Michael J.K. Kratsios, acting undersecretary of defense for research and engineering, delivers remarks as the Defense Advanced Research Projects Agency holds a virtual Electronics Resurgence Initiative Summit and Microsystems Technology Office Symposium linking government, defense, academia and industry to collaborate on technical advances, Aug. 18, 2020.
Transcript
Now joining us on the virtual stage. Director of the DARPA Microsystems Technology office, Dr Mark Rostker. Welcome to the 2020 DARPA Electron ICS Resurgence Initiative Summit and Empty Oh symposium. This year, of course, we are conducting the summit virtually. But even though we will not be in person, we’re excited to have the opportunity to relate to you the goals and progress of your eye and empty Oh, and to explore the future of microelectronics innovation with you along with 1600 other members of the community. I’ll have a chance to tell you more tomorrow. But for now I would like to restrict my comments to a few words about this meeting. Two of the highlights of the summit will be the Technical Leadership Panel discussion, which is later today and the Executive leadership panel tomorrow, both with a diverse set of technology leaders. Other highlights include two fireside chats. The first of these is today and will include the Microsoft CEO, such a Nadella enacting Undersecretary of Defense for research and engineering, the honorable Michael Cretaceous. The second will be on Thursday where Armed CEO Simon Segers will chat with Mr Stephen well be the executive director and CEO of the Eye Tripoli and former assistant secretary of defense for research and Engineering. The plenary of sessions this year Organized around are four key themes. First five G and future R F Communications, second microelectronics, security and access. Third Artificial Intelligence, autonomy and processing and forth heterogeneous three D microsystems design, fabrication and packaging. Each session is structured with a plenary speaker, followed by empty Oh program managers and performers highlighting here I accomplishments and technology transitions in that particular area. There is also an exhibit hall with various virtual posters and demos as well as opportunities for discussion. And finally there will be over a dozen workshops, each giving a lot more detail on content about our progress and future vision. We look forward to a great event and future partnerships with you as we work together to innovate the fourth wave of microelectronics. And now it is my great pleasure to introduce my boss, the acting director of DARPA Dr Peter Item. Thank you, Mark. I promise computing systems for real world applications have occupied much of my professional life both commercial and the national security. Personally, DARPA was there for me in neighbouring a small university team to make chips with Moses and later providing high performance computers AI in the early 19 nineties. Now doctors working to ensure that the department defense always has ready access to leading edge secure microelectronic devices. DARPA launched er i in 2017 committing to spend $1.5 billion over five years in 2020 re out the midpoint, with outstanding progress to report. Results so far include the first integration of photonics with state of the art F PGA, the first three D posa made in domestic foundry and three orders of Magnitude performance. For what? Improvements in radar posting and push button open source. RTL toe layout software. This couldn’t be done without working with all of the major universities leading edge microelectronics companies, the dearly national labs and the many companies that is small, that make up the defense industrial base and, of course, great support from the administration and the Congress, who recognized the urgency and importance of continuing to ensure that the Department of Defense always has ready access to the leading edge secure microelectronic devices. Now it’s my distinct pleasure to introduce the honorable Michael Crisis in his White House role. He has long been an outstanding champion for US technology today and for the industries of the future. We look forward to even closer collaboration in his new and additional role as the acting undersecretary of defense research and engineering. Thank you, Dr Heineman, for that introduction. I am thrilled to be here. The Third Electron ICS Resurgence Initiative Summit. As this audience knows, DARPA has played a unique role advancing the semiconductor industry, both in funding groundbreaking research in convening the community to collectively support US technological leadership. National defense strategy issued by the secretary of Defense in 2018 particulates that in this era of great power competition, we must supercharge or innovative capabilities in every way possible to maximize or economic military dominance and promote peace and prosperity for all Americans. The semiconductor industry is at the root of our nation’s economic strength. National security and technological standing and semiconductors influence many other industries. The future as well. The nation that leads in mice Electron ICS development will determine the rules of the road for artificial intelligence, five g quantum computing, autonomous vehicles and more. As a result, the department defense has made Michael Optronics one of its highest priority technology efforts. American companies lead the world in semiconductor production with 47% of the global market share, and the U. S has industry dhobley ahead in semiconductor design and R and D, even a semiconductor manufacturing is migrate overseas. United States is one of only three countries currently capable of domestically manufacturing at the leading edge. But we cannot allow the continued attrition of our manufacturing base. Nor can we imagine that just because we lead the pack and designed today, we will continue to do so in the future. To maintain its current leadership position in the United States, commercial and our security sectors are forming meaningful partnerships that address share challenges and needs in both the short and long terms. To meet immediate needs, the department defense is focusing on enhancing security in the MIC Elektronik supply chain. The department is moving towards a new quantifiable