The Department of War launched GenAI.mil to 3 million employees in 60 days, with over 1 million unique users in the first 30 days, by architecting secure data flows that prevent classified data from entering public AI model training pools.

(00:00:00) The military buildup in China is the biggest military buildup in world history and so there's a real urgency on our side to ensure that we are ahead but we stay ahead and that's going to take a different level of investment and different type of thinking than we've had in the last 20 years. (00:00:14) In the 80s there were 50 defense contractors and they got merged so they're only about five. (00:00:19) There's a lot of room for new entrants. (00:00:20) It's crazy to me that SpaceX, Anderil and Palantir all had to sue the Department of War for their first contract. (00:00:28) So the idea is you don't have to sue anymore. (00:00:29) Come through the (00:00:32) because people are not going to fight you. (00:00:33) We're now excited about lower cost, faster, more sophisticated options. (00:00:40) Hi, listeners. (00:00:40) Welcome back to No Priors. (00:00:42) We're here today with Emil Michael, the former chief business officer of Uber, White House fellow, and currently CTO of the Department of War. (00:00:50) Emil, thanks for joining us. (00:00:52) Welcome to No Priors. (00:00:53) Good to see you guys. (00:00:54) It's been a long time. (00:00:55) Congratulations on the new role. (00:00:57) Very exciting news. (00:00:58) Could you describe a little bit more about what that role is? (00:01:00) And what's changed at the Department of War to sort of create this new war, this new role, this new momentum, new initiatives that you all are focused on? (00:01:09) Yeah. (00:01:09) So for a long time at the Department of War, there was one organization called Acquisition Technology and Logistics. (00:01:20) And it was all bunched up into one thing. (00:01:23) And then about eight years ago, they said, well, tech is moving faster. (00:01:27) on new kinds of weaponry and defense systems than on the old system. (00:01:32) So we're going to split out acquisition from research and engineering, they call it. (00:01:36) So I'm the Undersecretary for Research and Engineering, which is cool, right? (00:01:40) Because... (00:01:41) I get to work on the stuff that I used to work on where I was in Silicon Valley, like with work with entrepreneurs, work with new companies. (00:01:47) I'm now responsible for DARPA, which is obviously super cool because it's some of the most advanced research that happens in the country, if not the world. (00:01:56) In the last few months, I took over as the chief AI office in the Department of War, which were 3 million employees, the biggest (00:02:05) organization with the biggest budget in the world, is not a small thing to think about how to do AI right for. (00:02:12) And then the Defense Innovation Unit, which is actually based in Mountain View, and it's supposed to be the link between the defense industry and the startup community that's building commercial products and may have dual use. (00:02:27) And then last is the Strategic Capability Office, which takes kind of existing (00:02:33) systems and tries to modify them in strategic ways or supplement them to get, for strategic surprise, they call it. (00:02:40) So all that, all has technology underneath it. (00:02:43) And the idea is to unify that across the department because we spend, $150 billion a year on tech in one way or another. (00:02:51) So (00:02:52) you want to avoid duplication, you want to bring things to market faster. (00:02:54) So that was a big announcement yesterday by the secretary at Starbase. (00:02:59) You mentioned DIU has been involved with providing early funding to a variety of promising new defense tech and hardware companies. (00:03:06) Could you tell us a little bit more about what they've done and how you've approached that so far? (00:03:10) Yeah, DIU has had some good success with Sironic and (00:03:15) and a bunch of other companies. (00:03:18) And now it's like, now it's time to do more of it. (00:03:21) So it's more of a focus. (00:03:22) And yeah, DARPA is just, they invented the internet. (00:03:25) You go to their gift shop and they have like a napkin that where someone drew out the internet sort of architecture and they sell that in the gift shop in case you guys come to DC, we can get you some of those. (00:03:37) This is like a, I mean, it's an amazing unification of like a bunch of stuff that's clearly important to America and to global security. (00:03:46) It's a very different