Description
The world has witnessed a phenomenal rise of Artificial Intelligence (AI). Scholars in International Relations (IR) who focus on the global governance of AI frequently use the example of how states regulated the use of nuclear weapons in the past through the Nuclear Non-Proliferation Treaty. Many scholars suggest similar ways of governing AI with the understanding that AI, like nuclear weapons, poses a significant threat to humanity. In this paper, I argue that commonplace ways of relying on nuclear weapons conventions and the NPT to govern AI are fundamentally wrong. There are three reasons for my claim. First, the revolution of nuclear weapons remained bound in its dependence on advanced scientific and technological skills. Regulating nuclear weapons was possible when a small group of advanced nuclear weapons states agreed on a solution. However, the revolution of AI is far more decentralized. More sophisticated training of AI algorithms is possible with semiconductors that are nearly a decade old. This decentralized nature of the AI revolution cannot be tamed like the way nuclear weapons were tamed in the past. Second, the negotiation of NPT worked in the period of decolonization, with much of the Global South forcefully silenced. The AI revolution has now peaked at the time of more ambitious voices from the Global South asking for greater redistribution of technology privileges. In the past even when India stormed out of NPT, African voices remained silenced. Today, the global governance of AI cannot take African voices for granted when semiconductor technologies largely rely on critical and rare earth minerals from the continent. Finally, the signature of NPT was yoked to the (unfulfilled) promise of disarmament. Today, with the distrust of the West, AI governance cannot proceed with empty promises. The global governance of AI needs a new model.
