What advances in artificial intelligence (AI) will mean for the future battlefield remains opaque. Mainline discourse is overly influenced by exaggerated forecasts such as that reflected in Russian President Vladimir Putin's often repeated 2017 declaration that “whoever becomes the leader” in the sphere of artificial intelligence “will become the ruler of the world” and a broadly similar claim he advanced as recently as May 18 when he insisted that without artificial intelligence, hypersonic weapons and other new technologies, “it would be impossible to secure the future of our [Russia's] civilization.” Whatever security benefits and military advantages they may bring, as I argue in this piece, the development and posturing of AI-enabled autonomous weapons systems (AWS) is likely to raise significant risks for crisis instability and conflict escalation in future confrontations between the United States and Russia. Policymakers would do well to consider—now—whether the potential risks are worth the hoped-for operational advantages.
With the exception of the advent of nuclear weapons, history counsels skepticism regarding the many recurrent predictions of the revolutionary impact of technological innovation on the character of warfare. Maintaining objectivity and historically-informed skepticism is no less warranted when considering the potential battlefield effects of AWS, or systems that, once activated, are intended to select and engage targets (PDF) without further intervention or guidance from a human. Still, as we enter the third decade of the 21st century, it has become clear that a growing number of countries are engaged in efforts to develop and field AWS and other AI-enabled military systems. These efforts are by far best funded in the United States and China, with Russia lagging behind in spite of all Putin's motivational rhetoric. Unsurprisingly, these efforts are motivated by a desire to secure military operational advantages on the future battlefield.
Motivations and General Aims Driving AWS Development Efforts
The Pentagon views AI and robotics, along with several other emerging technologies including hypersonics and directed-energy, as key to offsetting the anti-access and area-denial capabilities and concepts developed by China and Russia over the past two decades, thereby regaining and sustaining U.S. conventional deterrence overmatch. Meanwhile, China's People's Liberation Army anticipates that AI could fundamentally change the character of warfare even as it fears the emergence of a generational gap between its capabilities and that of the U.S. military. It thus seeks to develop AI and other “strategic front-line” technologies in future military competition with the United States.
Nominally at least, Russia's vision regarding the aims of exploiting AI for military purposes are not dissimilar from those of the United States and China. It supports research (PDF) in a number of AI application areas; in just the past three years, the Kremlin has declared its intent to establish six new national initiatives dedicated to AI research and development, including the Advanced Research Foundation (ARF), Russia's analogue to the U.S. Defense Department's Defense Advanced Research Projects Agency (DARPA). As recently as April 21, ARF's deputy director boasted to RIA Novosti that as a result of the foundation's research, “Living fighters will gradually begin to be replaced by their robotic 'brothers' who can act faster, more accurately and more selectively than people.” Despite what Putin's bold assertion might otherwise seem to suggest, Western experts generally agree that Russian AI development significantly lags behind that of the United States and China. However, in stark contrast to U.S. AWS development efforts, if only marginally less so than China's, Russia places great emphasis on the development of AI for information warfare aimed, as a recent comprehensive report by my RAND colleagues documents, at causing political and societal damage to the target state. As argued below, instead of seeking to gain military operational advantages by competing to match U.S. AWS developments, Russia is much more likely to emphasize and invest in two other military capability areas.
Operational Advantages of AWS
U.S. military planners believe AWS promise to provide a number of significant operational advantages on the future battlefield. Less dependent on human control and decisionmaking, AWS are anticipated to be far faster and more agile than today's manned weapons systems undertaking thousands of complex and highly coordinated decisions at machine speeds. Today's remotely piloted drones rely on communications links to distant command and control and information analysis centers, which can result in inevitable communications delays and leave them vulnerable to jamming or denial through electronic warfare techniques. AWS, on the other hand, will be designed to operate on their own, if necessary, in such information contested environments (PDF), conferring degrees of stealth, decisional agility, persistence and survivability not enjoyed since the earliest days of stealth aircraft, if then.
AWS are expected to operate in large swarms comprising hundreds, or even thousands of small, relatively low-cost and expendable vehicles with networked sensors and on-board computing. Fighting in swarms, AWS hold the promise, as Paul Scharre notes, of effectively returning mass to the battlefield as they will be capable of rapidly and autonomously coordinating their own tactical movement in response to enemy actions. As a consequence, AWS are expected to provide the capability of synchronized attack or defense that enables one to fight inside the enemy's decision cycle or OODA loop, overwhelming and defeating (PDF) the enemy's abilities through mass, intelligence, coordination and speed. In conflicts against opponents lacking similar capabilities, U.S. military planners maintain that the combination of such characteristics will provide decisive operational advantage.
