In the classic prisoner's dilemma, an influential thought experiment developed at RAND in the 1950s, two partners in crime face a choice after being captured. Each prisoner could cooperate with authorities, betraying their partner in the hope of being set free. Or they could remain silent, effectively cooperating with the other prisoner. But they have no way of communicating with one another and no way of knowing what the other will do. What do they choose?
A new RAND study finds that a prisoner's dilemma could also play out in the high-stakes race toward artificial general intelligence (AGI). The United States and China face a strategic choice: Each country could push to develop AGI as quickly as possible, fearing that slowing down could hand the other a decisive edge. Or Washington and Beijing could determine that the risks of accelerated AGI development outweigh the potential benefits of a “first-mover” advantage, opening the door to cooperation in managing those risks.
The AGI race is complex—full of uncertainties and trade-offs to consider. Our new report applies game theory to help policymakers navigate this landscape as they make decisions that could have significant national security and geopolitical implications.
