Christopher Kirchhoff on Military Innovation and the Future of War

Conversations with Tyler is produced by the Mercatus Center at George Mason University, bridging the gap between academic ideas and real-world problems. Learn more at mercatus.org. For a full transcript of every conversation enhanced with helpful links, visit conversationswithtyler.com. Hello everyone, and welcome back to Conversations with Tyler Cowen. Today I'm here with Christopher Kirchhoff. He wrote a book that I thought was great. It is Unit X: How the Pentagon and Silicon Valley Are Transforming the Future of War, and that is co-authored with Raj M. Shah. This book has a rave review from the Financial Times and on Twitter, also from Chris Blattman. To give you a full biography of Chris, he is an expert in emerging technology. He helped to create the Defense Innovation Unit,, and he's also worked as an assistant to Eric Schmidt. Earlier in the Obama administration, he was the director of strategic planning for the National Security Council and also the senior civilian advisor to the chairman of the Joint Chiefs of Staff. Chris, welcome.

Thank you, Tyler.

I have many questions about your topic. How permanent is the ascendancy of drone warfare? So it's the thing now. Is this the next 100 years, the next 5 years?

What? It's certainly the next generation of military technology. And, you know, we've seen this coming for a while, right? EU Defense Innovation Unit, the topic of the book, in 2017, started the military's first commercial drone unit and began experimenting with both offensive and defensive uses of drones. And now we see the full coming of the circle of the future that some people in the military saw then, which is that, you know, just last month in Ukraine, the US military has had to ask the Ukrainians to remove from the front all 31 of the M1A1 Abrams battle tanks. That we gave the Ukrainians because a quarter of them had been destroyed by Russian kamikaze drones. So not only is it the ascendancy of an era of drone warfare, but it's probably the end of man-mechanized warfare as well on land. And we can also talk about similarly epic changes at sea and in the air and in space.

But some of the drones already are not working that well. Like the Turkish drones seem less powerful than a few years ago. What is one possible way that defense against drones might catch up?

Well, I think right now a very open question. I think drones are primarily offensive dominant in the sense that, you know, there's right now almost 300 companies in Ukraine that have faster innovation cycles at this moment than all of the Western military companies fielding drones. And so if you go to the front lines today in Ukraine, I got a chance to visit there in October with my co-author, Raj Shah, you will find predominantly Ukrainian drones operating. And the reason why is that they're literally able to go back to garage shops all over the country and overnight change the algorithm in the drone to respond to rapidly evolving Russian electronic warfare tactics that change the—

So that's the counterjamming in part. Yes.

Yes.

Yeah. Can you imagine the Pentagon having a process like that? I don't mean at this moment, but if matters were somehow more dire for us, is there a way we could do something like that or just it would never happen?

Well, you know, the Pentagon is trying. And a few months ago, they announced something called the Replicator Initiative, led by Defense Innovation Unit, which is a push to develop autonomous swarming drones for in the air and on sea and undersea also. But, you know, an important caveat to that, you know, this is an initiative that was announced more than a year after the start of the Ukraine War, which tells you we're playing catch-up. And because it was announced outside the budget cycle, they've had to go around and bureaucratically pull money from other programs,, which is a whole extra step. And if you add up all the money that the Pentagon is spending on Replicator, it's less than a percent of the procurement spend of the Pentagon. So I think the question is, that program is directionally right, but it certainly is not right in size or scale in comparison to the changes we're seeing, not just in Ukraine, but in the Middle East and other places around the world with autonomous weapon systems.

But say our forces had greater immediate combat responsibilities than is currently the case, though there are some now, as of course you know. Is it laws that would stop us from speeding up tinkering with drones, or is it cultural inertia in institutions? What's the relative balance there?

Oh, it's completely cultural inertia.

So there's no laws in the way. We're just not good at it.

There are parts of the military that are good at it, like DARPA, like Defense Innovation Unit, that have a phenomenal track record of—

But not like in a 3- to 4-day time window.

Well, some of DARPA's cyber programs and some of DIU's work does, in a month, actually produce things. And you can imagine, in the context of a longer conflict and ecosystem on the US side quickly developing. But the reality is that if you look at the Pentagon spend, even now, 2 years into the Ukraine conflict, it's largely, to be honest, along historical trend lines. So that tells me the military really isn't changing, even as many of the military senior leaders recognize how quickly war is changing.

Now, Ukraine is using drones for defense, of course, but on average, Does current drone technology favor offense or defense?

What's really remarkable about the Ukrainian drone ecosystem is, you know, they're not just building simple kamikaze drones. They're building surveillance drones that will automatically go fly, surveil the battlefield, return, upload images that can then be uploaded to, you know, killer drones that then go out. On top of this, you have people using open-source software to build fusion systems that take sensor data from EW detection systems, feed live feed from drones and put them all together on a controller that an operating unit going forward actually uses to mount a coordinated attack using multiple drone systems. So this is to say, in effect, Ukrainian startups have replicated the entire battle management system of the US Air Force. They have effectively AWACS drones, attack drones. It's quite remarkable what they've done with open-source software and hardware that is pennies on the dollar compared to any system in the Western arsenal.

If we think about Azerbaijan versus Armenia, right? Turkish drones helped Azerbaijan much more than anything helped Armenia. Is there any kind of generalization looking forward? Oh, drones are good for Israel. Drones are good for countries willing to tolerate a lot of disruption. Drones are good for fill in the blank. What would you say?

