Pentagon, Va. --
The following is the edited transcript of an Airman magazine interview with Chief Scientist of the United States Air Force Dr. Victoria Coleman at the Pentagon, Arlington, Virginia.
Coleman serves as the Chief Scientific Adviser to the Secretary of the Air Force, Air Force Chief of Staff, and Chief of Space Operations. She provides assessments on a wide range of scientific and technical issues affecting the department’s mission. In this role, she identifies and analyzes technical issues, bringing them to the attention of department leaders. She interacts with other principals, operational commanders, combatant commands, acquisition, and science and technology communities to address cross-organizational issues and provide solutions. Dr. Coleman also interacts with other services and the Office of the Secretary of Defense on issues affecting the Department of the Air Force’s technical enterprise. She serves on the Executive Committee of the Air Force Scientific Advisory Board and is the Principal Science and Technology Representative of the Air Force to the civilian scientific and engineering community and to the public at large.
During the interview Dr. Coleman discusses the need to reestablish sustainable and trusted microelectronics development, manufacturing, testing and supply chain based in the U.S., the benefits of more focused and frequent testing of systems and the power of diversity in Science and Technology success.
Dr. Victoria Coleman:
So you did one of these with Dr. Joseph here? (Dr. Richard J. Joseph, former Chief Scientist of the United States Air Force)
Exactly the same spot, although he was sitting there facing that direction a little taller than me. Yeah. Interesting guy. It was interesting to see that interaction with everything that Dr. Roper (Dr. Will Roper, former Assistant Secretary of the Air Force for Acquisition, Technology and Logistics) was doing with agility and how that enhanced the research field.
Dr. Victoria Coleman:
This is a really interesting job in many ways, because you have zero authority, but you do have knowledge and expertise. So you need to find ways of doing things that are not direct.
Sometimes it's the best day and sometimes the worst day. Once in a while you get something important done, so you can't complain too much.
Each time leadership changes in this office, is it a stop/start thing for various projects, or do you find that a lot of the previous priorities continue to be supported?
Dr. Victoria Coleman:
There was one project that Dr. Joseph and I worked on, which was the 2030 Air Force Science and Technology Strategy. He and I were co-chairing the executive review panel for the work. In fact, this is how I ended up coming here. So there was a continuation of that for sure.
The position is the most senior technical position in the Department of the Air Force, and your job is to give advice for so many different things. How can you do it all as one person? You can be very smart, but you can't know everything. The range of technologies that we work on here is so broad.
Early on, I figured out that we need to get some more people into the office. This office is the smallest two-letter in the whole department. You know what the two-letters is, right? The closer you get to the top, the less letters you have.
I report directly to the chief (Gen. Charles Q. Brown, Jr., Chief of Staff of the Air Force). There are three billets in this office: myself, the EA (Executive Assistant) and my Military Assistant (an Air Force colonel). That's it. Can't get a lot done by ourselves. So we started this process actually building on what one of my predecessors had done with bringing in more special assistants. And now we have a whole gaggle of them, which means that we can do more things because each person brings expertise in a different domain. So now we can have multiple projects. We are running a number of projects in a very deliberate way.
My background is commercial, where we planned and executed. We use the same kind of planning discipline inside this office for our various projects.
What were the special area areas of expertise that you were interested in bringing in?
Dr. Victoria Coleman:
I wanted help on micro-electronics and help on AI (artificial intelligence). Not that I didn't know these things. I do know these things, but you can't be everywhere all the time. My goal was to fill my gaps, but then also get the priority areas from the department and put them together and see what that intersection looked like.
So we have somebody, for example, that understands hypersonics really well. We're doing a bunch of work in nuclear science and technology. That's something that I still know very little about, but luckily I know people that know more than I do. So we're trying to bring in all this kind of talent to support these areas that are not just gaps for me, but also areas of priority for the department.
Does any of your work support free research or is it all targeted?
Dr. Victoria Coleman:
From my perspective, the only kind of science that I ever was interested in was based on solving a problem. I don’t mean to say it's the only science that people should be interested in, but if you drive it from a problem perspective versus a technology perspective—what can my technology do for me?—I think eventually what you end up doing is breaking research silos and seeing how one person’s research can be applied other places.
However, silos exist because it takes a very long time to A - develop competency and B - be able to solve problems that are really hard. Other problems that are encountered along the way to solving the primary problem but which are, in many ways, separated from the problem you are trying to solve.
Unless you have those silos, you never create enough body of knowledge to apply to the effort.
But then if that's all you have when the time comes to solve a problem that matters to somebody, you're not able to do it. So you need both, kind of this push and pull. You need silos, and then you need the force that says, “I need to solve this problem. What can you do for me?”.
So that's when you begin to bring in the various kinds of disciplines that brings you closer to a solution. Sometimes it's not even a solution, sometimes it's just a way of thinking about how you'd go about solving a problem that kind of completely changes your solution.
I'm not a social scientist, but I have read many scholarly articles that describe in detail how much more effective diverse teams are. Diverse in terms of not just disciplines, but ethnic backgrounds, race, gender. You just do a much better job if you have all these perspectives together.
I have got to tell you, it's one of these really wicked problems. How do you create and maintain and operate a team that is diverse enough to give you these additional IQ points, if you like? In a volunteer force, well, it's not just around the Air Force, I think in STEM (Science, Technology, Engineering and Mathematics) overall, we have a fairly diverse body in terms of gender as you get started in your career.
But, as people become more and more senior, you see that the ranks of women, for example, are thinning out. It's the same with underrepresented groups. If you look at the at the early career stages, you will find people of color in much greater numbers than in more senior levels. The same with the Hispanic population. And that's a real, real problem for us.
Solving it is of course far from simple because the way you need to look at it is both in terms of pipeline and retention.
