Artificial intelligence is changing society faster than we can write the rules. The people building AI admit there is a real chance it goes catastrophically wrong. Those odds seem higher than we would ever accept from planes or power plants. Yet governments struggle to find neutral, credible evidence about the risks, because most warnings come from companies or campaigners with a stake in the answer.
Michael Noetel builds the evidence that lets governments act. With MIT, he helps run the AI Risk Initiative, including a database of more than 1,000 ways AI can cause harm, cited in the International AI Safety Report. He also led one of Australia’s largest surveys of attitudes to AI, which found the public wants AI to be as safe as air travel, far safer than experts expect.
Michael honed this approach in health, combining 218 trials to show exercise can treat depression as well as medication. CNN, the BBC and Good Morning America carried the results to millions. Now he is comparing AI risks with the risks we tolerate in other industries, asking experts which fixes would actually work, and tracking whether AI companies keep their safety promises.