It is 2025. Protests in the ethnic Russian enclave in Riga, Latvia have NATO on edge. Russian units in the Western Military District are on alert conducting snap exercises involving autonomous ground and air attack systems.
The Russian president makes a speech promising to protect ethnic Russians wherever they are with military forces if necessary. In response, a U.S. Army brigade combat team bolstered by intelligence, air defense, and aviation support elements from U.S. Army Europe deploys. Their mission is to reassure Latvian forces, deter Russian aggression, and if necessary conduct a mobile defense.
The task force processes petabytes of unclassified social media posts. Machine learning software agents isolate images of potential Russian covert elements agitating protests, cross referencing cell phone pictures posted on social media with police traffic cameras, and more sensitive collection platforms. U.S. forces provide these images to the Latvians along with a projection of likely Russian activities over the next 48 hours.
The Latvians distribute the images on a cellular alert network that lets concerned citizens turn their cell phones and other personal devices into a civil defense sensor network. This civil defense network acts as a cloud, helping cyber defense apps secure critical infrastructure and conducting predictive models of where possible Russian cross-border insertions might occur based on historical data, weather, terrain, and news reports.
The technology in this future battlefield is already driving a wide range of commercial applications. From Amazon figuring out what book you want to buy next to Google optimizing the ads you see while searching, we live in a world defined by “big data” and artificial intelligence applications that identify patterns in our consumer habits and daily life. These applications have the potential to change the character of warfare. The first nation that adapts accordingly and integrates artificial intelligence across the force will have a generational advantage on the battlefield.
The U.S. Army needs to develop a strategy for integrating narrow artificial intelligence applications into the force. Existing Department of Defense investments in artificial intelligence tend to emphasize future autonomous systems such as tanks, robot soldiers, and planes that can operate with minimal human input. An alternative approach is to experiment with predictive models and big data to increase the combat power of the current force.
What is Artificial Intelligence?
Artificial intelligence is commonly defined as the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. It can range from weak forms, such as narrow artificial intelligence, that processes big data to answer basic questions and generate predictions (e.g., think of Waze helping you drive home or ad placements online) to strong forms such as “Artificial General Intelligence” and “Artificial Super Intelligence” that exceed human intelligence, creativity, and adaptability. Military applications range from finding optimal human-machine symbiosis, the centaur Deputy Secretary of Defense Robert Work speaks about, to using increased computer processing power to replace people on the battlefield with autonomous attack swarms.
Short of the promise of driverless cars and robot servants, narrow forms of artificial intelligence like machine learning are starting to change sectors ranging from healthcare to logistics. With respect to public health, Flowminder, a Swedish NGO, uses narrow artificial intelligence to predict the spread of diseases. In logistics, machine learning is helping companies make supply chain adjustments, optimize delivery routes, and design warehouse systems. Vehicle developers, such as Volvo, apply artificial intelligence to improve predictive maintenance. Volvo collects data with smart sensors on their vehicles and applies machine learning techniques to conduct diagnostics that reduce down time for services and better inform the resupply.
Artificial Intelligence for the Military
While many commercial applications of artificial intelligence are based on identifying patterns and trends using big data, most military applications focus on autonomous systems. Existing artificial intelligence programs in the Department of Defense include Navy unmanned undersea and aerial vehicle programs such as the Low-Cost Unmanned Aerial Vehicle Swarming Technology (LOCUST), and Air Force/DARPA ventures such as the Gremlin anti-surface-to-air missile drone program. The Gremlin program involves launching a large number of unmanned systems from a transport plane that swarm to attack hard targets such as mobile SAM platforms and C4ISR nodes. Within the Army, labs are using artificial intelligence to experiment with autonomous vehicles. Concepts range from larger logistics convoys composed of one manned vehicle and a large number of autonomous vehicles to combat formations mixing manned and unmanned platforms.
