It’s already too late to stop the AI arms race—We must manage it instead

Abstract

Can we prevent an artificial-intelligence (AI) arms race? While an ongoing campaign argues that an agreement to ban autonomous weapons can forestall AI from becoming the next domain of military competition, due to the historical connection between artificial-intelligence research and defense applications, an AI arms race is already well under way. Furthermore, the AI weapons challenge extends far beyond autonomous systems, as some of the riskiest military applications of artificial intelligence do not select and engage their own targets. This article draws on the history of AI weaponization and arms control for other technologies to argue that artificial-intelligence and robotics researchers should cultivate a security culture to help manage the AI arms race. By monitoring ongoing developments in AI weapons technology and building the basis for informal “Track II” diplomacy, AI practitioners can begin building the foundation for future arms-control agreements.