Finally, like today's drones, AWS will reduce the numbers of casualties and body bags in conflict. But, in contrast to current drones, AWS are expected to lead to cost savings as the frequently overlooked need for large rotating teams (PDF) of remotely located and highly trained pilots and operators would be eliminated. Such cost-savings could be substantial if realized not only from the advent of AWS, but from the diffusion of AI and autonomy to other military functions such as logistics and transportation, battlefield healthcare, cybersecurity and combat simulation and training—a benefit that would reduce the demand for personnel and could lead to smaller militaries.
Russia Unlikely to Seek to Match U.S. AWS Developments
The economic outlook for Russia, which remains dependent on exports of oil and gas for much of its revenues, is grim at least in the short-term. COVID-19 forced Russia to agree to a deal with OPEC to significantly cut production and exports, which it initially rejected in March, and sent oil prices down in what is bound to reduce budget revenues and cause economic contraction. Russia is now forecast to experience a GDP contraction of 5.5. percent and an increase in unemployment from 2.5 million to 8 million workers this year. Moreover, continued dependence on exports of commodities whose prices Russia cannot control, depopulation (PDF), and a host of other structural factors indicates that the economic outlook for Russia will remain bleak in the longer term, too, in absence of deep reforms. As a result, for the foreseeable future, Russia's federal budget is likely to be constrained. If for no other reason than that, the Kremlin will be unlikely to respond to U.S. AWS advancements by substantially increasing investments in the area, at least in comparison to two other capabilities it is more likely to prioritize well ahead of AWS as discussed immediately below. But there is yet another reason. In brief, such symmetrical responses to capability differences have seldom been part of Russia's playbook in the post–Cold War era. In this respect, it is too early to identify just where Russia may be heading in terms of the character of AWS ground, air, or naval capabilities it is likely to field, let alone to discern any associated operational concepts. The time doing so is likely to prove time misspent. Instead, Moscow is much more likely to look to two other capability areas in search of comparative military operational advantage with respect to the United States and NATO.
First, the Kremlin can be counted upon to continue with its customary strategy of underscoring and even increasing its reliance upon nuclear weapons both for deterrence and possibly even warfighting, a cost-effective strategy, comparatively speaking, that Vasily Kashin and Michael Raska economically refer to as “countering the Third Offset Strategy with the First Offset Strategy (PDF).” In this regard, Putin's unveiling of the five nuclear “superweapons” in his March 2018 nationally televised speech to the Russian Federal Assembly can be seen as exhibit one of this strategy. Still, Moscow's continued reliance on nuclear weapons by no means suggests that it will be willing to cede leadership in emerging disruptive technologies entirely to the United States.
In fact, the second military capability area Russia is likely to emphasize is that of hypersonic missiles, one of the few weapons realms, outside of nuclear weapons, in which it is building a track-record for both developing and fielding apparently functioning, sophisticated 21st century weapons systems. In just a few short years, it has developed and deployed two hypersonic missile systems, and is currently testing a third. Deployment of the Avangard hypersonic glide vehicle has already begun, as Russian officials announced in late December 2019 that one missile regiment on the border with Kazakhstan was armed with the hypersonic glide vehicle. The Kh-47M2 Kinzhal air-launched ballistic missile appears slated to be deployed by 2024 by a MiG-31K regiment operating from the VKS base in the Siberian city of Kansk. While a deployment date for the 3M-22 Tsirkon anti-ship hypersonic cruise missile is still uncertain, a Russian navy frigate fired the weapon from the Barents Sea against a ground target in a purportedly successful test in January of this year.
Implications for Crisis Stability and Conflict Escalation in U.S.-Russia Confrontations
While holding out the promise of significant operational advantages, AWS simultaneously could increase the potential for undermining crisis stability and fueling conflict escalation in contests between the United States and Russia. Defined as “the degree to which mutual deterrence between dangerous adversaries can hold in a confrontation,” as my RAND colleague Forrest Morgan explains, crisis stability and the ways to achieve it are not about warfighting, but about “building and posturing forces in ways that allow a state, if confronted, to avoid war without backing down” on important political or military interests. Thus, the military capabilities developed by nuclear-armed states like the United States and Russia and how they posture them are key determinants of whether crises between them will remain stable or devolve into conventional armed conflict, as well as the extent to which such conflict might escalate in intensity and scope, including to the level of nuclear use. AWS could foster crisis instability and conflict escalation in contests between the United States and Russia in a number of ways; in this short essay I will highlight only four.
While holding out the promise of significant operational advantages, AWS simultaneously could increase the potential for undermining crisis stability and fueling conflict escalation.
