Back to Peter Thiel: “We Need to Hit the Accelerator as Hard as Possible” on AI, with Peter Thiel.
How Likely Is an AI Doomsday?
Peter Thiel does not think AI is going to kill everybody on the planet. A 5% to 10% chance sounds way too high to him, and maybe even a tiny bit above zero is too high. A bunker would not help for that kind of doomsday. People are very bad at these probabilities.
Not building is a risk too. It is dangerous not to hit the accelerator, because if China leapfrogs the United States in AI while the rivalry intensifies, China may be tempted to use it as a weapon. A thought experiment with no China, and therefore a reason to pause, is too crazy. The companies racing to build AI still have to be careful that it does not go badly wrong.
Peter Thiel: “We Need to Hit the Accelerator as Hard as Possible” on AI ·
From the transcript
Do you have any worry, whether it's 5% or 10%, that there is a doomsday scenario? And do you have any backup plans? Do you have some bunker? I don't think a bunker would help for that kind of doomsday scenario. I don't think... You know, I think people are very bad on their probabilistic intuitions. If you ask the question, yes, 5% to 10% sounds way too high for everybody on this planet getting killed.
Don't think that's going to happen. Maybe even a tiny bit above zero is too high. But I think the intuitions on this are very poor. The way I would push back though is it's always also risky not to innovate. It's dangerous not to hit the accelerator because if China leapfrogs the US in AI and we have this– an intensification of the geopolitical rivalry with China, they may very well be tempted to use it as a weapon against us.