You need to have the people to hire in the first place. For that, you need to have people be interested in being part of the STEM community and be interested in being part of the DoD community and the Air Force community. That's step one.
Then you need to bring them in a way that is systematic and cognizant of the disparities and what they bring and where they're from.
Then you need to make sure that they are given the opportunity to contribute and succeed that, again, will not look the same for everyone.
In the Air Force, we know that we have challenges. One of the things that am personally very proud of is that there has been, in this administration, a real focus on diversity.
I tell you that from our perspective, it's been truly a godsend because it allowed those of us that are deeply familiar with this challenge to actually make progress. And there's nothing like having a victory once in a while to keep the morale high.
We were able, for example, to create the first ever UARC (University Affiliated Research Center) in the Department of the Air Force (DAF). There are 14 of them that support the DoD broadly, but the Department of the Air Force had never had one before. In fact, there was lots of opposition to it.
We put a cross-department team together. Everybody came together. We figured out what problem we’re going to try to solve and we crafted an instrument that would get after it. It happened to be a UARC.
We're expecting the first proposals of the first ever DAF UARC which will be led by an HBCU (Historically Black Colleges and Universities). The consortium, hopefully, will be focused on tactical autonomy, which is one of the biggest challenges that will have.
Once we kind of formulated that concept, it was actually really easy to get it through. I went to the Secretary of the Air Force with a group of people and we pitched it. I was so nervous. And he said, “What took you so long?” I thought, “Yes!” Once in a while you have a home run. My philosophy is that in order to affect global outcomes, you start with local ones. For us, this is a small victory that affects the bigger picture of things.
(Editors Note - On Monday, January 23, 2023, the Department of the Air Force announced that it has selected Howard University as the first historically black college or university to lead an Air Force university affiliated research center. The center will be focused on tactical autonomy technology for military systems and Howard University will receive $12 million per year for five years to fund research, faculty and students.)
You mentioned that further down the line, you start to see diversity drop off. What are the reasons for that? Is it simply losing people to the private sector, or is it something else?
Dr. Victoria Coleman:
I think that, again, it's a very complex picture. I don't think there's a single answer for a question like that. There definitely is competition with the commercial sector in many of the areas that we work in. There isn't a company on the planet today that would not want to hire the kind of talent we have and most of them can offer a lot more money than we can.
Some of them also can offer a compelling mission, but we compete. I believe we're competitive. I think there are specific issues around retention that pertain to underrepresented groups within the force. Whatever applies to the white male population applies to that (underrepresented) population, only many times magnified.
The way that female leaders are perceived is very different from the way that male equivalent are perceived. Eventually people begin to peel off because it becomes difficult to succeed.
I come from The Valley (Silicon Valley), VC (venture capital) land, everything's about the pitch. If you present pitches to investors that have names on them, you get a very different set of outcomes.
If you present (business plans) with names on them and it's statistically complete and utterly valid, women will get many fewer investors than people perceived to be white males just on the basis of names.
There is bias. And that bias is real, and it's felt, and it has consequences. I think for us, the challenge is to understand it, manage it and fix it. It's very difficult. However, it doesn't mean because it's difficult we don't try. One of the things that we're doing very differently in the department is when we interview people, we ask the same kinds of questions to each candidate.
We spend a lot of time crafting the questions. And then this is how we go through the interview process. We make sure that we deliver the same opportunity to provide input to us, to the panel and to every candidate.
We also make sure that as we make decisions, we do that completely on the basis of the package that we have in front of us, the written package and the interview, not prior knowledge or comfort with a candidate.
If I happen to know you, you come to interview and I know you are really, really good at something, should I be able to take that into account? What if another candidate is just as good or better than another, but I don't happen to know them? That is bias creeping into the process. So we work very hard not to do that.
We are very explicit that even if you know something that is either positive or negative about the individual, if it is not contained in the package, it is not a factor in the hiring decision. Is this 100% error-proof? No, it isn't. Is it better? It is. We try step by step to make to make the process more blind, less biased, and therefore, hopefully bring us better outcomes.
Let's talk about the Defense Advanced Research Projects Agency (DARPA). You were the director there. Did working at DARPA prepare you for coming here? How is your current position different?
Dr. Victoria Coleman:
Just to tell you a little bit about myself, obviously you can tell about the accent that I wasn't born here. I'm an American by choice, not by accident. I was born in Greece.
I will talk a little bit about STEM and why people choose to pursue it or not.
In my case, I was nine years old when there was the first landing on the moon. My dad went out and bought our first ever TV set, this grainy black and white thing. I still remember watching this. It changed my life.
That was the day I decided I was going to be a scientist. I knew I wanted to be part of that. Down the pathway, I ended up a computer scientist. I started my career as a professor. I taught computer science in University of London for a number of years.
I then came to the US. I've worked at the Stanford Research Institute out in California doing primarily DARPA work, almost exclusively DARPA work. Then I went on to corporate. I spent a dozen years working in companies like Intel, Samsung, HP (Hewlett Packard).
I have built mobile phones for a living. I have also run large software services. I was the CTO (Chief Technology Officer) of the Wikimedia Foundation, for example, which is the nonprofit that supports Wikipedia. Very different things. I've been very fortunate to be able to do lots of things and that suits my temperament. I think I get bored if I do the same thing for a long period of time.
Throughout this though, the thing that always was constant, was my wonder and my affection and my interest in all things military technology. Even as a professor, the first airplane I ever worked on was the F-14. You look at that machine and it just blows your mind. My mind has stayed blown all this time.
Even though I was working other jobs, commercial jobs, I always had an advisory nighttime job for the DoD. There comes a point in your career, where I was getting into trouble with manager after manager.