environment than a lot and I have seen you in the past, right? (00:03:52) You've worked with technology companies and even within them as a leader had this reputation for being high speed, aggressive deal maker, if I'm allowed to say that. (00:04:04) How does that work with the Department of War? (00:04:07) which like amazing impact, importance and scale, not the strongest reputation in previous decades and administrations for speed. (00:04:16) Yeah, well, I mean, hopefully part of the reason I got chosen for the job and certainly the reason I took it is to inject some of that urgency speed, you know, being a little bit impervious to barriers and trying to run them over as opposed to being stopped by them. (00:04:33) And it works. (00:04:36) Leadership in organizations is somewhat transferable, right? (00:04:40) If you can show that kind of urgency in leadership and create a culture in a company, you could certainly make some progress in a bureaucracy or a government agency. (00:04:50) It might not be as fast, but I also have way more resources, right? (00:04:54) And I have a way more, you know, in some ways, important mission to humanity. (00:05:00) So those two things are different in some ways sort of (00:05:03) give me the same abilities to sort of drive fast and people respond. (00:05:08) I mean, I'm getting a lot of people who want to join my team or other teams at DOW and across the government. (00:05:16) And yeah, it's harder, but in some ways it's more impactful. (00:05:21) So I'm just as motivated. (00:05:22) Yeah, it's been interesting to watch all the entrepreneurs kind of gathering in DC in a way that I really haven't seen in a while. (00:05:28) So it's been impressive in terms of the types of talent that you all have been able to recruit. (00:05:32) One thing that I think is related to moving fast is prioritization. (00:05:35) And I think in another interview in the past, you said something like 14 priorities mean no priorities at all. (00:05:40) And you've kind of really honed down into a handful of key areas. (00:05:44) Do you mind walking through sort of those key areas from an overall broader innovation focus? (00:05:48) And then maybe we can touch on one or two of them as we go. (00:05:50) Yeah. (00:05:51) I mean, so just to analogize to Uber or any company that we've worked on together, (00:05:57) Imagine you went to a management team and you're like, what's your, how many product, how's your product line going? (00:06:02) You're a series B company. (00:06:03) You're like, well, I've got 14 product lines. (00:06:05) You'd say, well, wait a minute. (00:06:06) Sort of uninvestable, right? (00:06:08) So when I got here, there were 14 critical technology areas. (00:06:13) So critical. (00:06:14) critical to our national security. (00:06:16) So I looked at them, I said, well, number one, if there's that many, they can't be critical. (00:06:22) But #2, hard for people to hold in their head that many priorities and wake up every morning. (00:06:28) knowing that they have to get up and execute against those priorities and make progress, because you get distracted, you get too much separation, and therefore not enough resource to any one thing. (00:06:38) So I cut them down to six after studying, and I made it more action-oriented, so we put sprints behind them like you would in the engineering team. (00:06:46) So off the top of my head, the number one is applied AI, because we're not building a foundation model at the DOW, because there's being hundreds of billions of dollars being spent (00:06:58) by the private sector on this. (00:07:00) So how do I adapt or use what's being developed in the private sector and apply it to the Department of War use cases? (00:07:07) So that's number one. (00:07:09) Number two is, I don't know if you've heard of hypersonic missiles, but they're sort of the newest variation of, we used to have nuclear weapons and ballistic missiles. (00:07:18) Now these hypersonics, Elon talks about this all the time, that could go five times the speed of sound, the minimum Mach 5. (00:07:26) And so they pose somewhat of a unique threat and they're maneuverable. (00:07:29) And so they're, it's the new sort of threat environment weapon that's entered the picture. (00:07:35) Few countries can build them. (00:07:36) They're Chinese, Russia, US. (00:07:39) But my critical technology area is to do that at scale and at a reasonable price, because they're all very expensive now, they're exquisite. (00:07:49) So how do you make them, in the way Anderil thinks about them, much more producible, much less exquisite, much