I think drones are good for anybody looking to defeat a conventional military operating in conventional ways. And I'll just list off a couple examples. So one, how was it that Hamas fighters bridged the border in Gaza? Lines, which is essentially a modern-day Maginot Line with multiple lines of defenses. The key tactical maneuver they pulled off on October 7th was using quadcopters to drop small charges on the generators powering the border tower surveillance systems. So similarly, you know, the Houthis have effectively, you know, harassed shipping now with a combination of loitering munitions, cruise missiles, and autonomous drones. And then in northern Israel, perhaps most strikingly, Hezbollah has depopulated the first 10 miles of northern Israel using loitering munitions and cruise missiles because the IDF can't effectively defend their northern border. And that has displaced 85,000 Israeli civilians. And this is, you know, the IDF is a very modern military.

Is there a forthcoming era of assassination by drone? Do we know this? Can we prepare for it? What will that look like? Presidents never step outside?

You know, to be honest, it's not outside the realm of possibility. I mean, you can ask the question today, what US military installation or important government site is adequately defended against an extremely advanced, unprompted, surprise drone attack? I would guess the number is very few, if not perhaps zero. And there are people now very actively, not only at Homeland Security, but in the Department of Defense, looking at how you defend against this. But it's very hard. Drones are small. We don't have early warning radars, in the same way that we missed the Chinese spy balloon, that are tuned to detect drones. So now people are asking questions. Can you use cell phone tower transmissions and use AI algorithms to process from that and detect flying small drones that we don't expect to be there?

US is somewhat behind in 5G compared to some other nations. For military issues, does this matter?

It's a fascinating illustration of how we, as a nation, as a state, have not made wise choices with our allocation of resources. And the story in a nutshell is that in the— Cold War, through the end of the Cold War, the military made enormous investments in a set of military systems that took up a significant part of the best 5G spectrum. So Navy radars, some other highly classified systems and covert systems, and the expense of shifting those systems away from the frequency that turns out to be best for 5G is enormous. And yet we don't have a mechanism in the government, right, that can take these sort of long-term looks at spectrum. Spectrum is an an example of where we're failing strategically. And so as a result, if you go to Beijing, 5G speeds are what, 50 times faster, 100 times faster? Because they're on a completely different part of the 5G spectrum.

But faster for what? Why should I care? I go to South Korea, everything's faster. But in terms of absolute time, I don't feel I save anything.

Well, I mean, if you think about what's coming next and what 5G will enable, you have self-driving cars, right? Self-driving cars are not going to be able to get better, along with a whole bunch of connected devices, without really fast 5G that penetrates really far. I mean, you can buy a Tesla, it's great if it drives around in the city, but what if you're trying to go visit a national park out west in a place where there is very little cell connectivity? So 5G really is a national innovation emergency that we have yet to find a solution to.

Now, I never understand what I read about hypersonic missiles. So I see in the media, China has launched the world's first nuclear-capable hypersonic, and it goes 10x the speed of sound. And people are worried. If mutual assured destruction is already in place, what exactly is the nature of the worry? Is it just we don't have enough response time?

It's a number of things. And when you add them up, they really are quite frightening. So hypersonic weapons, because of the way that they maneuver, don't necessarily have to follow a ballistic trajectory, which means we have very sophisticated space-based systems that can detect the launch of a missile, particularly a nuclear missile, but right then you're immediately calculating where it's going to go based on its ballistic trajectory. Well, a hypersonic weapon can steer. It can turn left, it can turn right, it can dive up, it can dive down.

But that's distinct from hypersonic, right?

Well, ICBMs don't have the same maneuverability, so that's one factor that makes hypersonic weapons different. A second is just speed. So with an ICBM launch, you have, you know, 20 to 25 minutes or so. This is why the rule for a presidential nuclear decision conference is you have to be able to get the president online with his national security advisors in, I think, like 5 or 7 minutes. So the whole system is timed to defeat adversary threats. The whole continuity of government system is upended by the timeline of hypersonic weapons. And oh, by the way, there's no way to defend against them. So forget the fact that they're nuclear capable, If you want to take out an aircraft carrier or a surface combatant or assassinate a world leader, a hypersonic weapon is a fantastic way to do it. And watch them very carefully, because more than anything else, they will shift the balance of military power in the next 5 years.

And do you think they shift the power to China in particular or to larger nations or nations willing to take big chances? At the conceptual level, what's the nature of the shift? Above and beyond whoever has them?

Well, right now, they're incredibly hard to produce. Right now, they're essentially in a research and development phase. The first nation that figures out how to make titanium just a little bit more heat resistant, to make the guidance systems just a little bit better, enables manufacturing at scale, so not just 5 or 7 weapons that are test-fired every year, but 25 or 50 or 75 or 100, that really would change the balance of power in a remarkable number of military scenarios.

So how much China has them now? Are you at liberty to address that? They just have 1 or 2 that are not really that useful, or they're on the verge of having 300, or—?

So what's in the media and what's been discussed quite a bit publicly is that China has more successful R&D tests of hypersonic weapons. Hypersonic weapons are very difficult to make fly for long periods. They tend to self-destruct at some point during flight. And China has demonstrated a much fuller flight cycle of what looks to be an almost operational weapon.

And where is Russia in this space?

Russia is also trying. Russia is developing a panoply of sort of Dr. Evil weapons. The most— the latest one to emerge in public is this idea of putting a nuclear payload on a satellite that would effectively stop modern life as we know it by ending GPS and satellite communications. You know, that's really somebody sitting in a Dr. Evil lair stroking their cat coming up with ideas that are game-changing, but they've come up with another number of other weapons that are quite striking. So supercavitating torpedoes that could take out an entire aircraft carrier group. Advanced states are now coming up with incredibly potent weapons.

