Military Tech Experts Suggest The US Should Give AI Control Of A Nuclear Warfare Button

Many of the nuclear deterrence strategies that the U.S. has today were developed during the Cold War era when military officials had significant time to detect an incoming nuclear attack and decide a course of action for a response. This mutually assured destruction worked, in that any enemy with nuclear weapons knew the U.S. would have time to fire its own nuclear response at them and neither would survive the outcome.

mushroom cloud

However, technology marches on and times have changed, with potential enemy forces around the globe making significant upgrades to nuclear arsenals -- upgrades such as hypersonic cruise missiles and other types of missiles that are difficult to detect. One type of hypersonic cruise missile gives the U.S. military only six minutes of lead-time to detect an incoming attack, devise a plan, and launch a counter-attack.

The fear is that America's aging nuclear command and control system, known as NC3, might not offer U.S. leaders the time they need to detect an incoming attack and decide on a response before American lives are at risk in a surprise attack. However, there are options that the U.S. has for escaping the compressed time scales that modern nuclear war would impose, and one of them is artificial intelligence.

An AI system would be capable of detecting an incoming attack, analyzing pros and cons, and then choosing from a list of pre-approved options to respond to a detected attack in enough time to give the U.S. back its capability of launching a retaliatory strike against any surprise attack. The obvious problem is, however, that an AI-powered system would eliminate humans from the decision making process of launching nuclear weapons.

AI experts also fear a rouge AI that could start a nuclear war that wasn't necessary, creating a Skynet situation. Supporters of the system point out that AI is already used for target identification, controlling autonomous platforms, pattern recognition, and other wartime tasks. Google's DeepMind has already worked on an AI kill switch that could stop a rogue AI.

b 2 b 52 bombers usaf

Another option likely to be wholly unacceptable to many Americans, is to simply absorb the first nuclear attack and then respond to the attack. Some experts think this capability could be sufficient to make the first surprise strike against America not worth the risk for enemies. Another option would be improving our pre-launch strategic warning systems, such as improved surveillance and reconnaissance as part of a preemption strategy.

In the end this all seems like the plot to a Hollywood movie... in fact, it sounds just like WarGames. WarGames was a hit blockbuster that came out 35 years ago starring Matthew Broderick. Its plot revolved around a supercomputer that was tasked with detecting and responding automatically to an incoming nuclear threat from the Soviet Union. Needless to say, things didn't exactly go as planned once a pesky teenager "hacked" into the system.

Image: USAF Tech. Sgt. Richard Ebensberger

Show comments blog comments powered by Disqus