assurance strategy based on the zero trust principles found in cybersecurity will allow us to quickly and safely build and deploy leading edge electron ICS technologies. This is a departure from the previous model security that severely limited our ability toe work with leading edge firms and demonstrates the department’s four looking approach to promoting security. We are committed to a robust and secure supply chain, advances American innovation and keeps the American people safe on the long term horizon. We must ensure that the United States is developing technologies that are leaps and bounds ahead of other nations is not good enough for us to be on par with our competitors. America must develop the very best of circuit design, manufacturing and packaging here. I is critical to this long term vision. Many of the initiatives investments are designed to advance the development of novel manufacturing capabilities and new architectures for chips that give our nation a strong competitive edge. The duty is dedicated to maintaining and accelerating the nation’s rapid pace of Mike Electron ICS innovation, one that keeps us ahead of our adversaries and of mounting technological challenges. I am encouraged that events like the ER I summit are working towards the short and long term goals which can only be achieved by enhancing partnerships outside of government. I’m further encouraged by the tremendous participation of our defense industry and the commercial sector partners building in transitioning exciting technologies more rapidly than our adversaries will determine the winners and losers in our new competitive era. It’s critical for the D. O. D in American industry to work together and meaningful partnerships to realize our shared leadership objective. I’m excited to be here to kick off the summit and look forward to the conversations to come. Thank you. Now, joining Mr Cross CEOs for a fireside chat The CEO of Microsoft Corporation Mr. Satya Nadella. Good afternoon. I’m so excited to be here with you, Sathya, for this this great conversation ear eyes, obviously a huge event here for us that at DARPA department defense and that we’re so we’re so lucky to have you. Um, I think one of things that I kind of wanted to kind of start kicking off on is this question off of partnerships. It’s absolutely critical the US government and in the private sector to really lean in on partnerships to drive critical breakthroughs in generally and technology. So I love to get a sense from you on kind of What is your view on partnerships? What have you guys been doing at Microsoft and wish we’d be thinking about here here in the U. S. Jeep. First of all, Michael, it’s a real pleasure to be with you. It’s really an honor to participate in this particular event. You’re absolutely right. I mean, if I look back, our entire Industri, our own company all exists because off the approach that institutions like DARPA and the Dogg took on, how you created even the modern Internet and all of what came after. So in some sense, I feel when we think about the partnership, especially spanning the public and the private sector’s, that’s I think the spirit we have to sort of get back to the reality is that the MAWR, that is technology that is moving freely across these boundaries, the better off both sides will be. It is. Sometimes people talk about how the commercial sector is making great advances, it mean, but at the same time, there are certain structural things on the private side or on the commercial side that are leading to that scale. I look back and even Microsoft’s own say data center footprint. In the last 45 years, I mean the fact that we have 60 plus regions all over the world with even the network infrastructure connecting, although that’s pretty unprecedented global scale. But that doesn’t mean we see everything that you all see and all of the hybrid computing requirements you have at the edge. Ah, that’s a use case. That’s pretty unique. So I think it’s that coming together off What’s that future workload? That is high ambition that’s going to drive computation in a different direction on how can technologies that were developed, perhaps for a different use case, be assembled together? So that’s kind of for me. Partnerships like the one that we have with the D. O. D. Are super important for us to, in fact, move the frontiers of technology forward. No, that that’s a That’s a very good point. I think we we constantly think about, you know, it is almost imperative on us here at the D. O. D to make sure that whenever we’re driving forward in any type of technology, that partnerships are such a core and key integral, integral part of it. She had shifting a little bit to the event here today, Um, through through your eye, it was important to realize is you know, we had a deal idea have underscored really. We all know the importance of microelectronics to to generally this theme of economic growth here in the United States. But it is also critically important for our military capabilities and also our national security. So can you maybe elaborate or give us a little bit of context for sure, Your perspective on this intersection between microelectronics and Nash security Domain? Yeah. I mean, the way I come out and Michael is a little bit about the overall system architecture, right? See, if you think about it. It’s lots of fascinating things that were confronted with, which is some off the cutting edge new software workloads, Right? Let’s pick what’s happening in natural language processing, right? The transformer models with many billions of parameters that we’re doing computations around for training. An inference require us to take a very different approach of the system level, right, especially in a time when, in some sense, Moore’s law is slowing down. Let’s say how do you start saying okay? But at the same time, we need more computation, not less. Ah, and then you say, OK, what what needs to be the innovation? What needs to be the innovation in silicon and in the microelectronics side. Then how do you pull that all to a system where the cop compute and storage and network all come together to support this new workloads on. Then let’s take the national security side how to do this all confidentially, right? And it’s got to be tamperproof