While the private sector focuses on narrow artificial intelligence applications that aid decision making and optimize business models, the military’s focus on autonomy arises from a technological bottleneck. A more advanced version of artificial intelligence than exists today is likely a necessary development before the military can field autonomous systems able to adapt to complex, changing environments. This evolution is especially important for land forces. Ground platforms have more variables to address than air or naval systems. Jets and submarines do not have to dodge potholes or jaywalkers. An autonomous system operating in a future megacity would have to do both under fire. The result is that autonomous combat vehicles will likely emerge in the air and maritime domain faster than the ground domain. Yet, that should not stop continued investment in and experimentation with artificial intelligence by the Army.
The Army needs apps like Waze, the popular driving program, for war. Instead of primarily focusing on artificial intelligence for autonomy, the Army should position itself to leverage commercial applications that optimize staff processes. Beyond self-driving convoys, which will take longer than predicted due to legal and safety concerns, the Army could use machine learning applications across its existing warfighting functions. Image recognition software could help intelligence analysts examine video feeds to identify IEDs and establish patterns of life. Analysis of data sets describing past enemy operations could provide probabilistic forecasts of enemy behavior. Software agents could anticipate supply bottlenecks before they occur and recommend mitigation options. Data already resident in the operational environment from both commercial repositories and data collected by platforms within Army formations (e.g. weather data from drones) offers opportunities to examine factors that may affect operations. Natural language processing programs could filter social media and local news outlets to identify common themes and messages. Many administrative functions could be automated to improve record keeping and medical readiness. In the extreme, artificial intelligence could produce a robot general staff capable of recommending an operational approach, options for allocating resources against different lines of effort, and ways to deceive the enemy.
Note that none of these functions would completely remove humans from the loop. Rather, the goal would be to super-empower soldiers. The increase in capability would mean that the Army could alter its staff organizations and force structure, producing an increase in combat power despite a decrease in the number of personnel. Smaller, streamlined staffs with new billets for data scientists could replace large organizations.
Using Artificial Intelligence to Gain and Maintain the Initiative
Seizing the initiative requires the U.S. military to observe, orient, decide, and act faster than our adversaries can. The operative question is how do you achieve surprise, maneuver, or mass effects in depth on a future battlefield that is likely to be urban, populated by a variety of threat groups and wired with cheap, connected sensors? Every civilian will have a cell phone able to collect and disseminate intelligence in real-time that the enemy will combine with increasingly cheap, precision fires. The enemy will then deny U.S. forces the ability to retaliate with its own precision fires by blending with the population.
In this battlespace, initiative will require increasing situational understanding. The paradigm of intelligence shifts from singularly identifying enemy capabilities and estimating their motivations, to assessing a changing environment and its likely impact on your operations. Commanders can better understand the sentiment and concerns of the local population through natural language processing programs that analyze news, patrol reports, key leader engagements, and social media. They will be able to confirm or deny their theory of victory and planning assumptions in real-time. With respect to tactical actions, this form of initiative through understanding creates new types of engagements. Systems like IBM’s Watson will analyze the correlates of past troops in contact instances to show how an adversary fights and potential vulnerabilities.
Using Artificial Intelligence to Enable Battle Command
The 1994 Force XXI Operations concept defined battle command as the “art of decision-making, leading, and motivating informed soldiers and their organizations into action to accomplish missions at the least cost to soldiers.” The military concept held that battlefield dynamics evolved as the rate of information available to soldiers increased the speed of decision- making. The force best able to capitalize on the increased volume of information would be able to set the conditions and define the tempo of operations.
Narrow artificial intelligence could reduce the number of soldiers required to conduct basic tasks and optimize resource allocation. Take a core task like medical evacuation: Narrow artificial intelligence agents loaded onto a command network will automatically identify blood types and the medical history of a unit that reports troops in contact. In milliseconds, the agent can analyze the routes, weather, local landing sites, and the anticipated rate of additional medical emergencies in the next 72 hours to determine the best form of casualty evacuation (e.g., air or ground, level of armed escort required, etc.). This basic predictive modeling would increase the efficiency of individual medical evacuations and ensure the staff does not culminate in its ability to provide care across the force in a high-end engagement likely to see significant causalities. In this example, narrow artificial intelligence applications can categorize events, identify patterns, and optimize staff decision-making.