Share on TwitterFirst, a state facing an adversary with AWS capable of making decisions at machine speeds is likely to fear the threat of sudden and potent attack, a threat that would compress the amount of time for strategic decisionmaking. The posturing of AWS during a crisis would likely create fears that one's forces could suffer significant, if not decisive, strikes. These fears in turn could translate into pressures to strike first—to preempt—for fear of having to strike second from a greatly weakened position. Similarly, within conflict, the fear of losing at machine speeds would be likely to cause a state to escalate the intensity of the conflict possibly even to the level of nuclear use.
Second, as the speed of military action in a conflict involving the use of AWS as well as hypersonic weapons and other advanced military capabilities begins to surpass the speed of political decisionmaking, leaders could lose the ability to manage the crisis and with it the ability to control escalation. With tactical and operational action taking place at speeds driven by machines, the time for exchanging signals and communications and for assessing diplomatic options and offramps will be significantly foreclosed. However, the advantages of operating inside the OODA loop of a state adversary like Iraq or Serbia is one thing, while operating inside the OODA loop of a nuclear-armed adversary is another. As the renowned scholar Alexander George emphasized (PDF), especially in contests between nuclear armed competitors, there is a fundamental tension between the operational effectiveness sought by military commanders and the requirements for political leaders to retain control of events before major escalation takes place.
Third, and perhaps of greatest concern to policymakers should be the likelihood that, from the vantage point of Russia's leaders, in U.S. hands the operational advantages of AWS are likely to be understood as an increased U.S. capability for what Georgetown professor Caitlin Talmadge refers to as “conventional counterforce” operations. In brief, in crises and conflicts, Moscow is likely to see the United States as confronting it with an array of advanced conventional capabilities backstopped by an interconnected shield of theater and homeland missile defenses. Russia will perceive such capabilities as posing both a conventional war-winning threat and a conventional counterforce threat (PDF) poised to degrade the use of its strategic nuclear forces. The likelihood that Russia will see them this way is reinforced by the fact that it currently sees U.S. conventional precision capabilities precisely in this manner. As a qualitatively new capability that promises new operational advantages, the addition of AWS to U.S. conventional capabilities could further cement Moscow's view and in doing so increase the potential for crisis instability and escalation in confrontations with U.S. forces.
In other words, the fielding of U.S. AWS could augment what Moscow already sees as a formidable U.S. ability to threaten a range of important targets including its command and control networks, air defenses, and early warning radars, all of which are unquestionably critical components of Russian conventional forces. In many cases, however, they also serve as critical components of Russia's nuclear force operations. As Talmadge argues, attacks on such targets, even if intended solely to weaken Russian conventional capabilities, will likely raise Russian fears that the U.S. conventional campaign is in fact a counterforce campaign aimed at neutering Russia's nuclear capabilities. Take for example, a hypothetical scenario set in the Baltics in the 2030 timeframe which finds NATO forces employing swarming AWS to suppress Russian air defense networks and key command and control nodes in Kaliningrad as part of a larger strategy of expelling a Russian invasion force. What to NATO is a logical part of a conventional campaign could well appear to Moscow as initial moves of a larger plan designed to degrade the integrated air defense and command and control networks upon which Russia's strategic nuclear arsenal relies. In turn, such fears could feed pressures for Moscow to escalate to nuclear use while it still has the ability to do so.
Finally, even if the employment of AWS does not drive an increase in the speed and momentum of action that forecloses the time for exchanging signals, a future conflict in which AWS are ubiquitous will likely prove to be a poor venue both for signaling and interpreting signals. In such a conflict, instead of interpreting a downward modulation in an adversary's operations as a possible signal of restraint or perhaps as signaling a willingness to pause in an effort to open up space for diplomatic negotiations, AWS programmed to exploit every tactical opportunity might read the modulation as an opportunity to escalate offensive operations and thus gain tactical advantage. Such AWS could also misunderstand adversary attempts to signal resolve solely as adversary preparations for imminent attack. Of course, correctly interpreting signals sent in crisis and conflict is vexing enough when humans are making all the decisions, but in future confrontations in which decisionmaking has willingly or unwillingly been ceded to machines, the problem is likely only to be magnified.
Concluding Thoughts
Much attention has been paid to the operational advantages to be gained from the development of AWS. By contrast, much less attention has been paid to the risks AWS potentially raise. There are times in which the fundamental tensions between the search for military effectiveness and the requirements of ensuring that crises between major nuclear weapons states remain stable and escalation does not ensue are pronounced and too consequential to ignore. The development of AWS may well be increasing the likelihood that one day the United States and Russia could find themselves in just such a time. Now, while AWS are still in their early development stages, it is worth the time of policymakers to carefully consider whether the putative operational advantages from AWS are worth the potential risks of instability and escalation they may raise.
Burgess Laird is a senior international defense researcher at the nonprofit, nonpartisan RAND Corporation. He is a contributing author of Deterrence in the Age of Thinking Machines.