Outside perspectives to inform understanding of the state of Artificial Intelligence and Machine Learning
(AI/ML) applications to improve Department of Defense operations
April 19, 2023
JOSH LOSPINOSO: Chairman Manchin, Ranking Member Rounds, members of the subcommittee, it is my honor to have the opportunity to testify before you today on the state of Artificial Intelligence and Machine Learning applications to improve Department of Defense operations.
While AI research is now many decades old, the field has accelerated at a blistering pace. From ChatGPT to self-driving cars, recent AI-powered technologies have again captured the public imagination.
I commend the subcommittee for treating this accelerating development with renewed urgency. In 2021, the National Security Commission on Artificial Intelligence’s message was clear: if trends continue, China will surpass us within a decade.
This subcommittee has asked whether we’ve made progress towards the NSCAI’s recommendations, what gaps exist, and where policy is impeded. In this testimony, I want to bring attention to two facts about today’s military weapon systems, AI, and cybersecurity:
Fact #1: Most major weapon systems are not AI ready
As data scientists are quick to say, “garbage in makes garbage out.” Data allows us to investigate, train, and monitor novel AI-enabled techniques. Without high-quality data, we cannot build effective AI systems. Unfortunately, today the DoD struggles to liberate even the simplest data streams from our weapon systems. These machines are talking, but the DoD is unable to hear them.
We cannot employ AI-enabled technologies without great data. This requires taking seriously the difficult, unglamorous work of laying strong foundations: clean, labeled, enriched, comprehensive data; sound, simple, decentralized, scalable data architectures; and straightforward, unambiguous metrics for measuring AI-empowered systems’ effectiveness.
America’s weapon systems are not AI-ready. While the Department of Defense’s intention is to integrate and employ AI capabilities across the Joint Force, the weapon systems themselves are not ready for them. We must implement solid, scalable edge-computing. We need to enable full-take data collection at the edge. We must solve the operational challenge of transferring terabytes of data from the field to the cloud, making them available to the AI-enabled technologies they will fuel.
Fact #2: The DoD cannot solve weapon system cybersecurity without AI
Without AI, the DoD will never be able to keep these weapon systems cybersecure – it has made little progress addressing the perils identified in the GAO’s 2018 Weapon System Cybersecurity Report. The DoD spends trillions of dollars fielding major weapon systems. Each one contains dozens – sometimes hundreds – of special purpose computers that perform every conceivable function. From the control surfaces on an aircraft to data radios on submarines, these systems are highly digitized. Unlike modern zero-trust architectures, our weapon systems are complete-trust architectures. The DoD needs AI-powered capabilities to detect anomalies and prevent cybersecurity intrusions.
The NSCAI is right. If we don’t act – now – China’s goal of surpassing us will be realized. Major weapon system programs, both new and old, need funding and requirements to make them AI ready. The good news is that between industry, academia, and government, solutions to these challenges exist today.
I look forward to discussing these matters with you, and continuing support to the warfighter.