AI Tools Surpass Human Planners in Air Force Experiment

The United States Air Force recently announced that artificial intelligence (AI) tools significantly outperformed human planners during a battle management experiment. This finding emerged from the Air Force’s latest initiative, known as the DASH experiment, conducted in the fall of 2023. The experiment involved military personnel from the US, Canada, and the UK tackling various hypothetical military scenarios using AI solutions from six different companies.

The Air Force evaluated the performance of both the AI tools and the human participants while addressing challenges such as planning airstrikes, rerouting aircraft, and managing intelligence data related to unusual electromagnetic signals. Col. John Ohlund, director of the Air Force’s Advanced Battle Management System Cross-Functional Team (ABMS CFT), noted that one AI algorithm produced more proposed Courses of Action (COAs) in a shorter time frame while also making fewer errors than its human counterparts.

According to Ohlund, “These machine-generated recommendations were up to 90 percent faster than traditional methods, with the best in machine-class solutions showing 97 percent viability and tactical validity.” In contrast, human-generated COAs took approximately 19 minutes to produce, with only 48 percent deemed viable and tactically valid.

Understanding the Experiment’s Design

The DASH experiment was structured to challenge human planners in unfamiliar territory. Ohlund explained that the scenarios were intentionally complex to push operators beyond their usual training limits. “We challenged the operators outside of their comfort zone,” he said, as they faced multi-domain problems requiring quick decisions.

The project aims to create a collaborative command system across all military branches, facilitating operations in air, land, sea, space, and cyberspace. Participants predominantly had backgrounds in Air Force operations, with some, like a junior enlisted airman, having limited exposure to electronic warfare strategies.

The experiment utilized unclassified data, as real command-and-control networks are classified, which contributed to the unfamiliarity for the military participants. While the AI tools also lack traditional military training, they processed the same pre-experiment briefings that included comprehensive guidance and data on various military assets.

Ohlund emphasized that the AI’s ability to retain information sets it apart from human participants, who often became overwhelmed by the extensive new information. “The computer does not forget,” he noted.

The Future of AI in Military Operations

The results of the experiment indicate that while the AI tools excelled in generating viable COAs, they are not yet ready for operational deployment. Ohlund clarified that these AI solutions are designed to evolve into “microservices” within a larger military command-and-control framework. In the structured environment used by the ABMS team, generating COAs is just one of several steps required to formulate an executable plan.

One participant, Lt. Ashley Nguyen, expressed initial skepticism about integrating technology into military decision-making but acknowledged the AI’s user-friendly and time-saving capabilities. “The AI didn’t replace us; it gave us a solid starting point to build from,” she stated, highlighting the potential for AI to enhance rather than replace human judgment in military planning.

As the Air Force continues to explore the integration of AI into military operations, the DASH experiment serves as a critical milestone in understanding the capabilities and limitations of AI technology in high-pressure environments. The outcomes may pave the way for more collaborative human-machine interactions in future military endeavors.