Why I’m Not Worried About AI Killing Us All

Summary

This opinion article argues that debates about AI killing humanity often conflate two separate questions: whether artificial general or superintelligence will ever be built, and what would happen if it were. The author defines superintelligence as an entity so far beyond human understanding that humans would not be able to control it, but separates loss of human control from the claim that AI would kill everyone. The author’s optimistic view rests on three arguments: superintelligence may not share human motivations, AI development is unlikely to be stopped by any individual, company, or government, and humanity already faces serious catastrophe risks without AI. The article cites the Cuban Missile Crisis and the 1983 Soviet false alarm as examples of how narrowly nuclear disaster was avoided, and points to nuclear weapons, pandemics, geopolitical instability, debt, inequality, and social fragmentation as existing risks. It does not advocate unrestricted development: the author supports safety testing and alignment work, while opposing confidence in centralized control. The author’s main concern is the period before superintelligence, when powerful AI remains under human control and could be used to cause substantial harm.