Because all of them would say to me, “What's wrong with you? Why are you spending so much time with the DoD versus your day job?” And it was like, “Oh, because I enjoy it”. And then one day you wake up and you think, “Okay, maybe I am a slow learner, but if that's the thing I enjoy doing, why don’t I do it for a gig?”.
My last job outside of government was a startup. I was the CEO of an AI startup that was a spinoff from Stanford when I made this decision to come to the government. I was fortunate enough to begin with the opportunity to go to DARPA. There's an affection that you have within an organization like DARPA, because it is very unique.
When I was a student, I remember reading articles, this is a long time before I even became a professor, that they could build machines walking on water. It's like, “What? How can you do that?” I mean, it was just fascinating.
Having worked with them or around them since, really, since 1998, I got to know the people. I got to know the mission. I got to know the work. I got to know the kinds of the mechanisms. I just felt such a great affinity for the organization. Having the opportunity to go and work there was probably the best day of my life.
DARPA is a very interesting kind of construct. It's extremely effective in what it does. People have tried to replicate it but failed to replicate the DARPA magic.
So I think kind of internalizing that, supporting it, enjoying it, magnifying it was what I was trying to do.
At DARPA, one of the things that you have to do as the director is approve programs and hire people. So for approving programs, DARPA has this very well-known process or checklist, the Heilmeier Catechism.
George Heilmeier was one of these phenomenal directors of DARPA who took the DARPA magic and bottled it in this catechism with certain questions that you have to answer.
George would ask people, “What are you trying to do? No jargon. How’s it going to make a difference? Who cares about it? How much it's going to cost? How do you know you're making progress?” All these are great questions. But then there are other things that matter that I added.
To answer your question, first we have to educate our audience, but people have this tendency to use a lot of jargon to tell you what it is that they're doing, and most of the time they fail to tell you what they're doing.
So I want an elevator pitch. This is something that we learned in The (Silicon) Valley; if you have like a minute with me, what are you going to tell me about this thing? If I know nothing about it, why should I care? So every elevator pitch really gets people to explain the essence of what it is that they do.
And by the way, an elevator pitch is hard to get right. You use the wrong elevator pitch in The Valley, you get thrown out of the room. So the day that you don't get thrown out of the room, you know that you got it right. I wanted them to do that.
I also wanted them to be very clear about the addressable and serviceable markets, which again, are terminologies that the VC would use.
If I go to a VC with an idea, they will say, “Well, how big is the market that you're getting after? Is it $1 billion? Is it a $100 billion? And what is the serviceable market, in that, if it's a $100 billion, how much of that can your solution actually service? How much of that is potential revenue for you?”
If you translate that as you would for an impact statement for a DARPA program, what I want to know is how big a problem are you solving, and how much of it are you solving? Because that's truly material to me.
If it's a really important problem and you're solving 1% maybe I'll decide to fund you, maybe not, but it's important to know those specifics. That was the second addition to Heilmeier.
The third one is, again, driven from my experience in the past of delivering things. If you build mobile phones and you run big services, outcomes are very concrete. There is no hand waving. It either is there and it works or it doesn't. So I wanted them to be very clear about what they would deliver and to whom.
If somebody comes and pictures a program and says, “I'm going to do this cyber thing and I want $40 million”. I want to know what are you going to do, exactly? Are you going deliver a standard? Are you going deliver a bunch of papers? Are you going to deliver code? Are you going to deliver an architecture? Who exactly is the customer? Because unless you have that, it’s impossible to tell if you’re making progress toward the goal.
Let’s talk about the shift within the Air Force towards failing fast, rapid prototyping and agile acquisitions. In a world where technological advancement is accelerating, things that used to take a decade to come about are now coming about sometimes in days. How does that affect the planning and execution of your department?
Dr. Victoria Coleman:
Well, you put your finger on it. I think that is our singular challenge. You know, it takes us a very long time to conceive, build and deploy technology. And it’s killing us.
In my experience, failing fast is absolutely essential. You look at testing; let’s talk about hypersonics for a moment. Maybe we’ll do a test every six months. A hypersonics test, a real flight test, is such a complicated process. We try to experiment with so many variables in one test that success is almost precluded. It’s crazy.
So what you’re saying here is the advantage in being able to test more often is that you can hyper-focus on one specific area?
Dr. Victoria Coleman:
You hyper-focus on one specific area, and you keep it small. You keep it small, because if you fail, you know what it is that you failed at. It was that small thing that you tested. And then you go back, and you fix that. Then you test and fix the next small thing and the next thing and the next thing.
It does not matter how many tests you do in a chamber someplace. It's taking that solution, sticking on a B-52, taking it up there and releasing it and seeing what happens. Unless you're able to do that time after time after time, you will never make enough progress. We're not able to get there because we try to solve too much of the problem at the same time.
It permeates acquisition. I think one of the reasons why it's taken us so long now to build and deploy systems is because we try to do too much at the same time. In business, we learn the hard way that you can't do that. If you put too much risk in projects like that, you just go out of business.
Companies that have been able to stick around and be successful for many years, the Apples of the world, they figured this out.
You know, we talk a lot about DevSecOps (development, security, and operations) and software factors in the DoD. Which is awesome, by the way; 10 years ago, we didn't even have that conversation.
But what we failed to appreciate is that it is, in many ways, the tail end of a much bigger process.
You may have a software factory and you may have this amazing pipeline, but what are you going to put through it? Who decides? That is a broader kind of discipline. It's understanding what is the minimum viable product (MVP) that I can field so that I can begin to A - solve a problem, but B - learn about the technology in real world use so that I can begin to build on the basis of that incrementally.
And how important is it getting feedback from the end user?
Dr. Victoria Coleman:
Exactly. A product manager in the world outside has a critical job to do. One is defining the minimum viable product. The other one is defining a roadmap of what it is that you're going to improve. But then they have this incredible responsibility of convening everybody.
