Foreign Policy Essay
Wars of None: AI, Big Data, and the Future of Insurgency
Editor’s Note: The rapid pace of technological innovation is changing the nature of warfare, and futurists are busy spinning out scenarios of a U.S.-China clash in twenty years involving nano-technology and fully autonomous weapons systems. Yet how will new technologies shape insurgency and counterinsurgency, which conjures up images of guerrillas hiding in Vietnam's jungles? My Brookings colleague Chris Meserole looks at two of the latest books on the subject and assesses how the balance between rebels and government may tilt.
When U.S. Special Forces entered Afghanistan in 2001, Facebook didn’t exist, the iPhone had yet to be invented, and “A.I.” often referred to an NBA star. Seventeen years later, American special operations forces continue to ride horseback in rural Afghanistan, but information technology has advanced rapidly. Recent breakthroughs in robotics and artificial intelligence (AI) have captured the popular imagination and prompted sober talk of an impending AI revolution. Yet surprisingly little of that talk has touched on the small wars and insurgencies that have dominated U.S. foreign policy in the 21st century.
The definitive work on emerging technology and insurgency has yet be written, but two recent books offer suggestions for how the era of big data and AI will affect the United States’ modern conflicts. Small Wars, Big Data: The Information Revolution in Modern Conflict, by Eli Berman, Joseph Felter, and Jacob Shapiro, offers few musings about the future of insurgency, but lays out a compelling theory about the ways in which information shapes insurgent violence. By contrast, Paul Scharre’s excellent new book, Army of None: Autonomous Weapons and the Future of War, offers little in the way of counterinsurgency strategy, but is wholly concerned with how artificial intelligence will reshape armed conflict. Taken together, they begin to sketch out a vision for how AI and big data might alter insurgent dynamics.
The core insight of Small Wars, Big Data is that insurgencies are ultimately competitions over information rather than territory or ideology. Since insurgents can readily blend in with their surrounding populations, regime forces cannot defeat an insurgency unless the local population identifies who and where the insurgents are. The challenge for the state is thus to convince local civilians to provide that information, while the challenge for insurgents is to persuade them not to. In Berman, Felter, and Shapiro’s telling, just about everything that happens in an insurgency—from building schools and hospitals on the one hand, to the indiscriminate slaughter of civilians on the other—can be read as an attempt to coax or intimidate civilians into divulging or withholding what they know.
Small Wars, Big Data is by no means the first to offer that argument. But it is unique in terms of the breadth and depth of the empirical evidence it marshals. From the pioneering research of Stathis Kalyvas in the early 2000s on, political scientists from Lisa Hultman to Laia Balcells have compiled an extraordinary body of empirical work on the “micro-foundations” of insurgent and civil war violence. As leading contributors to that literature themselves, the authors do an admirable job of surveying its findings.
Of particular note is chapter four, which offers a thorough overview of the debate regarding the effects of developments in information technology on insurgent violence. As the authors note, on the one hand, information technology may reduce insurgent attacks by making it easier for states to gather intelligence about the insurgents. On the other, it may instead increase attacks by enabling insurgents to better coordinate with one another. By comparing the rollout of cell phone networks with insurgent violence in Iraq, the authors show that cell phones—by offering low cost, anonymous ways of supplying information about insurgents at little risk to the informant—do appear to reduce and disrupt insurgent activity. That the Taliban in Afghanistan and Boko Haram in Nigeria have both repeatedly targeted cell phone towers, despite the improved communication they enable, suggests that even insurgents themselves fear that information technology has tilted the balance of power to the state.
For all its strengths, Small Wars, Big Data is not without its flaws. One is that it gives short shrift to the kind of “brute force” tactics that Jacqueline Hazelton discussed in a controversial article last summer. The book would have been stronger if it had discussed at greater length why exploiting information technology is more effective at defeating insurgencies than draconian policies like mass incarceration, mass resettlement, or even mass killing. But the other critique is that the book’s title hints at a topic it never addresses: It is primarily a book about the role of information in insurgency, rather than information technology. The title makes for great marketing, but it’s a bit of misnomer.
Army of None, by contrast, more than lives up to its billing. Scharre has spent nearly a decade framing the early debate over autonomous weapons in D.C. and the Pentagon, and the experience shows. The book plainly and masterfully lays out the major questions that AI and autonomous weapons raise for the future of armed conflict. Although written for a popular audience, even well-informed academics will find it worthwhile as an introduction to the technical, ethical, and strategic issues that AI-infused weapons systems will introduce.
Scharre’s argument about the coming ubiquity of big data and AI has profound implications for the future of insurgency. Very roughly, two futures are possible. In one, AI and autonomous weapons are both distributed and commoditized, such that insurgents can afford weapons systems that are nearly as capable as those of any given regime. Think of the commercial off-the-shelf drones that the Islamic State has deployed in Iraq and Syria, but with low-level intelligence and object-detection baked in via Tensorflow. Since many of the cutting-edge AI projects are open source and publicly available, it may well be possible to build makeshift lethal autonomous weapons systems that are nearly as good as state-of-the-art systems but at a fraction of the cost. In such a scenario, the balance may shift slightly to insurgents; one shudders to imagine the carnage a Mumbai-style attack could produce if the attackers had lethal drone swarms at their disposal.
However, if the future of artificial intelligence is one that favors scale and centralization, then AI may give the upper hand to regime forces. In this world, regimes with access to the products and infrastructure of a great power, like the United States or China, may rapidly unravel insurgent networks. We might think of this as the “wars of none” scenario, since the regime’s overwhelming informational advantage may come to limit the need for violence altogether.
Recall that the central point of Small Wars, Big Data is that insurgencies are primarily contests for information about the identity and location of insurgents, and that even relatively simple technology—such as a text-based tipline—appears to make it much easier for states to gain that information. What happens when tiplines are replaced with real-time surveillance systems equipped with facial recognition? As the cost of ubiquitous surveillance drops, we may see the emergence of a new counterinsurgency strategy: Whereas the “hearts-and-minds” approach tries to coax information out of civilians and the “brute force” approach attempts to coerce it, the “big brother” approach may bypass civilians altogether. There’s little need for informants if you have enough sensors, cameras, and processing power to recognize and track everyone, everywhere.
Although such a scenario may seem far-fetched, early versions are already feasible. In the United States, Anduril Industries, the latest security startup funded by Peter Thiel, is fast at work building an “electronic” wall that has already proven remarkably effective at detecting and monitoring unauthorized border crossings in Texas. In China, meanwhile, authorities have responded to deadly attacks in Xinjiang by widely deploying facial recognition and mass surveillance technology there. And the capabilities of Chinese authorities are only set to grow: Beijing recently announced a massive new investment in SenseTime, an AI startup whose next generation product aims to identify objects and individuals across 100,000 live camera feeds simultaneously.
Whether AI and information technology will empower states or insurgents remains unclear. Most likely it will do both, with AI and information technology supercharging insurgencies in failing or weak countries while constraining the space in which insurgency can occur everywhere else. What seems certain is that the information revolution is poised to revolutionize insurgency soon too.