Thinking of robotic systems coupled with artificial intelligence (AI) and robotics as an existential threat to humanity as (for example) Elon Musk, Stephen Hawking and other scientists have done recently is at best premature, and at worst that thought process threatens to cede U.S. military advantage to hostile and aggressive state competitors.
Of course, objections by scientists to the militarization of technological advances of all types has a rich—if somewhat misguided—history. But as military professionals, it is incumbent to discount breathless reports of our imminent extinction at the claws of our silicon superiors and understand the concrete reality of robotic systems and AI on the battlefield to solve specific military problems.
It is incumbent we think through the military possibilities of these technologies given that China is racing to field a dominant military largely based on artificial intelligence and Russia is well aware of the potential strategic implications of AI-based military advantage. Avoiding them or prematurely prohibiting military experimentation will not make them go away. As Elsa Kania notes in her recent study on China’s military AI investments, “the PLA may leverage AI in unique and perhaps unexpected ways, likely less constrained by the legal and ethical concerns prominent in U.S. thinking.”
Recommended: 5 Places World War III Could Start in 2018
Recommended: How North Korea Could Start a War
Recommended: This Is What Happens if America Nuked North Korea
The DOD is also looking at the military uses of AI, embarking on a multiyear plan to both restore the warfighting readiness of the Joint Force and to build a more lethal, decisive conventional force going forward. Secretary Mattis, reflecting on the urgency of military innovation, noted that “it’s an equal obligation for me to make certain that the Secretary of Defense after next has the same competitive edge that I enjoyed growing up.”
Central to a more competitive, lethal, Joint Force will be the technologies of the third offset strategy—most notably robotic systems, machine learning, artificial intelligence, and human-machine teaming—and the associated joint and service concepts that will bring these technologies to life in wartime.
Robotics and AI at War will Drive Future Force Design in New Directions
In an earlier paper on the future of infantry combat, I noted that “successful armies are able to bring together units that are each distinct in the mix of mobility, protection, firepower, and awareness they provide on the battlefield.” Although this dynamic expresses tradeoffs in unit designs large and small, another set of metrics should be considered as we apply robotic and autonomous systems to the Joint Force and think through the tradeoffs between the traditional joint forces and the parts that will be robotized and “cognitized.”
So what does an “offset” force taking that fully embraces AI and robotic systems actually look like? The answer is not simply more and better technologically advanced systems and capabilities. Military change is always about understanding tradeoffs between many desirable but often incompatible design options. We will need to think through what attributes of robotics and AI are most important for future force design under different and varied conditions. It should set up a discussion about how we might employ a Joint Force that is smaller, connected and more autonomous than it is today.
The implications of the adoption of robotic and AI systems suggest that the component weapons, platforms, and units of the Joint Force must shrink, link, and think in some combination, and that future Joint Force commanders will be responsible for varying these components in response to adversary tactics by orchestrating the behavior of these smaller, connected, and sometimes artificially intelligent forces on the battlefield during operations.
The first attribute in force design that AI and robotics will allow is to reduce the physical size, observable cross section, and overall cost of individual platforms. The incentive to shrink each component of the Joint Force comes from rapidly advancing adversary reconnaissance strike, the advantages available in the miniaturization of integrated circuits but also emerging and potentially dramatic advances in the mass customization of parts and sub-microscopic arrangement of materials leading to better sensors, antennas, communications, and coatings.
The trend toward increased capability per unit of mass (and per dollar) goes up it should also mean more—not fewer systems on the battlefield. However, the Joint Force is not moving quickly enough in this direction. The dramatic difference in combat potential between today’s air wing, armored brigade, or carrier task force compared to those in the pre-digital age dramatically illustrate this trend. For the United States and the West, however, this has been realized through more expensive systems that have dramatically reduced number and variety of systems available to them.
The lethality and combat power of large and expensive platform continues to improve—but commensurate reductions in cost are not being realized. Taken to its logical extreme, the trend toward expensive, multi-mission platforms could (in principle) leave us with future “joint” force consisting of a few very expensive—but very powerful—platforms capable of doing everything, but also risking everything due to accident or a single lucky shot.
To do this, combat platforms will shrink. As robotic systems become cheaper and more capable, they will augment, support and defend expensive ships, aircraft, and land vehicles and augment individual platforms in both offensive and defensive modes. In the near-term future, large capital platforms—such as warships, tanks, submarines and aircraft—may be followed and surrounded by integrated (and often expendable) surface, subsurface, and aerial robots to escort and shield them from attack.
In the longer term, as industrial-scale additive manufacturing takes off, future militaries will further rebalance away from forces built around the few, highly capable, and expensive, and double down on capabilities that are “just capable enough to do the job.” In time, the capital platforms themselves may fade away, with the bulk of the capacity resident in the swarm. Together, the vast number of very small autonomous systems combined will severely complicate the sensor picture and the ability of commanders to make sense of the battlespace as well as overwhelm projected defenses by their sheer numbers.
