Jacqueline Tame | CalypsoAI Advisor and Former Deputy Director of the United States Department of Defense, Joint Artificial Intelligence Center (JAIC)
Last month I left the Department of Defense (DoD) after almost 15 years of federal service. It has been a fascinating exercise to listen in at virtual symposia and panels, read press releases, and watch my friends and former colleagues reflect on the continued ongoings, just as I have for the past decade and a half but now, from a position a bit further removed. That space – both geographic, following a cross-country move to the west coast, and emotional – has prompted a lot of introspection. Now, back in the civilian sector, I find myself thinking – in particular – on the needs of our warfighters, what that term really means, and how we – as a Department and a nation – can do better. As shown in the FY22 budget, released just as I was transitioning out of federal service this summer, DoD continues its relentless (and necessary) focus on modernization, but its leaders must constantly ask themselves: what capabilities will both enable our warfighters and determine their success on the battlefield? My time at the Joint Artificial Intelligence Center (JAIC) confirmed that artificial intelligence (AI) is one such critical capability, but also one wherein the speed, way, and scale at which we adopt and use it will determine our success.
It is not enough to rapidly adopt, integrate, and scale AI and AI-enabled capabilities; we must do so reliably and responsibly. In addition to the $874 million the DoD intends to spend on AI in 2022, the Department also released a memorandum outlining its inaugural roadmap for implementing “Responsible AI” (RAI) at scale. A memo like this has special power. In it, the Deputy Secretary of Defense reaffirmed the Department’s commitment to its 2020 AI Ethical Principles, which ensure the Nation’s values are reflected in the military’s design, development, deployment, and use of AI capabilities.
While I was both encouraged by the release of this memo and personally cognizant of and grateful for the tireless work that went into its drafting and publication, I find myself reflecting – almost daily – on the fact that the DoD must now implement it. Without Responsible AI and the security and assurance protocols to underpin its design, development, test, and deployment we risk forgoing a temporally and politically fragile opportunity to harness its power to secure U.S. military advantage and safeguard our Nation.
Fortunately, this memorandum already has significant momentum: Responsible AI enjoys bipartisan support on the Hill and is a demonstrably, mission-critical priority that has transcended administrations. Building upon the consensus that surrounds RAI will enable us to accelerate deployment across the entire national security community as well as other federal agencies and several of our allied partners.
Further, the recently-released National Security Commission on Artificial Intelligence’s final report recommendations call upon the JAIC – the statutorily-designated AI coordinator and implementation arm for the Department – to ensure justifiable warfighter trust in AI systems by establishing a test, evaluation, verification and validation (TEV&V) framework that “integrates real-time monitoring, algorithm confidence metrics, and user feedback to ensure trusted and trustworthy AI capabilities.”
Metrics, milestones, and measures of effectiveness, in particular, are critical to tracking AI accountability. When warfighters and DoD users understand the algorithms they’re leveraging, they can begin to trust their AI systems and glean valuable insights from real-time monitoring of autonomous systems. It is equally important, as captured in the memo, that the American public and our international partners build trust in these same systems.
And yet, despite this widespread recognition, the Department is simply not yet equipped with the science nor technology to comprehensively test and evaluate AI-enabled systems. While it is improving its ability to validate and verify technical performance metrics, other facets of operational performance remain hard to quantify, which presents a barrier to progress that the Deputy Secretary’s memorandum strives to address. I saw this firsthand while serving in the JAIC, where it was clear that the DoD still lacks an iterative approach to AI development, testing, fielding, and sustainment.
U.S. private industry innovation currently outpaces the federal government’s internal research and development efforts for AI and machine learning. As such, to effectively implement the Deputy Secretary of Defense’s guidance, the Department must embrace technology partnerships with industry leaders, and prepare to implement reforms that enable rapid acquisition and fielding of AI capabilities.
Specifically, the Department must continue its focus on broadening acquisition opportunities for forward-leaning companies that will help the DoD achieve its Responsible AI goals. This will require particularly close coordination between the JAIC and its sister components across the Department’s acquisition and research and development arms, as well as with the Chief Data Officer. Together, these critical components must embrace and champion RAI within the DoD, paving the way for enterprise AI adoption and setting a precedent for other U.S. Government and international partners.
DoD cannot build confidence and justifiable trust in AI performance alone; it requires the entire AI ecosystem and, more specifically, critical partnerships with the Intelligence Community, Industry, academia and Congress. Over the next six months, these key AI stakeholders must work in tandem to track financial investment in, Congressional hearings on, and legislative provisions supporting Responsible AI, because deliberate cooperation heightens the urgency to establish standard practices and applications.
Warfighters learn lessons of ethics, law, and morality within our society before they ever deploy. Throughout their service, those lessons are reinforced and strengthened by the military community’s values and warrior ethos. As the DoD increasingly adopts AI capabilities, our intelligent machines must be as responsible and reliable as our human warfighters have been trained to be, and are held accountable for being.
Acting on the imperative set by the Responsible AI memorandum is an opportunity we simply cannot afford to miss. Now is our chance to harness the power of AI to responsibly secure the U.S. military advantage, protect our Nation, and empower our warfighters on tomorrow’s battlefield.
Jacqueline Tame is an advisor for CalypsoAI and most recently served as the Acting Deputy Director and inaugural Chief Performance Officer of DoD’s Joint Artificial Intelligence Center. Previously, she served as a Senior Advisor to the Under Secretary of Defense for Intelligence and Security; Senior Staffer on the House Permanent Select Committee on Intelligence; Chief of Customer Engagement at the Defense Intelligence Agency; Advisor to the Chief of Naval Operations; and Policy Advisor to the Deputy Director of National Intelligence. Jacqueline holds a Bachelor of Arts degree in French with a minor in classical vocal performance and a Master of Public Affairs degree from the University of Texas at Austin, as well as a Masters in National Security and Strategic Studies from the U.S. Naval War College’s College of Naval Warfare. She is passionate about growing and promoting women leaders, having founded Command After Next – a small group dedicated to the mentorship, professional development, education, and championship of women in national security.