IT/Networks

DoD’s top IT official talks AI, JEDI, and how to win over the commercial tech community

When Pentagon leaders talk about the tools they will need to win future wars, two capabilities consistently come up: the cloud and artificial intelligence.

The official leading those efforts is Dana Deasy, the Defense Department’s chief information officer. The former top IT executive at JPMorgan and BP, Deasy is leading what some might call a transformation at the Pentagon to modernize enterprise IT and better arm the war fighter for the future fight.

He spoke recently with Jill Aitoro, executive editor of C4ISRNET and its sister brand Defense News, about the tech priorities for the department — and how that Amazon protest of the department’s enterprise cloud contract award impacts his plans.

C4ISRNET: Since you’ve started, what areas have needed your undivided attention?

Dana Deasy: It started way back when I first interviewed with Secretary Mattis, and then Deputy Secretary Shanahan. They kept referring to this national defense strategy and gave me a copy of it to read. There was something that was going to need to be done obviously around the lethality space.

I realized, "Hey, cloud was going to need to be a substantial part of our future.” It could handle the need for speeding up the speed of procurement of infrastructure.

The other thing I noticed was the way that we were building software was very much the way that we had been building software for the last 25, 30 years. Cloud really forces you to think differently about how you want to build applications. So I saw, hey, there was an opportunity for speed of procurement. There was an opportunity for speed of delivering applications. Then finally, it really hit home for me when I took my first trip out to Afghanistan. I always knew that cloud could impact the tactical edge, but it wasn’t until you went out to the tactical edge and you sat down with the war fighter and they showed you what they had to do and the various systems, the data that they had to pull together.

I came home from that trip and basically said, “We have to double down on cloud. We have to do this as fast as possible.” The fourth reason I thought cloud was going to be important, and I kind of went back to my private sector days, for AI, artificial intelligence, for it to really work, it needs three fundamental things. It needs massive compute power. It needs the right data, and you need to have the right algorithms.

As I looked across Department of Defense, it became very clear there was no large scale infrastructure that was available for all the services to quickly take advantage of that they could start to port either experimentation of AI onto, or fielding AI onto.

C4ISRNET: So how disruptive is the current state of your cloud procurement JEDI?

Deasy: We’ve obviously been through this very long drawn out process now with the protest. We’ve gone from GAO to federal claims court, to the IG. In all cases, we’ve come out in a very positive position, which we always felt very strongly we would. It has been really exciting now to finally be able to focus on [starting] to build the cloud capabilities. We recognize that we have to do this alongside of the protest that’s going on with Amazon right now. I’ve told my team, “Let’s not get distracted here.” We have a set of people that are focused on working on the protest, and we have a set of people that wake up every day now and are just getting ready to stand up our unclassified environment, which we hope will be ready by mid-February.

Then our secret environment is approximately six months thereafter. Then our top secret … a lot is going to depend on not only how the protest plays out, but how well and quick and smoothly the unclassified comes up and then the secret comes up.

C4ISRNET: What I’m hearing from you is more, “We need to grasp it now and we need to figure out where it fits and how it can be rolled out in an appropriate way.” Kind of kicking the tires.

Deasy: We are spending a lot of time now getting ready to stand up our first “unclassified” JEDI environment. It’s going to have a whole variety of services. What do we need, what’s the bare minimum we need to be able to stand up a real capability that the services can use come mid-February, but it won’t be the full complement. We will add on more and more services as time goes on. This is a really good example of not waiting until you have a whole complement of services done, but let’s get a capability out there that the services and the co-comms can start to use. We’ll learn about what they need second, third order, fourth order. AI’s the same way. We are putting out there first generation, second generation solutions where the data’s going to get more refined. We’re going to bring in better data and we’re going to refine the algorithms but we’re not trying to get the high degrees of perfection. That is new for the department.

C4ISRNET: I think there’s been confusion in terms of how JEDI ultimately will be used, and whether this is primarily enterprise technology or also for military on the ground. What’s your vision?