However, a mass of systems that cannot work together is not a unit, but a mob. Moreover, the tradeoff in the small and the cheap is that individual units are perhaps more vulnerable to high-, low- and no-tech counters (not to mention sheer accident and chance). For example, high-powered microwave systems or other directed energy systems can damage small components and the miniaturized (and likely unshielded) integrated circuits and controllers that run them. Thus, the small must also be varied, and distributed over many wide areas of the battlefield to reduce risk to operations as a whole.
The implication of being able to mass produce small platforms means that many systems will be able to work together in concert to achieve military objectives. Perhaps the most important military implication of the information revolution has been that capabilities working together can be physically separate over ever increasing distances. This brings us to the second key attribute of future AI and robotic force design—the imperative to link individual nodes in a future combat network widely. Although the individual parts may be less capable than today’s high-end units or platforms linking them together results in robotic formations that are far more than the sum of their individual parts.
The increasing density of links connecting robotic units will further encourage the disaggregation of physical systems (sensors from shooters, jammers from aircraft, protection systems from vehicles, etc.) but to be connected in space and time when required to achieve a particular battlefield condition or effect. This will require them to communicate together. As platforms, units, and systems shrink, they will still have a great deal of ground to cover, and they must continue to work towards common objectives and adapting the formation to conflict conditions as they evolve.
Characterized as a “large, armed, nervous system,” these future robotic formations must bring together a menagerie of dissimilar land, sea, air, space, and cyber capabilities against objectives both in parallel and in complex sequences against adversaries. From a command-and-control perspective the need to distribute operations and the ability to persistently communicate among units suggests that today’s predominant “one crew/one platform,” model into something more akin to “one crew/one swarm,” and beyond with one crew may potentially be capable of directing and commanding a single cross-domain swarm.
The tradeoffs in networking robotic forces, however, is that they will rely on radio connections, microwave relays, radar apertures, and any other receivers that allow the units to coordinate and connect with the world around them. Any sensor or aperture may introduce vulnerable points through which the systems, circuitry, and software upon which they rely—can be engaged. This leaves them vulnerable to cyber and modern electromagnetic warfare techniques that allow them to be penetrated and the swarm disaggregated. Future military operations will work to disrupt the coherence of networked forces in peacetime, competition, and conflict by threatening, contesting, interdicting links in a variety of ways.
The implication of machine-speed electronic and cyberattacks on a robotic swarm’s data links suggests the need for individual platforms to understand context and carry on when disconnected. Too many Hollywood films have portrayed the consequences of too much centralization, networking, or EMP vulnerability on a robotic or networked force. The requirement for platforms, systems, organizations and individuals to understand context, make decisions, and act on these decisions independent of direct control from superior command elements.
The constituent robotic elements of the future Joint Force may be cut off electronically or through cyberattack, and not always have the capacity to reliably communicate with the rest of the force. Although individual units may be disrupted in this way, the robotic systems must not cease to operate or even simply return home when command guidance is lost. They must be able to stay in the fight. When future links cannot be established—forces must have a measure of autonomy and the capacity to operate independently until reconnected to the logistics and information structures of the remainder of the force. The future robotic force must think on its own and continue to pursue military objectives on a fast-moving battlefield.
Here, AI enabled systems will require the ability to independently sense and react to the environment. The future of command and control will be concerned with how to provide boundary conditions and guidance on the behavior of increasingly intelligent automated systems and platforms forces on the battlefield—including cyberspace.
The need for systems to think independently will also be driven by battlefield engagement and adaptation speeds that will far surpass the reaction times of human beings. The appearance of hypersonic missile systems, computer-driven electronic warfare capabilities, advancing cyber operations, and laser- and high-power microwave systems will all serve to accelerate the pace of military operations. The need to integrate sensor warfare, manage radio-frequency systems, and coordinate the array of effectors in drive the need for more automation, especially artificial intelligence for battle management purposes.
Don’t Fight the Robots
Let’s not be so quick to dismiss AI and robotic systems because of ethical and moral concerns that are, at best, premature. As capable as robotic and AI tools are, they remain just that—tools that are dependent on humans to employ and maintain them within rather narrow fields and in discrete operational conditions. Both are already far along in addressing difficult problems for the military, most of which are rather remote from terminator-like visions of human extinction. They require across the board human direction and support to work in the real world. Ignoring the promise of robotic and AI systems will not avoid a future of warfare in which the exquisite, scarce and dependent on continuous human direction are readily defeated. Victory in the future requires a force consisting of the many, small and smart. The United States and its Joint Force needs to get there first, and when it does, it needs to be aware of any advantages—and limitations—these new capabilities will provide.
Jeff Becker is the Chief Futurist supporting the Joint Concepts Division at the Joint Staff J-7 and their deep futures study, Joint Operating Environment 2035. The views here are his alone and do not necessarily reflect those of the Joint Staff or the Department of Defense.