cheaper, more mass. (00:07:57) So I call it scaled hypersonics. (00:07:59) Another one is scaled directed energy. (00:08:01) So directed energy, (00:08:04) is high-powered microwave or lasers to take down missiles, like iron beam, we think about what the Israelis did. (00:08:10) And now with all the drone warfare you see, it's much cheaper to take out drones with an energy zap than it is to send a missile at another, at a drone, right? (00:08:20) A $5 million missile at a $50,000 drone or something. (00:08:22) So directed energy, again, is the technology exists, but how I do that at scale. (00:08:27) I'm a little bit curious on the drone area in particular, one of the (00:08:32) things that I think a lot of people believe is that we're gonna move from sort of big iron giant systems into more distributed fleets of autonomous drones or vehicles. (00:08:42) And how do you think about that in the context of some of the programs or purchasing behavior of the Department of War? (00:08:49) We're still buying aircraft carriers, we're still buying very big sort of systems and there's all sorts of uses and needs for those, but what do you view as the shift to autonomy and the shift to drones that's gonna happen over time? (00:08:58) And what proportion of (00:09:00) Hardware do you think goes there? (00:09:02) I think that starting with the Russia-Ukraine war, you've got this sort of renaissance in drones or robots, think about it as robots, as the new front line, right? (00:09:15) And say in a territorial battle, if you're fighting over land, you're less costly in human life if you have robots fighting first and then before humans come in. (00:09:24) Without the drone warfare in that area of the world, you're probably seeing way more human casualties. (00:09:30) It's already devastating. (00:09:31) It would have been way more devastating. (00:09:33) So I think for territorial battles, you'll see that happen real fast, where the shift, the shift mix toward robots or drones will happen much faster. (00:09:43) When you talk about carriers and sea, there's still projection of power. (00:09:47) There's still protecting sea lanes. (00:09:49) There's still all those things which will continue for some time. (00:09:53) But the mix shift for is still also going to move to autonomous, whether it's autonomous submarines, autonomous boats, (00:10:00) autonomous sort of big drones, essentially autonomous airplanes, that mix will happen, that shift mix will happen over time. (00:10:07) And you'll see every defense budget year over year, more and more will be allocated to those kinds of things as AI gets more prominent and being able to control these things and sense the environment around them and create autonomous action. (00:10:20) So, you know, in 10 years, it wouldn't surprise me if 20, 30% of the budget is spent on those kinds of systems. (00:10:28) which are also cheaper. (00:10:29) So for 20, 20, 30% of the budget, you get way more firepower than you would for the other systems that are not designed to be, you know, for that case without humans in them, with humans in them. (00:10:42) Given the scope of the different services, right, and the scope of the organization, like what does the prioritization exercise look like for you operationally? (00:10:52) You know, you go talk to the different services, you have all the people reporting to you, like how do you decide if it is these hypersonic (00:10:58) missiles or it's electronic warfare or space or something else that is a priority for like, you know, urgent undersecretary this year versus next. (00:11:08) If you're at the Pentagon and you're working for the Secretary of War, then your job or my job is to, as a representative for him in the technology area, is to look across the services, Army, Navy, Air Force, Space Force, and say, what are the common needs? (00:11:26) the enablers or the tech enablers, if you will, or the enablers that are useful in a broad way. (00:11:31) If it's just useful for one service and one use case, that's probably something that service can handle on their own because they have their own labs and research and development dollars and ability to procure and so on. (00:11:42) So I'm looking for commonalities where you get economies of scale where there's things that could be useful for all of them. (00:11:49) And so that's why it took me a couple of months to get through these 14 technology areas. (00:11:54) So to go get a sense of what's really important, to scale them down to six that have some common thread to them. (00:12:00) that could be more useful across the