Nuclear weapons are key to preventing AI from taking hold

An AI researcher has called on the countries of the world to use the threat of nuclear war to prevent the rise of artificial general intelligence. In an op-ed for TIME, AI theorist Eliezer Yudkowsky said that pausing research into AI isn’t enough. Yudkowsky said that the world must be willing to destroy the GPU clusters training AI with airstrikes and threaten to nuke countries that won’t stop researching the new technology. ADVERTISEMENT Yudkowsky’s op-ed was a response to an open letter calling for a six month moratorium in the study of AI by asking the world to shut down all its AI research. In Yudkowsky’s thinking, the pause isn’t long enough. For him, it is not a matter of if AI will kill humanity, but when. “Not as in ‘maybe possibly some remote chance,’ but as in ‘that is the obvious thing that would happen,’” he wrote. Yudkowsky believes that the only way to stop the creation of AGI—machine intelligence that matches or surpasses humans—and therefore the destruction of the entire human race, is to take actions that would inevitably cause a wider war and lead to millions, or even billions, of dead humans. In his piece for TIME, Yudkowsky said that the world must shut down all the large GPU clusters and put a ceiling on the power draw used by a computing system training AI. “No exceptions for governments and militaries,” he said. “Make immediate multinational agreements to prevent the prohibited activities from moving elsewhere. Track all GPUs sold. If intelligence says that a country outside the agreement is building a GPU cluster, be less scared of a shooting conflict between nations than of the moratorium being violated; be willing to destroy a rogue datacenter by airstrike.” ADVERTISEMENT Yudkowsky is asking the audience to be more afraid of the hypothetical possibility that AI will kill humans than the likelihood that his prescriptions would cause a nuclear war. “Frame nothing as a conflict between national interests, have it clear that anyone talking of arms races is a fool. That we all live or die as one, in this, is not a policy but a fact of nature,” he said. “Make it explicit in international diplomacy that preventing AI extinction scenarios is considered a priority above preventing a full nuclear exchange, and that allied nuclear countries are willing to run some risk of nuclear exchange if that’s what it takes to reduce the risk of large AI training runs.” The U.S. and China are the relevant examples here because both their military and civilian infrastructures are racing towards developing AI. These are the countries, in Yudkowsky’s scenario, most likely to air strike a data center in the heart of their rival. To be clear, either country conducting that kind of airstrike would start a war, full stop. Tensions are already high between China and the U.S. and both are armed with nuclear weapons. China can’t float a balloon above the U.S. without the Pentagon shooting it down, it’s hard to imagine it responding to an airstrike on a data center in San Antonio with anything less than the full force of its military. China, likewise, likely wouldn’t suffer an incursion and bombing in its airspace. Yudkowsky made clear that nuclear threats are on the table and that the world’s great powers should not stop short of a full nuclear exchange to prevent the rise of AI. China and the U.S. entering into a full nuclear exchange would kill billions of people and permanently alter the climate of the planet, starving billions more. He is, in his moral calculus, willing to kill potentially billions of people and possibly doom the planet for a generation to prevent The Terminator from happening.