Deasy: What the Department of Defense has never had is an environment where you can stand up, unclassified, secret, top secret data. This will be the first and only environment inside the Department of Defense. It’s not three distinct environments. We’re building what’s called cross-domain solutions, and this is the ability to move data, ensure it sits in its right classification. Remember that problem, what I saw in Afghanistan, where they were trying to bring in secret data, unclassified data, top secret data from these disparate standalone siloed solutions? To the war fighter, all they want is what it takes for them to make a decision.

C4ISRNET: Were there any lessons learned?

Deasy: People grabbed hold of [the claim] that this was a 10-year, $10 billion contract. Which has never been the case. This was a two year initial contract of which the only money we put out initially is $1 million. The rest is all dependent on how fast Department of Defense starts to consume the capability.

I’ve seen cloud technology in the last 10 years dramatically change every two to three years. It became very clear to me we needed these breakpoints in the contract to be able to reevaluate.

By the summer of 2018, we [released] the DoD Enterprise Cloud Strategy, which said all of this mess, all of these individual clouds were not a winning hand. I have said from day one we are a multi-cloud environment and we will continue to be a multi-cloud environment.

C4ISRNET: How will there be opportunities for other companies?

Deasy: There is no such thing as one technology that can serve all the diverse needs of the Department of Defense. So there are cloud contracts that have been stood up already inside the Department of Defense. Navy has one for their SAP environment. Air Force has one that they call Cloud One. There are ones that we’re evaluating right now with other vendors that we’ll be making awards on that will all be serving specific needs. We are going to need more than one cloud provider to satisfy the needs of our unclassified environment. It’s too big, it’s too broad for any one vendor to handle.

C4ISRNET: What’s your role in having a more productive relationship between the tech community and DoD?

Deasy: Over the years, we’ve seen a tremendous acceleration of innovation take place in the private sector, to the point where the DoD clearly recognizes that this is not something that we can do on our own, ignore the outside world and just build great innovation.

[The Joint Artificial Intelligence Center] is working with R&E, working with the academic world, working with large established technology firms, and working with small startups.

If you look at the projects that we have underway right now, some of the origins of the algorithms come from the academic world. Some come from small startups, and some of them come from the larger technology players. So I think AI is going to be a space that’s going to accelerate and help to mature how we bring all of this together.

C4ISRNET: Rather than a lot of the small business contracts, can’t DoD put more money into fewer? Invest in 10 rather than 50, for example, but give them bigger contracts to build on?

Deasy: That could be a little shortsighted, because as soon as you start to do that, you may be ignoring three or four or five brand new ones that are coming up that could be doing something really special that the DoD could truly leverage.

The things that government as a whole needs to learn how to do better is how do you bring in small startups. How do you in the procurement processes make sure that you don’t find yourself in these elongated protest periods? How do you get them in the door, on the books so to speak, and allow them to start to innovate? Then finally, even if we could bring them in quicker, and we can set up the right environment for them to innovate faster, we have to solve for how quickly we can field. And field the next and the next. AI is going to force us to solve for that, because you don’t roll out a solution to the field and then walk away and go to the next thing. You are going to constantly reiterate with new data and new algorithms. Small companies are very, very good. They are very adapted to this highly automated rollout, and we’re going to have to learn from them more and more.

C4ISRNET: Can we go back to what you saw in the National Defense Strategy?

Deasy: Inside the NDS is a really interesting statement that says artificial intelligence could change warfare. Under Dr. Griffin in R&E, they are doing some amazing things in basic research of AI. So are a lot of the services. But when you look across the DoD as a whole, what you didn’t see was fielding being done other than in small pockets. There was no attempt to do fielding at large scale and to take that fielding and be able to scale it up across multiple services.

So that is why we created JAIC. If the department is really going to be serious about building and being a world leader on AI, then we have to practice and learn how to accelerate fielding of AI and be able to do that at scale.

C4ISRNET: How is that moving along?

DEASY: We’ve gone from an organization that you could count on two hands to about 110 people today. That is now delivering its second and third sprints, meaning its second and third iterations of the HADR and the preventative maintenance that we talked about before. We are ready to start our first lethality project in the joint war fighter targeting space. We’re looking to do something in the health space as well.

