Insider Brief
- The National Geospatial-Intelligence Agency is preparing to issue a call to industry for over $700 million to enhance AI-driven computer vision systems.
- The initiative includes a significant focus on the NGA’s Maven program, which plays a critical role in military intelligence by helping AI systems accurately identify targets of interest.
- According to Vice Adm. Frank Whitworth, the director of the NGA, this will be the largest data labelling request for proposal ever issued by the U.S. government.
- The agency also announced a new pilot project, the Accreditation of GEOINT AI Models (A-GAIM), to evaluate and certify the reliability and robustness of AI models used in geospatial intelligence.
The National Geospatial-Intelligence Agency (NGA) is reportedly preparing to issue a significant call to industry, worth over $700 million, to enhance AI-driven computer vision systems. This initiative includes a major focus on the NGA’s Maven program, which plays a critical role in military intelligence by helping AI systems accurately identify targets of interest.
Vice Adm. Frank Whitworth, NGA’s director, highlighted that this will be the largest data labelling request for proposal ever issued by the U.S. government. Data labelling, a key part of training AI models, involves humans identifying and categorizing objects in a way AI systems can understand. This process is crucial for ensuring that AI can distinguish between friend and foe and between combatants and civilians—an essential requirement under international laws like the Geneva Conventions.
The NGA gathers vast amounts of imagery from satellites and aircraft, turning this data into geospatial intelligence (GEOINT) products for use across the U.S. government. However, as the volume of data grows, so does the challenge of processing and analyzing it effectively. AI and machine learning are vital to meeting this challenge, but these systems require extensive training to recognize military targets and detect abnormal activities.
The agency also announced a new pilot project, the Accreditation of GEOINT AI Models (A-GAIM). This initiative aims to evaluate and certify the reliability and robustness of AI models used in geospatial intelligence. Through A-GAIM, the NGA seeks to expand the responsible use of AI across the GEOINT enterprise, supporting the warfighter and generating new intelligence insights. This program aligns with the Department of Defense’s guidance on ethical AI and includes a specialized training program for AI developers called GEOINT Responsible AI Training (GREAT).
Image credit: NGA
