Integrating Artificial Intelligence and Machine Learning in the Marine Corps – War On The Rocks

Every day, thousands of marines perform routine data-collection tasks and make hundreds of data-based decisions. They compile manning data on whiteboards to decide to staff units, screenshot weather forecasts and paste them into weekly commanders update briefings, and submit training entries by hand. But anyone who has used ChatGPT or other large-scale data analytic services in the last two years knows the immense power of generative AI to streamline these processes and improve the quality of these decisions by basing them on fresh and comprehensive data.

The U.S. Marine Corps has finally caught wind. Gen. Eric Smiths new message calls for the service to recognize that [t]echnology has exponentially increased informations effects on the modern battlefield, making our need to exploit data more important than ever. The services stand-in forces operating concept relies on marine operating forces to integrate into networks of sensors, using automation and machine learning to simplify decision processes and kill chains. Forces deployed forward in littoral environments will be sustained by a supply system that uses data analysis for predictive maintenance, identifying which repair parts the force will need in advance.

However, there is a long way to go before these projections become reality. A series of interviews with key personnel in the Marine Corps operating forces and supporting establishment, other services, and combatant commands over the past six months reveal that the service needs to move more quickly if it intends to use AI and machine learning to execute this operating concept. Despite efforts from senior leaders to nudge the service towards integrating AI and machine learning, only incremental progress has been made.

The service depends on marines possessing the technical skills to make data legible to automated analytic systems and enable data-informed decisions. Designating a Marine expeditionary force or one of its major subordinate commands as the lead for data analysis and literacy would unify the services two-track approach by creating an ecosystem that will allow bottom-up creativity, scale innovation across the force, and speed the integration of these technologies into the fleet and supporting establishment.

New Technologys Potential to Transform Operations, Logistics, and Education

AI, machine learning, and data analysis can potentially transform military education, planning, and operations. Experiments at Marine Corps University have shown that they could allow students to hone operational art in educational settings by probing new dimensions of complicated problems and understanding the adversarys system. AI models, trained on enemy doctrinal publications and open-source information about troop employment, can use probabilistic reasoning to predict an enemys response. This capability could supplement intelligence red teams by independently analyzing the adversarys options, improve a staffs capacity for operational planning, or simply give students valuable analytic experience. And NIPRGPT, a new Air Force project, promises to upend mundane staff work by generating documents and emails in a secure environment.

Beyond education and planning, AI and machine learning can transform how the Marine Corps fights. During an operation, AI could employ a networked collection of manned and unmanned systems to reconnoiter and attack an adversary. It could also synthesize and display data from sensor networks more quickly than human analysts or sift through thousands of images to identify particular scenes or locations of interest. Either algorithms can decide themselves or enable commanders to make data-informed decisions in previously unthinkable ways. From AI-enabled decision-making to enhanced situational awareness, this technology has the potential to revolutionize military operations. A team of think tank researchers even used AI recently to rethink the Unified Command Plan.

But, achieving these futuristic visions will require the service to develop technical skills and familiarity with this technology before implementing it. Developing data literacy is a prerequisite to effectively employ advanced systems, and so this skill is as important as anything else the service expects of marines. Before the Marine Corps can use AI-enabled swarms of drones to take a beachhead or use predictive maintenance to streamline supply operations, its workforce needs to know how to work with data analysis tools and be comfortable applying them in everyday work settings.

Delivering for the Marine Corps Today

If the Marine Corps wants to employ machine learning and AI in combat, it should teach marines how to use them in stable and predictable garrison operations. Doing so could save the service tens of thousands of hours annually while increasing combat effectiveness and readiness by replacing the antiquated processes and systems the fleet marine force relies on.

The operating forces are awash with legible data that can be used for analysis. Every unit has records of serialized equipment, weapons, and classified information. Most of these records are maintained in antiquated computer-based programs of record or Excel spreadsheets, offering clear opportunities for optimization.

Furthermore, all marines in the fleet do yearly training and readiness tasks to demonstrate competence in their assigned functions. Nothing happens to this data once submitted in the Marine Corps Training Information Management System no headquarters echelon traces performance over time to ensure that marines are improving, besides an occasional cursory glance during a Commanding Generals Inspection visit. This system is labor intensive, requiring manual entries for each training event and each individual marines results.

