PerspectiveThe Big and Urgent Task of Revitalizing Nuclear Command, Control, and Communications

Published 4 October 2019

A few weeks ago, Adam Lowther co-authored an article proposing the use of artificial intelligence in nuclear weapons launch decisions. The article was met with pointed criticism. “Our sole concern is achieving the desired end-state — strategic stability and American supremacy, which are not mutually exclusive. We believe that trends in technological development will ensure artificial intelligence plays a central role, but the exact shape of the future is yet to be determined,” Lowther writes. “I would ask those who are serious about ensuring the survivability of the American nuclear deterrent to join us in thinking about new approaches to guaranteeing that its adversaries never, even for a moment, doubt that the United States can command and control its nuclear forces under any set of conditions. Whether it is through an artificial intelligence-based NC3 system or some other means will be decided over the coming decade — and only after several technologies reach maturation.”

Perhaps we were naïve, Adam Lowther writes in War on the Rocks, saying that he and hiscoauthor did not expect their article in War on the Rocks“America Needs a ‘Dead Hand’” —  to generate such great and mostly critical interest. Giving the article a provocative title was an effort to draw readers into a topic — nuclear command, control, and communications (NC3) — that is often overlooked. “And even though we chose the title, the argument we made was far more nuanced than just those few words,” he writes, adding:

Unfortunately, from the critiques that followed in the weeks after publication, we were unsuccessful in our objective of generating real discussion about America’s outdated NC3 system and how best to overcome the challenge presented by new nuclear weapons developments — Russian, in particular. Instead, our article was turned into a straw man and burned at the stake for heresy.

We certainly don’t mind an open debate, but hope to clarify some key points that we worry were lost in the minor controversy that followed.

Artificial intelligence certainly has hurdles to overcome, Lowther writes, but across many scientific disciplines, the once impossible is now reality. When the Apollo lunar module landed on the moon on July 20, 1969, less than seven years had passed since President Kennedy challenged the nation to go to the moon. “We expect artificial intelligence and related technologies to follow a similar path of rapid advancement. The reality of our time is that we are on the cusp of major innovations that will challenge our current approach to nuclear operations and war more generally.”

Lowther writes:

Our sole concern is achieving the desired end-state — strategic stability and American supremacy, which are not mutually exclusive. We believe that trends in technological development will ensure artificial intelligence plays a central role, but the exact shape of the future is yet to be determined.

I would ask those who are serious about ensuring the survivability of the American nuclear deterrent to join us in thinking about new approaches to guaranteeing that its adversaries never, even for a moment, doubt that the United States can command and control its nuclear forces under any set of conditions. Whether it is through an artificial intelligence-based NC3 system or some other means will be decided over the coming decade — and only after several technologies reach maturation.