enterprise. (00:12:02) So GenAI.mil, I think, was like a surprise in how quickly it happened. (00:12:10) Can you tell us the story of that and what the goal is? (00:12:13) So GenAI.mil, because if you're on a Department of War network, even the unclassified networks are secure, right? (00:12:22) And different levels of security as you go up. (00:12:26) So you still have to figure out how to architect using AI into those networks. (00:12:31) It's not simple as going to chatGPT.com and signing up because those things are sort of generally prohibited because you have to have restrictions on how to use them. (00:12:39) So we had to figure out what are the policies, how do you architect it in the network so nothing goes (00:12:45) back into the pool of data that ChatGPT or Claude or any of these companies have, because we certainly don't want, no one in this country wants our data getting out into the general public, right, into these models. (00:12:57) So you have to architect a different data flow. (00:13:00) So, but we moved really fast. (00:13:02) I had some like Databricks engineers, former Databricks engineers, former meta engineers, former AWS people, tiger team, 60 days, good, great collaboration from Gemini, from Google's Gemini, who already works at Department of Wars, had some familiarity with our architecture and our systems, and got that launched to 3 million people. (00:13:23) And we've had over a million people uniques use it in the last 30 days, which is kind of awesome. (00:13:28) We've got one third of the enterprise on one model that's 2.5, not 3.0. (00:13:33) 3.0 is coming in a couple of weeks. (00:13:35) Then we're going to do Grok with XAI, and we'll see where we get to with the other two. (00:13:39) And then we do it across classification levels. (00:13:41) So it is going to fundamentally change the way we work here. (00:13:45) And we were able to do it fast because we've got fast people. (00:13:49) That's amazing. (00:13:49) Could you talk about more broadly what are the set of applied AI initiatives that (00:13:53) you all are focused on. (00:13:54) I know it's a major thrust and it's one of the key areas that you talked about. (00:13:57) Yeah, so instead of white papers and theories and committees, we said, okay, how are we going to deploy AI? (00:14:05) So we simplified it. (00:14:07) and say, there's three sort of broad categories we're going to deploy it on. (00:14:11) One is enterprise use cases, like any company would use it for efficiencies, right? (00:14:15) The second is intelligence. (00:14:17) We get a lot of intelligence from satellites, from all the things the United States does across the enterprise that previously used human analysts for. (00:14:27) And so if you could (00:14:29) improve the leverage of the human analyst by keying up to them, here's 10 things we saw that you should look at. (00:14:38) Man, you're going to make that analyst way more efficient, right? (00:14:41) And then, well, if I could take this type of intelligence and fuse it with this type of intelligence, I'm going to make that analyst really capable because usually one type of analyst looks at one type of intelligence. (00:14:51) So the more data I can aggregate from all these sources, the more intelligence we can use, which is a great deterrent because we have the best intelligence (00:14:59) collection in the world. (00:15:00) And then third for war fighting. (00:15:02) So war fighting is a complex sort of thing. (00:15:05) You're doing war gaming, you're doing planning, you know, there's a lot of formalities and how you execute orders, how you simulate what might happen if you do different scenarios. (00:15:16) So across those three areas, enterprise, intelligence, war gaming, we'll have a few pace setting projects within each of them. (00:15:25) which are designed to sort of both break down barriers to make sure we get their data in the system, but to show people in the department, holy cow, you realize this is what AI could do. (00:15:35) So we picked them purposefully so that we could, they could be both demonstrations or pathfinders, we call them in our bureaucraties or, and across these dimensions. (00:15:44) So we hit all parts of the department of war. (00:15:48) And we think that's a way to get people thinking about AI for every use case possible and then deciding not to only when it's like clearly not useful. (00:15:58) One question that we've asked people from some of the major labs. (00:16:01) researchers working on the technology and others, and they often don't want to answer the question because it's hard to