Colorado reservist granted $2.5M to develop artificial intelligence prototype

  • Published
  • By Tech. Sgt. Devon Cole and Tech. Sgt. Frank Casciotta
  • 310th Space Wing Public Affairs

The 310th Space Wing’s senior intelligence officer has been granted $2.5 million to develop an AI software prototype through the Air Force’s Intelligence Surveillance and Reconnaissance, Modernization, Automation and Development (IMAD) program to improve space domain awareness using predictive intelligence analysis through artificial intelligence (AI).

The IMAD program, which is managed by Air Combat Command, launched in 2020 with a key focus on machine learning (ML) technology as part of the Next Generation of Intelligence Surveillance and Reconnaissance (ISR)  Dominance Flight Plan. U.S. Air Force Lt. Col. Nicholas Demakakos was one of 18 whose proof-of-concept presentations were selected to begin prototype development.

Demakakos is developing a program he calls Space Targeting Analyses and Reporting Celestial Information System (STARCIS) that he hopes to implement across the ISR enterprise. He says, through LLMs, STARCIS can provide ISR analysts with a common intelligence picture that improves intel collaboration and close-by direct operational support gaps within analysis, collection, targeting, military exercises and wargame scenarios.

“When a tactical report gets generated, the LLMs can generate the assessment, the analysis, the predictive analysis, and allow the analyst to leverage what the AI and LLMs are giving them,” said Demakakos.

Large Language Models (LLMs) are programs trained to take in cumbersome sets of data and refine them into a system that can be easily managed by the Intelligence enterprise, allowing ISR operators to apply tradecraft to more abstract problems that deal with complex threat scenarios against US Space Operations.

Between satellite data retrieval, security layers and analysis it typically takes analysts three hours to generate a tactical report, but Demakakos showed STARCIS could condense this process down to 15 minutes in a proof-of-concept presentation. 

By the time leadership gains all the information needed to make key decisions; one or more components have changed. Intent is hard to determine through data of adversaries and their assets. At the moment, what happens in space highly monitored and greatly debated. STARCIS and AI in the military are motivated to change that. 

“I want to be able to help a space operator, understand the threat, understand the anomalies that they might get with resident space objects,” said Demakakos. “When you include that ISR layer, you can look at an object and go ‘Hey, we have moderate to high confidence that this object is not a threat,’ or ‘we have moderate to high confidence that in the next three days, this object will be a threat.”

Historically, victory through conflicts are measured by who had higher confidence in intel, minimal response time to threats/attacks and greater resources. 

In today’s contested space domain, adversarial forces often jam signals and impede imagery collection by U.S. assets in a race for space superiority. 

“We’re seeing satellites that actually can grab another satellite, grapple with it and pull it out of its operational orbit,” said U.S. Space Force Gen. Bradley Saltzman, chief of space operations, as the keynote speaker at the Space Symposium in Colorado Springs, Colorado in April 2023. “These are all capabilities they’re demonstrating on-orbit today, and so the mix of these weapons and the pace with which they’ve been developed are very concerning.”

Demakakos says prototype development will begin Oct. 1, 2024, and will roll out STARCIS no later than Sept. 30, 2025. The prototype will be dispersed to initial group of users amongst nine squadrons under the 310th SW, Space Operations Command Deltas 5 and 7 and Space Training and Readiness Command Deltas 10 and 11.