Here’s an insightful episode of the Defense Unicorns podcast with Artem Sherbinin, a Lieutenant in the US Navy who was integral to the first applying machine learning to effect surface warfare kill chains called Task Force Hopper.
He discusses the numerous difficulties encountered along the way where he learned how hard it was to get an Authority to Operate for simple things like data labeling software, encountering pushback from resource sponsors to get funded through the POM, injecting feedback from users, and finding a place for Navy software billets and software factories.
There’s a lot there so definitely listen to the whole thing. It’s most interesting because you have an obviously bright and enthusiastic young Lieutenant who clearly knows about his monetary value in the private sector. What are the aspects of Navy culture at the intersection of technology and culture that make good people want to stay? That was essentially Rob Slaughter’s question to Lieutenant Sherbinin:
Very simply, the shortest answer I will give to this whole conversation is: Less work on the problems we’re most passionate about — with other people who, I hate to say it, but move fast and break things. Every command that I’ve been to that’s a traditional ship — very old way of doing things that aren’t exactly that most exciting. Going to sea, that’s exciting, but the business stuff is not.
At a place like Hopper, the sky was the limit. I could make whatever I wanted and we felt like we were pushing the envelope of what was possible. If I could come into work every day and have that excited feeling, that I was going to break something and that was going to be rewarded, I would stay in this career for the rest of my life.
I think what he’s saying is this: I will accept lower monetary returns in exchange for a system where I can contribute to dramatically accelerating surface navy kill chains. Indeed, the opportunity to make that impact for national security is relatively clear, as Sherbinin said the capability of AI/ML to perform certain tasks is known to be technologically feasible. But the enterprise tooling to enable it is still being worked on and under-prioritized.
Sherbinin notes that the Navy’s premier software factory The Forge is located under PEO IWS, an acquisition office. He contrasts that to the Army and Air Force. Of course, the Army’s software factory in Army Futures Command.
But I’m not sure if there’s a single pattern for the Air Force’s several software factories. Kessel Run seems to have transitioned to PEO Digital in the Air Force as its budget has matured into a Program of Record. Platform One seems to be a direct report to the Chief Software Officer. Kobeyashi Maru was Section 31 of SMC, so like the early Kessel Run it was part of a systems command and not a specific PEO. Ski CAMP is at Hill Air Force Base in Utah.
Back to the point — The point of budget reform is to increase discretion of those at the lower levels to use their local knowledge to impact change. But senior people may scoff at such “true” delegation of program choice. They say the lower levels don’t have the capability or incentives. But I think it is clear that the United States military has access to smart, dedicated, and honest people. The current processes are stopping them from fully expressing their potential.
Be sure to check out all four of Sherbinin’s article in the US Naval Institute.
Leave a Reply