GAO says US military playing too loose with AI development management
In the U.S. government, it is the job of the Government Accountability Office to poke holes in bureaucrats’ spending plans, and GAO analysts have some strategy-level thoughts on military development of AI.
It sounds like top defense officials would do well to adopt some warfighting tactics: thoroughly plan, coordinate and create an accountable chain of command.
A recent pair of reports paint a picture of incomplete reporting — and possibly incomplete organization — which could hamper department-wide development of effective systems including computer vision and facial recognition.
Earlier this month, the GAO published a report, mandated by the House of Representatives, critiquing the Navy’s 2021 development framework for uncrewed surface and undersea vessels. It is likely that biometric surveillance will be part of the department’s uncrewed future.
The report found that the Navy is mismanaging its $4.3 billion, five-year program to add 21 robotic vessels.
For example, GAO analysts found that that price tag does not include a digital infrastructure, such as data repositories, to support planned autonomous ships.
They reported that there is no cohesive management of the program, either. Adopting a portfolio management approach instills the discipline needed to align investments throughout the program match strategic objectives and resources.
Critically, the GAO says, Navy officials have not defined any criteria needed to evaluate prototypes. In fact, the Navy has yet to create improved prototype schedules.
The DoD currently has some but not all of the data and computing power needed for facial recognition, according to the assessment.
Seven executive-level recommendations buttress the report, and together they boil down to more openness with Congress, sturdier schedules and implementing portfolio management.
The story told in a March GAO report about the Defense Department’s overall AI efforts is very similar.
Analysts wrote that “AI-related strategies could be more comprehensive.” The department also has not “issued guidance that clearly defines the roles and responsibilities of components that participate in AI activities.”
Navy officials have not developed full descriptions of resources and investments required of their funded development plans. Nor have they identified risks associated with deploying AI, including facial recognition.
Key metrics are not fully identified in the department’s AI strategies and plans, according to the GAO, including milestones and performance measurements.
Department officials have done a better job addressing ethical values, goals and objectives and workforce needs.