The team comes together, and they look at the work for that day of the week. But then at the end of every sprint, everybody gets together—the product manager, the designer, the sales person, the software manager, the hardware manager—they’re all in the same room, and they say, “This is what I did. What did you do? I did this.” And you know, it's amazing how much you learn from having all the stakeholders in the same room.
Now if you do that, week after week, you solve it within weeks. You don't solve it six months or a year or years later when the thing gets deployed and the operator says, “What? I didn't ask for this. What I really want is that.” It's really fundamental that we change our way of building these things.
But then we also have to change the way we acquire these things. We talk about agile software development, which we can do, but can we buy software in agile increments?
How do you make sure that as capabilities been developed, you incrementally bring it in and you deliver it, you give to folks and you pivot if it's not the right thing to do.
Our contracting vehicles, our acquisition process has not been designed to be agile. My colleagues in acquisition don’t want to be slow. They don't, they're going to be as fast as any other person. But the whole process within which they work has been designed for a different era.
And it's so interesting because when those same people who support a slow, plodding process pick up their phone, they are holding a perfect example of the marvel that is feedback and iteration on a product.
Dr. Victoria Coleman:
The entire globe is swimming in agile acquisition and minimum viable products. Every time you go to bed at night, your apps update and it's seamless. Yet we can't seem to get around that corner when it comes to an ever-increasingly digital military.
Dr. Victoria Coleman:
I think that acquisition reform is really a fundamental need that we have. It's something that many people have taken on. Occasionally I will go back and look at the rationale and the process and the outcomes of attempts at acquisition reform. It’s wickedly complicated.
I think it's like those problems with diversity and inclusion, we were talking about earlier, it's a wicked problem. Doesn't mean to say that you shouldn't try. I think we are ripe for figuring out how, in this increasingly agile world, we adapt our acquisition processes and our regulations and our training and the risk tolerance of our acquisition professionals to be able to move faster.
If you look at what happens with a commercial Chinese jet, it takes them less than a decade to get it out. Some cases far less than that. It takes us two, three decades sometimes to build and field something. That is not acceptable.
And how much of that is the benefits of an authoritarian regime that says, “do this and here's all the money you need to do it” versus a democratic process with oversight and seemingly endless continuing resolutions in Congress?
Dr. Victoria Coleman:
I have got to tell you, I believe it's all about feedback. So in the commercial world, the feedback comes really quick. If you get something wrong, you are going to know about it quickly, and maybe you get another chance, but you're not going to get a third chance. It's over with. In government acquisition, we just don't have that market feedback, right? You don't have somebody that comes in midstream and says, “No, this is a piece of garbage. I'm not going to buy it. Stop it”.
We need to find ways of artificially injecting that feedback into our process to make sure that we're cognizant and that we are held accountable to making the right choices, making choices at the right times versus letting things kind of evolve from micro mistakes to macro mistakes.
Has seeing hypersonic weapons dropping on Ukraine encouraged a change in our processes in order to catch up?
Dr. Victoria Coleman:
Certainly the orbital test that the Chinese did about a year ago concentrated our minds. It demonstrated not only what their capability is, but what their intent is. And it certainly accelerated our own thinking about how we should be perceiving this.
It's kind of interesting because in the case of China, almost everything that they do is the first time. The rate at which the PRC is building capability approximates or is at a war footing, something that we haven't seen I think for generations in our country.
We have a ton of capability. So everything that we do is done with respect to that existing capability. With the Chinese it's the first time that they build something, so there is no inertia. We ask ourselves many questions to make sure that what we deliver is not only the right thing but is the right thing in this broader kind of portfolio of things that we have.
You mentioned developing capability as if on a war footing. The building that we're sitting in right now was designed and built in 18 months in the middle of a global war.
Dr. Victoria Coleman:
That's right. We can absolutely do it if we put our mind to it. Nobody can beat us in this game. Our government, the DoD, is committed to manage the competition with the PRC. I mean, the last thing we want is for it to glide into conflict that nobody wants.
At the same time, we see their aspirations and we need to be prepared for that. And it's not, of course, just the Chinese. The United States is a presence across the globe. We represent core values that are very unique. So we need to be ready for a number of things so we don't just kind of build capability with respect to the PRC. It's really fundamental that we move faster because we're going need to solve problems across the world.
But we've seen the enemy and it's us. Hypersonics is a great example. Testing once every six months, that is not going to get us there. I want to be able to test every week.
We can leverage, in many ways, strengths that the opposition doesn't have. In this country, there is an incredible kind of innovative base that sits outside the defense industrial base. We need to work with commercial partners in order to accelerate what we do.
There's least one company that does commercial hypersonic test flights. I’d love to be able to take some of our payloads and systems and not just rely on the one B-52 that we have out at Edwards to do this testing. I want to work with them both to get the cost down and get the opportunity to fly many more times than we're able to do today. If we think more broadly about the resources that we have available to do problem solving, boy, we can move mountains.
Speaking of China, they have a foreign and economic policy geared to not only secure rare earth minerals and other physical resources for their own consumption and military development, but also to create supply chain difficulties for us and our allies. How is that challenge influencing our ability to test and develop capability?
Dr. Victoria Coleman:
You ask a really good question. Let me just start off by correcting myself. We'll talk a lot about China, but we shouldn't be talking about China. We should be talking about the PRC (People’s Republic of China). These are very two different concepts. You know, as a nation, we value our friendship and respect for Chinese culture and people. Everything that we say and do in the department is with reference to the PRC, not China. So we should always be careful to make a distinction between the government and the people.
So let's talk about the PRC for a moment. They have certainly emerged as a fearsome competitor. Why is that the case? They have their strengths and their weaknesses, and so do we.