How bad would an EMP attack be? Let's just say set off once, but over an inhabited area. Pick a smallish country, maybe an island, it happens. What's next? Cooking stoves. How many people die? Let's say there's a million people. On the island? How many do we— Advanced economy.

Yeah, well, that's the thing. I mean, with neutron weapons, you're just frying electronics. So the electricity grid will probably go down in whole or in part. If you're in the center of a strong EMP blast, a lot of microelectronics will stop working. This means that your computer is a brick, your car is a brick, gas station pumps are a brick. So EMPs are, you know, another weapon from the Cold War, right, that were tested in very sophisticated ways in the 1950s. '90s, and thankfully have not been used by a major power.

And do you think of it as a greater threat against armed forces in the field, or something you would use against a place with civilians?

I mean, we can only hope, you know, in the laws of war that most violence in war, if we have to have an armed conflict in the first place, would be constrained to the battlefield. But as we're seeing more and more, as homelands become more reachable through cyberattacks, through sabotage, through terrorism. I mean, just think about the mayhem the DC sniper caused in this city years ago, right? One person essentially effectively shut down a city. So it's hard to imagine a large-scale conflict playing out without large elements of domestic upheaval and sabotage.

The current potential for an AI arms race, at a philosophical level, do you think we have any choice other than just to try to win it? Is there some other path we could think about taking? Or just it's another arms race, we tried to win the ones before, we've got to win this one too, maybe it's more important than usual? Or is there something else that can be done?

Well, the human rights community has, at a number of junctions, tried to propose alternate pathways. And I'm very attuned to the arguments they're making, because I think they're arguments that should be listened to. And by the way, when I was a Pentagon employee through the— I think it's called the Combined Federal Campaign, we could donate a portion of our salary to a number of nonprofit charities.. And one of the charities I donated to was Human Rights Watch, which is sort of wonderfully subversive for a Pentagon employee because they actually produce incredibly great research about what's happening on the ground in conflict areas around the world that every national security official, I think, should be reading. And out of Human Rights Watch and another community of computer scientists led in part by Stuart Russell, there is this movement called Stop Killer Robots that emerged about 10 years ago, foreseeing the shift in AI, foreseeing the shift in the rise of smart micro drones. And asking us not to go down that path. I actually got to sit in a couple meetings with Stuart at the White House, and it was interesting to watch him make his argument. Although I respect him incredibly as a technologist, what was interesting to me about the way he made the argument, it was without any sense of the literature on strategic stability. Out of the Cold War, we have this fascinating part of political science that has studied what is the consequence of the spread of nuclear weapons among states.? And to the best that you can study this question in the world, the consensus in the literature is nuclear weapons have actually lessened great power conflict in the modern world, because nobody actually wants—

Because there's a frequency versus intensity issue, right?

Oh, sure. And you wouldn't want everybody in the world to have access to a nuclear weapon, but it shows you that sometimes technological parity can have a stabilizing effect in international relations. And so the question about killer robots is, I think, very similar. If we were to, for instance, unilaterally disarm in a field of technology where you really can't have verification that the other side is also disarming— and you can do this with nuclear weapons because of the specialized technology that goes into them. You can do it a little bit with chemical and biological weapons. You certainly cannot do it with software. It's just too difficult, and so I think had that group of people been successful in, for instance, pausing military development using AI, I think we would be even more imperiled in Ukraine today.

And that's because of the difficulty of monitoring software compared, say, to nuclear weapons. That would be one factor.

I think that autonomous weapons are here to stay. This is how wars are likely going to be fought. And if you decide not to pursue autonomous weapons, you're going to lose.

Right now, the US has a lead in at least many forms of generative AI. To the extent that lead is sticky, do you think that comes from chips, from talent, from data, from something else? We have a better private sector, or what's the source of that lead?

Well, it's really remarkable. I got a chance to serve on the National Security Commission for Artificial Intelligence, which studied this issue, particularly the AI competition between China and the US, quite rigorously. And China, as a state, of course, is making enormous investments in AI. Through the doctrine of civil-military fusion that Xi announced, he has essentially directed any technology and company in China to ensure it's open, its technology is open and available for use for the People's Liberation Army. So you have a lot of things happening in China that are a deliberate strategy to make them the most competitive nation they can be on technology on a variety of fronts. And yet today, particularly with generative AI, the United States does seem to have a breakout advantage. My best guess for that is, in the end, we have structural advantages in a far freer market, far more freedoms, and none of the suppression that Xi has exerted over the tech industry at arguably his political peril the last few years in his quest to suppress dissent China-wide.

My view is that They're quite afraid of AI, that they understand it will disrupt a lot of social relations. American leaders, in a sense, are too asleep to even see that. So we just let things happen, and that's for the better in the long run. But they throw their CEOs in jail, and they're afraid it will take away power of the CCP, and they're just not going to catch us.

Well, but that's certainly true for the population facing AI. So the stories that we've all read this past week about generative language models in China being held up to make sure that 60,000 censored matters wouldn't come up, right? By delaying their deployment. But that is a system of enforcement and censorship that is focused on the civilian population. There are also large institutions in China that are focused on solving military problems and advancing AI in the military that are not operating under those constraints. And so it's important to split off what we see happening with the Chinese government's attempt to control its own populace and what they're doing to try and make the PLA a fighting force that can meet the military objectives that Xi has set for them.

But as you well know, there's a lot of research by accident, so to speak, that from transformers have come a lot of advances that have military relevance. And if you try to split them off and say, we're only going to get AI that's relevant for the military, that over time you'll just be much weaker for AI in general.

