How the military can get more out of artificial intelligence

The National Security Commission on Artificial Intelligence (NSCAI) has been busy in recent months, as it should be. The stakes are high when you look at the role artificial intelligence will play at nearly every level of national security in the years ahead. To underestimate the impact of AI on our nation’s safety and security is to do so at great risk. The biggest risk would be to neglect the recruitment, retention and training of elite human warfighters who will drive the successful deployment of AI.

Like many in the fields of operations research, analytics, and data science, we have been closely following the work and recommendations of the NSCAI  with a keen and specific eye as veterans. We have served as both officers in the operational forces (Navy helicopters, Army air defense artillery and Navy surface warfare respectively) and as analysts. All of us participated in the military services’ development and execution of analysis to guide national defense strategy as well.

Now we observe from outside the working analytical levels of government and strongly assert the importance that all sectors of society engage with the coming age of AI.

The military, most of all, must remain educated and become the masters of AI instead of training to be its servants. To accomplish this, there needs to be enough proficient talent in uniform to understand how operation and theory of current and future systems work, and how they fail. Many AI systems found in commerce are built, deployed and developed incrementally as more experience is gained. In the defense construct, these systems will truly be tested for the first time under fire. These service members will have an astonishing amount of leverage as to the outcome – success or failure – of future wars. The impact of a single AI expert in the future may eclipse even that of the combatant commander.

Who are the ‘real warriors’?

The coming revolution in decision centric warfare requires a thorough re-examination of real warrior ethos. The military services are cultural organizations, and there is, as one might expect, resistance to the notion that so-called ‘techno nerds’ share prestige with jet pilots and infantrymen. The sooner we get past this, the better off – and more secure – we will be. Without cultural acceptance, the top talent will continue to disappear regardless of financial and other incentives.

Currently, there is unbounded support in the Department of Defense for “all things AI.”  Paradoxically, there is little support for focused strategies to attract and retain required talent. Without creating promotion paths to senior levels (generally flag/general rank), the career paths within the current system will ultimately fail.

Although the services have the technical mechanisms currently in place to create uniformed warfighters in data analytics and AI, we see a steady decrease in the number of operationally experienced officers sent to these programs. Reduced production has increased the demand on these individuals within the Service, and their marketability in the civilian sector. There is a need to make military and civil career fields more porous in the sense that AI experts need to be able to move in and out of government and a digital service academy might help with that.

Three offsets and two career paths

It might be useful, when thinking about manpower for AI, to consider the ways that ‘offset strategies’ have changed the Department of Defense as a whole. Considering the relationship between government and industry, the first offset is inherently governmental, with non-government industry playing a very small role, if any. The second offset was government-led, with industry following closely behind. The third offset is different in that industry is, at worst, coequal with government, and in many areas likely ahead. This difference must apply not just to the technologies themselves, but also the people involved in developing and operationalizing them. Therefore, to have effective AI leadership, uniformed members may need to spend time both in and out of uniform.

Create analytic flags

The uniformed services are very much like the rest of society in that success follows success. In academia, this is measured by promotions – positional or rank based – and publications. In business, success is measured by both technical and non-technical promotions and revenue.

In the military, success is also measured by promotions, specifically to flag rank (general or admiral).  Admittedly, each of the service analytic organizations have flag officers and/or senior civilians who are responsible for them and act nominally as the advocate; however, it is rare – but not unheard of – that the advocate for analysis is themselves an analyst. The creation of analytic flags who are both the product and proponent of their community will show both the viability of the career and provide top-level mentorship for the community as a whole. This will create individuals who are the de facto leaders of their community with the horsepower to shepherd it.

Some steps in the right direction

In early 2019, the Army activated the AI task force, attempting to follow the recent recommendations from NSCAI, to provide a group of organic, professional level developers and reduce reliance on contractors.

The Navy’s own attempt at this came in the form of the Operations Analysis Specialty Career Program  which began in 2011 and effectively ended in 2019.

The U.S. Air Force recently moved their operations research analysis specialty code from the acquisition and financial management field to a newly created 15A specialty code positioning officers to provide direct support to operations and senior operational commanders.

What else needs to be done

These are solid steps, but to reach AI’s full potential, organizational changes to capture, value and advance the 21st century warfighter must be made. That warfighter must understand the underlying mathematics behind the decision algorithms, their strengths, limitations and assumptions; and if necessary, be able to adjust those algorithms to meet emerging challenges. Likewise, they must be able to attack adversary AI systems and defend our own from misdirection. As we move forward into the robotics age of warfare, warfighting systems requiring humans will become not only less relevant but will also pose a risk to the operational commander when employed. Therefore, the minds of the 21st century warfighter need to be nurtured, educated and advanced.

Jeff Kline is a professor of practice of Military Operations Research at the Naval Postgraduate School and the director of the Naval Warfare Studies Institute. A 26-year veteran of the U.S. Navy’s surface warfare community, he commanded two ships and served in a variety of senior operational positions. He is also a member of INFORMS, the largest trade association of Operations Research and Analytics professionals.

Scott Nestler is the academic director for the MS in Business Analytics program in the Mendoza College of Business. Before joining the University of Notre Dame in 2015, he was a U.S. Army officer in the Air Defense Artillery and served as an operations research & systems analyst. He previously served as the chair of the INFORMS Analytics Certification Board.

Harrison Schramm is the president-elect of the Analytics Society of INFORMS and a principal research scientist at Group W. Prior to joining industry, he was a helicopter pilot and operations research professional in the U.S. Navy.