How Societies Can Succeed in the AI Era

Deanna Lee

Hey there, it's Deanna. Before we get to today's show, did you know that RAND has a graduate school? The RAND School of Public Policy is actually the only graduate school embedded within a global research organization. And you can apply now.

Students at the RAND School learn from leading experts who work on critical policy challenges and bring that real-world expertise into the classroom. In fact, our guest today is one of those experts.

So what you're about to hear might give you a small taste of what it's like to be at the RAND School, learning from the world's top policy minds. You can study in D.C. or L.A. and earn your master's degree in as little as nine months. Scholarships are available for qualified applicants.

If you're interested, you can learn more and apply today to start this fall at rand.edu. That's rand dot edu. Okay, here's the show.

Deanna Lee

You're listening to Policy Minded, a podcast by RAND. I'm Deanna Lee. AI could reshuffle the deck of global power and determine the fate of nations across the world. But what exactly will it take for a country to succeed in the AI era?

Our guest today is just the person to answer that question. In a recent RAND paper, Michael Mazarr looked beyond the technological competition of the AI race, arguing that national success may hinge instead on the social impacts of AI.

In other words, the countries that thrive in the AI era may not simply be the ones with the best AI models or the largest data centers. They may be the nations that are best able to come together as a society to absorb AI's shocks and harness its benefits. Mike, thanks for joining us.

Michael Mazarr

Great to be here.

Deanna Lee

Before we dive in, tell us a little bit about your background, what brought you to RAND, and why you study what you study.

Michael Mazarr

Yeah, well, I've worked in the national security field for a long time now. Um, and what brought me to RAND was basically the interest in continuing to work on those kinds of issues in a research capacity, especially one supporting decisionmakers.

Um, the AI field is relatively new. A number of us at RAND were brought into it a couple of years ago—two and a half years ago or so—to bring perspectives from a broader national security and international relations background to thinking about the implications of AI in a more strategic way in addition to a lot of the technical questions.

Deanna Lee

Great, and we're gonna talk about a lot of that today. In the paper that we're going to be discussing, you write that the world stands on the cusp of a defining technological revolution, that AI is "potentially the most wide-ranging and influential general purpose technology in human history."

Can you walk us through what that might look like? Maybe help us understand what it means not just for tech companies or governments, but also ordinary citizens.

Michael Mazarr

Yeah, sure. I mean, when we think about historical parallels, we look at, like, for example, the Industrial Revolution, you see a variety of early and late technologies in that process from sort of the early textile capabilities and eventually electricity, railroads, all that. There's a whole series of what they call general purpose technologies, which is things that don't just have one purpose, but are applied to multiple societal purposes and have wide-ranging effects.

Well, AI is sort of the apotheosis of that kind of a thing that has the potential to revolutionize a whole bunch of fields at the same time in ways that none of those technologies did before. Like, electricity had a dramatic effect on society in many ways, but it did not necessarily revolutionize the practice of education, for example. It made a huge difference to some industries and less to others, especially at the beginning.

The issue with AI, of course, is that it has applications basically across the board. So, industrial manufacturers are already using it to improve the efficiency and productivity of their activities. Universities are having to figure out how to integrate it into what they do and whether, in fact, the current model of education as we understand it will continue. Medicine is being transformed in terms of discovery, but also the patient-doctor relationship.

So, there's almost no social, economic, political activity in society that doesn't have the potential of being dramatically changed by this new technology. So that's why so many people think of it as this kind of unprecedented, transformative tool.

Deanna Lee

Absolutely. And you write that it's not just about the technology, though, even though many of our discussions about it sort of focus on that. You know, we hear a lot about chips, training runs, data centers, that sort of thing.

But your core thesis is that the competitive challenge is primarily social, not technological. Can you tell me what you mean by that?

Michael Mazarr

Yeah. And I think that's, I mean, that's partly a lesson we draw from earlier, historical techno-industrial transformations like this, which is that, um, the countries that get the technology in and of itself are not necessarily the ones that prosper the best 50 years later.

France had scientists that were pushing the boundaries of a lot of industrial era innovations just as much as Britain. But it did not become the world leader of the Industrial Revolution in the way that Britain did because of a number of characteristics of British society that made it more amenable to entrepreneurs and innovators that had more of a private-sector reservoir of capital available to investors that created, that had sort of networks of engineers and implementers who took the new technologies and experimented with them.

And then eventually, Britain had to go through a period of 50 or 75 years of reforms to shape the effect of the technology—reforms in terms of child labor and labor standards and workers' rights, environmental reforms, political reforms to reflect the changing dynamics that were coming with the new technologies.

So ultimately, the argument is, and this draws on earlier RAND work that we did for the Office of Net Assessment that asks the question, in a long-term U.S.-China competition, what are the sources of advantage? Who's gonna do better and why? And a lot of the historical lessons speak to those social characteristics or qualities that countries have.

And another great historical example that speaks to that kind of thing is the Cold War, where the United States and the Soviet Union in the 1950s, many people thought Sputnik and all that, that the Soviet Union was getting ahead in basic science. But ultimately, it was the nature of American and allied societies, more dynamic, more innovative, more open, more adaptable, that meant that they prevailed.