project. (00:16:06) So if you feel the same, feel free to demur. (00:16:09) But if you think ahead 5 to 10 years and you ask, what are the big areas where AI is going to most transform how the Department of War works, how you think about autonomous systems, how you think about warfare in general and defense in general, what do you think are the big shifts that are going to happen? (00:16:22) Or what does that future world look like? (00:16:24) I mean, I think that autonomous systems is the clearest (00:16:29) because we're already thinking about that with drones. (00:16:32) And then now you take that to other things like Sermonics building or Anirals building or whatever, and you apply AI to them and more sophisticated AI over time that could do sensing and change how it behaves. (00:16:47) Those are absolutely going to be some of the things that happen in the next few years that AI will have a dramatic impact on, which is why Boston Dynamics (00:16:55) didn't make as much progress as Optimus did in the AI. (00:16:59) The things that started later on robotics just worked better because they were more AI infused. (00:17:04) So I think. (00:17:06) we're actually in a good moment for Department of War to start taking advantage of those things. (00:17:11) And that's where we see some big improvements. (00:17:13) And then the enterprise piece, people, you're like, it's sort of boring for Silicon Valley people. (00:17:18) You're like, okay, of course. (00:17:20) But you guys are immersed in that because you're living in like the frontier of AI use cases in the enterprise. (00:17:28) We're just starting over here in DC. (00:17:30) So I think people are going to go, holy cow, you don't have to write 20 page PowerPoints, 100 page (00:17:36) congressional reports, in the same way that you did with all the manpower you did. (00:17:41) And that'll be more exciting for people to work at too, right? (00:17:44) It's because you're not doing this much drudgery. (00:17:47) So you'll attract better talent as a result. (00:17:50) So I think those are two clear areas. (00:17:51) In the next few years, you'll see a lot of advancement. (00:17:55) You mentioned that one thing you have to have to get these changes to happen is like some (00:18:01) talent from Silicon Valley or just talent that is excited about this pace? (00:18:05) Like, who are you recruiting? (00:18:06) Who do you want to be part of the Department of War? (00:18:08) One of the big untold stories about Doge, which is, yeah, did some aggressive actions in the beginning that were controversial and all that, but the talent they brought in to every department could then be deployed on other projects like the AI action plan that we did with Gen. (00:18:28) AI. (00:18:29) So because they were already tech, like Elon Cho's technically sophisticated people, they weren't all engineers, some were lawyers, some were had from, but they all had a mentality of a fixer builder mentality. (00:18:41) So I borrowed a lot of those people. (00:18:43) And then they attract more people on the way in. (00:18:45) And then we have this US tech force that Scott Kapoor from Andreessen Horowitz is deployed, and we were hoping to get thousands of people out of college for a two-year stint, sort of make it, you know, this is your service to the country as a technologist rather than as a soldier. (00:19:02) So now we want both. (00:19:04) And we're gonna try to make that a badge of honor so that you go back in the industry with some additional credibility and be proud of your service here. (00:19:13) So those are the ways I'm attracting talent and I'm dialing for dollars. (00:19:16) I call everyone who's had just left a job and I'm like, hey, you have a year to spare doing the coolest stuff you could possibly imagine. (00:19:23) Recruiting Tuesdays, I call them. (00:19:25) And I call my friends like you guys and say, do you have anyone who's on the bench right now? (00:19:30) And we have so much interesting stuff to do that I'm picking a few people off every month. (00:19:34) You and Secretary Hexeth have been on this Arsenal of Freedom tour. (00:19:38) Can you tell us about it or anything you learned or that the American people should hear about? (00:19:43) It's been amazing. (00:19:44) And I'll tell you why. (00:19:45) So the Secretary's first reform or transformation, if you want to call it, was restoring the warrior ethos. (00:19:52) It was bringing objective (00:19:56) gender neutral, race neutral standards to military service members, right? (00:20:00) These are fighting units. (00:20:01) This isn't, you know, this