I think we need to understand their strategic strengths because that tells us what we need to get after. Also, we need to understand their strategic weaknesses for respective reasons.
If you look at the competition, for example, over Taiwan they have a strategic advantage in that they will be fighting close to home if they choose to fight. We are very far away. That's the strategic advantage that they have.
During the Cold War, we had a very a very well-articulated and executed vision to bankrupt the Soviets. The PRC has very deep pockets. It's going to be really hard for us to bankrupt them. So that's another strategic advantage. Their economy is so intertwined with ours and global prosperity that saying that we're going to bankrupt them makes no sense. So we have to be much more strategic about that.
The other thing is that we have fought multiple campaigns. They have not. There are pluses and minuses in this, because we fought multiple campaigns, we have combat experience, but they’ve had plenty of opportunity to observe how we fight in the environment.
We always had air superiority in the past. So at the time of our choosing, we could put resources and assets in many, many places on the planet. They observed that. Then they went off and they built a whole bunch of defenses that make it difficult for us to fight the war we used to fight. That's another strategic advantage that they have created for themselves.
We see them being fearless in using their economic might to affect outcomes across the world. So their Belt and Road Initiative is a classic example of that.
They have been careful but aggressive in creating those strategic outposts so that you could trace the known globe with their investments. We have failed to do so. We have programs that are longstanding, but I don't think that we're using them nearly as strategically as they use theirs.
If you think back to the Marshall Plan, what a difference that made. Think about it. That basically brought peace and prosperity in Europe for 60 or 70 years. So, when we do it right, it has massive outcomes. So I think there are opportunities for us to be more responsive in that competition. Sorry, bit of a tangent there.
No worries. I was just really kind of trying to get at the supply issues, how they're cornering the market on a lot of necessary resources.
Dr. Victoria Coleman:
Maybe we'll talk about chips a little bit.
That's near and dear to my heart. I'm still waiting for my new refrigerator because of the electronics shortage.
Dr. Victoria Coleman:
I think we've always known that the microelectronics supply chain was overly dependent on the Far East and the PRC in particular. But throughout the pandemic that became even more pronounced.
With the supply chain for microeconomics in particular, you certainly have rare earth minerals and so on that are part of the bigger picture. But that's just the start.
The supply chain is so complex for microelectronics that even having visibility into that supply chain is a challenge. We talked about AIML (Artificial Intelligence/Machine Learning); I wish somebody would go off and do some of that to figure out what that complexity actually looks like.
But we do know that over the last 20-30 years we've let major portions of that supply chain leave the continental United States. We did that because the economics were favorable and also because, as a nation, we have this aversion to industrial policy interfering in the free market.
Well, it turns out other people don't have such concerns. With massive subsidies (from foreign governments) and so on, a lot of that industry and links in that supply chain went overseas.
Now we're in a situation where this old notion of “we'll keep the high-value parts of the value chain here at home and we'll let the others go overseas”, that's coming back to bite us, big time. It's impossible to maintain the front-end, high-value design part here at home if the expertise for building these things is all overseas.
Andy Grove - I used to work at Intel - when it comes to the semiconductor industry, he was the most brilliant business strategist that has ever walked this planet. (Andrew Stephen Grove was a Hungarian-American businessman and engineer who served as the third CEO of Intel Corporation.) He was very clear that we walk on a really thin line if you believe that you can do the highly skilled things here and other things we do can go elsewhere. Eventually you lose that workforce that allows you to innovate in the first place.
If you think about it, if you are a fresh graduate and you've just spent four years spending a lot of money and getting very educated, now you have a choice. Why would you go to MIT and pay even more money to do a Ph.D. in microelectronics if you know that aren't any jobs here at home?
You would pick something that is very popular that you could be highly compensated for here at home. When you look at the statistics of the graduate students that we educate here in this country, a very significant majority go off to China and other places, because that's where the jobs are.
So this warning from Andy Grove, it’s beginning to come true.
Today, if you look at the leading-edge production of semiconductors across the globe, the United States owns 12% of that production. The wake-up call that led to the CHIPS legislation (The Creating Helpful Incentives to Produce Semiconductors and Science Act of 2022), is that if we were to do nothing, 10 years down the line our global manufacturing share would shrink further to 9%.
If you look at the sum total of the semiconductor supply that we and our allies buy, do you know what that percentage is? You'll be really surprised. It's over 70% and only 9% of that will be manufactured here at home.
Now they have a hold on us, right?
This is why CHIPS was in many ways an obvious piece of legislation. At the same time, it was a really incredibly difficult piece of legislation to get right, because, A - we have this aversion to industrial regulation interfering with free markets and B - because, for the most part, we don't seem to be able to agree about anything. Within the CR (Continuing Resolution - legislation to continue to fund the federal government before a new fiscal year begins), the fact that everybody came together and passed the CHIPS legislation, and the appropriators also decided to fund it, I think it's a minor, no, a major miracle, actually.
It will allow us, for the first time, to begin to rebuild the capacity here at home; to have a supply chain that’s domestic, not all of it, but enough of it so that we can satisfy our own demand. Maybe working with allies, we can get us to 12-15%. It's still not going to be 75%, but the Europeans, for example, are making similar investments in their supply chain. By being very strategic, we can come together.
I think overall it's a critical technology for us. In the DoD, especially in the Department of the Air Force, we depend on technological advantage to win. If you're in the Army, you can march. If you are in the Navy, you can row. In the Air Force, you cannot just fly. Everything we do depends on technology. That technology depends on microelectronics. Having a trusted, assured supply of components that we need for our systems, both for our legacy systems and the future ones, is really integral to our success.
I want to ask about other priorities of your office, but when it comes to micro-electronics, you talked about the complexity. What makes this process so complex?
