Hosted by John Raine and Sir Richard Dearlove
Guest
Paul Scharre
Paul Scharre is executive vice president at the Center for a New American Security (CNAS) and an expert on military AI. His work centers on how artificial intelligence and autonomous systems are changing warfare and the competition between the United States and China. In 2023 TIME included him on its list of the 100 most influential people in AI.
Before CNAS he worked in the Office of the Secretary of Defense, where he had a lead role in setting policy on unmanned and autonomous systems and emerging weapons technologies. He helped draft policy guidance for the 2012 Defense Strategic Guidance, the 2010 Quadrennial Defense Review, and secretary-level planning guidance. Earlier, he was a special operations reconnaissance team leader as an Army Ranger, with multiple tours in Iraq and Afghanistan.
His first book, Army of None: Autonomous Weapons and the Future of War, won the 2019 Colby Award and was named by The Economist among the top five books for understanding modern warfare. He followed it with Four Battlegrounds: Power in the Age of Artificial Intelligence, which also won awards. He has written for the New York Times, Wall Street Journal, TIME, Foreign Policy, Foreign Affairs, and Politico, and appears on CNN, MSNBC, NPR, and the BBC.
He holds a PhD in war studies from King's College London, and an MA in political economy and public policy and a BS in physics from Washington University in St. Louis
Episode Summary
Between cheap-to-produce drones and artificial intelligence, the future of warfare may already be here. Paul Scharre, former US Army Ranger and executive vice president of the Center for a New American Strategy, joins former head of MI6 Sir Richard Dearlove and co-host John Raine to discuss how drones and AI are reshaping military conflict and militaries themselves. Scharre explains why cheap drones are beating expensive armies, how the cost asymmetry between offense and defense is reshaping the battlefield, and why the US may be losing its technological edge faster than it realizes. He also assesses Ukraine's drone revolution, the race between the US and China to adopt military AI, and why keeping humans in the loop may matter more than any weapon system.
In this episode:
(00:00) Cost Equation of Modern Warfare
(02:39) Defining the Present State of Warfare
(04:18) How Militaries Are Dealing with AI
(06:24) Technology Reshaping US Military Education
(09:11) Evolution of Drone Technology
(14:45) Cost of Countermeasures
(18:06) Likelihood of Protracted Warfare
(20:38) Redefining Victory in Modern Warfare
(22:20) Adversarial Distillation
(25:15) Strategic Balance Between US and China
(27:32) Culture of Superior Firepower
(29:55) AI and the Ukraine War
(33:06) How Drones Change Ground Combat
(39:28) Outlook For Expensive Tech
(43:00) Keeping Humans in the Loop
(44:44) Drone Swarms
(46:55) AI and Naval Warfare
(48:52) Drones and Non-State Actors
(54:32) One Decision: Xi-Trump Summit and AI
(56:36) Takeaways and Wrap Up
Hosted by Sir Richard Dearlove (former MI6 Chief) and co-host John Raine (International Institute for Strategic Studies)
Listen & Subscribe
- Watch on YouTube
- Listen on Apple Podcasts
- Listen on Spotify
Transcript
COLD OPEN
[00:00] SCHARRE: If you shoot down a thousand-dollar drone with a million-dollar missile, you're losing every time. It's not a winning proposition. And we've seen in recent conflicts a number of countries be on the wrong side of this cost equation. Certainly that's true for Ukraine, where they face a lot of situations where they have a limited number of Patriot air defense missiles, for example. And it's not worth shooting down Russian drones. They have to save these scarce missile interceptors to go after Russian ballistic missiles, the higher-order kind of threat.
INTRODUCTION
[00:30] RAINE: Hello and welcome to One Decision. Conversations about intelligence, statecraft and the forces that drive world politics. With people who've been in the room and on the ground. I'm your co-host, John Raine.
[00:50] DEARLOVE: Hello, and I'm Sir Richard Dearlove, the former Chief of British Intelligence. And John, it's really good to see you and to have another really important podcast to discuss.
[01:04] RAINE: And great to see you too, Richard. We have a rich interview for you today. To keep up with all our content, subscribe wherever you get your podcasts. And we'd love it if you could like this episode or drop us a comment on YouTube. That will help others find us.
Right, let's get to the business in hand. Today we're looking at nothing less than the future of war. The Russia-Ukraine war and the conflict between the US and Iran have a common thread. They are telling us that we are now entering a new era of warfare where cheap drones and commercial tech can rival expensive military platforms costing tens of millions of dollars. And this feels like just the beginning. As artificial intelligence matures, the battlefield as we've known it will shift even more dramatically.
Our guest this week has spent the last decade considering where this shift might take us and in particular how the United States should prepare for the transformation. Paul Scharre is Executive Vice President at the Center for a New American Security in Washington. He's worked on autonomous weapons policy at the US Department of Defense in the Bush and Obama administrations. And before that, importantly — Paul completed multiple combat missions in Iraq and Afghanistan as a US Army Ranger. He is the author of Four Battlegrounds: Power in the Age of Artificial Intelligence. Paul, welcome to One Decision.
We're talking today about nothing less than the future of war. So it's an enormous subject and great pleasure to have you here to guide us through it. I thought we might start by just asking you what characteristics do you see as defining the new drone-centric warfare, and maybe give us a little bit of a feel of how we got there.
WHAT DEFINES THE NEW DRONE-CENTRIC WARFARE?
