The evolution of DoD policy and the role of AI in modern warfare

Story

June 21, 2024

Drew Thompson

Sealevel Systems, Inc.

Integrating artificial intelligence (AI) into military applications presents complex and multifaceted challenges, encompassing technological advancements, policy frameworks, strategic considerations, and ethical concerns. To keep pace with the rapid speed with which AI technology continues to improve and evolve, the U.S. Department of Defense (DoD) has established a unified adoption strategy aimed at improving the organizational environment within which DoD leaders and warfighters will be able to make rapid, well-informed decisions by expertly leveraging high-quality data, advanced analytics, and AI for enduring decision advantage.

The use of AI in military and battlefield applications is a relatively new phenomenon, but limited use of the technology is already benefitting warfighters. In this nascent form, AI is largely being used in information gathering and processing capacities. For example, AI is much more efficient than humans at sifting through large amounts of data and images or monitoring feeds for meaningful information.

One of the more significant current applications is threat identification and recognition, especially in air combat. This can be done through locating unique radio or radar signals that are emitted by individual aircraft or aircraft types. In the past, this was an enormous undertaking, requiring a wide variety of sensors operating and recording information from many different sources and frequency ranges. This sensor data was then analyzed by specialists to locate and identify various signals associated with individual aircraft or aircraft types. Today, the identification work conducted by many individuals, over dozens or hundreds of hours, can be done in milliseconds to seconds by AI systems.

The hardware powering military AI

There are several different hardware components used in military AI applications.

High-performance computing (HPC): The majority of military AI applications require powerful computing resources to process large volumes of data and perform complex calculations in real time. HPC systems, including but certainly not limited to supercomputers and clusters of high-end servers, provide the necessary computational power.

There has been a tremendous amount of discussion and scholarship around the location of these HPC resources. There is one school of thought that argues the HPC components are more appropriately located in a central area, far away from a battlefield. The alternative view is that all of the computation should be pushed to the edge.

Performing most of the intensive computation in a central location allows a much larger amount of, and a much larger range of types of, equipment and components. However, this makes the networks or “pipes” a much more crucial component of AI applications.

Alternatively, field-deployed edge hardware is more size-constrained than externally located hardware. Edge hardware is limited by size, while external hardware is limited by the security and strength of the pipes.

Graphics processing units (GPUs): GPUs are not strictly necessary but are often used for accelerating AI computations, particularly when utilizing machine learning and deep learning algorithms. GPUs are a major benefit in applications that rely on parallel processing. Military AI systems often utilize GPUs for tasks such as image recognition, object detection, and autonomous navigation.

Software for AI algorithms and UI

AI algorithms and models: Military AI applications rely on advanced algorithms and models to perform tasks like image recognition, natural language processing, decision making, and predictive analytics.

Simulation/training software with large data sets: To train AI systems and simulate various scenarios, specialized software platforms are used, which enable realistic simulations of military environments, tactics, and equipment. To best train military AI, these simulations require applying massive data sets – the more data, the better.

Integration software: Military AI systems need to integrate with the existing infrastructure and interact with other systems seamlessly and intuitively. Soldiers in the field cannot be expected to navigate challenging user interfaces within software platforms.

U.S. Department of Defense AI policy

The U.S. Department of Defense (DoD) has been strategically incorporating AI and machine-learning (ML) technologies through various policies and strategic documents over the past few years. The DoD’s published “2018 DoD Artificial Intelligence Strategy” laid the groundwork for developing a centralized infrastructure, integrating new technology, and achieving international leadership in AI ethics and safety. Subsequent strategies, such as the “2020 DoD Data Strategy” and the creation of the Chief Digital and Artificial Intelligence Office (CDAO), further emphasized the importance of data-centric approaches and the optimization of AI capabilities across the DoD.

The current guiding policy, outlined in the “2023 DoD Data, Analytics, and AI Adoption Strategy,” builds upon the previous policy documents and places a major focus on speed, agility, learning, and responsibility. It emphasizes decentralized authority and the creation of tight feedback loops between developers and end-users, aimed at enhancing the decision-making processes within the DoD. The 2023 strategy outlines a foundational, guiding approach to AI rather than a step-by-step guide.

Key components of the 2023 strategy include the AI Hierarchy of Needs (Figure 1), which prioritizes high-quality data as the foundation for insightful analytics and responsible AI development. The strategy also advances the need for user-friendly infrastructure and continuous refinement of policies and processes to ensure responsible AI implementation.

[Figure 1 ǀ The DoD AI Hierarchy of Needs prioritizes high-quality data. Image courtesy U.S. Department of Defense.]

Deployed AI solutions

There are a wide variety of manufacturers and contractors that are currently incorporating AI into military applications. These manufacturers and contractors range from large established companies like Boeing, General Dynamics, Lockheed Martin, Raytheon, and Northrop Grumman to upstarts such as Anduril.

Shooter-detection systems: While not a strictly military application, shooter-detection systems have evolved an AI-integrated solution that enables first responders to pinpoint the exact location of gunfire. The system uses a series of acoustic and infrared flash-detection sensors integrated into video, access-control, and mass-notification systems. The data collected by the sensor system is fed through an I/O module directly into an AI-powered software platform, which can determine if and when a gunshot happens, pinpoint its exact location, notify authorities, and send mass notifications, all within 0.5 seconds.

Tactical Intelligence Targeting Access Node (TITAN): The Tactical Intelligence Targeting Access Node, or TITAN, is a scalable and expeditionary intelligence ground station that will accelerate and simplify the Army’s ability to access and process massive volumes of intelligence, surveillance, and reconnaissance (ISR) data.

Physically, the TITAN (Figure 2) is a mobile data center with integrated power, heating and cooling, redundant communications, and computing platforms, all built into a large truck-based platform. The vehicle-mounted expeditionary ground stations will use AI to provide deep-sensing capability that will enable long-range precision fires for the modern battlespace. Using AI, the TITAN will perform data integration, fusion, processing, and analytic capabilities using AI and ML to automate and assist the Army in shortening sensor-to-shooter timelines.

[Figure 2 ǀ Shown is the TITAN ALPHA working concept vehicle. Photo credit: Palantir.]

The Sealevel Relio R1 Rugged embedded computer is at the heart of the TITAN system. The Relio R1 Rugged monitors TITAN’s overall health and performance. The small-form-factor computer hosts multiple software applications and interprets data from a wide variety of internal sensors.

AI going forward

The integration of AI into military applications represents a significant advancement in modern warfare, offering enhanced capabilities in information processing, threat identification, and decision-making processes. The evolution of AI technology is linked with the development of robust technological infrastructure, guided by strategic initiatives and policy frameworks set forth by organizations like the CDAO. While AI presents immense potential for improving military effectiveness, it also raises important considerations regarding responsible development, deployment, and the implications of autonomous systems in conflict scenarios. Continued collaboration between manufacturers and developers, policymakers, and warfighters will be crucial in ensuring that military AI applications enhance operational capabilities and responsibly contribute to global security.

Drew Thompson is a technical writer and marketing specialist for Sealevel Systems, the leading designer and manufacturer of embedded computers, industrial I/O, and software for critical communications. He holds an M.S. in global studies and international affairs from Northeastern University. Thompson can be reached at [email protected].

Sealevel Systems • https://www.sealevel.com/