Why Nations May Race to Build Superintelligent AI Despite Risks

A recent essay explores why governments might pursue superintelligent AI (ASI) — AI far smarter than humans — even if it could pose an existential risk. The author argues that in an anarchic world, nations constantly fear each other. The US and China, for example, see each other as existential threats. Gaining a decisive strategic edge — like ASI — could ensure survival. So even if ASI is risky, the risk of being dominated by a rival may seem worse. This mirrors the nuclear arms race: countries built enough weapons to destroy civilization multiple times, and still couldn't agree to stop.

The author notes that nuclear disarmament largely failed, but nonproliferation (preventing more countries from getting nukes) succeeded. That gives hope: if the US and China can cooperate to slow down ASI development and prevent others from racing ahead, we might avoid the worst. They might publicly agree to build safe ASI and share benefits, even if cheating is likely. The key is making the risks of ASI clearer, so nations take them as seriously as they take nuclear war.

But the essay is pessimistic: as long as there's no world government, nations will compete for power. Technology has often provided decisive advantages, so the race for ASI may be unstoppable. The author suggests that clarifying the strategic situation — helping leaders understand the dangers — is valuable. If the risks become obvious, like they are for nuclear weapons, maybe nations will cooperate enough to avoid disaster. The hope is that the US and China, despite mutual distrust, can cooperate on safety and nonproliferation, just as they did with nukes.

In the end, the essay doesn't offer a clear solution, but it frames the problem: we need to make ASI risks as vivid as nuclear risks, so that self-interest aligns with caution. Otherwise, we may stumble into a race we can't control.