Well, this is similar to a conundrum that we face ourselves. On the COVID of Unit X, we have two images. One is of the iPhone. The other is of the F-35 fighter, the most advanced fifth-generation stealth fighter in the world. What most people in America don't realize is that these technologies are produced from completely separate technology ecosystems, and they work very differently under very different incentives. The design of the F-35 was actually frozen in 2001 when the Pentagon awarded the production contract to Lockheed Martin. The garden. It didn't become operational until 2016. That's a long time in technology years. Look at the iPhone. Its processor is changed every year. So now we're in a situation where a $2 trillion program of record is producing jets whose processors are slower than what we're all carrying around in our pockets. This matters because if it turns out that 2027 AI is going to beat 2026 AI on the battlefield, then you had better be getting your AI from the commercial system of production, not from the bespoke military-industrial complex system of production.

Now, some people argue the major AI labs are not secure enough against, say, hostile foreign powers stealing American AI developments. Are there changes we could make or should make that would address this, but without killing the goose that's laying the golden eggs?

I think we have to be very careful. And I've seen— I'll bring up one historical parallel and then take a shot at one argument that I hear circulating. So the historical parallel is actually with cybersecurity, where we ended up in the 2000s and 2010s with this extraordinary situation of realizing how vulnerable critical infrastructure was to cyberattacks. This was hard for the national security establishment to wrap its head around because this infrastructure is owned and operated by the private sector, and yet it's so critical to our security. Guess what? Initially, none of the operators of that infrastructure and none of the CEOs and CTOs of those security companies even had security clearances. They couldn't talk to the parts of our intelligence community who knew about the kind of exploits that our adversaries were fashioning to go after our infrastructure. Now, eventually we solved that through, among other things, a really extraordinary public-private cooperation mechanism called the Enduring Security Framework, where we got everybody clearances, and we met, and information was exchanged. There's one example that's been— to classify this— extraordinary that I'd be happy to talk about that made safe US computers from an attack an adversary had ready. The argument I would take a shot at, though, is there is this sense that the AI models that are being produced today by the frontier labs are so important that somehow the government should sort of nationalize them. Or maybe there have been some arguments made in the last couple of weeks by people that have been employed at places like OpenAI that perhaps the national labs should step in and take a role. I think that's a mistake. Our national labs do some things very well. On the other hand, they're a place where good science goes to die. If you want to see the difference between the tech sector and the national labs, drive from Lawrence Livermore to the campus of Google and see what's going on. They are very different ecosystems. One is ruled by the Civil Service Act, the other is ruled by the free market. I think we have to be very careful exerting government control. Over breakout technology, but that doesn't mean that the law enforcement agencies like the FBI and our intelligence agencies can't cooperate very closely to help protect certain private firms from the nation-state threats they are most certainly facing. Would you reform our system of security clearances?

Is it not flexible enough or too bureaucratic, or is it just right?

I'll tell you a story. So I'm going up for my TS/SCI clearance for the first time.

Tell us what that is.

This is top secret specialized compartment information. It's the James Bond clearance you get before they let you see the spy satellite pictures and all that. You basically need it to go into any meeting on Iran and the Pentagon where I had just been appointed a young aide in the Obama administration. The security manager in the Secretary of Defense's office where I was managing got all flustered because my partner at the time, First of all, we were a gay couple, so that was one issue for him. Then the second, my partner at the time was an Indian economist, and he did not like the fact that I was cohabitating, as the Security Clearance Forum calls it, with a foreign national. So he decided that I should have a polygraph test, which is very unusual, a counterintelligence polygraph, which by the way, don't work, right? Polygraphs are demonstrative proof, you can't put them in court. So my chief of staff gets upset. What? And you know, it says, "Why are you polygraphing my political appointee just because he's in a same-sex relationship with a foreign national?" And so she calls the head of the Defense Intelligence Agency responsible for granting our security clearances, and he ruffles around a bunch of papers and looks my file up and says, "Well, his girlfriend is a dual citizen." And Sandy said, "Excuse me, boyfriend." Right? So no, I mean, our security clearances will pick up very obvious irregularities if you have extreme debt issues, if you have mental health issues, I mean, the system will pick that up, but it probably won't pick up much else.

Over the next 10 years, most countries in our world won't be able to build their own decent AI systems, but they might need to use AI even just to run ordinary governmental operations and perhaps parts of their military. How are they going to cope with this? Are they going to trust the US or trust China or What does that equilibrium look like? Does the US have a lot more power because we can withhold the AI or we'll know everything they do?

That is certainly a part of the world that we're heading in. But I think an even more fundamentally tricky part of that world is just the dual-use nature of powerful AI. So a model that's really powerful for figuring out military solutions might well also be able to cure cancer. And so we're caught in these dilemmas, which are common, actually, with technology of, is it better is the net assessment of technology. Encryption is one of these, to just let it out there because we know on the whole it makes the world a better place, even though we know with encryption there are going to be moments when terrorist cells or nation states are going to take advantage of it to successfully prosecute a small attack. So this is the dilemma that we face with AI on a grand scale.

And what do you think we should do?

If you sit in the Pentagon and it's your your profession as a national security professional to prevent 9/11s, to prevent wars from happening, you are very security-minded and you are very biased towards restricting technology because that does, in the end, help the security equation. But unless you have an argument about what to do with a particular technology with economists and trade experts and the rest of the government, you're not going to reach consensus that is going to be best for the nation. So these arguments are going to have to be fought out in the Situation Room with a variety of expertise. And having sat in both places, on the National Security Council and in the Pentagon, I am firmly in favor of dual use to the absolute extent we can, because I think the net assessment of that is for the best.