Dr. Victoria Coleman:
One of the facts about the semiconductor business is that because the systems are very complicated, it’s very, very expensive to make them. To make them we use fabs.
What is a “fab”?
Dr. Victoria Coleman:
It's a colloquial term for a fabrication facility. So that's where chips are made in large quantities. There is a distinction between what we call a “fab” and what is called a foundry.
Intel runs fabs. That’s where they make the microprocessors. TSMC (Taiwan Semiconductor Manufacturing Company Limited), which is the world's largest chip manufacturer, runs a foundry.
A foundry is a little different because instead of building chips that only one company designs and sells, they build chips for many, many different companies. Back in the day, anybody who was making microprocessors had the engineering, the design, and the fabrication in one company. This, the so-called IDM, integrated device manufacturer, model.
The only company around the world that is still in the IDM model is Intel. The rest of the world has bifurcated into companies that do the design and foundries, like TSMC, that actually build things.
In the case of the of the DoD, because our attempts in the past to have our own specialized fabs have failed, we need to have access to commercial parts for the systems that we depend on day-in, day-out.
Attempts at having our own fabs failed because of the economics. It's very expensive to build them. You need to have both the money and the volume to keep pumping through, otherwise, the yield goes down and then they become unworkable. So there is no business model for government-owned fabs, which means that we have to rely on commercial foundries for the parts that we need.
Now, this is the conundrum. If we buy a computer or a mobile phone, it's obvious that commercial parts are good enough. But if you're building an F-35, for example, is it okay to have a commercial part in that system? If you're building a satellite, is it okay to have a commercial part in that system?
The reality is that we are in a situation because of the economics, that we have no choice. We have to use commercial parts. So for us, having access to those commercial parts, having access to our supply chain here at home, having a trusted and assured supply of microelectronics is critically important.
That's why the passage of the CHIPS act was a landmark. Now we know that we will have a much better chance of accessing that critical technology for our systems here at home. Is that enough? It's not enough. We also need to make sure that, you know, that these commercial parts are trusted enough for us to put them into our systems.
When you say “trusted,” are you talking about chips coming through a secure pipeline where there's no chance of foreign actors altering those chips?
Dr. Victoria Coleman:
Well that's part of it. But that's not a part that I personally, nor many of my colleagues, worry about. Because it's really hard to make a performance microprocessor. It's almost impossible, frankly, for an adversary to inject functionality.
It is a very, very unlikely way of subverting our ability. It’s much more likely that these systems failed because they are very hard to design in the first place. Our approach, and it is shared by the OSD (Office of the Secretary of Defense) is a principle of quantifiable assurance.
In other words, building a risk profile with respect to each commercial part that we take from a commercial supply chain, and we bring into our systems. That means developing ways of evaluating by testing and other methods to convince ourselves that we have reached a tolerable risk level when we bring that device into a system.
Everything has a risk level; every component, every alloy, every rivet that you put inside a fighter jet comes with a risk level attached. And microelectronics are no exception to that.
This is our strategy here. It's to support a strong, viable domestic supply chain so that we have access to those commercial parts. Then at the same time, develop the standards, these quantifiable assurance standards that give us ways of building with enough confidence in the system so that we can deploy them.
Is there any new research in microelectronics or nanoelectronics that you see that could open new possibilities for the military?
Dr. Victoria Coleman:
There’s a ton of that. Absolutely. People often talk about the end of Moore's Law…
For background, could you explain Moore's Law?
Dr. Victoria Coleman:
Moore’s Law is all about really being able to get more performance out of the same microchip design by shrinking it and cranking up the frequency. Imagine I'm able to do five things in an hour today. If tomorrow I put instrumentation together that allows me to do 10 things every hour, all of a sudden now I can get much more done, right? This is precisely the principle of Moore’s Law.
If we clock up the frequency, it means that you can actually complete, within the same amount of time, many more instructions than it could before. That gives you increased performance.
How do you get that by shrinking? We create one microprocessor and then over the next year shrink it so that we can build the same microprocessor on a smaller geometry. We can fit more processors in the same space. It's a great gravy train to be on. You can sell them for a lot more money.
Unfortunately, there comes a point with the shrinkage that the physics stops working. That’s the end of Moore’s Law.
You need to find the ways of continuing to deliver more performance. So microprocessor companies started embellishing designs. This business of just shrinking the same thing no longer worked, right?
So that has led down the path of a lot more innovation. Things like 3-D stacking, for example, being able to build devices not just in a two-plane assembly, but in 3-D assembly so that you can dissipate heat and you get more ports where you can put information in and get information out. That's one of the most promising new ways of designing microprocessors.
Others are things, like neuromorphic architectures, where you move away from the traditional model of designing microprocessors. They are actually beginning to emulate how human neurons work. In some cases, the results that we get are phenomenal.
I'll give you a very concrete example. In the Air Force, we like to be able to match assets. Weapons that we have are matched with targets that we need to get after. So, you would want to make sure that you send the most appropriate weapon to the most appropriate target.
There are many, many combinations of doing this. For example, if I have 10 by 10, 10 assets and 10 targets, there are 25 billion combinations to finding the right match. A human being cannot do this. A computer maybe could do it. Classical computers can do 20 by 20, but they max out a 20 by 20.
A neuromorphic architecture can do 20 by 20 in about two milliseconds. So the capability that you get from the ability to build neuromorphic architectures completely changes the game.
Then there are also material advances. Using photonics, for example, for switching, you generate a lot less heat than if you switch voltages.
Quantum computing is another area where a lot of a lot of work has been done. We are seeing some applications already, like quantum navigation, for example, becoming really fundamental. Quantum navigation means that I can navigate without the benefit of GPS, which would be an incredible capability to have.
