Operationalizing Project Maven

The U.S. military has deployed an advanced artificial intelligence system to the battlefield for the first time.

In April 2017, Deputy Defense Secretary Robert O. Work established a new Algorithmic Warfare Cross-Functional Team for the Department of Defense. Dubbed “Project Maven” and led by Lt. Gen. John N.T. “Jack” Shanahan, director for defense intelligence, warfighter support with the Office of the Under Secretary of Defense for Intelligence, the initiative explores ways the military could use deep learning and neural networks to extract insights from intelligence data, support warfighters, and ultimately defeat ISIS. Within two months, the team received funding from Congress. In December, after six months of research and development, the group deployed to the Middle East its first mission-ready product: an object recognition algorithm for identifying features in video footage from ScanEagle reconnaissance drones.

Imagery labeling and sorting is a lucrative task for data analysts, but also a tedious one that can lead to burnout and fatigue-driven errors. By automating the object detection process, military operators can analyze larger quantities of data faster and maintain accuracy while redirecting human energy to more abstract areas. 

A week after combat trials began, the algorithm was able to identify people, vehicles, and different building types with an accuracy rate of about 80 percent, Defense One reported. The Maven team has paired its algorithm with a Navy and Marine Corps tool called Minotaur, which geo-registers an object’s coordinates and displays its exact location on a map.

As testing continues in the next few months, the algorithm will be refined and deployed to more U.S. Special Operations Command teams for use with larger tactical UAVs and eventually ISR satellites. The system will also be introduced to other data types (such as radar) for use across more operational contexts.

While the applied use cases are still narrow and require careful oversight, Project Maven’s early success is a sign that AI may soon play a key role in military combat. Military competitors such as China and Russia have taken interest in the technology as well, reinforcing the idea that the future landscape of war will rely heavily on human-machine teaming.

Photo Credit: U.S. Department of Defense