The Centaur and the Cavalry: A Future Battle on NATO’s Eastern Flank

September 29, 2016

To see the future of ground combat, look no further than the tactics of Russian-backed forces in eastern Ukraine. Full-spectrum electronic surveillance quickly identifies Ukrainian forces by their emitted signals. Once identified, a remotely piloted vehicle or helicopter verifies the Ukrainians’ position. Shortly thereafter, massed artillery and rocket fires destroy the Ukrainian force.

The war in Ukraine is not an aberration. The U.S. Army’s adversaries have carefully prepared to defeat America’s preferred tactics and systems. In the next 20 years, the Army should expect its adversaries to become adept at achieving effects based on fewer intercepted electronic signals, with greater speed and lethality, over greater distances. The old ways of warfare will largely be doomed.

Lately there have been many suggestions for tactical or strategic solutions to challenges like these that the Army will face on the future battlefield. While these are useful, they address only two sides of a three-sided problem — the U.S. military must also improve how it fights at the operational level of war, the critically important time and space in which commanders arrange tactical actions and purposefully align them to strategic objectives. The future is approaching fast, and the Army cannot ignore the challenges and opportunities that divisions and corps are likely to face.

Now imagine the U.S. 1st Cavalry Division two decades from now. Thanks to defense modernization and professional soldiers, the 1st Cavalry Division will be devastatingly powerful, able to deliver massive firepower whenever and wherever its commanding general chooses. The 1st Cavalry Division will find itself standing shoulder-to-shoulder with NATO allies, ready to move into the attack. Their shared purpose: to expel aggressor forces from NATO territory, restore the international border, and save thousands of civilians caught behind enemy lines. On this imagined battlefield, the 1st Cavalry Division must take back a vital transport hub, much like the one in Małaszewicze, Poland [map]. In the process, the 1st Cavalry Division must liberate 4,000 locals in the neighboring village and secure a forward transfer and distribution point to sustain NATO’s push over the remaining five miles to the international border.

Surveying a digital picture of the battlefield updated in real time at the division’s forward command post, the 1st Cavalry Division’s commanding general will put the division in motion. But where, when, and how should the attack proceed? To answer that, we must return to 2014 and 2015, and revisit two momentous and largely overlooked announcements.

First, in 2014, IBM Research revealed a microchip unlike any other in history. We may one day look back through a quarter-century of history and point to the debut of the SyNAPSE chip as the day computing changed forever. Neurosynaptic processors such as the SyNAPSE chip not only pack more transistors into a square centimeter than ever before, they do it without any conventional programming. Instead, neurosynaptic chips learn as mammalian brains do. Someday an enterprising scientist will string together enough of them to create a machine as smart as a human brain. That could unlock deep-learning capabilities beyond anything possible using conventional microprocessors.

Second, in 2015, Deputy Secretary of Defense Bob Work revealed defense investment in “human-machine collaboration and combat teaming.” This combined concept is the biggest of Work’s bets for the third offset strategy, an ambitious effort to maintain the U.S. military’s technological superiority well into the future. Human-machine collaboration and combat teaming, in Work’s view, will solve vexing tactical and operational problems. Work has not been shy about asking the services to explore the concept.

The word “and” in the phrase “human-machine collaboration and combat-teaming” properly implies that this idea is in fact two separate concepts working together. Human-machine combat teaming is the cooperative, tactical partnership of humans with unmanned systems. Interestingly, Arizona State University recently found that so far autonomy is driving weapons to replace humans instead of team with them. Human-machine collaboration, by contrast, “is using machines to help decision-makers make better decisions,” per Work. Because every level of war requires decision-making, human-machine collaboration can happen at any level of warfare.

If the Army seizes the opportunities presented by neurosynaptic processors and the concept of human-machine collaboration, the 1st Cavalry Division’s forward command post of the future will be alive with activity like nothing we see today.  The dozens of men and women manning scores of radios and computers will be gone, replaced with planning processes that develop realistic options based on fast-paced machine computations and nuanced human understanding. As the 1st Cavalry Division’s tanks move toward their assembly areas, the division staff will actively collaborate with the most advanced computers on earth to do with a few dozen people what it takes hundreds of humans to do today.

In present Army doctrine, planning is both an art and a science that results in situational understanding, a vision of a desired future, and description and analysis of ways to turn a sub-optimal present into the desired future. The 1st Cavalry Division of the future will off-load the scientific aspects of planning and analysis to its neurosynaptic computers, thus freeing the division’s humans to focus on the art of using “their skill, knowledge, experience, creativity, and judgment” to assess ends, ways, means, and risk (Army Doctrine Publication 5-0). The lives of 4,000 civilians and nearly 20,000 American soldiers will rest upon how well humans and machines translate their commanding general’s intent into an actionable plan.

Conceptually speaking, operational-level human-machine collaboration hinges on the role of the human. Today’s humans use machines to analyze data, but rely on their own thoughts to build courses of action. In the future, neurosynaptic computers will both analyze the data and build courses of action. It will then be up to humans to enable effective machine analysis by asking the right questions.