s. So it’s not just about coming up with the system, But every part of the system needs to have the verifiability off the supply chain to the runtime capability. Even we take some inspiration from what we did, even on the Xbox. In fact, we learned a lot, you know, through the multiple generations of X box, to say, How can you make the X box a Tampa proved device, especially on the X box one? I mean, there’s some fascinating things, right? Which is we know, like, Okay, how does the phone where software not have access to the keys? Or how do you keep the entire got stack from the operating system all encrypted. We even measure, for example, the temperature of the CPU in order know that there’s nothing, you know, crazy going on so that taking that systems view all the way to even say what we’re doing with as your confidential computing right, which is where in use data is encrypted. So it’s not just a grass store in transit, but in use. You have encryption. So that’s the thing that I feel like if you have a system architecture that starts all the way from whether use brother application domain. But you can go all the way to the supply chain of your microelectronics. And can you say everything is tamperproof? Everything is confidential. That’s, I think, the approach that we will need to take Yeah, no after absolutely I think that’s that’s critical on is something that I think DARPA, especially your I r. Come on the front lines of sort of, of sort of thinking about it, kind of just double clicking a little bit into that into that kind of question. There’s there’s a lot of thought around this question of trust and ensuring sort of trust in in what the D. O. D. Is actually using in its weapons and everything else we build in the system. And I think that sort of reminds me and kind of allows me kind of the bigger question of trust. writ large and emerging technologies and how we can bring on not only sort of trust relating to the sort of very critical defense related operations that we do, but but broadly, this question up of engendering trust, emerging technologies that can actually make a real difference in changing the direction, economic growth and everything else for our for our country. So what are your views on kind of that trust that that trust indexer questions? Yeah. I mean, I mean broadly, if you sort of the way you sort of Fraser is actually absolutely the place to start before we get into the intricacies of any given technology solution, trust ultimately leads to sort of lower transactional costs, right? I mean, when you can trust something, you can use something with confidence, whether it be in the, you know, the D o D order, not consumer lies in both contexts. I think transaction costs at a societal level go up when you can’t trust something. So then the question comes down to technically, how do you build back trust and one of the principles and some of the principles off? Whether it is, you know, like transparency, can things be ratified. Can things be audited? Can things be explained? Especially even when we talk about a I do. We have the explain ability that goes with how these algorithms were. So I think that the core fabric, some of the system architecture points are made in the previous question are also important. But they’re not sufficient. I think we also need ah set of principles and a set of policies that allow us to be able to have that level of transparency that engenders trust, at use and in fact, another aspect. Like take even a I safety. Michael, I’ve been thinking a lot about it. There are two sides to it, right? I mean, even if you take all the things that were debating on facial recognition, there is the design time aspect of it. And then there is the runtime aspect of it, like in the design time. How can you trust that the data on which you’re training he’s verifiable. It has not been attacked by some adversary with some adversarial attack. So that’s sort of one thing you need to ensure, and then the runtime is that you’re not using something for the wrong purpose. And so both of these air different considerations that both need to sort of have both sort of policies and engineering processes and design. I could not agree more. I think we you know, we as administration have been thinking You’re careful around this, this question of a regulations for for a number of years and both on certain that that the global level on also, domestically, there’s been this very important conversation being had about you know, what kind of rules of the road we need to be put in place. At least that’s what even the federal state level around questions off trustworthiness, safety, robustness explain ability. And I think generally in the policy role you’re seeing is there is a unified consensus on achieving those goals. But as you say, I think the question is it’s in the details. You actually need to have a technical engineering solution to explain ability. So if we woke up and said, you know, going forward any time using our AI algorithm, it must be explainable. That presents it in an onerous burden on industry that they could never they can never achieve. So I think what we have trying to do. And hopefully a lot of a lot of the folks watching have been tracking, you know, even at at the president signed executive order a few years ago relating to to artificial intelligence. And one of the ask was for our Office of Management and Budget to put out regulatory guidance relating to air power technologies and industry. US has actually been that the first in the world, Teoh to kind of put out a framework which is now Alfred comment and hopefully have finally in the next few weeks or months. And this is a good opportunity to show that if we need to lead the world in artificial intelligence, it’s critical that we are four leading in the rules of road. So we can empower engineers to be working on the right types of of engineering questions and Continental loops back to the incredible work that DARPA is even doing here. I mean, they are doing incredible work, explain ability and lots of other programs for of ai next. And until we make those breakthroughs, we can’t actually deliver on a lot of the ultimate policy goals. No, absolutely might. In fact, that is one of the questions I was going even ask because