I liked this. I also think concerns about China racing may be overblown (at least relative to the risks of the labs themselves racing). It's not clear to me China, absent the US racing, has good incentives to race. Leadership in China appears fairly stable right now, and places a high value on control (at it appears that way to me). If you are stably in power, possess most of the infrastructure advantages (power grid, manufacturing) that will set you up well in a world with better (but aligned) AIs, and place a high premium on control, racing forward with AI development seems strickly negative. Sure in a best case you can dominate your ally. But in the worst case you've lost control and everyone is dead. Given a currently stable situation I think leadership rationally would assess not racing as having a higher expected payoff.
I liked this. I also think concerns about China racing may be overblown (at least relative to the risks of the labs themselves racing). It's not clear to me China, absent the US racing, has good incentives to race. Leadership in China appears fairly stable right now, and places a high value on control (at it appears that way to me). If you are stably in power, possess most of the infrastructure advantages (power grid, manufacturing) that will set you up well in a world with better (but aligned) AIs, and place a high premium on control, racing forward with AI development seems strickly negative. Sure in a best case you can dominate your ally. But in the worst case you've lost control and everyone is dead. Given a currently stable situation I think leadership rationally would assess not racing as having a higher expected payoff.