TRANSCRIPT:

Kahl: You're listening to World Class from the Freeman Spogli Institute for International Studies at Stanford University. I'm your host, Colin Kahl, the director of FSI.

Today we're talking to my good friend, Dr. Amy Zegart. Amy is a Senior Fellow at FSI, the Morris Arnold and Nona Jean Cox Senior Fellow at the Hoover Institution, a senior fellow at Stanford's Human-Centered AI Institute, and a professor by courtesy in our political science department.

She's also the author of five books and one of our country's leading authorities on U.S. intelligence, emerging technologies, and national security. In addition to her terrific scholarship, Amy has advised senior officials about intelligence and foreign policy for more than two decades, having previously served on the National Security Council staff and as a presidential campaign foreign policy advisor. Amy, welcome to World Class.

Zegart: Colin, thanks for having me. It's so great to be here.

Kahl: Awesome. So much to talk about. I don't think that this is going to be particularly controversial, but I think it's fair to say, internationally, that we are living in the most turbulent period since the end of the Cold War. You might even argue that it's the most turbulent period in international affairs since the period between the two world wars. I don't know where you fall on that pessimism spectrum.

In this context, I know you're finishing up a book manuscript on navigating political risk. You've written on this previously as well. And I'd love just to start with you giving our listeners a little bit of a feel for the core argument in the book you're writing and where you shake out on where international order—or maybe we should say international disorder—is headed.

Zegart: Well, first, thank you for using the verb “finishing” a book. As you know, we're always writing books, but it's rare that you can say you're finishing a book. So that's a nice moment to hold.

So, this book I'm co-authoring with Condi Rice. It's our second book on political risk, and it's a really different book. You've lived in this world even more than I have in your job when you were in Washington.

The hurricane is not blowing over. We're at the beginning of this disordering period, and it's not going to go back to the international order that we are all so familiar with of the past 80 years.

And so the first part of the book is what's happening and how did we get here? It's sort of a political sciencey take. And then the last part of the book, really two-thirds of the book, is about if you're a business, whether you're a small company or a big company, what can you learn from the world of national security?—because national security folks think about this kind of stuff for a living—that you can apply so that businesses can navigate this tumultuous geopolitical environment.

And I'll just say a thing about the first part, which is what is happening and how did we get here? I think we all kind of took for granted that this international order was the natural state of the world when it wasn't.

And so we argue in the first part of the book that there were three factors that underpinned the international order. The first was the Bretton Woods Institution’s rules and norms really about trade, which if you're a political science junkie, you know that well, where countries really came together and established these new institutions to avoid the interwar period and what happened there.

But the second two pillars of that order I think have been taken for granted. And they are, first: U.S. economic policies. So, the U.S. had a large share of GDP in the world after World War II. And we deliberately crafted economic policies—Marshall Plan, rebuilding Japan—to let other countries rise. Not out of largesse, but out of a cold-eyed view that that was the way to combat communism. So it was good for the United States and good for the world.

And then the third pillar is U.S. security guarantees. So, the freedom of navigation operations, which you know and love well. Most Americans don't think about global commerce doesn't just happen. The U.S. Navy actually secures the seas against pirates and all sorts of threats, so that that happens. And those three pillars have all come undone.

And we can talk about why that is, but we're not going back to the way things were

Kahl: Do you think that it's been undone in the way that we're rewinding, a kind of back to the future moment? That actually this period since the end of the Second World War is what's abnormal and what we're transitioning back to is a more normal state of politics? Maybe that's the kind of realist or real politique view of things. Or do you see it as something else?

Zegart: I’d be curious to know what you think about this. So I'll give you my take, and then I want to ask, professor, what you think.

I think it's both. It's just like the Ukraine conflict is trench warfare with modern capabilities. It's this weird hybrid of historical and futuristic. That's the order we're in. So we do see this every country fending for itself, me, mine, mine, and resurgence of realism in the order.

But layered on top of it is this transformation in technology. And technology is really driving new sources of national power, the integration of economics, technology, and security in ways that we have never had to experience before. What's the AI race? You and I have talked about this before. Well, there are multiple dimensions of multiple races that affect economic development, that affect technological development, and that affect military capabilities.

