The Defense Advanced Research Projects Agency was created in 1958 in the aftermath of the Sputnik launch, part of a determined effort by the Pentagon to ensure that the United States and its military always have access to cutting-edge technology.
The agency now has an annual budget of close to $3 billion, which over the years it has used to develop such major technologies as synthetic biology, carbon nanotubes, and the Internet itself.
Darpa’s director, Arati Prabhakar, a former venture capitalist and Darpa program manager, hosted The Chronicle in her office recently to discuss the agency’s interest in working with university researchers not only in traditional realms such as engineering but also in newly expanded initiatives into areas that include the biological and social sciences.
In the second part of our interview, below, Director Prabhakar discusses issues related to cybersecurity.
TRANSCRIPT OF THE FIRST VIDEO
PAUL BASKEN: Hi. We’re here today Darpa headquarters in Arlington, Va., with Arati Prabhakar, the director of Darpa since 2012. Darpa funds a lot of research, both basic and applied, certainly in the university community. I was wondering, Dr. Prabhakar, if you could tell us what basically is the relationship between Darpa and research universities this country?
ARATI PRABHAKAR: Darpa is reliant on research universities as one part of this huge ecosystem that we get to work with in pursuit of our mission at Darpa. Our mission traces back to our founding. In the wake of Sputnik’s surprise, Darpa was created to add to this growing community of federal R&D that was building in the post World War II era, of course. But Sputnik was the recognition that we needed a place whose day job was to prevent that kind of technological surprise.
And now in nearly six decades of creating technological surprise in pursuit of our mission, at Darpa we find over and over again that the only way that we can do that is to draw from the deep foundational research, almost always at places like great universities. And so they’re a vital part of our community.
PAUL BASKEN: And what, if anything, are universities doing in terms of the relationship with Darpa that you maybe could see them doing differently or better, that would be more productive both for them and for Darpa?
ARATI PRABHAKAR: Well, I think the relationship that we have, I think is strong and vibrant. That’s important to us to keep it that way. The parts of universities that have traditionally worked with Darpa, I think historically it’s been more focused, more exclusively focused on mostly engineering disciplines. I know, certainly, my background in electrical engineering and applied physics, in those areas and computer science, those are places where I think faculty know Darpa.
Even if they’re not working with us, they know what role we play in the ecosystem. And most importantly they know that when they’ve got – when they’re chasing a really big idea that could be transformative, they know that we could be a tremendous partner for them. And my hope is that we will build that understanding in some of the areas that have not been traditionally part of Darpa’s portfolio, but that I think today are actually very, very important.
A couple big examples of that: One is in the biological sciences. So in biology, and bioengineering, and biomedical engineering, and the medical research arena. Those are places where today we have a biological technologies office. They’re out working with, but also trying to get the message out, to that broad community. Again, not for everything that goes on in that research area, but when people are chasing a research idea that could become the basis of a big new technology capability, we hope that they’re going to come knock on our doors.
So that’s one area. And another that’s even newer is what’s – I think the opportunities for new technology and capabilities for and coming out of the social sciences is going to be very interesting area, as well. We just recently started a program in that area called Next Generation Social Science. We’re right in the middle of getting great ideas in the door for that. But I think those are examples of communities that I hope will come to understand Darpa in their ecosystem.
PAUL BASKEN: OK, as opposed to some of the agencies like NIH and NSF that university researchers are maybe more familiar with, Darpa is a bit different in that its program managers have a lot more power it seems, to decide a project, to set up a project. Can you sort of tell a university researcher what that means for them and how they might approach it differently than they might approach someplace like NIH or NSF?
ARATI PRABHAKAR: Right. The way Darpa works with all of our performers, including university researchers, again stems back to what our mission is. And so when we go out with a program, what we’re trying to do is advance technological capability. Often we’re taking something from fairly early research to the point that we can actually show a technical capability. And the reason we’re doing that is because we think that it’s an area that, perhaps in the near term, perhaps a long term, could have a transformative effect on national security.
And a consequence of that mission-driven research that we do is that the programs are very actively managed. Our program managers have great latitude in crafting these programs. We really look to them to define programs. And Darpa’s a very bottoms up organization. Every program that we run is the invention and the creation of an individual program manager. And then as that individual starts running that program as well, they’re going to be building a community that involves – often we find we’re able to attract the most amazing minds in the field. That’s always our objective. And that program manager’s job is to gently pull that community together and help them achieve something that’s phenomenal.
