February 22, 2022
Dr. Jared Dunnmon is the Technical Director for Artificial Intelligence (AI) & Machine Learning (ML) at the Defense Innovation Unit (DIU). In this role, he brings a technical perspective to problem curation, vendor sourcing and evaluation, and project execution. Prior to DIU, Jared was an Intelligence Community Postdoctoral Fellow in Computer Science at Stanford University, where he was advised by Prof. Chris Ré. His research interests focus on combining heterogeneous data modalities, machine learning, and human domain expertise to inform and improve decision making around such topics as human health, energy & environment, and geopolitical stability. Jared holds a PhD from Stanford University (2017), a B.S. from Duke University, and both an MSc. in Mathematical Modeling and Scientific Computing and an M.B.A. from Oxford, where he studied as a Rhodes Scholar.
Why were you interested in working at DIU?
I chose to move to DIU from a successful startup because it offered a unique opportunity to combine my technical skillset and industry experience in support of national security. I was also attracted by the incredible team of folks from industry, technology, and government, from whom I can learn an amazing amount every day.
What is your favorite DIU experience?
This is a really hard question – I love what I do every day. If I had to choose, my favorite experience was making the trip to Guam to deploy an augmented reality microscope that could help the pathology team at the Naval Medical Hospital more rapidly identify cancer. It is rare that one gets to see a project from development through deployment, and in this case it was particularly satisfying to be able to observe exactly how the cutting edge technology involved in this project could be used to improve medical care in low-resource environments.
If you could solve any Department of Defense (DoD) problem tomorrow, no matter how big, what would you tackle and why?
Changing the budget process such that DoD can more rapidly integrate emerging technologies into its operations. At present, when we have a successful prototype that was built with *research* funding, it can take years before the *sustainment* funding is allocated, leaving potentially transformational technologies on the sidelines. We have to move at the speed of relevance, and the funding cycle is a major blocker. We continue to make this case at DIU and find ways to move the ball forward.
What emerging commercial technologies are you most excited about?
One of the most exciting parts of working at DIU is that, while I formally sit on the AI portfolio, I work very closely with folks in our other portfolios like Space, Energy, Cyber, Human Systems, and Autonomy. On the AI front, I’m particularly excited about leveraging knowledge graphs to drive how we think about people, places, and things, and how we combine continually improving approaches to MLOps with low-power edge computing hardware to change the way that we deploy AI systems. Outside of AI, I am particularly bullish on leveraging technologies like small remote sensing satellites, renewable energy systems, and advanced manufacturing to simultaneously improve operational capability, enhance resilience to climate change, and reduce the public sector carbon emission footprint.
What are you reading right now?
Over the holidays, I managed to catch up on most of the International Conference on Learning Representations papers that had been on my reading list. I just finished Aftershocks by Colin Kahl and Thomas Wright, an excellent book about the effect that the pandemic has had and will continue to have on geopolitics. I’m just getting started on The Dark Forest by Liu Cixin (the sequel to the Three-Body Problem).
What defense challenge/commercial solution are you working on right now?
In my role as Technical Director for AI, I have anywhere from 10-15 individual projects that I am tracking at any one time. Applications range from predictive maintenance (with the Air Force) to knowledge graph construction (with the intelligence community) to combatting illegal fishing with computer vision (with the U.S. Coast Guard), and I have spent a lot of time on this last one – including a good number of coding hours – via the xView3 machine learning prize challenge that just concluded this month. At a higher level, I also try to address common issues across our programs and provide some strategic direction for the type of problems we spend our time working on. In November, for instance, we released the DIU Responsible AI Guidelines, which represented nearly a year and a half of work to provide actionable guidance on how DIU aims to operationalize the DoD AI Ethical Principles. I am also currently working to put forth an architecture for MLOps within the DoD that strategically leverages the latest commercial technology to improve the efficacy with which DoD can rapidly iterate on machine learning models it aims to deploy.