Defense Department Must Invest in the Right Kind of AI for Future Conflicts
Written By: Brandon Tseng, Shield AI Co-founder
The release of yesterday’s National Security Commission for AI report by a distinguished group of bipartisan officials including academics, former military officials, and tech leaders rightfully signals a clarion call for the United States to come to grips with understanding AI’s near-ubiquity and sheer omnipotence. If sentences themselves could magically blink red, the NSCAI commission writing “Americans have not yet seriously grappled with how profoundly the AI revolution will impact society, the economy, and national security… the United States remains unprepared for the coming era” would certainly be fitting examples. As a former Navy SEAL and co-founder of a national security-focused AI tech firm, I was invited to meet with the writers to offer my perspectives on the intersection of AI and national security.
In the era of AI and autonomous systems, my sobering assessment is that our military is unprepared for future conflicts because collectively, as a nation, we have failed to invest in and integrate the AI and autonomy that matters into our armed forces at any degree of scale. We’ve also treated AI and autonomy as a monolithic form of technology, focusing on low-risk easy wins, and have not invested in the most relevant aspects of the technology to counter growing threats posed by foreign powers, namely China and Russia. Not all AI and autonomy are the same and not all AI and autonomy can provide effective deterrence against our adversaries.
To confront strategic and operational vulnerabilities in both our legacy and future unmanned platforms, policymakers and DoD officials must urgently prioritize the integration of AI and autonomy software that enables intelligent swarming operations of unmanned systems in GPS- and communications-denied environments. This AI and autonomy stack must be tightly coupled with scalable simulations supported by reinforcement learning techniques that enable continuous, near self-directed improvement.
Technologically, the commercial sector is doing all of this today, and it is most apparent with self-driving cars. These autonomous vehicles on our streets can already function without reliance on GPS or communications because tech firms needed to plan for obvious eventualities, such as ensuring that their vehicles can still safely operate while inside tunnels and during network outages. Self-driving vehicles don’t have the luxury to reach back to the cloud or to an operations center to decide whether to brake for pedestrians; instead, they must decide in real-time on the edge, on their own. Elon Musk has famously stated “Any cars that are being made that don’t have full autonomy will have negative value. It will be like owning a horse. You will only be owning it for sentimental reasons”
It may sound surprising, but the hard truth is that an “autonomous” military drone dependent on GPS for waypoint navigation and prolonged communications-links for real-time decision-making does little to protect our national security interests; it will be cannon-fodder against a peer adversary with advanced electronic warfare capabilities. Unfortunately, collectively, with a few exceptions, this description represents the current state of our military’s arsenal. Our adversaries know it. And we know it. Therefore, DoD must immediately establish a threshold requirement for our unmanned arsenal to be able to continue functioning in GPS- and communications-denied environments, and to be operationally capable of level 5 self-driving autonomy. Furthermore, DoD should push itself to rid its unmanned arsenal’s reliance on GPS and communications by end of 2023 – a daunting but achievable task with the power of AI.
Some might argue that creating such new advanced systems means that the Biden Administration must therefore increase the size of DoD’s budget. But spending more just for the sake of spending is not the answer. Published reports indicate that in 2019, the US spent almost three times on its military than China did for its armed forces. But as the NSCAI report notes, “In some areas of research and applications, China is already an AI peer, and it is more technically advanced in some applications. Within the next decade, China could surpass the United States as the world’s AI superpower.” DoD’s budget is large enough and more importantly flexible enough to enable the diversion of meaningful resources in a surgical way for AI and autonomy.
Two days ago, the Biden administration ordered airstrikes in Syria, retaliating against Iran-backed militias and standing up for those who can’t stand for themselves. But Syria and Iran are not technological peers. If America continues its path of incremental improvements to legacy technologies used during the Cold War and Global War on Terror, our national security strategy of deterrence through strength will be in for a rude awakening. Events like the invasion of Crimea by the Russian Federation and China’s brazen crackdown in Hong Kong and the buildup of islands in the South China Sea will continue to spiral if respect for America’s technological military edge is not restored. We must lead in AI and autonomy in order to protect our own national security and those of our allies.
Only in the past 70 years has the US military emerged as the preeminent stabilizing force in the world. The NSCAI report makes clear that to continue living in a world of relative stability, the US military must urgently integrate AI and autonomy into its DNA – and it must be the kinds of AI and autonomy that will deter our adversaries and help prevent state on state armed conflict from ever taking place.