you know whether it is the White House work you date or now at the D. O. D. One of the key things that you’ve always talked about is how do you bring what’s happening obviously in the government, what’s happening in academia and what’s happening commercially on have a set of policy guidelines as well as cooperation and partnerships that move us forward as a nation. You know what you sort of thoughts on that, like, What is your idea? Like you asked me the question on partnership. What’s your ideal partnership? Sort of construct? Yeah, absolutely. I think for us over the last few years we’ve been spending a lot of time and energy trying to design sort of federal level national strategies on achieving US leadership. In our core, emergent technology has been things like a I want a computing five G, for example, And in each of those that the general track we’ve taken is that in order to drive US leadership, you need to have sort of essentially four lines of effort, and each of those have a partnership flavor to them. So I think it’s kind of valuable kind of walk through them quickly. So I think that the first one which I think is probably most relevant to the arms we have today, is this this extreme focus on research and development. That is the number one thing we must do. You must increase budgets to support the next generation or Indian these critical fields and put in the hands of scientists engineers who will ensure that the next great breakthroughs happen here in the United States. And that goes back to the coordination question. You know, as as a in the federal family, we deploy capital either extra mural lee or intramural E. We have amazing national labs that deal. We. We have an incredible set of 60 plus laughs here at the here at the Defense Department. Really groundbreaking work is being is being done there. DARPA is doing incredible work. National Science Foundation, Department of Energy. So the role of the White House often takes is trying to coordinate these efforts to make sure that we’re trying to achieve understanding what ultimate goal is and driving R and D towards towards that direction on that requires partnerships because you’re doing it with university drink with the private sector, and they need to come to come together. The second line of efforts always around workforce, that I know that’s something you and I have talked about and you guys take things so seriously. It’s critical that we have a pipeline of American talent that can go in and be the next generation of researchers and scientists, Whether it’s our programs here at the d. O D around grants and fellowships and scholarships towards folks focused on STEM were supporting initiatives, department education. To do this, it’s critical that we kind of build build that pipeline. The third, which we touched on, is this is this regulatory question We need to remove the barriers to innovation in these particular emerging technologies and create frameworks. Allow these technologies to succeed. Sometimes I think of the world of emerging technologies is often a subset of technologies that are born in captivity. These are technologies that unless the government does something that cannot be commercialized and cannot be used by by the American people, think drones and autonomous vehicles think supersonic flights unless dio TNF a take action. Those those those particular devices are grounded. So for us trying to find ways to remove those beers and free those technologies require intense partnership with the private sector, understanding the nuances of those decisions. What the spillover effects will be and how we can help minimize those and last is around is around international partnerships. And I think this goes back to what we were talking to about about trust and about working with our allies, for the US to succeed and continue to be a leader in his emerging technologies. We must have our allies on board and a core pillar of our national defense strategy here at the D. O. D. Is around centered around our partnership with allies. And as we work through our issues with artificial intelligence, regulation is critical that our partners on board that train as well, having a difference that rules the road for our allies and us is not a passport for achieving leadership among Democratic nations. So we’re very keen on that, and I think like that, that framing of using kind of those lines of effort to pursue national objectives on emerging technologies a good, good way to go? No, absolutely, and in fact, in a lot of those policies are also going to, in fact, to your previous question. Also helped build more trust, especially things like the guidelines are on A I across both of the United States and its allies, I think, is only going to help more that I safety becomes something that is much more people can trust it. And so I think that’s fantastic to see. So one thing we I’m always curious to ask people who are much smarter than me is about is about emerging technologies in generally where the technologies or breakthroughs that you think are making the biggest impact in the world today. What were you sort of most most excited about? Your DVD? We have 11 modernization priorities that cover everything for May I Quantum five G. But I’m curious. Kind of what? What is on your mind, The use that could make the biggest difference. American people. Yeah, so maybe I would sort of say, three layers. So the first one I always sort of start with being a systems company at some level is what’s that next generation manifestation off distributed computing right, which is the cloud and the edge, However, to how is the distributor fabric getting even laid out Like I mean, I talk a reference that fact that we now have whatever 60 plus regions all over the world with all this connectivity between the region. So that’s a lot of compute capacity. But obviously it’s not just what’s happening in the cloud, but what’s happening in the edge, especially with five G rolling out with low latency compute. Being brought to Theo Edge, I think, is going to be revolutionary. In fact, the fact that you have this continuous fabric and your applications can take advantage of it. I mean, I go think about his homeland’s or the I’ve ass devised in the field being able to rendezvous with an