Establishing and analyzing performance standards from these events could identify which units have the most effective training regimens. Leaders who outperform could be rewarded, and a Marine expeditionary force could establish best practices across its subordinate units to improve combat readiness. Automating or streamlining data entry and analysis would be straightforward since AI excels at performing repetitive tasks with clear parameters. Doing so would save time while increasing the combat proficiency of the operating forces.

Marines in the operating forces perform innumerable routine tasks that could be easily automated. For example, marines in staff sections grab data and format it into weekly command and staff briefings each week. Intelligence officers retrieve weather forecast data from their higher headquarters. Supply officers insert information supply levels into the brief. Medical and dental readiness numbers are usually displayed in a green/yellow/red stoplight chart. This data is compiled by hand in PowerPoint slide decks. These simple tasks could be automated, saving thousands of hours across an entire Marine expeditionary force. Commanders would benefit by making decisions based on the most up-to-date information rather than relying on stale data captured hours before.

The Marine Corps uses outdated processes and systems that waste valuable time that could be used on training and readiness. Using automation, machine learning, and AI to streamline routine tasks and allow commanders to make decisions based on up-to-date data will enable the service to achieve efficiency savings while increasing its combat effectiveness. In Smiths words, combining human talent and advanced processes [will allow the Marine Corps] to become even more lethal in support of the joint force and our allies and partners.

The Current Marine Corps Approach

The service is slow in moving towards its goals because it has decided, de facto, to pursue a two-track development strategy. It has concentrated efforts and resources at the highest echelons of the institution while relying on the rare confluence of expertise and individual initiative for progress at the lowest levels. This bifurcated approach lacks coherence and stymies progress.

Marine Corps Order 5231.4 outlines the services approach to AI. Rather than making the operating forces the focus of effort, the order weights efforts in the supporting establishment. The supporting establishment has the expertise, resources, and authority to manage a program across the Marine Corps. But it lacks visibility into the specific issues facing individuals that could be solved with AI, machine learning, or automated data analysis.

At the tactical levels of the service, individuals are integrating these tools into their workflows. However, without broader sponsorship, this mainly occurs as the result of happy coincidence: when a single person has the technical skills to develop an automated data solution, recognizes a shortfall, and takes the initiative to implement it. Because the skills required to create, maintain, or customize projects for a unit are uncommon, scaling adoption or expanding the project is difficult. As a result, most individual projects wither on the vine, and machine learning, AI, and data analysis have only sporadically and temporarily penetrated the operating forces.

This two-track approach separates resources and problems. This means that the highest level of service isnt directly involved in success at the tactical level. Tactical echelons dont have the time, resources, or tasking to develop and systematize these skill sets on their own. Whats needed is a flat and collaborative bottom-up approach with central coordination.

The 18th Airborne Corps

Marine Corps doctrine and culture advocate carefully balancing centralized planning with decentralized execution and bottom-up refinement. Higher echelons pass flexible instructions to their subordinates, increasing specificity at each level. Leaders ensure standardization of training, uniformity of effort, and efficient use of resources. Bottom-up experimentation applies new ideas to concrete problems.

Machine learning and data analysis should be no different. The challenge is finding a way to link individual innovation instances with the resources and influence to scale them across the institution. The Armys use of the 18th Airborne Corps to bridge the gap between service-level programs and individual initiatives offers a clear example for how to do so.

The 18th Airborne Corps fills a contingency-response role like the Marine Corps. Located at Fort Liberty, it is the headquarters element containing the 101st and 82nd Airborne Divisions, along with the 10th Mountain and 3rd Infantry Divisions. As part of a broader modernization program, the 18th Airborne Corps has focused on creating a technology ecosystem to foster innovation. Individual soldiers across the corps can build personal applications that aggregate, analyze, and present information in customizable dashboards that streamline work processes and allow for data-informed decision-making.

For example, soldiers from the 82nd Airborne Division created a single application to monitor and perform logistics tasks. The 18th Airborne Corps Data Warfare Company built a tool for real-time monitoring of in-theater supply levels with alerts for when certain classes of supply run low. Furthermore, the command integrates these projects and other data applications to streamline combat functions. For example, the 18th Airborne Corps practices integrating intelligence analysis, target acquisition, and fires through joint exercises like Scarlet Dragon.

As well as streamlining operational workflows, the data analytics improve training and readiness. The 18th Airborne Corps has developed a Warrior Skills training program in which they collect data to establish a baseline against which it can compare individual soldiers skills over time. Finally, some of the barracks at Fort Liberty have embedded QR codes that soldiers scan to check in when theyre on duty.