So I think it's both elements of the past and certainly novel elements of the future that are both colliding in this moment.

Kahl: We just celebrated Indigenous People's Day here in the United States. And of course, it used to be called Columbus Day and maybe still is in parts of the United States. But there's this old adage: don't fall prey to the Christopher Columbus problem, which is just because it seems new to you doesn't mean it's actually new. The New World existed before Columbus “discovered” it.

My own sense on this is that there's parts of this that feel like a back to the future moment: the resurgence of major power competition, balancing and hedging, and the reemergence of protracted interstate conventional conflict. These are things that were perennial problems in the post-Westphalian international system. They produced two world wars at the first half of the twentieth century. So, the geopolitical part of it, there are aspects of that that seem perennial or familiar.

But I'm where you are. I do think that there's a technological layer. And it's not new that technology has changed the world. We had the Industrial Revolution. We've had technological revolutions before.

I think one of the things that is different or feels different now is the pace and scale. But also the fact that it's vesting enormous concentrations of power in the private sector. We had the British East India Company and we've had large oil companies and other multinational corporations. But I don't think there's a period of history where we had corporations that not only had global reach and enormous resources, trillions of dollars, in the case of our large technology companies, but actually almost near sovereign control over the critical domains, the digital domains mostly that we experience reality through.

And that strikes me as a kind of power that is meaningfully different or qualitatively different, and that is oftentimes vested not in states, but in private corporations. So that feels different.

And then the other thing that feels different is that, interdependency is not new; we've had economic crises, we've had global pandemics. But I think our international system is so entangled that there's just a lot more transnational shocks, whether it comes from a pandemic or climate change or externalities from technologies like AI, that can just affect billions of people a lot more rapidly than was true before.

And so when you layer kind of old school geopolitics with new school techno politics and these existential externalities, that kind of feels qualitatively different to me. But that said, I'm living in the moment, I'm not living, you know, a hundred years ago.

Zegart: What you've hit on though is the most important question we should be asking, which is what is old or enduring in this moment and what is genuinely novel? And how do we separate those different factors?

So I'll give you my take on what's old and what's new. What's old? Geopolitical competition, rising states, falling states, even competitors of nuclear weapons. We've seen this movie before. It's a hard movie, but it's a familiar one. That's old.

What's kind of old? What you alluded to is companies with lots of power. But the scale, speed, and scope of that power is, I think, unprecedented. I hate to use the word “unprecedented” because maybe there is a precedent somewhere deep in the bowels of history. But it feels really qualitatively different when one man, Elon Musk, can decide where battles are fought in the world, for example.

And then what's really genuinely novel—and you and I have talked about this—is that the sources of national power are changing. So it used to be it was tangible things like territories, militaries, oil, ships, and steel that determined great powers and the capabilities they had. Those things are still important.

But now it's also intangible things like data and knowledge and algorithms. And those naturally diffuse. You can't put those genies back in the bottle. And so we're grappling with what does that mean for economic competition? What does that mean for militaries? What does that mean for export controls, which you and I arm wrestle about.

So I think the genuinely novel part, the hard part to really wrap your arms around is that technological layer, and what does it means for power, and who has it.

Kahl: And a super good transition to what I wanted to ask you about next. One of the manifestations of old geopolitics roaring back is that the world is now in the midst of a number of large-scale conventional wars. And that just was not a thing. There actually were not that many interstate wars during the Cold War. The post-Cold War was characterized by plenty of civil wars, ethnic and religious struggles, humanitarian crises. But really not wars between countries.

But now we have the largest land war in Europe since the Second World War. We have on again, off again wars raging in the Middle East along the U.S.-Israel-Iran axis and the Houthi-Saudi axis and other issues. And one of the aspects of these wars that I think are most notable is the role that emerging technologies are playing in them.

And that you see these wars as a mix of kind of old school war with new school technologies, not the least of which being drone warfare. And I know this is something you've thought a lot about, written a lot about, even before it was cool.

