“Geoffrey Hinton suggests that a major AI incident, analogous to the Chernobyl disaster, might be necessary to compel society to invest adequately in AI safety.”