But you can go even further than that. One of the areas that I have been interested in for a very long time is what I call molecular computation, where you can actually take DNA structures and use DNA as a computational medium. DNA is a particularly good substance for computing.
Honestly, the only constraint that we have is putting enough money into the various research groups that can pursue these things.
Something that I care deeply about is our ability to prove out those innovations.
So we talked about CHIPS, but I want to put in a plug for one particular component of the CHIPS Act, actually the only component that we will be executing here in the DoD, which is the Micro Electronics Commons.
This is something that I've been working on for some time, certainly way before I went to DARPA. One of the things that has happened to us, in this desire to let the market forces rule, is that in order for us to prove out an innovation in microelectronics at the system level, we have to go overseas. We can't do it here at home.
So I talked just now about neuromorphic architectures. The NSF (National Science Foundation), through one of their expedition programs, has spent well over a $100 million on one specific project at Stanford University on neuromorphic architectures. Great work. Phenomenal. Smart, brilliant people.
In the university lab, maybe you can build three or 10, maybe 100 instances of a new device like that. In order for me to show the computational advantages that I just talked about, pairing targets to assets, I need to build millions. If I am going to go to a venture capitalist or an investor and say, “Please give me $20 billion to go and build a new fab to make this new thing,” why would they give me this money? I need to show to them that the computational benefits that I claim are achievable with this new structure have been proven.
Where do we go for that today? We go to IMEC (Interuniversity Microelectronics Centre) in Belgium, a facility that the Belgian government sponsors, or we go to the PRC.
We worry a lot about IP (Intellectual Property) being exfiltrated. We already have a ton of restrictions, and we know it still leaks. In the case of really advanced research, like neuromorphic architectures, we put that IP on the plate and we present it to the PRC.
That means that we're not only dependent on them for manufacturing our stuff, we're dependent on them to let us prove out our innovations and hope they don’t copy it. Think about that for a moment. Andy Grove said that eventually you lose your ability to innovate. That's exactly what he was talking about.
So there is one piece, and it's a small piece, a $2 billion appropriation in the CHIPS Act, to build out what we call the Micro Electronics Commons, which is a network of facilities here at home to allow our researchers, small companies and large companies to keep that proving out here at home without having to go overseas.
To put a fine point on it, the importance of microelectronics is that they are a part of everything; they are in everything. Is that what we're talking about?
Dr. Victoria Coleman:
That’s about right. You know when Mike Griffin, the previous USDRE (Under Secretary of Defense for Research and Engineering) came in, he had seven priorities. He had electronic warfare, he had hypersonics, he had, Lord knows, I can't remember all of them. But, every single one of them was dependent on microelectronics. There is nothing we do today or will do in the future that does not depend on microelectronics, not only in the DoD, but in life in general.
That sounds like you have your elevator pitch for microelectronic funding.
Dr. Victoria Coleman:
I hope you cut me a check.
If I had it, I would. Can we talk about another area of expertise that you brought up, Artificial Intelligence? Can we discuss that from the problem-solving perspective, such as ABMS (Advanced Battle Management System) and JADC2 (Joint All-Domain Command and Control). The Air Force is really good at collecting data. The bottleneck comes in collating it, disseminating it, and making it quickly actionable. Maybe let's talk about the importance of developing artificial intelligence and advanced computing methods which will enable the war fighter to communicate and react upon data much more quickly.
Dr. Victoria Coleman:
You touched upon one of the key priorities that we have in the department, Operational Imperatives directed by Secretary Kendall; one of them is indeed battle management.
Battle management is the ability to close the kill chain—going from understanding the situation, sensing a target, tasking an asset to apply effect and then assessing whether the effect desired was achieved.
We are fundamentally challenged in the Department of the Air Force, and, I think, in the DoD more broadly, where there are multiple efforts. Ours is called ABMS. Over the years we have struggled to even figure out what that is and what it looks like. How would we know it if we had it?
In order to affect this outcome of closing the kill chain, you need to have comms, first of all, so we can talk to each other. We need to have fusion, so we can understand what it is that we’re seeing and we can communicate it. We also need to do things like distributed dynamic targeting, because otherwise we will have multiple assets getting after the same things.
It is a multilayered solution that we need to build all the way from connectivity to building sensor fusion: building the sensors for sure, but then fusing the sensors so we know what it is that we're seeing, communicating intent, and then letting the units at the end to do their job and do it productively and efficiently.
One of the things that we don't want to do is burn a batch of rounds that we don't have to. If something is already dead, why would we kill it again? So how do you solve that?
At the core, it is actually a very computer-sciencey kind of problem, which is this distributed agreement. We do need to solve it in a particular layer in order for us to be able to do efficient, dynamic, distributed targeting.
ABMS is so wickedly hard because you need to get every one of these levels designed right and built right. It's building the whole stack in order to get the capability out there. Then we have to do it within our silo because we need to be able to task our own assets, but we also need to integrate with the sister services because nobody's going to go fight alone.
Then at the next level, we integrate with our allies because we will fight with allies. So it really is a wicked hard problem. But I feel like we have to make enough runs at it; this is what you have to do.
DARPA knows this really well. At DARPA, you will start something, sometimes it will work, sometimes it won't, and then you kill it and five years later you'll come back and you'll try it again. In five years, many things change, including other core technologies that you didn't have when you first started it.
I believe it's the same with ABMS. I think now we're getting to the point where many of these constituent technologies are actually there, and our own understanding of how you go about layering solutions so that it all comes together and has the desired outcomes.
Our understanding and our competency is getting to the point where we're beginning to see the light at the end of the tunnel.
We talked about microelectronics. It takes microelectronics to make the radio pods to communicate. It takes microelectronics to put in the sensors. It takes microelectronics to do the fusion. It certainly takes microelectronics to guide weapons. There is nothing that we do that does not include microelectronics in order to do this advanced weapons pairing.
