Pentagon data centralization is crucial to the future of military artificial intelligence (AI) applications, an AI entrepreneur told a congressional hearing on July 18th.
“AI systems are only as good the data they are trained on,” Alexandr Wang, founder and CEO of the San Francisco-based Scale AI, told the House Armed Services Committee’s (HASC) cyber, information technologies and innovation (CITI) panel. “The DoD creates more than 22 terabytes of data daily, most of which is wasted…Scale [AI] fully supports the CDAO [the DoD chief digital and artificial intelligence office] and its legislative mandate to establish a centralized data repository to enable DoD to harness the power of data with AI.”
The Pentagon has “all the fundamentals to have an insurmountable advantage [over China] because the DoD generates 22 terabytes of data–far more than the PLA [People’s Liberation Army] generates on a daily basis, so if we can instrument this data into one central repository, we can come out ahead,” Wang said in response to a question during the hearing from Rep. Ro Khanna (D-Calif.), the CITI panel’s ranking member.
The idea of a central Pentagon data repository dates back to at least 2017. At a Defense Innovation Board hearing in April 2017, now-retired Air Force Lt. Gen. Jack Shanahan referenced the 22 terabytes of data–what he said were the equivalent of two times the printed holdings of the Library of Congress–but published reports on the hearing differed as to whether the 22 terabytes of DoD data generation daily referred to DoD as a whole or just Air Force intelligence, surveillance, and reconnaissance.
Before his retirement in 2020, Shanahan served as the director of Project Maven and then as the inaugural director of DoD’s Joint Artificial Intelligence Center starting in December 2018.
The National Geospatial-Intelligence Agency (NGA) took over the Pentagon’s Project Maven this year, an effort to develop an AI tool to process object data from full-motion video (FMV) collected by unmanned aircraft, such as General Atomics‘ MQ-9s; decrease the workload of intelligence analysts who may spend hours sifting through FMV; and reduce targeting to firing timelines from hundreds of minutes to single-digit minutes–a capability that Pentagon officials have said is needed for fleeting, “time sensitive targets.”
Vice Adm. Frank Whitworth, director of NGA, said on May 22 that combatant commanders are routinely using Maven, which may become a program of record this year, to satisfy their operational needs (Defense Daily, May 22).
U.S. Central Command’s (CENTCOM) fiscal 2024 wish list points to what NGA has said is a primary longer term goal of AI–the prediction of adversary behaviors that a human analyst may not have envisioned.
The CENTCOM fiscal 2024 wish tabulation–one of a number of DoD’s incongruously named “unfunded priority lists” (UPLs)–includes $34 million for Palantir Technologies‘ [PLTR] Maven Smart System (MSS) (Defense Daily, May 23).
“USCENTCOM’s military advantage is increasingly reliant on data-driven technologies and the integration of diverse data sources, elastic computing to the edge, and rapid software deployment,” the CENTCOM UPL said. “Information sharing and data analytics are principal mechanisms for understanding China’s encroachment, illuminating Russia’s activities, disrupting Iran’s UAS/Advanced Conventional Weapons (ACW) supply chains, and spoiling ISIS and AQ [Al Qaeda] external operations.”
“[The $34 million in] funding provides USCENTCOM with an enterprise license for users of MSS,” the CENTCOM wish list said. “MSS allows users to explore and visualize the battlefield while increasing ability to anticipate, monitor, and respond to destabilizing activities from strategic competitors. Leveraging MSS as a strategic asset provides new ways and means to monitor, respond, and deter aggression at the speed of operations.”
MSS is the software for Project Maven, which kicked off in 2017 with the oversight of the office of the undersecretary of defense for intelligence.
At the Intelligence and National Security Association’s spring symposium in March, Phillip Chudoba, NGA’s associate director of capabilities, said that Maven is the “only performant computer vision, AI/ML capability in the DoD.”
DoD has said that Maven has deployed to dozens of sites for Army, Navy, Marine Corps, and Air Force tactical drones and medium altitude MQ-1C Gray Eagle and MQ-9 Reaper drones to detect and geo-locate people, vehicles, and buildings; track moving objects; and collect data to improve identification and tracking algorithms. The Pentagon has said that it wants to leverage the employment of Maven in the CENTCOM AoR for use in countering high-tech adversaries through the detection and tracking of tanks, artillery, aircraft and missile launchers.
An April 2018 Congressional Research Service report said that DoD initially incorporated Project Maven AI tools for 10 sites.
Google [GOOGL] was the prime contractor for Project Maven but dropped out in 2018 after receiving pushback from employees about the company’s tools being used for an AI drone imaging effort.
At the July 18th hearing of the HASC CITI panel, Wang said that DoD has not had a pathfinder AI project start since Maven and that it should create new AI pathfinders to drive Pentagon AI innovation. While the DoD budget request represents a high point dollar-wise for AI, the Pentagon is spending three times less than China on AI, which the PLA is positioning for use in autonomous drone swarms, adaptive radar systems, and autonomous vehicles.
“AI is China’s Apollo project,” he testified.