We're seeing drones not only increasingly used on the battlefields of Ukraine or used by the Iranians or the Houthis to wreak havoc against critical infrastructure across the Middle East.

But we're also seeing it in Russian hybrid warfare against NATO: buzzing NATO countries with drones. The PLA is using drones all the time to harass Taiwan and force Taiwan to come up and police the skies around the island.

This is something you've been thinking about for a long time. I wonder how you see drones not just really changing warfare, but coercion—the way in which states signal each other, try to deter each other, try to compel or coerce each other.

How are you thinking about drones in that context?

Zegart: So, first, thanks for reminding everyone how old I am, because I've been doing this for a long time. And second, I just want to take a minute: the reason I got interested in drones before drones were cool had nothing to do with me. It had to do with CISAC.

When Bill Perry was still very much a presence at CISAC, he said, “I want to take a bunch of faculty to Creech Air Force Base, to see drone operations, to understand them in Afghanistan, to taste, touch, and feel what it's like with these new technologies. And that's what got me on this now 15 year journey in how drones could revolutionize not just warfare, but coercion. So, hat tip to him for really herding the cats of us and taking us on a field trip, which really changed the trajectory of my scholarship.

So, I think there's a lot of attention, as you point out, to the revolutionary capabilities of drones in war.

But the question I keep coming back to is: if Tom Schelling—father of deterrence theory, modern game theory—if he were alive today with the capabilities that drones offer, how, if at all, would his theories of coercion change? And I think the answer is that drones offer two novel ways to coerce adversaries. Because of course what you want to do is have weapons that prevent the fighting in the first place and then can win if you have to fight.

And typically we think about the severity of punishment. The nuclear escalation ladder. We're going to start shooting, and who knows where this will lead. And so that prospect of nuclear annihilation keeps rivals in check. Drones don't do that, obviously, but they offer something different, which is two things. One is the certainty of punishment. It's not the severity, it's the certainty.

I think IR theorists often don't look at other relevant literature about deterring bad behavior, but criminology is one of those areas. And research in that area finds what deters criminals? Certain punishments. The red light camera stops you at the red light more than a serious fine. And that's true for violent crimes as well. So drones offer certainty of punishment that can actually enable them to coerce.

The second mechanism is sustainability of punishment. Right now we are being deterred in a lot of areas. And our deterrence isn't working in the ways that we thought it would. Why? In part because our adversaries know we can't stay the course, even if we want to. We don't have the munitions, we don't have the cost curve to be able to keep the fighting on as long as it takes.

The Houthis know this. So they keep firing Shahed drones that cost $20,000, $30,000. And we're taking them out with several million dollar interceptors.

So what does this mean for coercion? I'm working on a piece right now on this for Foreign Affairs. And the answer I've come up with so far is we need three things: large quantities of unmanned systems that do a lot of different things, not just cheap quadcopter drones or short range drones that the Ukrainians tend to be using, but they do a variety of things. We need to relentlessly drive down their costs. Right now they're still too costly, even in the Pentagon. And then we need to think about creative signals.

I always say: all parents of teenagers know deterrence theory, even if they don't realize it, right? The challenge we have with our kids is you could threaten to take away the phone or the car, but your resolve to do that is something that they question. So, how do you convince your teenager, or in this case an adversary, that you really will do what you say you're going to do?

One answer to that for drones is not the tip of the spear brandishing the weapon, it's the spear factory. It's your ability to keep producing these munitions at low cost for as long as it takes. So I think signaling will no longer be confined to sending the carrier battle group or putting troops on the ground. The defense industrial base itself and our ability to keep producing can become a signal of resolve in drone deterrence in the future. That's at least what I'm noodling on right now.

Kahl: I think it's super interesting. First of all, for listeners who maybe don't speak Stanford as much, CISAC, which Amy mentioned, is the Center for International Security and Cooperation, which is one of our flagship centers here at the Freeman Spogli Institute. And Bill Perry, the former Secretary of Defense, such an important leader at CISAC for so many years.