The same logic applies to logistics. By increasing efficiencies associated with projected rates of ammunition, fuel and water usage, narrow artificial intelligence applications could reduce the number of soft target convoys moving on the battlefield. Maintenance platoons could follow the lead of organizations like Amazon and increase the efficiency of storing and distributing parts while decreasing the number of people required. Better resource allocation at the campaign level enables commanders to take an economy of force posture across the majority of the battlespace and weight the main effort with significant capabilities to set the conditions and sustain a tempo the enemy cannot match. Getting staff processes right can generate new options for commanders and reduce the burden of increasingly large headquarters. Narrow artificial intelligence has the potential to help the larger Army reshape its tooth-to-tail ratio and free up additional soldiers for critical areas like combat arms and intelligence, surveillance, and reconnaissance.
A Blueprint for Experimentation: The Louisiana Maneuvers
Commercially proven narrow artificial intelligence applications can be rapidly integrated into the Army through large-scale, sustained experiments. There are historical precedents to this approach in the Louisiana Maneuvers. In 1940 to 1941, the U.S. Army conducted a field exercise and war games with over 400,000 troops in Louisiana. During the exercises, they tested new concepts from fighting against armored units to sustaining the force with C-rations.
In 1991, Gen. Gordon Sullivan resurrected the idea. Faced with two significant changes in the operational environment — the end of the Cold War and the rise of the Information Age – Sullivan used simulations and field tests to examine new concepts and equipment. These initiatives resulted in the Force XXI Campaign to digitize the force using the 4th Infantry division as a testbed.
The modern Army can build on these previous large-scale experiments, and current initiatives such as Force 2025 and Beyond Maneuvers, to develop new concepts and capabilities based on narrow artificial intelligence applications found in the business world. Centers of excellence could propose AI concepts based off existing commercial applications. New concepts could also be crowd sourced through hackathons that bring together rising military leaders and technology experts. The best concepts to emerge would be tested through war games. Simulated battles where the staff was smaller, but artificial intelligence-enabled would be compared with traditional staff performance. New warfighting concepts would be tested in force-on-force experiments at places like the National Training Center, where rival artificial intelligence-enabled units could clash and see how pattern recognition and predictive modeling alter the deep-close-security fight. Deployed units in places like Afghanistan and Iraq could test new applications to gain situational understanding. For example, applications that map influence networks and identify significant themes and messages could be used in Afghanistan to win the battle for the narrative. Theater exercises such as the U.S. Army’s Pacific Pathways offer opportunities to employ artificial intelligence concepts in a variety of deployed environments with the real challenges of expeditionary conditions. Such a series of experiments would help the Army more rapidly develop concepts and integrate them into the force using proven technologies.
The Army can field narrow forms of artificial intelligence found in the business world to increase its combat power. The definition of this emerging capability and the spectrum of available and developing technologies is more than just the promise of talking robots and self-driving cars. This article does not attempt to understate the complexity involved with integrating these technologies and concepts into the force, and one should not conclude that these will cure all ills associated with the future operational environment. Fog and friction will still reign in war. However, through a sustained series of experiments, similar to the Louisiana Maneuvers of the past, the Army can identify ways to adapt commercially proven narrow artificial intelligence applications. If the Army fails to integrate existing commercial technologies, it may find itself waiting for futuristic autonomous platforms while its adversaries harness existing, off-the-shelve capabilities to increase their lethality.
Benjamin Jensen is a Major in the US Army Reserve currently serving as a Fellow in the Office of the Chief of Staff of the Army, Strategic Studies Group. Outside of the military, he is a Donald L. Bren Chair and Director of the Brute Krulak Center at the Marine Corps University. He is the author of Forging the Sword: Doctrinal Change in the U.S. Army, 1975-2010.
Ryan Kendall is a Lieutenant Colonel in the US Army currently serving in the Chief of Staff of the Army, Strategic Studies Group. He is an Aviation branch officer with previous assignments in the 101st Airborne Division (AASLT) and the 3d Armored Cavalry Regiment.
Image: U.S. Army