So it’s very active program management. That means year to year when we see – a program manager sees things that are making great progress, I will put more resource on them.
If we’re in a dry well, and it’s just not really panning out the way we and the researcher hoped, we’ll stop. And I think that’s just very, very different than, if you like, sort of the classic grant model. That’s a super important part of the ecosystem, and we are often building on top of research that has been funded by NSF or NIH. Our job is different.
PAUL BASKEN: Do you find university researchers generally get that and are able to sort of adjust their expectations and their way of working to fit that? Or do you find that because of the difference between what you do and other federal agencies do that it kind of makes it hard for university researchers to figure that out?
ARATI PRABHAKAR: I think expectations is the key word. And we try to be really clear about what our expectations and our needs are going in. And I would tell you that when that relationship works with a university or any other performer, it can be the most exhilarating time in your entire professional life. And of course, there are times when it doesn’t work. And I think it really comes back to being clear in our communications and expectations with each other, up front.
And, again, I think in my mind nothing that Darpa does should be viewed as a substitute for the much different kind of work that’s just about making sure that we have this robust research community. Our interest is really finding the individuals, the ideas, in that broader research community where we can team up and go forward in a huge way and start building technical capabilities.
PAUL BASKEN: Given your relatively high turnover in program managers, can you describe what you do to recruit from the ranks of universities? First of all, how many of them generally come from universities? How do you find them? What does a university researcher who might be interested in doing it, do to approach you?
ARATI PRABHAKAR: Yeah. Darpa is completely reliant on this rich research and technology community for our program managers. And these are folks who come to Darpa typically for three to five years. They come – sometimes they’ll inherit a program that’s been started by someone else, but they typically come in to start a program that’s their personal passion. They do that by listening to the great ideas that are in the community. They do that by understanding what the deep national security and military needs might be in the near term and in the long term.
But then they really are the ones who craft these programs. So because we’re so centered on having phenomenal people, that recruiting issue is core to every – I think about that every day, and my office leaders think about that every day. We’re out working it all the time. And it means about 30 new people a year will come into Darpa, and about that number will be graduating and going off to their next adventure.
A subset of those folks are coming to us from universities, and we are able to bring people in basically on leave from their university so they can maintain tenure and maintain their position. Of course, because of the requirements of the program manager responsibilities of Darpa, it’s not a trivial thing to make it work, because they do have to put their research on pause and hand it over to other players.
So I think it’s not a trivial commitment of time for somebody. But again, we’ve found over and over again – first of all, we just deeply value the perspective and the depth of research capability that these folks, especially from universities, can bring to us. But likewise, I know for many, many people who’ve come, it’s been a life changing slice of their professional lives.
PAUL BASKEN: And do you typically go find them, or do they come to you and – what’s the more ...
ARATI PRABHAKAR: Both. But it happens all the time in both directions. We sometimes will have people who show up and say I really want to go bite off something big as my next step. I’ve done a lot of great things that I’m excited about, now I have a big ambition ago perhaps drive in a particular direction. It’s more than I can do in my research group. I want to be able to have a scale of impact that I haven’t been able to have before.
And conversely, as we are always out looking for program managers, we will look for people who are doing interesting things or imagining interesting things for the future.
PAUL BASKEN: I’ve seen some talk that maybe you are considering having maybe longer term, you talked about three to five years being the average length of a program manager. There’s some talk to maybe letting some people say longer than that. Are you thinking of changing ... ?
ARATI PRABHAKAR: It’s really a mix. And what’s very important to us as an institution is that we – again, back to our mission. We are designed to be a place that looks around the corner at the next technology that’s going to be highly disruptive. We want to not be surprised by it. We want to drive it. If we’re going to do that, we need to make sure that are cadre of program managers, typically about 100 people in the agency at any moment in time, that entire group is always flowing through at a pretty high clip.
But, again, we don’t have arbitrary rules, and so there are times when a person will stay longer Occasionally someone will come for a shorter time. Three to five is typical.
PAUL BASKEN: As far as their to break down between basic and applied that you find roughly, how much of it is basic? And of the part is not basic, what’s your experience with universities being able to handle applied research? I know some maybe have difficulties in applied areas when it comes to military-related applications. Are you finding that a problem?