But yeah, the fact that not only is conventional warfare back, but protracted conflict is back. I mean, I think, you know, there are a million differences between Ukraine and Iran. But one of the things they have in common is that a great power believed that they could win a short, sharp contest to achieve pretty maximalist ambitions: territorial conquest in the case of Ukraine, regime change, or at least profound compellence in the case of Iran.

And those aspirations proved infeasible, and these have turned into protracted wars of attrition where people are very much looking at the munitions stockpiles that the United States has for counter-drone technologies or missile inter interceptors or long-range strike missiles that oftentimes are quite expensive.

So I think the ability to fight long wars is itself a deterrent for the reasons you mentioned and maybe just for old fashioned reasons about if you expect wars to be long, then you size up your adversary on their ability to engage in them with some degree of protraction.

I like your certainty of punishment issue and the resolve issue. I wonder whether there's also an angle on this that your colleague at Hoover, Jackie Schneider, on some of the war games she's run—and some other experts have found this, too—that people are willing to take different types of risks when they use uncrewed systems, right? Because pilots aren't there or sailors aren't there

And so you may be able to conclude that countries will be more willing to use their drones if they have that capacity, because it doesn't create the same type of political costs if those drones get shot down or destroyed.

I don't know if you buy that or think that that's relevant.

Zegart: Absolutely! And so the question is: which way does it land in what circumstances? Drones could make initiating conflict more likely, but it also, depending on how you threaten and what you threaten, could make preventing conflict more likely. So the devil's in the details.

But yes, I agree with you. If a weapon doesn't have someone who could perish on the battlefield and come back home with families, you're going to be more likely to use it.

Kahl: So, I want to transition from drones to some other emerging technologies. It's fair to say you made your first big intellectual mark in studying reforms in the intelligence community and reorganization after 9/11. You did some pathbreaking scholarship there.

In recent years you've written a lot about what the intelligence community needs to do to better integrate emerging technologies, especially artificial intelligence and big data analytics, into what they're doing. I think at the moment you're teaching this extraordinary class here at Stanford called Spies, Lies and Algorithms. You have a book with the same title. More than a hundred students take it every year. It's super popular,

I wonder if you could talk a little bit about how well you think the spies are integrating the algorithms into what they're doing. And if they're not doing enough, what more they should be doing?

Zegart: Thanks for the lovely comments about the class. I have found really the only secret to having a popular class is you let Stanford students rule the world, which they do in a multi-day simulation. So, we have a North Korean crisis and students get to actually play the role of intelligence officials and policy makers and they seem to like that even though it terrifies them. So they get to walk in your shoes for a few days, Colin.

So, I'm very concerned about the adaptation challenge for intelligence agencies today. Not just in adopting AI, but in tracking who's ahead and who's behind in different technologies and what does it mean for advantage.

As you know, our intelligence agencies are very adept at looking at weapon systems: who has more fighter jets, who has more tanks, where are we behind? But because these emerging technologies don't emerge from the Pentagon, they come from the private sector.

And because they're moving so fast, it's much harder for our intelligence agencies, and because many of them come from the United States, where our intelligence agencies don't collect on Americans, it's hard for them to be able to answer that question: who's ahead, who's behind, and where does it matter?

Which means that I'm very concerned about the risk of what I call strategic technological surprise. So think Pearl Harbor only in a technological area where China suddenly leapfrogs the United States so much that we can't catch up.

DeepSeek would be a small indicator of that kind of a surprise. When that model came out, it surprised markets, it surprised many in the intelligence community, it surprised many policymakers. But the insight that I got from being here at Stanford was DeepSeek was not a surprise to my colleagues at the Human Centered AI Institute, the computer scientists. So surprises are not surprises to everyone.

So the question for the intelligence community is: how can they get a finger on the pulse of these crucial technologies? And it's not just AI, think semiconductors, quantum computing, bioengineering, material science. How can they know what's around the corner? They have to get outside the classified world to be able to know that, to get that awareness. That means they have to be able to talk to people in labs at Stanford and other universities and in companies. 

