ST. LOUIS—Months after assuming responsibility for a largely classified program that applies artificial intelligence and machine learning (AL/ML) to video and imagery to detect objects for warfighters, the technology behind Maven has enjoyed major improvements and is being applied routinely to meet the operational needs of Combatant Commanders, the head of the National Geospatial-Intelligence Agency (NGA) said on Monday.
Maven is also helping NGA and its partners make use of the rapidly increasing volume of data being collected by imaging sensors, Vice Adm. Frank Whitworth, director of NGA, said here during the annual GEOINT 2023 Symposium.
Maven transitioned in January from the Office of the Secretary of Defense to NGA and the agency has worked with the Combatant Commanders “to integrate AI into workflows, accelerating operations and speed to decision,” Whitworth said during a keynote address. “This benefits maritime domain awareness, target management, and our ability to automatically search and detect objects of interest. We’ve increased fidelity of targets, improved geolocation accuracy, and refined out test and evaluation process.”
There have been “more than 18 million detections from Maven” that have gone into NGA’s “analytic workflow,” he said.
Mark Munsell, deputy director for Data & Digital Innovation at NGA, told reporters later that the workflow begins with the Maven algorithm being applied to any number of object detections on Earth, from thousands to millions, and “distilling those down into something useful.” This has been done “several times for different commands for different intelligence platforms,” he said.
Whitworth, in the same media roundtable, said Maven is driving a lot of excitement from “real world” users about the growth of the platform.
NGA’s mission is to bring geospatial-intelligence, or GEOINT, products to the warfighter. This can be maps, different types of imagery such as electro-optical and radar, and the analytics that enhance the raw data.
The explosion of data that has come with more collection systems, including commercially-owned and operated imaging satellites, and increasing work around space domain awareness, has increased the importance of using AI/ML technologies to “fuse” all this data that is being gathered “across disparate datasets” and result in “meaningful answers,” Whitworth said.
“So, we’re embracing this new technology, working to automate significant portions of dynamic collection, imagery, exploitation, and reporting workflows to rapidly exploit data and anticipate activity,” he said.
Maven is being applied to more than just targeting, Whitworth told reporters.
“But certainly, the distinction of behaviors to the distinction of objects and the importance of being right drives us to a certain standard that is also consistent with the standard we will need for indications and warning, and infrastructure study, and even elements of mapping,” he said.
Targeting is the hardest challenge and the initial focus because “a lot of the participation” by the Combatant Commands is going toward targeting but, for NGA, Maven “transcends all of it,” he said.
While NGA has the pivotal role for Maven given its expertise around geospatial data, Whitworth said that new users outside the intelligence—and even the GEOINT—space will want to participate and potentially use the same graphical user interface.
NGA is partnering with the Defense Department’s Chief Digital and Artificial Intelligence Office (CDAO) to expand Maven throughout the Combatant Commands, Munsell said. On top of that, he said that given the CDAO’s focus goes beyond GEOINT, the partnership will enable the office to build on the maturity of Maven and expand it to other users and applications in the next two years, he said.
Whitworth said that users are realizing that Maven enables interoperability between applications and workflows.
The NGA officials were asked about any lessons learned in the conflict in Ukraine for Maven, and Munsell answered that DoD hadn’t “trained our AI models a lot on destroyed equipment,” suggesting the technology can be useful in battle damage assessment of vehicles and other battlefield items.
Previously called Project Maven, the AI/ML effort is on track to become a program of record this fall, Munsell said.