Can AI Beat the Pentagon’s Bureaucracy?

Jordan Schneider (0:00)

Garrett Berntsen, last seen on ChinaTalk in his role as Deputy CDAO at the State Department. He went to the then Department of Defense for a little bit, now makes a triumphant return in his new role as Chief AI Officer at Accenture Federal Services. We're going to talk what the national security community needs, besides just API access to wonderful models in order to really change how it does its mission. All right, welcome back. Thanks for having me. And thank you to Accenture for bringing us this episode. All right, U-2 flies over Cuba, 1962

Why is this a parable we should recall as we're thinking about how to deploy AI in government?

Garrett Berntsen (0:49)

So the U-2, I'll tell the quick sort of Cliff Notes version of the story of the U-2. So built in a classified program with government buyers, government money, giving money to industry and basically unleashing a whole bunch of innovation around, you know, sensors and high altitude, you know, and they build everything new, like only all the pieces of this dish together, incredibly innovative and done in a secret sort of lab, right?

Fast forward and done very quickly, also sort of famously, if you fast forward to the Cuban Missile Crisis, suddenly we have overflight imagery of missile sites in Cuba. We're getting really good raw intelligence in a way that, you know, I think the Russians didn't know we were getting and was kind of incredible and miraculous to US decision makers. They're actually seeing changes on the ground over time. So, you know, you hear about this technology at the plane, which by the way, I was looking at some recently, still in service, actually, incredible. It was that sort of revolutionary and remains really valuable still. I think they keep talking about decommissioning it and they decide not to, you know. So the technology, of course, incredibly important and the innovation to put it all together. I think the reason I like this metaphor, though, and why I think it's relevant for today, why we're going to use it to talk about AI, is it's not just the technology, not just the plane. It's the way they bought it was important. The innovations around acquisition reform, the way they acquired it was important.

The way industry and government worked together to do that. And then the way that the information coming out of this plane were actually analyzed, institutionalized through intelligence analysis, right? They built a whole new intelligence sort of service, the NPIC, which eventually turned into NGA, right? It was like the National Photographic Interpretation Center turned into NGA later. How they built that whole chain of judgment and then probably most importantly, like put it in front of the president, right? And other decision makers. That is kind of the miraculous piece, I think, of the story and now sort of pivoting to the AI moment. I think we're slightly in this moment where the technology has actually moved so much faster because so much of it is from commercial, not managed in a government program. It's from industry. It's moved so far ahead of our ability to sort of integrate it into the actual systems, into the workflows, into the decision making processes that we now are kind of flat footed in this reality of this technology.

And that's a lot of what I work on my company. I worked on it when I was in government is trying to think through those other pieces that need to change, not just the AI. So that's why APIs alone are not enough.

Jordan Schneider (3:39)

Yeah. So let's spend a little more time on the U-2, right? So what were the technological unlocks that you needed to take that photo and realize that the missiles were coming on to Cuba? You needed new lenses, you needed a new canister, you needed this plane that...

Garrett Berntsen (3:57)

High-altitude flight, right? I think at the time, no one had ever flown that high. So even just the equipment on the pilots, I think they also accelerated at the time.

Jordan Schneider (4:09)

Can you even see it from the ground? Like, no. Yeah, because Eisenhower was doing it for a while over the Soviet Union before they even realized. Yeah.

Garrett Berntsen (4:17)

So, you know, all of those... I think the other piece about it that's so incredible, again, we have to remember, like, pace and speed are...

Doing things in pace and speed in an engineering way is, like, actually very hard, right? We're much more used to, hey, we do a first test, and then we do a second test, and then we collect data, and then we go back and refine it, right? And so, and that's... I'm not saying that's the wrong way to do it. That is, you know, that's how you do engineering, and especially if you're gonna put pilots into life-threatening environments, you have to do that kind of stuff. But the ability to move more quickly through that and to take more risk, you know, is a part of the magic of that technology. Again, back to now, I think that's why we are seeing such a focus on doing experimentation, pushing experimentation as far as we can, and senior leaders saying, hey, I'm willing to accept some risk here because the mission is so important and we need to take a little bit more risk to try these technologies out in more operational environments.