AI will now have input in Air Force decision-making

AI programs will create predictive models for the Air Force as the first stage of Project Quantum’s push to bring smart software to military leaders.

Algorithms will now bring artificial intelligence into the Air Force’s planning, programming, budgeting, and execution process. According to the Defense Innovation Unit Experimental (DIUx), this step marks the first stage of Project Quantum, an effort to use machine-learning resources to enhance decision-making capabilities among top military leaders. 

The first step is a contract with SparkCognition, an artificial intelligence-focused company based in Austin, Texas. The machine-learning programs applied to the Air Force decision-making data will be common to those of the company’s SparkPredict programs, according to a press release.

“We are customizing an AI engine for the Air Force to provide actionable insights and behavior predictions,” explained Tim Stefanick, Director of Federal Operations at SparkCognition. “SparkCognition will apply cutting-edge technology that will benefit our troops and our national security.”

According to SparkCognition, the focus of the Air Force algorithm program will be on recognizing and modeling higher-order patterns. The goal of the project is to process data corresponding to a wide variety of Air Force planning and programming practices in order to derive a model capable of informing future decisions related to these spheres.

Higher-order model generation algorithms are often knowledge-based programs, which are separated from typical algorithms by the fact that they contain more refined task divisions within the applications. Rather than dividing the application into parts for data and program, knowledge-based algorithms contain a part for data, a part for problem-solving information, and a part for the control program, which actually processes the data. Applied to the Air Force, this type of algorithm would process past data and identify more complex patterns.

Model generation with knowledge-based algorithms involves recognizing these patterns, so that they can be classified for the correct analysis method and matched with the relevant program assumptions, or parameters, according to the publication. For example, the program could be able to predict the impact of the removal of an entire airframe from the Air Force, according to Stefanick.

According to the DIUx, the algorithmic analysis will be conducted on both the micro and macro levels. The work will be conducted in phases, with the first phase scheduled to last 60 days. Later phases will focus on strengthening the model with more data and testing it using trial scenarios, reported SparkCognition.

SparkCognition will be working with the Air Force, through DIUx. Created in 2015, DIUx has served as the liaison between technology start-ups and the Department of Defense.