ARATI PRABHAKAR: Darpa historically, and also today, as it has over many, many decades, has been very roughly half about building radical new military systems.
Historically, think about Stealth aircraft and the technologies that allowed us to achieve precision strike. That half of Darpa has created those capabilities. The other half has always been, and continues to be, about core enabling technologies. And of course, historically, this is where material science emerged as a discipline from early work we did in late ‘50s and early ‘60s.
Think about what the very early work that led to the VLSI revolution and MEMS technology. Think about artificial intelligence, and of course ARPANET and Internet. So that core-technologies part of what Darpa has always done has traditionally been the place where we’ve had the most intensive university involvement, a very natural fit to the kind of breakthrough research that happens in universities.
PAUL BASKEN: Thank you Doctor Prabhakar. I enjoyed the time today.
ARATI PRABHAKAR: Thank you.
TRANSCRIPT OF THE SECOND VIDEO
PAUL BASKEN: This is the second part of our interview today at Darpa headquarters in Arlington, Va., with the director, Arati Prabhakar. Welcome, Dr. Prabhakar. Just thinking about what Darpa does, I mean obviously military, but increasingly we’re seeing a lot of the – you mentioned them, some of the applications they come out with end up having commercial applications.
ARATI PRABHAKAR: I’ll say.
PAUL BASKEN: Obviously, a lot.
ARATI PRABHAKAR: Huge, off-scale ones.
PAUL BASKEN: And maybe more so, I guess, what I’m getting at is, to what degree is that suggested that Darpa, as it moves into the future, will somehow become more of a mixed mission between military and nonmilitary? Or is that not in the offing? Or how does it somehow – how does Darpa somehow deal with the fact that so much of what it does really does have a nonmilitary use?
ARATI PRABHAKAR: We celebrate that. And in fact, frequently for us to achieve our national security mission, the commercialization of the technology has to happen. A really great current example is in the area of cybersecurity. Just as a recent example, we have – one of our many cybersecurity programs funded a lot of very creative university research. Really trying to ask questions about, can we come up with fundamental new approaches that would start taking whole classes of cyber vulnerabilities off the table?
One of those research projects at a university led to a spin-out company that had a product that now has been purchased by – that product has now being used by Hewlett Packard. And so now when you buy a printer, one of the very wide open doors for an attacker has been closed because of that university research finding its way into the printer. It’s a small piece of a much bigger cyber problem, but as we knit that attack surface closed, that’s going to help cybersecurity for anyone who buys a network and a printer. That’s what DOD relies on for our own cybersecurity as well, because we buy the same kinds of equipment.
So, I see this very natural symbiosis between the work we do in funding core technologies for national security purposes, and then frequently the commercialization path that they take with vastly more private investment that follows after we light the spark. And if you look historically, of course, in the information-technology realm, that has created not just products, but businesses and whole new industries. When I look forward, I think it’s really interesting to see, not just some of the work that we’re doing today in information technology and cybersecurity and data analytics and big data tools, but also increasingly now in areas like bio and biotechnologies.
Again, you can sort of see the same pattern emerging where we’re trying to do things that are a little crazy, but we get them far enough that they can spark commercial opportunities. So I see all of that as completely knit together around our national security mission. And because we’ve had such a strong track record of delivering on those core technologies, as well as military system,s by keeping a tight focus on the national security mission, my expectation – I think it’s actually very important for us to keep our focus on that.
PAUL BASKEN: You mentioned cybersecurity. That’s obviously a big category of attention among U.S. research universities these days. They’re teaching students left and right about cybersecurity. I guess the question I have for you is whether, one, you think they’re going about it in big picture, the right way? Or is it maybe a little too cookie cutter, some of the way that universities go after cybersecurity? We spoke with one of your counterparts at the Pentagon not too long ago, and they thought maybe we just really needed to sort of break out of the box a little bit more in the way we’re looking at cybersecurity. Maybe start almost on the ground level a lot more with the way computers work. I’m just curious what your take is on that.