So that's where this argument comes from, that the countries that integrate technological innovations with societal characteristics and qualities and potentially reforms for a healthy, dynamic, coherent social context—those are the ones that are really gonna prosper in the long run, even if a wider range of countries has access to the actual technologies themselves.

Deanna Lee

I'm glad you raised this idea of societal characteristics, because I want to get into that a little bit. And you gave some examples already that sort of showcase that. But can you tell us what are the core elements that define competitiveness—economic strength, military power, the ability to innovate?

What are those elements that make up whether a country is going to come out on top or not?

Michael Mazarr

Yeah, so again, in this work we did for the Office of Net Assessment, the things you just mentioned we think of as sort of outputs or outcomes. And we looked at the qualities of societies that tend to produce those things, that tend to be responsible for innovativeness and military power and stuff.

And we did a lot of research and eventually came up with a list of seven and I won't sort of go through them in detail, but just as some examples, the first one is national ambition and willpower. So the successful countries have a sense of sort of destiny about themselves that extends from the political leadership through the population to the business sector, to the scientific sector, to the cultural sector. It has a large number of ambitious people trying to make a mark and shape the world. And it gives that sort of driving energy. So that's one quality.

Another one is shared opportunity. So, this is kind of a human talent issue of the more people in society that have more opportunity to express their potential, their talent, their capacity, the more strength a society is going to have as opposed to one that is keeping major segments of its population kind of under wraps because of bias or different kinds of discrimination or whatever, or just lack of economic opportunity.

And another one that is sort of my favorite, it's very abstract. We called it a learning and adapting society. And that is societies that have a great deal of intellectual energy that have many components of society, whether it's, you know, senior officials in government or, scientists or business leaders, military leaders who are really interested in pushing the frontiers of knowledge—thirsty for new ways of doing things—and then change and adapt their institutions, their practices based on what they learn.

It's, it does, it sounds abstract, but you kind of recognize it when you see it in a lot of the classic, more dynamic societies throughout history. So those are three examples. There's a number of others, but there is, I think, a definable set of things that distinguishes the countries, especially great powers, that have this kind of long-term dynamism, solidarity, and energy that makes them competitive.

Deanna Lee

Okay, and looking across those traits, maybe using some of the examples you already cited, how does AI fit in? Where could AI strengthen a society in those ways? And on the other hand, where might it weaken one?

Michael Mazarr

Yeah. That's exactly the sort of analysis I was doing in this monograph was basically taking that framework and saying, okay, how does AI affect these things? So just to take the first one as an example, national ambition and willpower.

An interesting case of that is when DeepSeek came out in China, the Chinese open-source model that suddenly was competitive—or at least in the realm of being competitive—with some leading American frontier models. And there was a lot of international attention to this. But within China, there was kind of an outpouring of nationalist pride and sense of lots of sort of writings about, you know, this shows that our ambitions for scientific and technological advance are coming true.

And so in theory, countries that are mastering this could see it really fuel this national ambition, willpower that's essential to a competitive advantage—because you've got this new tool, you see it being applied, you see your business is getting more productive. You see your kids being more effective in school and enjoying it more, whatever.

The flip side though, and in each of these cases, there's worries that AI could, as you say, can help or undermine this. The flip side is the phenomenon that people are calling cognitive offloading, where people sort of lose their ambition under the shadow of AI, because first of all, they think it's gonna take their job, they think it's gonna take over control of society, if they do believe that. In specific cases, it may be disempowering and in a given business that's laying people off or something like that.

And people ... there's evidence that people can come to rely on it to kind of do their thinking for them, hence the term cognitive offloading. So there's a way that AI could empower a very small number of people in society and potentially through its economic and military effects, make that society more powerful by some measures while draining the larger society of ambition and willpower.

And that gets to sort of a fundamental theme of kind of where I came out, which is: it's all about the choices we make to make this technology empowering and increase the agency of citizens in the society—as opposed to stealing their agency and transferring it to some kind of abstract entity. That's the fundamental thing.

And I think you see that across all of these characteristics. I mean, opportunity is another great example. Just very briefly, you know, it can take people's jobs away. It can substitute for them. It can make it appear as if economic and individual mobility is a thing of the past, or it can become a co-pilot and make people feel much more empowered.

For example, if you've got an AI model that allows you to, you know, whatever, fight some local government agency more effectively and demonstrate legal principles or things, you are empowered in the face of larger institutions as opposed to disempowered. But how that plays out is going to take choice.

And that's one of the things that concerns me, is right now we are having zero dialogue. I mean, there's a lot to talk about in general, it's economic effects and is it gonna get out of control and lose alignment and all of that. We are having almost zero dialogue about the choices we need to make to ensure that at the end of the day, Americans at least, feel that their agency has been enhanced by it. And we need that, we need to have a lot more of that discussion.

Deanna Lee

Okay, so let's talk about some of those choices. Is there clarity on what—and we'll focus on the U.S. specifically—is there clarity on steps that Washington could take to make American society more competitive and do well in the AI era?