As the Cold War ended, it seems in the defense sector, we saw a lot of consolidation, much smaller number of very big suppliers. Was it a mistake to let that happen?

Bill Perry, who had his famous last supper and asked for that consolidation at the time, did it because he was terribly concerned that the so-called peace dividend, the lowering of the military budgets, it was not going to be able to sustain the larger defense industrial base that we had in place at the end of the Cold War. In his seat at the time, that was the call he made. It's turned out to be a disaster for the nation. It's essentially created 5 large companies that are so burdened by the system of requirements and the auditing and accounting requirements that go along with them to produce modern weapon systems, that the way to think about large defense primes is more akin to a utility company than it is a private sector company. The result of that, of course, has been very slow innovation, a culture of cost overruns, even as some phenomenal technology is produced. You have now, and what's really important, is a rise, thanks to Ash Carter starting Defense Innovation Unit and starting off a whole new cycle of venture capitalism, you have companies like Anduril and Shield AI and other defense startups. That are joining Palantir and SpaceX, that are becoming a new breed of product-oriented defense companies that are not burdened by the system of requirements that have turned our defense primes into electric utility companies, to be honest. I think the real strategic question now before the nation is how quickly can we swing more of the weapon systems we build to this new generation of product companies that are better at are faster, and are far less expensive for the taxpayer.

What do we need to do to keep DARPA dynamic? Well, DARPA's always been very dynamic.

But institutions atrophy, right?

You see this almost everything in the world, and often the better it is, the bigger your later problem is, because it becomes self-congratulatory.

Well, one of my great frustrations as a young aide in the Pentagon was I got a chance, both working for the Deputy Secretary of Defense and then for the Chairman of the Joint Chief of Staff to be their liaison to DARPA. So I got to be read into DARPA's entire set of programs, and they were doing incredibly impressive work. What was maddening is they were so far ahead of the military services, the Air Force, the Navy, the Army, in seeing the future battlefield. And over the years, at least since I've dealt with them starting in 2009, they've been right way more than they've been wrong, but very few of their visions of operational concepts concepts and the technology to back them have moved to the center of the systems that the services have adopted. There's a giant adoption program going on. This has to do with the fact that DARPA is a separate institution from the Pentagon, and its focus is on creating new technology. There has been a leadership failure going on for years along the E-ring that has prevented more of the technology DARPA and other of the leading labs have produced into the mainstream of the military.

How do we overcome what is sometimes called our valley of death problem? There's some intermediate stage where an idea gets started, but no one is picking up the ball and running with it. There's not yet a product that you can sell. How do we fix that?

Well, there's a couple ways to do it. Congress took a great step earlier this year by not only codifying Defense Innovation Unit in law, but also granting it for the first time a really significant budget of just under $1 billion a year. And there's a really big difference about what you can do in the defense innovation ecosystem with a billion dollars versus $100 million. So for a billion dollars, you can seed programs for long enough to not only demonstrate their success, but to ensure a high likelihood that they will successfully transition to one of the services. You can, in other words, create a springboard for a new technology.

So to spend more money, in essence, is the solution?

No, no, no. I think we need to be spending way less money on legacy systems that are demonstrably defeatable today. So, you know, anybody that's been watching Ukraine in the last two months should decide that tanks are not a great investment for a modern military. Similarly, anybody that's read even news articles about hypersonic weapons should decide that buying more aircraft carriers is not a good thing. But we do need some of those resources shifted to this new defense ecosystem that's very experimental, that's building swarming weapons. And, you know, the Air Force is very commendable under the leadership of Secretary Frank Kendall, they now have what's called the Combat Cooperative Aircraft. This is a 10,000-aircraft buy of supersonic stealth drones who will fly alongside manned fighters in the largest change of Air Force doctrine since really the Wright Military Flyer. Oh, by the way, guess who the two finalists are for the production contract for this massive program? They're Anduril on the one hand, and another non-traditional defense contractor Boeing, and later General Atomics. The traditional primes lost to those two companies for this revolutionary contract.

How is our system of inspector generals holding up?

Tyler Cowen, thinking back now that I've been in the private sector for a number of years to my tenure in government service, it's remarkable how little executive authority in government anyone has anymore. And I'll just give you a couple examples. For me to go out to a think tank, to come to the Mercatus Center for a lunch and listen to a lunch talk, when I was on the National Security Council, I would actually have to file a waiver with the lawyers to declare that I wasn't going to a lunch where I would accept an unethical gift, right?

That's because we're feeding you or because you walk in the door?