Much like today, the 1st Cavalry Division of the future will spread across a battlefield of hundreds of square miles. Manned vehicles will cover the majority of the division’s footprint, along with unmanned systems both on the ground and in the air. Unlike today, however, each of these platforms and many of their operators will carry sensors that connect back to the division command post. This dense forest of sensors will transmit vast quantities of data about friendly forces, the adversary, and the environment.

Where the links converge, neurosynaptic computers will consolidate the data to create awareness of the battle and combine it with enormous libraries of tactical system capabilities, simulation outcomes, geospatial facts, and planning factors. The machines will thus have all they require to generate courses of action in a fraction of the time it takes scores of human experts to pool their stove-piped knowledge and conduct planning today.

When the division’s links fail — such as when enemy jamming disrupts friendly communications — the computers will make assumptions based on the last available data and flag those assumptions for validation by human operators. By comparison, only the most deliberate and careful of today’s all-human planning teams actively record data shortfalls or critical assumptions for further consideration later. While this evolution of awareness will not end the “fog of war,” the big data-handling capabilities of the 1st Cavalry Division’s neurosynaptic computers will generate a richer, fuller, and more dynamic picture of the battle than a division enjoys today. The change will be a pivotal evolution in how the humans of the division better understand their environment, akin to the introduction of color film or the addition of soundtracks to movies.

While the machines will provide the awareness, validated assumptions, and courses of action needed to complete an approved plan, only humans can blend experience and creativity to shape very clear and specific criteria and questions for the machines. Here, the 1st Cavalry Division’s staff officers will get involved, and asking the right questions will become critically important. The machines will then analyze data based on the division’s criteria and develop options to answer the questions posed by the commander or staff.  To do so, the 1st Cavalry Division’s headquarters will have to hone effective human-machine collaboration over months of realistic training and education.

In a future of human-machine collaboration, asking broad general questions — as so many commanders do today — will no longer be effective because the power of a neurosynaptic computer lies partially in unrestricted analytical excursions. Imagine if the 1st Cavalry Division staff asked, “How do I take that hub and village?” A cascade of unsuitable answers might spill from their machines: level every structure, besiege the entire area, employ banned weapons, and more. Without context and human guidance, the machines could spend days on such dead ends, while the 1st Cavalry Division will have only hours to go into action.

The interaction of machines and humans will become more effective as human creativity and perspective shapes the machines’ thinking along a deliberate path. The humans must therefore clarify for the machines what they should be looking for in the first place, how to evaluate what they find, and what intangible constraints will affect the commander’s final choice. The more a human operator can tailor her questions to the specific situation, the better.

In the broader context of U.S. and NATO missions on our imagined battlefield, a better question for the 1st Cavalry Division to put to its neurosynaptic computers might be, “Can I take that hub in the next two days and the village three days after that with the forces currently arrayed under my control? If so, how can I take them with minimal risk to civilians and American troops and with minimal expenditure of resources?” The 1st Cavalry Division’s neurosynaptic computers will have the computational power to analyze such a question at superhuman speed. Thanks to effective human-machine collaboration, the machines’ computations will empower the 1st Cavalry Division to gather a precise number of valid options and consider them carefully.

As the planning process proceeds towards completion, the 1st Cavalry Division’s staff will further aid analysis by specifying how the machines should weigh the criteria they continuously receive from soldiers and unmanned systems in the field. The machines might apply all criteria evenly, favor one above the others, set one criteria’s importance lower than all others, or any other combination. The 1st Cavalry Division might even empower their machines to flag a plan as impossible or reject courses of action due to factors beyond the division’s control, such as weather or insufficient supplies.

With an exciting new form of computer and well-trained soldiers defining the right questions, criteria, and guidelines, the 1st Cavalry Division will create a defined number of relevant courses of action for its commanding general. A short burst of electrons from the division’s forward command post will broadcast the plan to those about to execute it. Although the villagers won’t know it, the 1st Cavalry Division’s collaborative efforts will have just assured a NATO victory.

Taken to scale, the concept of human-machine collaboration could have significant positive impacts on the Army. One possible impact is that staff work might never be the same again. What would a staff look like in size and composition if it was designed to develop questions, criteria, and guidelines instead of generating courses of action? How many staff sections will have access to neurosynaptic processing? How many of today’s machines could each neurosynpatic computer replace, and at what cost? We must ask and answer key questions as we put human-machine collaboration through its paces for the benefit of future division commanders.

Empowering human creativity, leveraging exabytes of data, and generating courses of action at extreme speed are the real opportunities of human-machine collaboration for operational decision-making. The safety of American citizens and U.S. allies abroad will rest upon the ability of humans in uniform to make the most of the third offset strategy by asking the right questions.


Maj. Theodore L. Zagraniski, MPA, is a Strategist in the Office of the Deputy Chief of Staff, G-3/5/7, Headquarters, Department of the Army. The views expressed are those of the author and do not reflect the official policy or position of any agency, organization, or the U.S. government.