And as you know, that's an unnatural act for our intelligence agents. They weren't designed to do that. Who owns the portfolio of technological economic competition in our 18 agencies of the intelligence community today? That whole structure was not set up for this geopolitical competition moment.

They're trying hard to adapt, but it is a really hard problem to solve. And so one of the things we're doing here at Stanford is trying to bring together engineers and social scientists to better understand what's happening in emerging technologies and sharing that with the intelligence community, unclassified information, so that we help provide that insight for them to make those assessments.

Kahl: Let me pull on that a little bit because you mentioned being surprised by a potential competitor or adversary. You mentioned China and the DeepSeek moment a year and a half ago or so. But there's also the possibility we get surprised by our own companies.

Zegart: Absolutely.

Kahl: So much of innovation happening at the frontier of science is not happening in the national labs. It's happening at OpenAI or Anthropic or NVIDIA or some leading biotech company.

And I wonder if you could talk a little bit more about that. A lot of how freaked out people are about AI, for example, depends on what assumptions people are making about how soon the models will essentially engage in completely autonomous or automated AI R&D.

The fancy term for it in the labs is recursive self-improvement, where the models are essentially improving themselves and you kind of get an exponential takeoff, a kind of intelligence explosion as the models have this flywheel of progress that spins faster and faster and faster until you get something like artificial general intelligence, superintelligence, something else.

Some people dismiss that. Some people embrace it. But I'll tell you this: the folks who work at Anthropic or Google DeepMind or OpenAI have much greater visibility into the degree to which they're using their own models for AI R&D and how close they are to something like RSI than NSA does or CIA does.

How should we grapple with getting the insights from our own private sector?

Zegart: It's such an important question. And I would agree with you that folks in these frontier companies—and they are companies, they're not labs, right? Let's be clear, they're for-profit companies. The technical experts there have visibility into this technology that most other people in government don't. That's absolutely true.

But they also have conflicted interests in saying, “Please, please, please stop everybody else from pursuing AI, and, by the way, reduce our liability with U.S. regulatory laws so we're not blamed if the AI goes rogue.” I can see the benefits to that position for those companies.

So, how do you harness the expertise about where AI is heading while balancing against what is naturally self-interest of these corporations to create a regulatory regime that most benefits them? That's where the rubber meets the road.

I think that the conversation has been really polarized in an unhelpful way. So it’s "let it all go,” and let's see what happens, or let's regulate, regulate, regulate. And I hope we get to an answer that's more thoughtful in the middle. And the middle path is where do we and how do we create independent capacity to be able to help judge with these companies what's safe—to the best that you can, because you're never going to reduce the uncertainty to zero—to release and what isn't.

By that I mean why don't we have national compute for the best computer scientists in American universities and the national labs to be able to do independent testing of these models? We would all be better off if we had no-kidding, real independent capacity to be able to understand what's going on inside these companies and make decisions about what should be released and what shouldn't be released, while still acknowledging that the rest of the world is racing on. So, if we slow ourselves down, we could be hurting ourselves nationally in the long run.

These are really important conversations to be had. I think we're moving toward those kinds of productive conversations, but the “it's all doom” or “it's all going to be great” doesn't really get us to where we need to be.

Kahl: We're both in the Bay Area, right? And so you go to a cocktail party and people ask your P doom, like how much of a doomer you are about AI or or anything else. I don't really think of myself as a doomer. I think of myself as a gloomer, in the sense that I am gloomy about our ability to move at the speed of relevance, especially on the government side, in the face of how disruptive these technologies are likely to be. And that doesn't mean they'll kill us all. I just think government is moving slowly.

There's been a little bit of a wake-up call this summer because of the OpenAI-Hugging Face incident and Mythos earlier in the year, and now some of the other conversations around pacing the frontier and other things. But we'll see.

I suspect, Amy, along the lines of what you've suggested, maybe there's space for bipartisan agreement around some combination of transparency from the lab so that we know more about what they're doing, some third party auditing and evaluation mechanism. Maybe it's a public-private thing. Maybe it's a government thing. Maybe it's something else.

And then some resolution on the liability question. I think that labs pushing for liability waivers is going nowhere. The politics on this on both the right and the left will be strict liability. Like don't release unsafe stuff, right?