[03:12] SCHARRE: Thanks so much for having me. We are in a moment of very profound change in warfare. The age of AI warfare is here. This is no longer something in the future, it's here now. We're seeing artificial intelligence being used in conflicts around the world — in Ukraine, in Gaza, in the US conflict with Iran. We're also seeing drones being used at a massive scale. Ukraine is producing four million drones a year.
One of the attributes of this new era is just the scale that's possible with cheap drones and missiles. Another example we saw of this was the barrages of missiles and drones that Iran has unleashed on the region more broadly in the Middle East over the course of the conflict with the US and Israel. So there's a physical scale to this era of machine-driven warfare, but there's also a cognitive scale that we're just starting to grapple with, where AI is processing information and helping people make decisions. There was just a report that came out this last week about an AI system hallucinating an intelligence report that almost had the US board a Chinese ship in the Middle East. So it's a wild time — the way that warfare is evolving very, very quickly with technology, and people are just struggling to catch up.
HOW ARE MILITARIES ADAPTING?
[04:39] RAINE: And Paul, everyone is coming to terms with pervasive AI and a new pace and scope to the technology that AI drives. In the military, which is a notoriously well-defined culture, how do you think that militaries themselves are coming to terms with this? Do they see this as just another weapon in the arsenal? Or do they think, as clearly you do, that this is a whole different era where everything gets redefined?
[05:17] SCHARRE: Well, militaries tend to be relatively conservative institutions when it comes to making changes and adopting new technologies. And the fundamental reason for this is we don't fight wars very often, thankfully. That's the good news. And so militaries tend to be cautious about changing the way that they're fighting. When it's adopting an improvement on an existing technology within a paradigm that they understand, militaries are pretty supportive. So a slightly better missile, a slightly better radar or stealth. But when it's something that just totally upends a paradigm for war fighting — all right, we're going to get rid of horses completely and we're going to use tanks — or we're going to take these airplanes and put them on ships at sea and use aircraft carriers and airplanes as the way to fight at sea instead of battleships — militaries are actually pretty slow to make those changes because they don't get the kind of feedback about what works the same way that, say, a company is getting in the marketplace. They release a product, they see whether people buy it or not. For militaries, they might go decades between major conflicts, and they need to know whether something is going to work. So within militaries you see a mixed approach with artificial intelligence. Militaries are excited about it, they're trying to adopt it, but they're not going to wholesale throw out the ways that they have been fighting. They're not going to say, drones are here, we're going to get rid of fighter jets and bombers. It's more like, how do we use this to complement what already exists in the force?
IS MILITARY TRAINING CHANGING?
[06:52] RAINE: I have to ask you, since you've been through US military training, is it impacting on what is taught in US military academies?
[07:03] SCHARRE: I think to some extent. Certainly there's excitement about using artificial intelligence. We're seeing the US military has pushed AI tools — large language models, chatbot tools like ChatGPT, Claude and Gemini — to both military and civilian personnel across the broader defense enterprise. So they're putting these tools in people's hands. But like a lot of industries, a lot of companies, people aren't sure what to do with AI yet. It's like, okay, I have this chatbot, I have this language model — how do I use this thing to increase productivity? It sounds like a great idea. It's actually tricky to figure out how to do this in practice. And I think we see that historically with new technologies, whether it's the advent of electricity and the electrification of factories during the Industrial Revolution, or introducing airplanes and tanks and submarines into warfare — getting the widget is only the first step. Figuring out what to do with it is really where the hard work happens, and that is going to take time.
BUILDING THE WARRIOR OF THE FUTURE
[08:12] RAINE: I have to ask too, Paul. Is military training still fundamentally about building the warrior, or is the warrior something new these days?
[08:23] SCHARRE: Yes and no — both, right? I think what we've seen over the long haul of human history in warfare is that the skills that you need warriors to have evolve quite a bit. I was an Army Ranger, trained as a sniper — marksmanship is really core to what we do. I spend a lot of time sitting behind a rifle, aiming and shooting, trying to hit a very small target. If you go back far enough, that wasn't a skill that people had. People were fighting hand to hand with swords and axes. And that's not really a skill that we prioritize — even infantry troops today maybe get half a day of bayonet training, but that's not where you focus a lot of your time. So I think we have seen warfare transition over the last 100 years or so to be much more technological. If you look at what a pilot does, or a submarine officer, they're managing a lot of complicated technology. And there's more and more of that. But it is probably a slow evolution over many generations — that shift towards a much more focus on using technology and knowing how to harness it effectively — and the sort of physical attributes, you know, being relevant in some cases. If you're a grunt out on the battlefield, you have to carry a bunch of gear around. But that is diminishing slowly over time in terms of importance.
THE ORIGINS OF THE DRONE REVOLUTION
[09:53] DEARLOVE: Paul, it's always been the case that warfare has driven and accelerated innovation. Needs must, under the pressure of warfare. But given this drone revolution, I'm curious — how did it actually happen? I mean, after 9/11, I was certainly very rapidly in the States and witnessed drones being used live in Afghanistan. So the Americans had already developed some capability. That's a long time ago. But it seems to me that the real acceleration of the technology has been in Ukraine, perhaps almost as an act of desperation, given that we all thought Ukraine would be so rapidly overwhelmed by the Russian army. But it just didn't happen. What are your comments on the roots and origins and the speed of this development? Because it has been around for a long time, but it didn't come to the forefront until Ukraine, really.
