🧠Hinton Says Humanity Has Under a Decade to Figure Out AI Risk

Your AI roadmap just got a deadline

Geoffrey Hinton, the Nobel-winning computer scientist who helped build the foundations of modern neural networks, says humanity will know within a decade whether it can manage AI's risks. He isn't a doomer. He thinks the upside could be a "wonderful future" with "radical abundance," but he's worried the people building this stuff are moving faster than they're thinking. Why should you care? Because the failure modes he's describing don't live in a research paper. An AI-enabled attack on the electricity grid could cause a massive brownout, and a financial system outage lasting a week is, in his words, possible. If you're running production systems on cloud infrastructure, your incident runbooks are about to matter a lot more than your model benchmarks. The numbers Hinton points to are blunt. He says an incident that could trigger a power or banking shutdown is plausible this year or next. He also points to the US FDA's drug approval process as a model for how AI should be tested before release, and notes that world leaders beyond Trump want mandatory testing of frontier models. What to watch: the UK prime minister wants to use the UK's G20 presidency next year to push a "single set of global principles and standards" for AI. Whether that becomes binding rules or another voluntary pledge is the real question.