C4ISRNET: Can you give any kind of preview on what we might see or expect to come out of the IT budget for the Pentagon?

Deasy: The spend is about $47 billion. What we’ve done this last year is we’ve put out guidance for the first time that specifically called out the digital modernization program and said in the matter of cloud, AI, C3, and cyber, these are specific things that we want you to put money aside on.

That’s been delivered to the field agencies. That’s been delivered to the services, the combatant commands. So it’s really exciting, because now for the very first time [we are talking] a common language about delivering against a comprehensive strategy.

C4ISRNET: Can you give an update on what DoD is doing to fortify security in the supply chain?

Deasy: When we find companies that we have clear evidence that they’re problematic from an adversary standpoint, we have the ability today to have them stand down. We are starting to take advantage of that. There is this whole complicated problem: how do ensure as you move from your first tier to your second tier- down that chain. Is there technology we can apply to that? I know that this is work that folks like DARPA are looking at. We’re in constant conversations with the NSA. How do we use their tools and their art craft to help us evaluate? We don’t have all the answers. This is the space that I think we’re going to have to spend a lot of time iterating and get smarter and smarter on.

C4ISRNET: There has been talk about DoD embracing more of an agile approach to technology. But others say procurement doesn’t necessarily reflect it. Have you made progress?

Deasy: For years, we have had this very well thought out rigorous process of how you define requirements, how you prototype, how you build, how you test, and how you deploy. When you start using the term and the phrase “agile” all that comes together in a much more compressed set of sprint approaches. It is a different mindset and it is a different set of technical skills you have to be able to use to deploy that.

We have great examples inside Navy Research, inside, for example, Kessel Run over in the Air Force, where we’re starting to show that we can move from small pockets to start to do this at a lot larger scale. We’re going to have to also solve for how can we test and field faster. So for me it’s a procurement issue, it’s infrastructure, t’s culture, it’s training. All these things have to be brought together.

C4ISRNET: Can you talk about the potential for 5G in supporting DoD infrastructure?

Deasy: This is now the third leg of the digital modernization command and control and communications. What you’re really trying to solve for is if the fight needs to occur and you’re bringing together all the services and the right assets, that you have to make sure we’re going to be able to fight through. If we’re spoofed, if we’re jammed, if you’re fighting through a highly contested environment, how do you make sure that’s going to happen? And how do you make sure that as you’re using new technologies where you’re going to use full motion video, where you’re going to use AI, where you’re going to use massive amounts of data, you’ve got to push that out to the tactical edge. You not only need secure, robust communications, but we’re going to need bandwidth.

This is why people are so excited about 5G. It’s the potential of improved throughput and bandwidth, but we’ve got to get the security right. But the use cases are really exciting, and this is why the work that R&E’s been leading through Dr. [Lisa] Porter [to define] the right way to start our experimentations [is so important]. So the work that we’re going to be doing with dynamic sharing, which is one of the pilots, and the work we’re going to be doing with the warehousing, really will start to allow us to understand — can we truly deliver on all the promises of 5G? What if any limitations are going to exist? What new type of security are we going to have to wrap around it that may be very different than we’ve used before?

C4ISRNET: Is there anything else that you feel is important to mention?

Deasy: Every time I talk about the digital modernization, people tend to focus on a part of it. Like the cloud or the AI or the cyber. I truly believe that people will look back on it and say it was truly the integration. It was the fact that we had the foresight to think about massive computing through a cloud, that we knew that AI was going to be a very powerful and very much a game changing technology. All of that doesn’t matter if you cannot get it out to the war fighter. So this next generation, command control communication through things like 5G is really important.

And we can never lose sight that we have a lot of technology that we have used for 20, 30 years that was never designed around a cyber world.

In a perfect world, we would’ve done cloud years ago. We would’ve had an enterprise cloud already in place. But we don’t, so now we’re needing to stand up an enterprise cloud while at the same time learning how to do AI at scale, while at the same time doing next-generation command, control and communication. Yes, it’s risky. But I firmly believe we don’t have a choice.

Recommended for you
Around The Web
Comments