[11:16] SCHARRE: I think that's right. There are these twin evolutions of drones in warfare. One is this sort of post-9/11 era of the US military and the CIA using drones to find and track terrorists. And that's a certain kind of drone. It's basically like a small airplane — a Predator or Reaper drone, something that flies at maybe a medium altitude, maybe 30,000 feet, has some missiles on it, a very high-end sensor that's used for both daytime and nighttime vision using infrared to find and track targets. And that type of technology has proliferated. We're seeing other versions from Iran, Russia, China, Turkey that are cheaper, more cost-effective versions of this that have kind of spread around the world. And then you've got this whole other evolution of drones, really coming out of the commercial space — things like commercial quadcopters, which really look more like a flying smartphone. A DJI drone, for example — DJI is a Chinese company — they're super small. And that's also a big part of what we're seeing in Ukraine. Both of these kinds of drones are being used very widely. And underlying all of this is really the technology coming out of the information revolution that's making all of this possible. It's cheaper sensors, cameras, computer networks, computer chips on board. The physical attributes of the drone are made possible by things like electric motors and batteries. But it's not the physical performance characteristics that's actually revolutionary here. It's really the digital guts inside it that are powering it that's making this whole drone revolution possible.
PRECISION, ACCURACY, AND COMMERCIAL TECH
[13:13] DEARLOVE: One of the things about the technology which I find most striking — and maybe that's an appropriate word — is the degree to which it can be accurately targeted at a massive distance. I was talking recently to the manufacturer of some of the chips used in some of the drone technology, and he was saying to me that now they can fly it through a car window at 600 kilometers. It seems to me that that sort of directional control technology is almost more important now than any other aspect of the development. At what point did those technical elements aggregate into what we're seeing today?
[14:06] SCHARRE: Well, that's a great example actually, because what's making that possible? It's not that the engines or motors behind drones have had some revolution that allows longer range — that technology's been around for a while, getting more cost-effective, sure. But what's really amazing here is the accuracy that you're describing. And there's a whole bunch of digital technologies that are making that happen. Ukraine's been conducting a lot of long-range deep strikes inside Russia, and to do that effectively you need to not rely on communications links because those can be jammed. So the drone is operating autonomously. How's it doing that? Well, one of the ways is you upload satellite imagery into the drone and map a route, and the drone has a camera looking down, comparing what it's seeing to what it's expecting in the satellite imagery, and using that to find its path. To do that, you need cheap commercially available satellite imagery — that's relatively new. You need computer chips that are pretty capable to do the onboard processing. You need high-resolution cameras. This is the same kind of guidance system that the Tomahawk cruise missile has had for decades, but it used to be this really exquisite defense technology that only the US military had, and now it's available anywhere for a couple thousand dollars. So this is really the democratization of information technology, making it cheaper and more widely accessible — that's what's enabling what you're talking about.
THE COST ASYMMETRY PROBLEM
[15:42] RAINE: It's a fascinating insight into how what we now see as military capabilities just rests upon the steady increasing availability and reduction in cost of the component parts. It leads to this extraordinary situation where you have the weapon itself costing very little and the countermeasure costing a fortune. Do the costs of countermeasures come down, or are we stuck with this curious asymmetry that plays into the hands of underresourced non-state actors and criminal groups? Is there a point at which it starts to level up?
[16:37] SCHARRE: I think the costs will come down, but the net effect is going to be to relatively benefit less capable actors — non-state groups, terrorists, smaller countries. Right now the cost of countermeasures to shoot down drones heavily favors the offense. If you shoot down a thousand-dollar drone with a million-dollar missile, you're losing every time. It's not a winning proposition. And we've seen in recent conflicts a number of countries be on the wrong side of this cost equation. Certainly that's true for Ukraine, where they face situations where they have a limited number of Patriot air defense missiles, for example, and it's not worth shooting down Russian drones. They have to save these scarce missile interceptors to go after Russian ballistic missiles, the higher-order kind of threat. In fact, one of the tactics that we've seen Russia employ is to send a wave of cheap drones to soak up some of the missile interceptors and then follow that up with ballistic missiles. And Iran did the same to the United States, severely depleting the US global stocks for missile interceptors over the last six months or so in the conflict between Iran and the United States. And Iran's been using pretty cheap drones and ballistic missiles. It's not just a cost factor here — for the US and Ukraine and others, there isn't some other inventory of these missiles available. They're working on ramping up production, but the supply chains have bottlenecks, and it's going to take years to refill that inventory. So that's a huge constraint. People are working on lower-cost interceptors and cheaper missiles and drones to shoot down cheap missiles and drones, so there are options coming that are going to bring the cost down. But in the long run, what we're seeing is warfare transitioning to this sort of war of attrition of cheap missiles and drones, where the economics of warfare matter a lot. What analysts call the cost exchange ratio between offense and defense — what is it costing one party to send more of these cheap assets to attack versus to defend against them. That's a different kind of game, where you're focused not just on how to make the most exquisite stealth fighter jet — which is kind of where the US military has been focused for the last 40 years or so — but on how to make things cheap, how to crank them out fast, how to make them affordable. And that's a ball game that tends to favor less capable actors.
A KALASHNIKOV MOMENT?
[19:17] RAINE: It feels a bit like a Kalashnikov moment, Paul, where suddenly somebody comes up with a weapon which is very cheap to produce and in an instant redefines warfare and combat. But I just wanted to ask you one more thing on this asymmetry. Does it also build in a protraction in warfare? Because one of the things that struck me about the Gulf conflict is the asymmetrically disadvantaged actors — the Houthis, the Iranians, all the other parties — they've managed to keep going because you don't have to deal a decisive blow. You just have to keep the conflict going, oblige your adversary to be on the defensive, to be burning up money, and eventually you win on the economics. Does it inherently lead to protraction or less decisive warfare?