What else do you need? You need AI, right? You need AI because human beings can only go so fast.
You can do it the dumb way, going through all the possible 25 billion combinations, or you can build models of what successful engagements look like and then use that versus this exhaustive kind of search. That is really important, because when you talk about efficient, distributed, dynamic targeting, you're talking about making decisions at the edge.
It's not always the case that in a contested environment that you will have perfect connectivity. So the F-16 over some place will know what is going on at the AOC, at the Air Operations Center. They need to oftentimes make a decision themselves with what they have. Machine learning gives you the opportunity to make those decisions locally but taking into account a historical global context that's reflected inside of the model.
It may not be perfect, but it's going to be just as good as you can possibly get it at the edge. I think it's a fundamental kind of capability. It's a capability that's been developed at DARPA through their ACE (Air Combat Evolution) program. And it's a capability that we begin to see our way towards deploying here at the Air Force through a triad of autonomy programs. One of them being the CCA (Collaborative Combat Aircraft) program. The other one being VENOM. Then the EOU (Experimental Operations Unit) team. Let me tell you about those things.
Dr. Victoria Coleman:
All right. So let's talk about autonomy and what a tactical autonomy means. Tactical autonomy means deployed in the mission. It's not something that sits in the lab. We understand that in order to fight and win, we will need to have groups of aircraft working together. We have multi-ship today, but we will need to be able to operate with unmanned aircraft in partnership with manned aircraft.
The Collaborative Combat Aircraft (CCA) Initiative is one of the core initiatives that we are pursuing in the Air Force. It’s the brainchild of Secretary Kendall.
So, how are we going to deploy a formation that is, you know, half-human, half-machine? How are we going to gain confidence in that? Will this unknown system behave as intended in all possible kind of operating regimes? How are we going to incorporate that capability in the squadron?
Squadrons today don't have autonomous machines. They have human beings flying machines. To get after deployable, missionized, fieldable autonomy, we have three system programs that we're hoping to get funded. One of them is CCAs.
The other one is the EOU team that is beginning to develop the tactics and procedures to incorporate that capability into a squadron. I can guarantee you a squadron that is half-man, half-machine is going to look very different than the one we have today.
Anticipating what that is really important—developing the tactics and procedures for incorporating the capability and operating it, understanding how the capability, if we had it, actually would be incorporated into the mission and how a squadron would practice it, train for it and deliver it.
Sitting in the middle is Project Venom, a bridge between a fully autonomous set of capabilities and a fully manned set of capabilities, which is where we are today.
With Venom, we will take six F-16s and we'll kit them out with an autonomous safety box, where we'll be able to drop in autonomy code and we'll have a human pilot experiment with it and fly with it to make sure that A - it works, and B - that it has the benefits that we expect the autonomy to deliver operationally.
The neat thing about this is you can do it really quickly. It usually takes a very long time to certify software to fly in airplanes. If you were to wait for that, maybe you would do one drop a year. Venom enables us to do tens of thousands of cold drops per week. So it's a really fundamental kind of capability that bridges and accelerates that transition from fully manned to this mixed mode. Human-machine teaming.
It's kind of sounds like where self-driving cars are now.
Dr. Victoria Coleman:
That's, that's exactly right. You know, self-driving cars didn't go from fully manual to fully automated, right? The Tesla’s and all the other electric vehicles, they've traveled millions or billions of miles where they learned and figured out how to interface with a human operator and to do so safely and securely. We don't get to skip that part in the Air Force. So Venom is precisely that capability that allows us to bridge the gulf from one state to the other. So we're very excited about it.
Okay. I'll give you one last question. Most people out there don't know what they don't know, myself included. So what should I have asked you about?
Dr. Victoria Coleman:
Oh, gosh. That's a hard question. I sit here as the most senior technology official in the Air Force. So I represent many thousands of incredible, talented people in the science and technology community.
I worry a lot about them working on real problems versus imaginary problems. It's really, really easy to sit in the lab, separated and disjointed from the warfighter, and work on really interesting problems, but we are not a university.
While we do and support long-term research, we do it because, at the end of the day, we're going to solve a real-world problem that affects the outcome of a mission that, frankly, affects the life of a fighter pilot or a crew member.
I want to make sure that while I'm here, and hopefully after I leave, there will be much better connectedness between the S and T enterprise and the war fighting enterprise, because that's what we're here for. We're not here to write papers. We're not here to do experiments. We're here to support our warfighter. For me, it's a key priority to bring together those scientists and engineers, to bring them out to the edge, make sure that they understand the real day-to-day hardships that a warfighter has to go through in order to fulfill their mission.
Last week, I had the opportunity to bring out the chief scientist group, we have about 75 of them. We get together a couple times a year. We always have a focus area. This last week our focus area was the Space Force mission. So we we’re in Colorado Springs. There’s a grave at the Air Force Academy: Lt Col. Morris “Moose” Fontentot, JR., an F-15C pilot, a graduate from our Weapons School (our equivalent of the Navy’s Top Gun) with 2,300 hours of flying the F-15C. He went down over West Virginia in 2014, at .83 Mach. He went into the ground at 680 miles per hour. Why? This is an instructor pilot that knew what he was doing. When I say that our pilots are X-Men, you have got to trust me. They’re X-Men and X-Women. So I want them to work on that problem. I took them to the grave and I told them, “This is what we need to work on. Nothing else we do matters. You save a life as an S&T professional, you’ve done your job. You’ve earned your keep".
So that is my exhortation to all of my peers in the S&T world: Get close to the warfighter. Understand not only their mission, but the context, the environment in which they have to pursue it and solve their problems.
Save one life. Just one. Just one.