I can say,  broadly speaking, the USG drives national strategic investments in AI to uplift the nation (American workforce, industry, policy and legal matters, economy etc) AND uses AI at individual project scales to derisk threats and advance new solutions against individual agency problem sets, such as foreign intelligence (NSA), accelerating scientific knowledge (DOE), and nuclear weapons security (DOE).  While AI is impacting all sectors, it’s impact on government may have the highest potential. Big data feeds AI/ML, and the USG has some of the largest datasets in the world. The USG is actively extracting knowledge from these data to stay ahead of adversaries, and  solve pressing challenges such as human health, nuclear security, and fundamental science. Big data + high risk critical challenges = awesome place for those wanting to build some AI muscle and put it to good use.

Let me share what it looks like working national policy and strategies. At DOE, together with partners at National Institutes of Health (NIH), Veterans Affairs (VA), and National Science Foundation (NSF), we worked to develop policy and programs to advance the US workforce, the direction of US R&D (research and development), and integrate AI/ML into solving agency mission specific challenges. To do this, a general process looks something like this:

(1) Focus on a top challenge for the USG, for example “healthcare for Veterans.” Knowing and choosing a top challenge is usually directed by senior leaders who are informed by congress, presidential priorities, and extended agency experience.

(2) Build recommendations  to support meeting the challenge.  Do this via national and global expert input. This was by far my favorite part of the job. The global scientific community opens up to you, because they know you represent the USG when calling. While there are many ways I gathered input, one standard practice is to use workshops. Set the topic, agenda, invites, location, dates, and the post-workshop writing team. I helped lead and participate in many workshops over my career at  DOE, here is an AI Example: 2019 AI for Science Townhall.

(2) Combine expert community findings, document in public report, and layer on top of that recommendations on what the USG should DO about it. Such as “fund research program X”. Working with a whole team, we used DOE-VA workshop to create “MVP CHAMPION”, partnering AI, big compute, and VA data.

(3) Gain support from decision makers. For example, here is congressional testimony I co-authored in 2018 for DOE Senior leader, Dimitri Kusnezov, to gain support from Congress. This  was written in the dead of night, last minute, and helped gain funding for AI efforts.

(4) Manage the support. If you are successful, now decision makers have given you (and the community!) what you recommended. Now you have to manage it. In one case, creation of a new DOE Office of Artifical Intelligence and Technology! Pat on the back to me for making a new box in the org chart of USG- all for AI. Or you might receive funds to support new investments across the laboratories and universities. Here is an AI one, I didn’t do this one-but I supported many like these.

If you are interested in working policy and national scales, I strongly recommend you consider the AAAS Science and Technology Policy Fellowship. Guaranteed they need EE and CS with AI experience and interests.

Working at NSA means I have to protect classified information and I, like most employees, err on the side of caution. So rather than me tell you -if you want the scoop at NSA, check out the newly released  NSA podcast “No such podcast”! Episode 5 on Large Language Models “The Cutting Edge of Classified: Research at NSA”