The Risk of AI Getting Out of Control

There was a call to stop the development of not AI as a whole, but to not train any even larger models than GPT-4. "I didn't sign it because I thought it was silly in the sense that what they're calling for is suddenly completely infeasible," he says. He also agrees with Sam Altman that if you think about the existential risk of these things getting out of control, given that you can't stop this stuff developing, should be doing experiments.