So, I think we'll have to see. But a precondition for any of that is government actually understanding what's going on inside the companies. And there I'm not sure we have great visibility.

Zegart: I would just say that I'll put my analytic biases on the table, which is I tend to think the better approach is to have healthy competition. You mentioned OpenAI and Hugging Face. OpenAI models hacked Hugging Face and it was a big news story.

But the part of the news story that people often don't know is what defended Hugging Face. Hugging Face asked those frontier companies for help and they said no. And so it was a Chinese open source model that actually came to the defense of Hugging Face. Now, I would not be advocating for having Chinese open source models defending our country, but the point is the more firefighters we have to rush to the rescue when these things happen, the better off we will all be.

I think the narrative has become everything from China is bad, open source models from China are bad inherently, and the reality is those models are already being used and can be forked, as you know, to be much more safe. And they help keep our ecosystem in line.

I worry about a world where there are only one or two companies, and you have to ask, “Mother may I?” for everything from will you defend me from this hack to will the Pentagon be able to use your models for defending the United States? I worry about that world.

Kahl: I think the worst outcome is probably one in which the only capable open source models are Chinese models. Because then researchers at universities like Stanford or MIT or Berkeley or wherever, are going to turn to these models because they can look under the hood, they can fine tune them, they can change them, they can learn from them.

And there needs to be a vibrant open source American ecosystem, and the computing resources to support the researchers using these models that aren't beholden to just a handful of companies while we're sensitive to the risks that all these models generate.

This is a good transition. I don't know if anybody's ever sent an email to a Stanford professor, but I would bet that we have among the longest signature blocks at the bottom of our emails, because all of us have like 57 different affiliations, right?

Zegart: [chucking] Yeah.

Kahl: Amy, your signature block is probably among the longest. Not only are you a senior fellow here at FSI, you were the co-director of CISAC, one of the centers we pointed to before. You're a senior fellow at Hoover. You're also a senior fellow at HAI, which is the Human-Centered Artificial Intelligence Institute. At Hoover, you oversee the Stanford Emerging Technology Review, which is an annual publication on the state of the art not just on AI, but commercial space and you know, energy and biotech and quantum and a whole bunch of other things. You run an AI Track 2 associated with that. At HAI, you're really engaged in trying to bring together communities across Stanford to focus more on AI and national security and geopolitical questions.

So, having us have impact along these dimensions is something you're deeply committed to. Pulling all this together . . . like, Dr. Zegart, what are you trying to do? And why are you trying to do it? And why do you think Stanford's a good place to do it?

Zegart: What's wrong with you that you have all these affiliations on your signature line is I think the underlying question. I think, Colin, to your point, we all have all these different hats that we wear. And I think Stanford is the greatest university in the world for that reaso

Lots of universities want to be multidisciplinary, having people work across party lines and different fields. We can actually do it. And so one of the reasons I say yes, maybe too much, to all these different things is I love learning from colleagues across fields. And FSI does a great job of this. And you're bringing a lot more scientific expertise into FSI, and I think that is an unalloyed good.

We need to have that team sport of the engineering faculty and the historians and the political scientists and the economists and the humanities professors and the law professors, because we each bring different analytic perspectives to the same problem.

And particularly given the geopolitical environment that you sketched out at the beginning of our conversation, this is how we come up with new ideas to improve the world. I think Stanford is incredibly exciting. It also helps we have good weather so we can walk from building to building and it's really pleasant. So, I walk to the computer science building for my weekly meetings. It's lovely outside. I meet with colleagues from all these different places. That's the secret sauce that is Stanford.

What does it mean for us in impact? You were serving at a very high level in government, and you took the ideas from Stanford . . . you took some your colleagues with you, not just their ideas! That reach back into the senior levels of policymaking is incredible what we can do here.

And then the flip side of that is our students. Being able to train and inspire the next generation, which is why I teach this course on intelligence to undergrads, is I want students to carry that mission with them, whatever they decide to do. That's impact.