edge node, you know, and then at the same time, through a satellite connection, have connectivity back to one of the cloud regions. What type of application functionality can be put on the front lines is going to be pretty breakthrough. So that’s one which is the evolution of distributed computing fabric. And every time somebody says, Hey, distributed computing is cool. You got to remember, you know, it’s also complex, you know? How do you debug a program? How do you sort of make it more fall tall. I mean, yes, you can make it more fault Tolerance, but debug ability is one of the fundamental challenges. So there’s a lot of good work that we’re doing it on the system side. If it The second aspect, of course, is you know, we are running through some limits on, especially the Moore’s law. So what’s that next generation system architecture that supports these AI breakthroughs? Right? We’re excited about these transformed models and what they’re doing with language, understanding and language generation are in fact, the multi mortal models, right? Which is you combined speech, vision and language on that, you know, it’s just, I think, breakthrough work or even simulation. Uh, you know, in fact, even one of the DARPA subterranean challenges was using our air. Sempra, you know, God capability. I think the CMU team, along with the Oregon State team, won that competition using the ability off this compute fabric to be able to do simulations. So very excited about what’s happening with the I writ large and then the last area that I’m also very you’re interested in, is seeing you’ve always computing at the end user side not being bound to any single device. See, if you think about it, we now need, You know, people you know when we talk about an operating system is not just about the hardware. In one note, it’s about all the devices in our life, eh? So what does it mean to have even an application which you can start with speech on one device, start become touch or none of the device? And in corn around of the device that multi sends multi device experiences? So those are the three layers. Experience is changing to be multi device multi sands ai becoming pervasive in all of our experiences and a distributed computing fabric that’s rich. So that’s it is how I think about it. Michael, maybe I should start to you. What did I miss? Well, to me, I think I don’t know if my answer is is particularly novel, but I think you know, living through and continue to live through this thing’s Corona virus. You know, epidemic. I think it sort of We, as as hearing in the federal government, have been thinking very carefully about sort of applying the next great emerging technologies towards combating this thing, this virus and No one project we worked on was sort of, you know, has really opened my eyes t to the power of Let’s say, a say I in health care is probably the general category. Won’t talk about his is a project we worked on with the Allen Institute with you guys with Georgetown and in many others and I h t create a single repository of all machine readable scholarly articles relating Teoh the covert family off of viruses and by creating this database were able to tap into an unbelievable amount of of scholars and scientists and AI researchers who didn’t even know the first thing about doing health care research. But we’re able to leverage this incredibly powerful database to find patterns to create search engines to provide tools to those who understand the vaccine. 11. Sift through this, and that’s just one small example. But it shows the incredible power of this particular technology, and it really opened my eyes to us thinking more carefully about how we can sort of aggregate critical, important data in order to help solve important solutions. So, you know, we do need to make advances on things like polymorphic encryption. Can we get to a point where folks, medical records and other material which we found very closely and think very seriously about them? But is there a way that we’re able at some point or another? 22 Aggregate information is safe and secure man and a drive better health care outcomes for Americans And to me, I think that is that’s an important future question. If we can make some critical breakthroughs and in some of this encryption stuff, I think we can get big advance for a country. Yeah, I know. I thought that’s a great sort of use out of use cases. I mean, the first one, essentially taking you and all these advances and mission reading and comprehension and applying it like you mean for a scientist today to keep up with what is the latest and greatest is one of the foundational challenges since, or the ability tohave, machine reading and comprehension help you form hypotheses that are novel and new, and in fact, improve the rate of science. That’s one side and the other side of having this longitudinal health record, but yet preserve privacy through even breakthroughs in home amorphous encryption on, then the ability to apply I at an aggregate level, I think, for precision medicine is going to be huge breakthroughs. I mean to me also, one of the things I’ve been observing as part of this pandemic is things that we have talked about always Michael, like, take even telemedicine. It’s amazing to see how much is the take rate. And I mean, there’s no going back, right? This inability to use any I triage tool up front with the telemedicine session after eyes going to fundamentally also help us even overcome some of the cost issues. I’ve been talking a lot to the V A and how they’re approaching even their care protocols on even reaching all the people who need the via services, even on an emergency basis and their ability to scale. It’s fantastic to see some of that. It’s really true. I think we’ve been able to accelerate a lot of our pathway stores, start integrating this important technology in our system through through what’s been going last couple of months. Well, thank you so much for taking the time it’s been it’s been such a such a great great time to speak with you. And I know the folks have Eddie R I and Darfur fit thrill to have you so thank you. And I look forward, Teoh Connecting soon again? Absolutely. Thank you so much. Michael is really, really great. Opportunity and an honor. Thanks. Thanks so much.