[20:14] SCHARRE: I don't know that it's inherently so, but it certainly can. Obviously the Ukraine war has been going on for four years now. But the Iranian conflict was a good example of this dynamic you're describing, where there was a huge uptick of strikes on both sides early on in the conflict — of Israel striking targets in Iran, the US launching some 13,000 strikes against Iran, and Iran launching waves of drones and ballistic missile attacks back against a variety of targets in the region. But then it's sort of transitioning over time to this longer tail phase of conflict where neither side could get a decisive advantage, where the US is chasing around mobile targets in Iran — mobile missile launchers, mobile drone launchers, military commanders who are hiding. All the big targets have already been hit. But the US can't stop Iran from launching more drones and missiles back. There are just too many of them. They're too good at hiding. They're able to reconstitute. And so the number of Iranian strikes sort of diminished but never went away. And that is, I think, a feature of this kind of conflict — it's a war of attrition in the sense of both larger numbers and the cost equation mattering. It can also be a more protracted conflict, one in which the key factor is who's able to reconstitute their forces faster, who's able to have an industrial base that can compete effectively, and who's able to just bear the economic and political costs. Who wants it more, who's willing to pay that price? And for the United States, that's a kind of warfare that the US military hates to fight. But that's some of the reality of warfare today.
NOT LOSING AS A STRATEGY
[22:10] RAINE: Seems to be war without the decisive maneuver.
[22:13] DEARLOVE: In a manner of speaking, it introduces the concept of — rather than victory and defeat — it introduces the concept of not losing. And if you don't lose — I would put the Iranians in that category at the moment, and the Ukrainians — it puts you in a relatively strong negotiating or political position. That has sort of shifted the gauge on how we fight.
[22:57] SCHARRE: Absolutely. There are certainly some conflicts where it can devolve into simply who's willing to outlast the other person or pay a larger price. But I think there's also this dynamic you're describing where, going into a conflict, oftentimes for the perceived militarily weaker party, just surviving can be beating the spread, and they can come out politically much stronger. And I think that's really clearly the case with Iran in this conflict, where the Iranian regime has continued to persist as a threat in the region. Militarily, of course, they are weaker than they were six months ago before the conflict began. But politically, they're much stronger because they've shown that they can weather these strikes from the US and Israel, and they're still here, still disrupting traffic in the Strait of Hormuz, still causing problems in the region and attacking bases in the Gulf. Whether the US and Israel like it or not, they are a force that they're going to have to deal with and the US is going to have to negotiate with in order to get stability in the region. And so that certainly makes Iran come out politically much stronger.
ADVERSARIAL DISTILLATION — AI THEFT
[24:07] DEARLOVE: Paul, in your Foreign Affairs article, you use this phrase adversarial distillation. Can you explain how that applies to the theory you've put forward about the commercialization of cheap warfare?
[24:33] SCHARRE: Okay, so it's a bit of a technical term. These large language models that people are using — Gemini, Claude, ChatGPT — are massive neural networks trained on billions, trillions of words. These neural networks are very large and expensive to run. Now, a technique that AI scientists use to take a bigger neural network and make a smaller version of it that is almost as good is called distillation. They basically take the outputs from the larger neural network and train a smaller one on it, so they can piggyback off all the effort and data and computing power used to train the larger one. That might be helpful if you want an AI model that is more cost-effective, cheaper to run, that could be run locally on your device — on a laptop or a smartphone. So that's an AI technique.
Now, adversarial distillation is when another AI company does this to somebody else's model. It's against the terms of use for those models. And a lot of Chinese companies are doing this right now to US AI companies, using it to draft behind frontier AI companies in their development. If you look at the most advanced AI models out right now, they're coming from US labs — from Google DeepMind, Anthropic, OpenAI predominantly. The Chinese companies are only a couple months behind, maybe about six months behind. One reason is they have smart engineers and scientists, and credit to them for that. They are also getting access to chips that they're not supposed to have — smuggled into China, or obtained through third parties, or accessing computing hardware in Southeast Asia. But also one of the things they're doing is they're getting information from these US companies, taking the outputs of those models and using it. I hesitate to call it stealing or intellectual property theft because it gets into some tricky questions. It's worth pointing out these AI companies themselves trained their models on a whole bunch of copyrighted data, which they said was legitimate. So that question gets a little dicey. But I will say it's against the terms of use for these models, and it is a big reason why China is so close behind the United States in AI right now.
THE MILITARY BALANCE BETWEEN THE US AND CHINA
[27:16] RAINE: That does raise the intriguing question: if there were a military conflict between China and the US, what is the current strategic balance between the two? Given that one of the factors is also how quickly you can cycle your innovation as well as what your inventory is — where does the military balance lie between the two commercial superpowers?
[27:51] SCHARRE: In the AI space specifically, there's a lot of excitement about which companies are ahead. I would say for military purposes, that doesn't matter very much. If the leading US companies are a couple of months ahead of Chinese companies, but it takes both the US and Chinese militaries a few years — charitably two to three years, maybe more like five to ten — to figure out how to use AI effectively, being a couple of months ahead on the commercial side just doesn't matter. The real race on the military side is in adoption, not development of the technology in the commercial sector. And that's a big challenge.