You said something in my class last year, Colin, which I use all the time. You said, “Our generation, our responsibility is to make sure problems remain solvable so the next generation can solve them.”

And I think that's the great blessing we get by being at Stanford is we can do both of those things.

Kahl: I like it, and not just because you quoted me back to myself. Stanford is overwhelmingly an engineering school in many respects. So many of our students are STEM kids. They do computer science or mechanical engineering or bioengineering or math. But they also then have a second field, which is often international relations or history or economics or something else.

And what I find so remarkable about our students is they have a lot of capacity to be kind of bilingual or multilingual across social science and actual science in ways that I certainly wasn't equipped with as an undergrad. But I think increasingly will be the leaders that we're going to need in 10 years to address climate change or the externalities from AI or how to grapple with advances in quantum science or whatever it is.

And I want to be in an ecosystem where we have colleagues across disciplines and we have students who actually want to be interdisciplinary in how they understand the world.

You have a hundred of these students right now, so I would love to hear your reactions.

Zegart: I completely agree with you. Actually, I have this joke that I give to my students: the glass is half full. So the optimist says the glass is half full, and the pessimist says the glass is half empty, but the engineer says the glass is twice as large as it needs to be. That's just a fun way of saying when you have students with all these different backgrounds coming to a class, they bring really different questions and perspectives with them.

I’ll just say one thing to your bilingual point. I couldn't agree more. And even though AI is making coding much more available for students who have no computer science background, it's the same way we talk about the importance of studying foreign languages. You still get a window into the culture. It's still a really important thing to do. That engineering background is still incredibly important.

Even with the increased role of AI, understanding the language and the logic of code and what machines are trained to do and what humans actually are not trained, but maybe inclined to do, where those intersect is going to be incredibly important for international security policymaking.

Kahl: What we're gonna need are humans that can that can basically at the front end have the experience, the insight, the intuitions, the taste to identify problems and then overseeing the machine in the middle that is increasingly doing the analysis, running the experiments, conducting the research, and then the judgment, the ethics, the moral compass, the personal leadership to take the outputs of that and actually translate it into something that makes the world better.

And someone who only does science or only does social science or only does ethics or only does religion, if you only do the one thing, you probably can't complete that process the right way.

Zegart: But I think oftentimes the solution to that is one person can't do all of those things, but a team can. So making it easier—and Stanford's really good at this—to bring the team together, I think is a really promising path forward.

Kahl: Awesome. Well, I'm really glad we're on the same team.

We always conclude our show by asking our guests for one article. I often joke, you know, Ezra Klein, who has this very popular podcast, he asks his guests for three books. And like I'm an academic and I don't have time to read three books. I'm jealous of people who do.

Is there one article, Amy, that you would recommend that our listeners check out to kind of understand this moment that we're living through in the world?

Zegart: Okay, so I was going to cheat and say I have a book and an article. So I'll tell you the book, which is the book I give to my students when they graduate. It's an oldie but a goodie, Tom Shelling, Arms and Influence.

Every time I read it, I learn something new. Remarkable mind. So that's my cheating book that you didn't ask me for.

The article that I keep coming back to that I just had not really thought that way before is an article by Evan Medeiros and Andrew Polk, and it came out last year in Washington Quarterly. And it's about China's development of its economic weapons. And so they label them, the term of art they use is precision-guided economic munitions.

And it's this wonderful early foray into how the CCP, the Chinese Communist Party, has really transformed its economic export controls and other policies to be far more sophisticated than they were even a few years ago. That really was an “Aha!” paper for me, because I'm not an economist, but it's important to understand what's happening on that side.

Kahl: Super interesting. And of course we saw that play out in real time as China weaponized our dependence on rare earths and other critical minerals. So go check it out. 

Awesome. Amy Zegart, thanks so much.

Zegart: Thanks for having me.

Kahl: You've been listening to World Class from the Freeman Freeman Spogli Institute for International Studies here at Stanford University. If you like what you're hearing, please leave us a review and be sure to subscribe on Apple, Spotify, or wherever you get your podcasts to stay up to date on what's happening in the world, and why.