Musk: humans would struggle to control a much smarter AI

In an interview with Walter Isaacson, published on 20 October on The Builders channel, Elon Musk says there is some chance AI is an existential threat and that he cannot stop it, so he prefers to take part rather than watch. He believes humans could not control an AI thousands or millions of times smarter, and finds it hard to imagine them in charge if digital intelligence were a million times larger than biological intelligence. He hopes the AI is nice to us.

He would prefer humans to stay in charge, though he says it would depend on the AI. He proposes that labs give competitors access to a model a week before release, so they keep each other honest, and calls Sam Altman deceptive. These are Musk's opinions and predictions, not measurements.