Now AI is only a piece of this, of course. There are other aspects to the military dimensions of the conflict. The biggest one is geography — if the US and China fought a war over Taiwan, that's in China's front yard, right on their doorstep. The United States has to travel across the entire Pacific. And China's made a lot of investments in long-range ballistic missiles, anti-ship cruise missiles, submarines, to slow down the United States and make it harder for the US to project power into the region. So that's going to be a big challenge for the United States. But a lot of these new technologies — AI, drones — are pretty widely available to everyone. In some ways they're like the opposite of stealth technology that came out of secret US defense labs and was very slow to proliferate. So the net effect, I think, is to relatively advantage other players instead of the US military. And that's a real hard challenge for the US to adapt to.
AI IN THE UKRAINE WAR — TARGETING AND COMMAND
[32:10] DEARLOVE: Paul, we talk about Ukraine, AI, the battlefield, but how more precisely would the Ukrainians now be using AI to coordinate their capabilities? Is it mainly about targeting, or is it more profound than that? And to what extent is command and control actually being handled by AI rather than humans? And I'm sure we remember that brief period in Ukraine when Trump switched off the intelligence feed from Wiesbaden — how does that all fit together?
[33:10] SCHARRE: So let me talk about ways that AI is being used today, and then what that might look like going forward. Today, certainly it's possible to use AI image classifiers to sift through drone video feeds and satellite images to identify objects. The US military is doing this. If you have lots of satellite imagery or drone imagery, you just don't have enough people to watch all that. There aren't enough human eyeballs to actually watch all of it. AI can do that and can find objects.
AI is also getting good enough to be miniaturized. The chips are capable enough to put some of that on the drones themselves. And there have been reports coming out of the conflict in recent months about both sides — Russia and Ukraine — having image classifiers using AI on a drone so that if the communications link to the human is jammed, the drone doesn't just hit a fixed target. It could fly around and look for something — maybe ground vehicles or some other kind of target — to then swoop in and attack. That's an autonomous weapon. That's a weapon now that is actually making its own decision, in a small degree, about what to attack on the battlefield. And it's certainly a small technological step, though I would argue from a moral or ethical standpoint it's a big line to cross — a profound one in human control over warfare.
Right now, things like command and control are still being done by humans. But if you switch to a different kind of conflict that's much more dynamic — for example, the US air campaign against Iran, where you've got mobile and relocatable targets, things happening in real time that are very dynamic — the US was using AI to help process all that information, feeding it into humans still making the decisions. But that enables things like an airplane in route to its target receiving new targeting information while it's flying — humans still all involved in that chain, but AI helping to do that faster and hopefully better.
DOES THE INFANTRY SURVIVE THE DRONE AGE?
[35:33] RAINE: Paul, the idea that drones will redefine warfare is gaining a lot of currency. But warfare's bigger than just the war between the weapon systems themselves. And I just wonder what are your thoughts on where this leaves conventional concepts of warfare — for example, in the British tradition, the poor bloody infantry have to go in and either take possession or suppress resistance. Does that persist in military circles, or do you think quite a lot of roles — both before and after, you know, preparation of the battle space — are now done on an automated basis? And are there fewer opportunities for medals for the US Rangers?
[36:28] SCHARRE: As you can imagine, I'm quite partial to the infantry here. I think that perception will still persist among infantry troops. But look, we're a long way from the days of the Swiss pikemen when the infantry was the vanguard of ground combat. It's not anymore. It hasn't been, at least since the days of the Blitzkrieg in World War II, where tanks really supplanted the role of infantry and cavalry in maneuver warfare.
What we've seen in the last couple of years in Ukraine is drones taking on a new dimension. If you rewind the clock a couple of years, very early on, the war in Ukraine stalled into this kind of World War I-style trench warfare. Over the last several years, drones have played a slowly incremental role in changing that. At first, drones were kind of incidental or acting like spotters for artillery. But if you look at where the conflict is today, it's actually quite different. The front lines have thinned out considerably. There are still ground troops on the front lines — infantry — but much fewer than there used to be, because you don't need them. The drones are actually what's detecting enemy advances. In fact, you want fewer troops at the front lines because they're just targets for drones. We've had situations where ground troops have surrendered directly to drones. The main threat on the front lines isn't machine guns now — it's drone warfare, the threat of being spotted by drones. It changes the dynamic on the ground considerably.
I think the paradigm shift is analogous to what machine guns did to ground warfare in World War I, where machine guns dramatically increased the lethality on the battlefield by about a hundredfold. It turned out that the old tactics of just marching forward in a straight line were not effective against machine guns. A similar transition is happening because of drones. The drones are always there overhead, always persistent. It creates a new kind of transparency and lethality on a battlefield that Russia and Ukraine are being forced to adapt to. And I think we'll end up in a place where people figure out how to unlock maneuver in that kind of world, but it's going to be a different kind of fighting. There'll still be grunts, but they're probably not going to be the vanguard. It's probably going to be drones in the future.
BREAKING THE STALEMATE
[39:37] DEARLOVE: Just following further on that thought — the situation in Ukraine seems to be that the Russians, strictly speaking on the offensive, haven't been able to aggregate their forces to break the Ukrainian line. Presumably because the moment you start aggregating, even if you're 60, 70 kilometers back from the front line, it's detected and you can send a swarm of drones to break up that concentration. How might maneuvering on the battlefield change? Because it seems that's what's become frozen — and it's much easier for the Ukrainians to defend than to go on the offensive and get back territory they've lost, although there have been areas where they've been reasonably successful recently. Is this a stasis? Are we stuck, or is there going to be an innovation somewhere that will change that?