It's because you're feeding me and I have to make sure that you're feeding me less than, I think, $22 or there's some amount that changes every year, right? No sushi, yeah. No sushi. Exactly. But here's a couple more pernicious examples. So one of my first jobs for the chairman of the Joint Chiefs of Staff, General Marty Dempsey, when he took over in 2012, a year of incredible change, technological changes, he wanted to go visit Silicon Valley, meet with venture capitalists, go to Google, meet with startups, and see the kind of commercial technology that people were telling him would make a difference in the battlefield. So we organized a phenomenal trip for him to Silicon Valley that really shaped how he as he was taking over the chairmanship, shaped his agenda as chairman. We get back from that trip and I get a phone call from General Dempsey's EA, his top officer, and he says, "Chris, can you write me a couple pages of justification for the Silicon Valley trip we just took?" I thought, "Oh my gosh." Sure enough, somebody had filed a complaint with the Inspector General that we had spent government dollars to take a boondoggle to California., and IG was deciding whether or not to investigate. So here's a much more serious example. Mike Brown, the second director of Defense Innovation Unit, the former CEO of Symantec, the largest cybersecurity in the world, somebody who really understands software, which is now the most important element of any military weapon systems, was in a remarkable act of leadership and foresight nominated by President Biden to become the Undersecretary of Defense for Acquisition and Sustainment, so the lead weapon weapons buyer in the Pentagon, somebody responsible for driving how nearly $300 billion of the Pentagon's budget is spent on weapons. Just days before his confirmation hearing was scheduled, in the news bubbled up this whistleblower complaint that had been filed at Defense Innovation Unit weeks or months earlier, maybe even years earlier, alleging that he misused hiring authorities to try and pay engineers in Silicon Valley close to market rate to so they could come to work for the government, which by the way, was the mission of Defense Innovation Unit, right? Was to bring technologists into the government. It also alleged, among other irregularities, that the office snack fund called Snacko had an insufficient basis of cash accounting. The person writing the complaint was a former comptroller employee. So receiving this complaint that their director had potentially violated Department of Defense policy, the Defense Innovation Inspection Unit General Counsel undertook an exhaustive investigation, published an 82-page report, found no substance to any of the allegations that had been met. You would think case closed. At that time, the Inspector General decided not to investigate. The Pentagon's General Counsel said, "Fine, you've looked into it. There are complaints all over the place. Case closed." On the eve of his nomination, this surfaces again. The whistleblower goes to the media again. It surfaces in a couple articles, and he had to end up withdrawing. He had done nothing wrong, but somebody weaponized the Inspector General system against him. So we're in this situation where good-minded public servants who are doing the right thing, who have done nothing wrong, are caught in this just bizarre process of persecution that runs at its own civil servant speed. I think that's one of many factors, including frankly the Solar Service Act, that is making the talent base of the US government erode. I mean, why is it that governments like Singapore are so much further ahead of the United States in terms of the talent they're able to bring to their governments? And it's not just salaries, although that's a big part of it. It's part of the burden that you undergo by becoming a public servant.

So more formally, if you're given the chance to restructure the status of whistleblowers, what would you change to make it better?

Well, look, whistleblowing is an important function, and there are cases of egregious behavior, and they need to be called out. And stopped. But at the same time, leaders in government, particularly those who have to work against the grain of their institutions, are going to ruffle a lot of feathers. My co-author Raj and I found out that firsthand in trying to start up Defense Innovation Unit. On our first flight to Washington, our government credit cards got shut off out of spite because somebody wanted to slow us down. We had to rebook our hotel reservations and pay out of our pocket. That was just one of many small examples of much larger political attempts to stop DIU. So I think anybody who is working against the grain in an institution to build something new should be given a lot of authority to prosecute that mission, to make that mission a success, and held accountable if they don't. Getting wrapped up in IG investigations that end 7 years after the fact is not going to make our system a public administration. Any more effective.

Now, the title of your book again, it is Unit X: How the Pentagon and Silicon Valley Are Transforming the Future of War. Just tell us, what's Unit X?

So Unit X is Defense Innovation Unit Experimental. And it was started in 2015, the year that Ash Carter became Secretary of Defense. And it was started off an insight that Ash Carter had when he was just Professor Ash Carter in 2001, when he wrote an article Keeping the Technological Edge, that was quite prescient about strategic trends. He noted at the end of the Cold War that federal research and development was essentially flatlining at the exact moment that the technology economy was taking off on a rocket-like trajectory straight up. By the time he arrived into office, and he forecast this in his 2001 article, the market capitalization of all the major tech companies, Apple, Microsoft, Google, Amazon, were each themselves individually bigger than the entire defense industry combined. So what does that tell you? That tells you that the locus of innovation has shifted away from military labs, away from DARPA, which used to be generations ahead in technology, to the commercial technology ecosystem. The problem is that the entire defense procurement system and the Pentagon's line of vision are tuned in to the military R&D labs and to the defense defense primes. But if they're no longer the ones coming up with the leading-edge technology, the Pentagon is not structured to become, in Ash's words, a fast follower of the commercial market. Defense Innovation Unit, the Pentagon's first ever Silicon Valley office, started by Ash, the first Secretary of Defense to visit Silicon Valley in over 20 years, was an attempt to pivot the entire US military to be able to more quickly adopt commercial technology, which was clearly starting to outpace the technology being built by the defense primes.

And just to clarify, your role in all of this has been?

You know, as an aide at the Pentagon, I worked on a lot of technology issues. So I was asked by Secretary Carter to run the working group that created the concept for a Defense Innovation Unit that he then launched. And just a few months after launching it, he asked Raj Shah and myself to come take it over. He, in fact, ordered me that I would moving from my then-present position on the National Security Council to California, which I promptly did. I had to buy more pairs of jeans. I had one pair of jeans and five suits, and so I went out and bought five more pairs of jeans and was ready for Silicon Valley. And starting in 2016, began a really exciting couple of years at the helm of Defense Innovation Unit, where we, among other things, pioneered a new way to buy technology that was much faster and proved through a number of very successful pilots that you could actually use technology on the battlefield in decisive ways.

And is there some reason in the incentives to think Unit X won't ossify, or is your view simply, for now, newness is enough? It may well ossify, but then we'll create a new structure yet again.

