Probably 0.5 percent or less. To be clear, even a 0.5 percent risk makes this one of the greatest potential known threats out there. I believe the first time AI does something truly troubling/deceptive/misaligned that causes real harm the public outrage will be unlike anything we've seen leading to strict controls that maybe don't have the political support today. I think it's likely that these initial harms will be small scale or at least not something that immediately threatens human existence. We can handle threats that unfold slowly whether it's this or rising global temperatures. I don't believe there will be a tipping point where all of the sudden AI becomes exponentially better. Or at least I don't believe we'll stumble into that tipping. We will see it coming.