[40:46] SCHARRE: There's clearly currently a stasis in the conflict between Russia and Ukraine, and neither side knows how to break free to maneuver again. I do think that will change, whether it changes in this war I think remains to be seen. But the dynamic is exactly what you're describing — in order to mass and concentrate forces to break through, it takes time, it's very visible, the drones can see you, and now they can attack you. How do you break out of that cycle?
You could imagine massing firepower through drone and missile attacks in a different way — waves of drones and missiles, deep strikes, to blind an adversary's communications, striking their command and control networks — that would then allow the space in the conflict and the time to start physically maneuvering. So it may be that moving ground forces, tanks and infantry troops and ground vehicles, is the second phase that happens after you first use drones and missiles and cyber attacks and electronic warfare to blind and disable your adversary. That could be one way out of it. But it's very possible we end up in a place where there's just a different tempo of conflict, where whatever breaks the stalemate is going to be faster than maneuvering with physical troops on the ground because of the dynamics you're describing.
THE FUTURE OF EXPENSIVE PLATFORMS
[42:35] RAINE: Paul, one of the things you've referenced a couple of times is the tech base that these weapons systems are coming from. The wide availability of a lot of that tech is somewhat breaking the monopoly that the defense primes have on weapon systems. I mean, an F-35 is around $100 to $120 million these days. At the other end of the scale, we're looking at maybe two to five thousand pounds a pop for the cheapest drones. And you can argue an F-35's a sophisticated platform that does a whole bunch of other things. But you've got these huge disparities. What is the outlook for these very expensive platforms as the staple of modern defense systems?
[43:43] SCHARRE: I think we're going to see large platforms still have a role to play, but balanced by cheaper, low-cost assets that are complementary. Some of the advantages to the really expensive, exquisite systems — whether it's a giant stealth bomber or an F-35 — is that there are going to be some special-purpose missions you still want that piece of equipment for. If you want to go in and take down an enemy's advanced air defense systems, you're going to want stealth aircraft to do that. That's going to be the best way to do it if you have that tool in your arsenal. And if you want a mobile airbase at sea, you need an aircraft carrier — that's valuable.
What's changing is we're going to see, for example, fighter jets with what's sometimes called loyal wingman — uncrewed aircraft operating alongside them. So you might have an F-35 that then has four or maybe eight drones paired with it that can carry more missiles, because there's only so many missiles the F-35 can carry inside its weapons bay, and to stay stealthy it has to carry them internally. You could have drones that are more like missile carriers. You could have some drones that are stealthy, pushed ahead with sensors on them and a communications link back to the F-35, picking up radar first. To use an American football analogy here — you might have the crewed fighter jet quarterbacking a larger team of drones of different shapes and sizes.
And there's still a role for vehicles that have people inside them, particularly because it's really hard to have long-range protected communications. Any of these environments are going to have jamming and electronic warfare. So I think you're going to have a need for people forward in the fight, but now it's going to look like this kind of mix of crewed and uncrewed assets, traditional legacy platforms — fighters, bombers, ships — complemented by a much larger number of drones that are helping to bring mass to the fight.
HUMANS IN THE LOOP
[46:25] RAINE: That's a fascinating example of how it might evolve. But to have a platform at the center of this carrier group in the sky suggests you see that humans stay in the loop and indeed quite a long way forward in the loop. The big concern is that they drop out of the loop. Is that just apocalyptic, or is it actually a possibility that they move out of the loop?
[46:59] SCHARRE: I think the relevant question is where are humans in the loop, or what role are they playing. I think we're still going to see humans involved in decision making on the battlefield, but that may not look like a human pressing a button authorizing every single target. It might look like a human directing a swarm of drones to go conduct some mission — to recon an area, to find and strike things. It might look like humans delegating even lethal decision making to drones in some circumstances. There's a concept called a kill box — a defined area where the drones are let loose, and anything that meets the targeting criteria, the drones will attack on their own. I think there's probably going to be time and space in warfare for where that's occurring.
I think the most effective militaries will still have humans involved because AI and automation often are not great at responding to novel circumstances. And there's a lot of that in warfare. You don't really know what your enemy's tactics are going to be. You can make some guesses, but they're going to try to trick you, they're going to try to come up with new things on the battlefield. You don't know where you're going to fight the next war.
WHAT IS A DRONE SWARM?
[48:15] RAINE: What you've just described — is that what a swarm looks like? Talk us through drone swarms, because there's something nightmarish about that image. What is a drone swarm in reality?
[48:33] SCHARRE: A swarm, like a true swarm, is more than just a deluge of drones — more than just a lot of drones. It's drones that are able to operate autonomously and cooperatively through various mechanisms. The drones can coordinate their behavior. So let's say they're attacking a target — they're able to come from multiple axes simultaneously. Say they're attacking a ship at sea, and there are drones coming to attack it. Rather than four or five missiles coming from one direction, the missiles are able to communicate, split up, come from multiple directions, and time their arrival to hit at the exact same moment. That's a much more difficult problem for the defender to react to. Now they're coming from multiple directions. And you can imagine very quickly getting to an era of cognitive overload for human defenders. There's only so much information that people can process all at once. One or two threats — maybe. But if you've got 50 or 100 drones coming simultaneously from different directions in waves of attacks, changing directions, maneuvering, responding — that's really hard for humans to respond to. And that's one where you have to look to AI to respond in kind.
To make another American sports analogy — if you think about baseball. Fly ball, two outfielders, who's going to get it? Somebody calls it. Humans can do that. Now imagine you've got 50 outfielders and 50 fly balls. You're going to drop a lot of balls. The humans can't coordinate like that that quickly, even if they're the best outfielders in the league. AI could actually do that. And so swarms could enable a degree of superhuman coordination on the battlefield that could be very hard for militaries to adapt to in response.