These examples demonstrate how a unit of data-literate individuals can leverage modern technology to increase the capacity of the entire organization. Many of these projects could not have been scaled beyond institutional boundaries without corps-level sponsorship. Furthermore, because the 18th Airborne Corps is an operational-level command, it connects soldiers in its divisions with the Armys service-level stakeholders.

Designating a Major Command as Service Lead

If the Marine Corps followed the 18th Airborne Corps model, it would designate one operating force unit as the service lead for data analysis and automation to link service headquarters with tactical units. Institutionalizing security systems, establishing boundaries for experimentation, expanding successful projects across a Marine expeditionary force, and implementing a standardized training program would create an ecosystem to cultivate the technical advances service leaders want.

This proposed force would also streamline the interactions between marines and the service and ensure manning continuity for units that develop data systems to ensure efforts do not peter out as individuals rotate to new assignments. Because of its geographic proximity to Fort Liberty, and as 2d Marine Division artillery units have already participated in the recent Scarlet Dragon exercises and thus have some familiarity with the 18th Airborne Corps projects, II Marine Expeditionary Force is a logical choice to serve as the service lead.

Once designated, II Marine Expeditionary Force should establish an office, directorate, or company responsible for the entire forces data literacy and automation effort. This would follow the 18th Airborne Corps model of establishing a data warfare company to house soldiers with specialized technical skills. This unit could then develop a training program to be implemented across the Marine expeditionary force. The focus of this effort would be a rank-and-billet appropriate education plan that teaches every marine in the Marine expeditionary force how to read, work with, communicate, and analyze data using low- or no-code applications like PowerBI or the Armys Vantage system, with crucial billets learning how to build and maintain these applications. Using the work it is undertaking with Training and Education Command, combined with its members academic and industry expertise, the Marine Innovation Unit (of which I am a member) could develop a training plan based on the Armys model that II Marine Expeditionary Force could use and would work alongside the proposed office to create and implement this training plan.

This training plan will teach every marine the rudimentary skills necessary to implement simple solutions for themselves. The coordinating office will centralize overhead, standardize training, and scale valuable projects across the whole Marine expeditionary force. It would link the high-level service efforts with the small-scale problems facing the operating forces that data literacy and automation could fix.

All the individuals interviewed agreed that engaged and supportive leadership has been an essential precondition for all successful data automation projects. Service-level tasking should ensure that all subordinate commanders take the initiative seriously. Once lower-echelon units see the hours of work spent on rote and mundane tasks that could be automated and then invested back into training and readiness, bureaucratic politics will melt away, and implementation should follow. The key is for a leader to structure the incentives for subordinates to encourage the first generation of adopters.

Forcing deploying units to perform another training requirement could overburden them. However, implementing this training carefully would ensure it is manageable. The Marine expeditionary force and its subordinate units headquarters are not on deployment rotations, so additional training would not detract from their pre-deployment readiness process. Also, implementing these technologies would create significant time savings, freeing up extra time and manpower for training and readiness tasks.

Conclusion

Senior leaders across the Department of Defense and Marine Corps have stated that AI and machine learning are the way forward for the future force. The efficiency loss created by the services current analog processes and static data (let alone the risk to mission and risk to force associated with these antiquated processes in a combat environment) is enough reason to adopt this approach. However, discussions with currently serving practitioners reveal that the Marine Corps needs to move more quickly. It has pursued a two-track model with innovation at the lowest levels and resources at the highest. Bridging the gap between these parallel efforts will be critical to meaningful progress.

If the Marine Corps intends to incorporate AI and machine learning into its deployed operations, it should build the groundwork by training its workforce and building familiarity during garrison operations. Once marines are familiar with and able to employ these tools in a stable and predictable environment, they will naturally use them when deployed to a hostile littoral zone. Designating one major command to act as the service lead would go a long way toward accomplishing that goal. This proposed command would follow the 18th Airborne Corps model of linking the strategic and tactical echelons of the force and implementing new and innovative ways of automating day-to-day tasks and data analysis. Doing so will streamline garrison operations and improve readiness.

Will McGee is an officer in the U.S. Marine Corps Reserves, currently serving with the Marine Innovation Unit. The views in this article are the authors and do not represent those of the Marine Innovation Unit, the U.S. Marine Corps, the Defense Department, or any part of the U.S. government.

Image: Midjourney

Link:

Integrating Artificial Intelligence and Machine Learning in the Marine Corps - War On The Rocks