Well, there is this Hegelian dialectic, right, in the history of defense agencies. And so DARPA was created originally to solve a crisis in spaceflight because the Air Force wasn't wasn't up to snuff. It actually spent some years in the wilderness before becoming the dynamic institution that we know it today. So far, the trajectory of Unit X has been very successful. Part of the reason for that is we have people that come there, kind of like DARPA, for a few years at a time on rotation. And they, our project managers, they're very focused on getting things done. And oh, by the way, we've pioneered a way to buy technology at Silicon Valley speed. So if you want to go buy a fighter jet or an aircraft carrier or a military radio, you're looking at using the federal acquisition rules, this sort of Old Testament-like system of regulations. Negotiating a contract typically takes 18 to 24 months. If you're a startup, that simply doesn't work, right? Because you have to show profitability before your next round of venture investment, which you're going to raise inside of that timeline, which is why when we arrived in the Valley in 2016, venture capitalists would actively pull funding from startups that tried to go after government contracts because they knew in the end it wouldn't be profitable. So DIU had to figure out a way to break through that, and we did it through the ingenuity of a single person. And this is a story, a heroic story of how one person can really affect change in the Department of Defense, a 3 million person organization. And her name was Lauren Daley. We met her when she was 29 years old. She's an acquisition specialist, stays up all night reading the newly released National Defense Authorization Act, another dictionary-sized tome of law. Our fourth day on the job, she comes to Raj and I and says, "Hey, I found this sentence in Section 815, and I think it's a giant loophole we can use to create a whole new system of contracting." Raj and I say, "This sounds fantastic. This is the biggest problem we have." We need to come up with a bigger system of contracting. If we can't do that, DIU will fail. Lauren says, "Well, great. Well, here's the 20-page white paper that I wrote about it." Quite literally, the next day, Lauren and I got on a flight to Washington and we met in rapid order with the head of acquisition policy for the department, with the head acquisition lawyer for the department, and with the Department of Defense General Counsel. They all agreed, and Secretary Carter blessed a whole new regime of using other transaction —this Apollo-era, little-used contracting mechanism—to let contracts in between 30 and 60 days. And so in 2 weeks, we got the department to declare DIU could use this new method. And in the 8 years since Laurien had that idea, $70 billion of technology has been acquired by the Department of Defense using the method that she pioneered.

Does China have something at all like Unit X? China has something different.

So in their system, they don't have one unit. That is specifically focused on acquiring advanced technology from the best startups and technology firms in China. They have, again, this doctrine announced in about 2015 of civil-military fusion that mandates that if you're producing technology that might be dual use, that might have a military application, that you have to notify the People's Liberation Army, and perhaps a PLA officer will come and join your board to watch that technology advance. If they see use case, they might decide to pull that technology into the military. So in the sense of their command economy, that is how they are deciding to ingest technology.

How well do they do procurement, in your opinion? You look at it, you think, wow, that's impressive, or you think, eh, they're doing well for other reasons, but their procurement is screwed up like ours?

Well, I think any large public institution or a large government institution, whether you're communist or any other capitalist is going to have inefficiencies in it. I think the Chinese have plenty of inefficiencies, but I think what's different is they have observed our military very carefully, really ever since the first Gulf War when the precision strike complex, as it's called, really became debuted on the global stage, and the whole world saw us on CNN dropping bombs down chimneys, right? So today, similarly, the entire entire complex of Chinese universities that study national security and military institutes are looking very closely at Ukraine. By the time the book was published, I think we had caught something like 120 papers, academic papers being published in the Chinese literature about the Ukraine war, very clearly seeing the decisive use of Starlink, calling for a Chinese version of Starlink, calling for Chinese kinetic and non-kinetic means of jamming low-Earth orbit space technology. What's different about their procurement system is, however good or bad it is, it's geared at producing the exact kind of technology that will find the chinks in our own armor and could, if we have to go to war in a Taiwan scenario, in another scenario, prove to be a potentially catastrophic awakening to the US military.

Is Unit X a dog-friendly office? Highly important question.

I arrived out in California with my golden retriever, and I was jealous of Secretary Leon Panetta, who both as CIA director and as Secretary of Defense brought Bravo, his golden retriever, to the office. Bravo, he's proud to say, was one of two canines cleared into the bin Laden operation, the other one being the German shepherd that flew on the helicopters into Abbottabad. I thought, here we are, we're in Silicon Valley, we're supposed to come up with an office with a different culture, let's make our office dog friendly. We did, but we had to go to war to do it. We were in a National Guard building and there was some colonel there in charge of— he was kind of our landlord. Even though there's no regulation against having a dog-friendly office in the Department of Defense, and we had published our dog policy, he threatened to evict us unless we ended our policy. His last name started with F, so we referred to him as Colonel Fear. And so—

And he's against the dogs. He's against the dogs.

He's against the AU. He's an old-school colonel. He doesn't like any of this. So how do we not get evicted? Well, we figure out he's retiring in 6 months. So what we do is we tell him, "Oh, you know, we don't actually have a dog-friendly policy. We have a service dog policy. And we're unable, based on the reasons our Air Force physician has declared, for each individual to have the 3 dogs in the office that we had through HIPAA reasons to tell you why, but we can assure you that HIPAA compliant paperwork has been filed declaring these service dogs. So we got our dog-friendly office and Colonel Fuhr went off to retire and not be heard from again.

What's the biggest problem with the dog-friendly office?