NAVAL WARFARE IN THE DRONE AGE
[50:42] DEARLOVE: One thing we haven't really focused on, and you've mentioned it in passing, is naval warfare. It seems to me that the vulnerability of capital ships means that that infinite dimension of sea, air, underwater is going to maybe transform naval conflict faster than land conflict. Am I making a correct assumption there?
[51:15] SCHARRE: I think that's right. When you look at what Ukraine has been able to do using small sea drones, they've been able to cripple Russia's Black Sea fleet at a fraction of the cost. I don't want to hold up the Russian Navy as the world's best Navy — they have maintenance problems on the ships, leadership problems. The US, British, French, or Japanese Navy are probably going to do a better job of ship defense. But this is a very real threat. It's low cost, it's accessible. These are low-profile vessels that are just hard to see and defend against. And you can make large numbers of them at low cost to swarm and overwhelm ships. So I think that's going to asymmetrically favor less capable actors and favor shore defense. Those assets are going to be limited in range — you're not going to take these small vessels and send them all the way across the ocean — but when you couple that with land-based drones and anti-ship cruise missiles, all of that is going to make it harder for navies to get close to a country. It makes amphibious assaults more difficult, projecting power more difficult. One way to think about this is that we are transitioning to an era of expensive power projection, where it's harder to project power and take territory — and it tends to favor the defense in relative terms.
HOUTHIS AND SWARM TECHNOLOGY
[52:50] RAINE: Paul, that graphic description of swarm technology — if that were in the hands of the Houthis or the IRGC, it would make it very easy for them to close critical seaways. How accessible is that technology? How close might they be, or other non-state groups, to acquiring swarm technology that is really difficult for ships to defend against?
[53:19] SCHARRE: Well, they certainly have uncrewed vessels — cheap boats, some of which are uncrewed. A lot of them are crewed. Iran's been using fast boats for a long time in the Strait to harass shipping. They certainly have drones and missiles. I don't think we've seen Iran or the Houthis at the kind of level of surface drone warfare that Ukraine has been operating at. But the technology Ukraine's using is not fundamentally challenging or secret technology, and I expect it to proliferate to Iran over time. The kinds of cooperative swarming that I was describing only really exist in laboratory settings right now. We're not even seeing the most advanced militaries yet employing this kind of cooperative autonomy. But that technology is going to be pretty accessible to others. What enables that? It's not some special piece of hardware or advanced material science. It's going to be software — some new software payload updated into the AI along with communications that allow the drones to communicate. So that's at a longer time horizon, but it's coming and it's going to be very accessible when it arrives.
[54:37] RAINE: Well, if we look at the speed with which those two organizations I've mentioned have acquired and innovated around new weapon systems, that may not be that far out.
[54:55] SCHARRE: Certainly. And they're already able to do quite a bit to disrupt shipping now, with the technology they have. Things that are very disruptive, hard to counter, that asymmetrically favor them. And when we look at where the technology is headed, there are even more things that are going to be widely available that are going to be very disruptive and really differentially advantage less capable states and non-state groups.
[55:24] DEARLOVE: The Houthi performance is rather extraordinary. This is almost a tribal militia from the mountains that's turned out to have a much more sophisticated military capability than anyone expected. Is that the result of IRGC training? Presumably it is to an extent, and presumably they're buying things commercially and adapting and developing them with the help of Iranian technicians. Would that be correct?
[55:58] SCHARRE: Yes, certainly there's a connection. But that is part of the feature of non-state groups — there are a lot of them. Across the Middle East, Hezbollah, Hamas, other groups get Iranian support. That's certainly one of the main ways that Iran likes to sow trouble in the region and complicate life for the Saudis, Israelis, Americans, and other countries. But also the technology tends to proliferate over time, and so you're going to get a lot of non-state groups that do have access to state technology either because they buy it or because they get some state-level support. That's probably part of the feature of this kind of class of actors.
HOLDING THE ETHICAL LINE
[56:48] RAINE: How in all of this, Paul, do the US and NATO hold their ethical nerve? How do we avoid being caught in a race to the bottom against adversaries who are unscrupulous in their use of AI?
[57:07] SCHARRE: Sometimes there's a tendency I'll hear in conversations about AI or autonomous weapons of: we can't have our hands tied because our adversaries aren't going to do it. And I think that's the wrong way to think about ethics and warfare. Just because terrorists are sawing people's heads off and lighting people on fire doesn't mean we should do that. That's not the right way to think about ethics in warfare. We should be doing things because either they're consistent with our ethical principles or because they make our military more effective as a result.
I think there are a lot of reasons to keep humans in control of warfare, some of which are moral, legal and ethical, and some of which are because it's going to make the military more effective. It is still the case, at least for now, that the most advanced cognitive processing system on the planet is the human brain. And certainly that's true when it comes to novelty. If you want to write computer code, have AI write it for you. But in a battlefield environment — what's the training data for that? The adversary is going to do something creative and clever and different. They're going to try to trick AI systems, and AI systems are not great at that. They certainly are not reliable. AI systems tend to make things up, as we're all well aware. There's a lot of value in having humans involved in that decision making. And it's not an either-or. You don't have to throw out people. We can use both — just the same way in your regular job you might use AI to help you with something, and then hopefully you read what the AI said and pay attention to what it's doing. We have to do the same in warfare: find the ways to harness the best of machine intelligence and cognition, but also the best of human intelligence, because there's still a lot that humans bring to this equation.