isn't ping pong. (00:20:04) This is, you know, this is war fighters have to be capable to do certain things. (00:20:08) And he was intent on restoring those things. (00:20:11) So that's sort of a fitness standard if you want to think about it that way. (00:20:14) Like what does a fitness mean for military? (00:20:17) And then there's the fitness of the industrial base and what you give the war fighter to execute (00:20:22) what tasks they're given. (00:20:24) That's weapons, it's defense, defense systems, it's cyber effects and so on. (00:20:29) We noted in this conversation, defense industrial base sort of wasn't ready to provide what's necessary for the threat environment. (00:20:36) And if you think about threat environment, what we've done in Iran, what we've done in Venezuela, (00:20:42) what we've done with the Houthis, just in one year, right? (00:20:45) Not to mention what might come in the years ahead. (00:20:47) You think about, were we postured for that with the defense industrial base? (00:20:52) And the answer was, we weren't as postured as we wanted to be. (00:20:56) So he said, okay, my next fitness standard is to go to industry. (00:21:00) and tell them that we want them and need them to produce. (00:21:03) We have their backs. (00:21:04) We're gonna bring all the resources to bear. (00:21:06) We're gonna cut out the bureaucracy. (00:21:08) We're gonna make it an open, wide front door for new entrants to come in and help us solve problems. (00:21:13) We're gonna change the way we buy, the way we do RFPs even. (00:21:18) For an RFP, it used to be, here's a 300-page requirement document, please fill it out and we'll decide if you win or lost. (00:21:24) Now we say, well, here's our problem, we're trying to achieve this. (00:21:27) Why don't you come back to us with a solution and you allow them to come in with different ideas about how to solve the problem. (00:21:34) And there's novel concepts out there. (00:21:35) So it's a different, so the primes will excel at filling out a 300 page questionnaire. (00:21:40) A startup won't. (00:21:42) So if you change it into here's the problem I'm trying to solve, well, startup may say, oh, I have ways of solving this. (00:21:48) So (00:21:49) That Arsenal Freedom Tour is about giving them permission and promoting and making sure they understand that we're serious about the fitness and the expansion and health of the defense industrial base. (00:22:00) So that's the Arsenal Freedom Tour. (00:22:03) So it's been a lot of fun. (00:22:04) We've been, gosh, six days on the road in the last two weeks and visited about 20 companies and it's been awesome. (00:22:11) Both the workers and the tech together, you know, again, these combined hardware, software systems, (00:22:17) It's really been inspiring to see the motivation in Southern California, in Texas, in North Carolina, and a lot of the places we've gone. (00:22:25) Amazing. (00:22:26) The other audience that I think is going to be inspired by the... (00:22:31) you know, leadership that they recognize or just pace of doing business like is entrepreneurs, right? (00:22:37) And you have talked about the need for more new primes and like people to solve some of the problems that you're looking at and are important to American dominance and competitiveness. (00:22:47) What should entrepreneurs, I think like it's wonderful that, you know, you've been so early in this and a lot as well, investing in defense technologies that has now become much more mainstream in terms of, and like that sounds ridiculous (00:23:01) to say, but it is acceptable and exciting to work on American defense now. (00:23:06) What should people look at in terms of opportunities and what should entrepreneurs know about doing business with the Department of War now? (00:23:12) I would say that, so a few things. (00:23:13) So number one, we're more open for business than ever before because there's a recognition that our adversaries who started from a cleaner sheet of paper (00:23:25) in the last 10, 15 years, like the buildup, the military buildup in China is the biggest military buildup in world history. (00:23:32) And so there's a real urgency on our side to ensure that we are ahead, but we stay ahead. (00:23:38) And that's going to take a different level of investment and different type of thinking than we've had in the last 20 years. (00:23:45) In the 80s, there were 50 defense contractors and they got merged, so there were only about 5 now. (00:23:52) So there's a lot of room for new entrants. (00:23:54) And (00:23:55) And it's crazy to me that SpaceX and...