ARATI PRABHAKAR: Yeah, I’m not an expert on the education side of how we’re training people in terms of cybersecurity. But absolutely the research focus of the work at Darpa in cybersecurity has been to get us beyond patch and pray. And really, that’s all we’ve had for a very long time now. We’ve been living in a world in which we all keep buying and using more and more information and information systems, because of how incredibly valuable the information is and the information systems are. But of course every time we do that, we just expose ourselves, give ourselves a bigger attack surface. Our question has been, can we take whole classes of vulnerabilities off the table? Can we come up with techniques that scale faster than the threat scales? And in fact, I think, we’re making huge technical progress with both of those.
Just one example. This summer, we’re going to be running a Cyber Grand Challenge. It’ll be one of the next big Darpa challenges. This particular one asks the question, if we build a league of their own for machines to compete in a capture-the-flag kind of competition – today, that’s a human game for amazing cyberwarriors essentially to do cyberwarfare in a box. We’ve created a league of their own for machines to do what is currently very intensive human reasoning about where vulnerabilities and flaws are and penetrations might be in a network, and we’re very excited to see what happens with that.
We’re going to run that competition the day before Defcon, where the big human competition happens, and I think we’ll start to see the rise now of automated systems that can help with cybersecurity. They’re not yet at human level quality, but you can easily see those new technologies starting to become ways that cybersecurity experts can dramatically amplify their own capabilities, and you start to see how you might be able to keep up with the pace at which the threat is growing.
PAUL BASKEN: How about foreign students at U.S. universities? The numbers keep growing. Is that a concern for you? Does it make it harder for Darpa to work with U.S. universities, or what’s your basic. ...
ARATI PRABHAKAR: Yeah. I think this is an issue for the country because it’s been a traditional, phenomenal strength of U.S. universities that were able to attract students from around the world. From a national security perspective, of course – first of all, for unclassified research, which is overwhelmingly what happens with our university funded projects, that’s not an issue. From a larger national security perspective, I think obviously there are real questions. Because so much of how our national security enterprises work is built around this post-World War II model where the U.S. was the unquestioned leader in every scientific field, and if we did it here, we were pretty sure no one else could do it. And of course, that’s not the world that we live in anymore. So I think your question is, in a sense, the tip of the iceberg of a much bigger global shift that’s gone on.
PAUL BASKEN: Right. Speaking of those kinds of questions, I’m curious if you have any thoughts on the whole thing that was going on between Apple and the FBI. Not that particular case, but just generally what it raises as far as the question in society of privacy versus the need for security.
ARATI PRABHAKAR: Yeah.
PAUL BASKEN: Is there a solution? Is there a technological solution to that dilemma we face as a society?
ARATI PRABHAKAR: We think that technology can help, but I think you’ve really put your finger on one of the most interesting aspects of what we work with here Darpa all the time. Whether it’s big data and privacy issues or synthetic biology and security and safety issues, almost inevitably, if you’re going to work on powerful technologies, and that’s our mission, if you do that, you’re going to encounter these very challenging questions that are really about how human beings are going to use this technology. And it’s something that’s front of mind for us, because it just comes up over and over again in the choices that we make about where we focus our investment.
And also because we frequently find that in the work that we’re doing, because of that research, we have a way of imagining what might happen in the future, ahead of the time that policy makers, for example, might be thinking about those challenges. In fact, I think it’s an important responsibility at Darpa. And actually, I think it’s true in the broader technical community, for us to think about those issues, to weigh them, to speak openly about them. One consequence of that is sometimes there will be technology elements to the solution. And I think you have to be very careful not to – you don’t want to say, well, we’re going to wave a magic wand and technology is going to solve the problem, because we know these problems are harder than that.
But specifically on the questions of privacy, we started a program in the last year of so called Brandeis, and it’s specifically about trying to build technology components that would – what we really want to do is change what is a very painful trade-off between security and privacy today. We want to find ways to give the owners of data much greater agency over their data out of a conviction that if they had that agency over their data, that they would want to actually to share it. For example, we’d probably all want our health-care records to be shared if we knew who was going to see them, how they were going to be used, how long they would be accessible, and know that they wouldn’t be published to the world. So, I think those might be some of the technical tools that allow for better policies and better practices in the future.
PAUL BASKEN: OK. Well, Dr. Prabhakar, thank you very much for taking time with us today.
ARATI PRABHAKAR: Thank you.
Paul Basken covers university research and its intersection with government policy. He can be found on Twitter@pbasken, or reached by email at paul.basken@chronicle.com.