How unchecked AI could trigger a nuclear war
How unchecked AI could trigger a nuclear war
When facing the possibility of nuclear war, human beings exhibit a level of thoughtfulness and compassion that AI might not possess.
Chinese President Xi Jinping and U.S. President Joe Biden agreed late in 2024 that artificial intelligence (AI) should never be empowered to decide to launch a nuclear war. The groundwork for this excellent policy decision was laid over five years of discussions at the Track II U.S.-China Dialogue on Artificial Intelligence and National Security convened by the Brookings Institution and Tsinghua University’s Center for International Security and Strategy.
By examining several cases from the U.S.-Soviet rivalry during the Cold War, one can see what might have happened if AI had existed back in that period and been trusted with the job of deciding to launch nuclear weapons or to preempt an anticipated nuclear attack—and had been wrong in its decisionmaking. Given the prevailing ideas, doctrines, and procedures of the day, an AI system “trained” on that information (perhaps through the use of many imaginary scenarios that reflected the current conventional wisdom) might have decided to launch nuclear weapons, with catastrophic results.
Something something wargames?
Fail Safe, even.
Chinese President Xi Jinping and U.S. President Joe Biden agreed late in 2024 that artificial intelligence (AI) should never be empowered to decide to launch a nuclear war.
Narrator:
Paging Dr. Strangelove.
might
I feel that would be an enlightening research project. Why speculate? I think some people love to play wargames. And there should be enough information available today about the Cuban missile crisis or the countless other near escalations. We could just feed the information a general or the president would have had, to an AI. Measure the exact rate at which it starts nuclear war. And not just guess.
(I believe people have already done similar things. Let AI play Diplomacy and so on. And I bet the defense industry is way ahead of me and did this already.)
Edit: And I wonder if it's better or worse than ChatGPT, that we currently have Putin, Trump and Kim Jong Un at the trigger...
If those guys, especially Trump, could personally launch with no chain of command or chance for a sanity check in between, I think I really might rather have nothing more than an LLM in charge.
I think all 3 of them have replaced most of the people around them with yes-sayers. everyone not in line, or too intelligent (dangerous) had to go. So even if there are procedures in place, I'm not so sure.
On the plus side they are some weird form of friends and they try not to oppose each other that way. And seems at least the military in the USA still has some normal people.