You know, there are some people that are allergic to dogs, and so that's why many major tech companies that are dog-friendly will have to have dog-free areas. But in my view, you know, especially now at a moment when offices are empty in most of the country, why not make them more fun? And boy, it was fun to watch Secretary Mattis come in, who himself loves dogs, and be greeted by my golden retriever. And boy, was it even more fun to see crusty 3- and 4-star generals that were very skeptical that Silicon Valley had anything to offer to first see my golden retriever and then see the woman who briefed our software demos whose hair was blue. By the end of the day, nonetheless, be really impressed by the technology we were putting in front of them.

Now, you wrote a political science dissertation at Cambridge on governmental commissions, right? Yes. What is it you learned from that dissertation that helped you do Unit X, military procurement, more innovation?

I'm really grateful I got the chance to go do a PhD, and in particular to do it in the British system rather than the American system. Why? It's shorter. That's always nicer, 3 years rather than 5. But rather than spend all my time studying for qualifying exams, I actually got to read the classic tomes of political science and sociology and understand them. I'm so grateful for that chance. I ended up studying breakdowns in the national security state. So what do you do after a catastrophic breakdown that the political system, the regular political system can't contain? I ended up studying government commissions in part because I had had I had the chance early in my career to serve around a couple after-action reports and investigations. The first, my first job out of college in 2003 was for the Space Shuttle Columbia accident investigation, a look at NASA's safety culture, not only what was the physical cause of the Columbia disaster, but what was the organizational cause behind the physical cause. My second job was a chance to go to Iraq and work on a team of historians that was writing an official history of the Reconstruction, another major breakdown in American foreign policy. In fact, I've gone on to specialize in disaster and a few other areas. I wrote the Obama administration's after-action report of Ebola. I used to joke that, you know, if I showed up in your office, you were really having a bad day. But what I learned is that, you know, expertise is often wrong. And settled notions of how business should be done is also often wrong. And those settled notions often kill people. And especially when they kill people, that's a strong signal that it's time to go back and take a fresh look at everything and to always be curious, always be skeptical, and always be as rigorous as you can because established ways of thinking don't often evolve in ways fast enough to accommodate changing circumstances.

What is it we need a commission for today where it might maybe plausibly succeed?

Well, the political science literature divides commissions into 3 types. The type that you don't want to be on is a disaster management commission. So this is where, oh, there's a giant political problem. We're going to name a blue ribbon commission that will never be heard from again to take the blame off politicians. The hardest kind of commission to be on is an agenda commission. So an agenda commission like the National Security Commission on Artificial Intelligence wasn't created after a 9/11 or a catastrophic breakdown, but nevertheless, AI Commission is a body that is going to have the bipartisan insights and support and the expertise to tackle AI, a really complex issue that spans government that, at least when the AI Commission was created, the government wasn't in a position to itself take on in a really rigorous way. The most effective commissions are what are called crisis commissions. The 9/11 Commission is a classic example. They happen in the aftermath of a giant political rupture. When everybody knows there's been a huge breakdown, where there's more openness to look freshly at it. And that commission went pretty well. A very high percentage of its recommendations were adopted, and more than that, it forcefully brought into the public light information that never otherwise would have made it out about who knew what and when. So I'm a strong believer in commissions as a form of government. The British government has them much more frequently than we do. They don't always work. I mean, an interesting historical anecdote, the first commission in the US was actually the Whiskey Rebellion Commission. It didn't work. George Washington had to get on his horse and ride west before the rebellion in Pennsylvania broke up. But as a form of gathering concentrated expertise for a short problem to dive deep on a topic, I think they're very powerful instruments.

And final question, what is it you want to learn about next?

Anything. I got a chance this spring to work at Anthropic in a policy residency, working both on AI safety and AI and national security. It was a short-term residency. I was only there for a few months, but I am incredibly enchanted by generative AI, and I cannot believe, based on my time inside a Frontier Model Lab, what is just around the corner. So I want to learn everything I possibly can about how to steer, democratize, and make safe this incredibly powerful new wave of technology that's about to crash down upon us.

And how would you describe what you think is around this corner?

On the national security front, there's been a heck of a lot of conversations about, is AI sort of as dangerous as the next nuclear bomb? And everybody at Frontier Labs went out last summer and watched Oppenheimer. It's interesting just culturally coming from the national security community where it was my day job to look at really awful stuff, right? To get up every morning and before the chairman came into the office or on the National Security Council, read the morning intelligence brief about all the things, all the awful things going on in the world every day. So I look at AI as a technology is not at all like nuclear. I never think it's going to achieve the sort of zero-to-one point technology, as we would call it, that nuclear is today, where you either have a bomb or you don't. It's going to be much more diffused. It's going to be much more a part of the commercial economy. It's going to be much more a question of, do you have an iPhone 15 versus an iPhone 12? Do you have the most advanced AI versus AI that's a little bit older, maybe from an open-source model. And because of that, as an international relations theorist, and as the literature would suggest, it's not going to be like nuclear weapons. It's not going to create sudden disequilibria in the international system. It's going to be much more shades of gray, I think, and gradations. And that's good because it means that we will have time among nations to find new equilibria without hopefully finding ourselves themselves in sudden disequilibria that could create an incentive for one nation to start a war against another.

Again, the new book is Unit X: How the Pentagon and Silicon Valley Are Transforming the Future of War by Raj M. Shah and Christopher Kirchhoff. Chris, thank you very much.

Thank you, Tyler.

Thanks for listening to Conversations with Tyler. You can subscribe to the show on Apple Podcasts, Apple Podcasts, Google Podcasts, Spotify, or your favorite podcast app. If you like this podcast, please consider giving us a rating and leaving a review. This helps other listeners find the show. On Twitter, I'm @TylerCowen and the show is @CowenConvos. Until next time, please keep listening and learning.