ONE DECISION TO WATCH
[58:58] DEARLOVE: So there's still a role for a Napoleon and his marshals.
[59:03] SCHARRE: For now.
[59:04] RAINE: Paul, we're drawing to the close now, so I'm going to ask you the question that we ask all our guests, which is: what's the one decision that you're going to be watching in the near future? And given that President Xi is visiting Washington this week, is there a decision there that might impact on what we've been talking about today that you might be following closely?
[59:31] SCHARRE: Certainly the Trump-Xi summit is a big one in the news this week, and we'll see what comes out of it — lots of issues on the table. What I'm quite interested in is where we end up in terms of dialogue on AI between the two countries. I suspect that we will continue to see some dialogue. We've already seen at the level beneath that, with the Chinese Vice Premier and Scott Bessent here in the US, agree to continue talks on AI. I'll be interested to see whether anything comes out of the Trump-Xi summit that talks about that and gives us a hint of the tone.
Neither Trump nor Xi is particularly focused on AI safety. And it's been a big topic of concern for people coming out of the leading AI labs, warning about the technology. We've seen now a whole spate of hacking incidents where AI agents have engaged in unauthorized hacking — on their own. These rogue AI agents that are breaking out of containers and accessing the internet and hacking without human direction — that sounds like science fiction, but it is real. It's happened already. So that's an actual problem, but I don't think Washington or Beijing are mostly focused on safety right now. They're focused on how to move faster and race ahead of each other. The logical answer here is well, maybe they could have some kind of agreement to focus on safety and not endanger all of humanity with this technology. I don't know that we're going to get that, but I think that's something I'll be watching for.
[01:01:10] RAINE: Let's hope that is the decision. Well, Scharre, it's been a real pleasure. Thank you.
[01:01:15] SCHARRE: Thank you. Thanks for having me. Really enjoyed this.
[01:01:17] DEARLOVE: Thank you very much.
POST-INTERVIEW DISCUSSION
[01:01:36] RAINE: Well, Richard, that was a fascinating and very engaging discussion on quite a bleak subject. What was it that really resonated for you?
[01:01:45] DEARLOVE: I think we've just had an explanation from a weapons expert as to why the war in Ukraine is stuck and why the US intervention against Iran hasn't really achieved any of its objectives. His description of the way that relatively cheap new technology can stop you being defeated — and the point that we discussed, that not losing now seems to be as important as winning, particularly if you're the subsidiary adversary, which Ukraine certainly is in the case of Russia, and Iran certainly is in the case of the United States. But I think he gave us a very good idea of the current dimensions of the battlefield and some view into how that might shift and change. At the moment I would say that both conflicts are pretty static — that doesn't mean that they aren't highly lethal. The one in Ukraine still is highly lethal, but they're not really going anywhere. But he seemed to suggest that there could be innovations over time to break that stalemate. I don't know whether you agree with that, John.
[01:03:21] RAINE: Yes, I do. I thought his comments on how we've seen the impact of the tech innovation, but the innovation in strategy and maneuver that hasn't caught up yet — but will come — were very insightful. And the second thing that stuck with me was that he clearly sees the human element as remaining critical in all the activities that go to make up a campaign. His glimpse into the future still had humans very much involved. So I thought that made it actually a very rounded analysis and vision. It wasn't overly based on the tech itself.
[01:04:15] DEARLOVE: Yeah, I think that is the conclusion I would personally share. I don't see the battlefield becoming totally dominated by AI and technological change in weapon systems. Obviously they can give you an advantage, they change the character of fighting. I mean, one area which we didn't really explore, which I think is going to be highly relevant, is the whole cyber dimension, which sits as it were above what we're discussing. Because one of the questions that I've often posed to AI experts is, what happens when you combine AI with quantum computing? When you crack the ability of quantum to make many entanglements and then work, as we understand computers working now, but at massively greater speed with massively greater quantities of data. So there's much more to come in this evolution. The evolution has changed what we're familiar with, particularly in terms of battlefield or naval conflict. But some serious brains, probably tucked away in DARPA and some of these research establishments, are thinking hard about that. And the other thing we didn't really talk about was the fact that, okay, in the People's Liberation Army you've got a society of engineers driving innovation, but they haven't fought a battle or a war for a very, very long time. Does that really conceptually affect how one should think of the People's Liberation Army, which has also been stripped of its senior officers by Xi, and might not be quite in the state even to invade Taiwan that we fear.
[01:06:33] RAINE: I think that's absolutely right, Richard. And that applies of course to all combatant parties — so countries who are adversaries are also learning quickly through experience. One particularly chilling scenario for me was the one that you raised over naval warfare, but I think it's also applicable to commercial shipping. There could be extraordinary vulnerabilities to shipping if we see swarm warfare falling into the hands of unreliable non-state maritime powers.
[01:07:22] DEARLOVE: And the other thing we didn't really discuss, but which worries me enormously, is the challenge to the Russian army of innovation and the fact that that has been a motivation for them to adapt their own tactics and their own technology. And in a way, you know, we're training them how to fight a European war in the future. And that is for me pretty concerning, the longer that the conflict in Ukraine continues.
[01:07:56] RAINE: Yeah, indeed. Well, thank you very much, Sir Richard. That's all for this episode. Join us every Thursday for new episodes. Subscribe to One Decision on YouTube so you never miss out on a conversation. Please also leave us a comment on YouTube or email us at hello@onedecision.com. You can check out onedecision.com for commentaries and explainers that you might have missed. I'm John Raine, thank you for listening.
