musk, he said they understand the risks but would rather have the money.