Michael Mazarr

Yeah, so partly, I mean, part of the problem is, there's not total clarity on that. I mean there are things we can begin to point to and at the end of the monograph I lay out some recommendations and in some of the other RAND work we've done on AI strategy, we've done that a bit, too.

One thing I would say is, it's not about Washington, primarily. One of the lessons of history, and I think it's very true today, is that when you have this kind of a broad-based societal—or as they're commonly called techno-industrial revolutions—the answer to that in the way that makes it socially empowering is not dictated from the top. It has to be kind of a grassroots and multidimensional pluralistic response.

Like in the British case, from the 1830s to the end of the century, you had a whole series of reform movements across a variety of issues that helped to shape the Industrial Revolution, not for obviously perfection—but for better than it could have been—and avoiding the worst outcomes. And on each of those, you typically had a combination of some kind of popular activists, some members of the elite and aristocratic classes that saw that this kind of reform was essential, some prominent cultural individuals, whatever, these sort of combinations of people.

I think that's kind of, rather than saying, you know, the federal government eventually is going to have to do some things, but the much more important things I think going to be done by state and local governments, by philanthropies, by individuals, yes, in advocating for certain kinds of legal remedies by private sector firms and the choices that they make.

So, just as an example, I think a leading case today is education, where we have no good final answers for what this needs to look like. We're beginning to see some evidence of what it shouldn't look like and some of the perils involved, and some other key choices are beginning to be defined.

So if this were Industrial Revolution, we'd say, okay. We'll have this upsurge of activism and interest and research and a 10, 20, 30, 40 year process of thinking it through, deciding on what works, implementing actual changes and ending up in a position where we've made some of the right decisions. We don't have that much time is part of the problem with AI.

So I think the answer to your question is, we can get into some specifics of, you know, some things that we might begin to do today. A lot of that though, has to do with experimentation rather than putting in place final answers to things.

You know, if I were to say, Hey, what should a large philanthropy do today to, to begin moving in the right direction? You know, I can give some answers to that, but the more general answer to your question is, um, people who are concerned about this in these different social sectors, need to begin bringing groups together to have conversations over the next couple of years really—not decades—to begin to figure out what those right answers are and do it from sort of the ground up.

And again, that process hasn't really started. Like I say, we can talk some more about some more specific actions that could be taken, but that sort of process is how we need to be thinking, as opposed to, hey, what should the federal law on this say to solve the problem?

Deanna Lee

So let me ask you, if we're talking about actions at the state and local government level, for example, and you mentioned this lack of conversations—the conversations that aren't happening that need to be happening—what are the risks of moving forward on AI deployment without those conversations and without kind of coalescing around this idea of understanding what might happen before it happens to us?

Michael Mazarr

Yeah. Well, I mean, the risks are considerable. I mean, on the one hand, you've got risks of loss of control or misalignment where AI models kind of go out of control and cause harm, which a lot of AI researchers are concerned about, and which Anthropic recently warned about—not only about that risk, but risks of misuse with cyber attacks, biological attacks, things like that.

But in this context, I think the risks are what we're already seeing, which is the vast number of people don't understand AI really at all. You know, there are different statistics about how many Americans have begun to experiment with it, but for most people, it is mysterious, far more mysterious than a railroad engine was in the industrial revolution, right? At least that, essentially, if you have any engineering knowledge, the basic principle kind of makes sense to you. In this case, how it works, how it's being applied, what the future looks like, is a complete black box to many people.

So there's immediately a sense of alienation. Then there's a sense of threat from loss of jobs. You know, one of the things we began to see is particularly young people, not only them, but a number of students believe that AI uses a ton of resources—electricity, water—and they don't want to be forced into using it.

So some processes that have begun, for example, in California, of universities trying to effectively mandate the use in order to make people comfortable with it, so that they learn more about it and can become educated users about it, run into this barrier of a certain proportion of people say, "You can't mandate this on me. I'm opposed to the use of it."

So you get that, you know, it's alienating. There's a perception of threat. And so to answer your question, to me, the risk is, one of the biggest risk areas that I found across a variety of these things is that this has the potential of really intensifying social divisions, social fragmentation, lack of trust that's already so rife in our society, that it just accelerates, you know, because one of the things is, as I argue in there, right, this is not coming in on a calm, sort of well-adjusted social situation.

We have a lot of instabilities and fractures that we're dealing with, and AI really has the potential to magnify those by adding yet another layer of perceived sort of loss of control. I mean, this is, you know, if you look back at like Brexit, right? The motto of the Brexit movement in Britain was take back control. And this is such a common theme of populist movements and people who are angry with what's going on, is we're losing control of what's going on in our society in one way or another.

AI is, you know, will just increase that by an order of magnitude. So to me, that's the biggest risk of this proceeding without having thought this through is that it is going to badly exacerbate flaws that we're already seeing in our society.

Deanna Lee

So would you say that awareness, kind of at a broad societal level and kind of a basic understanding of the technology is important to addressing this? And if it is, then how does that happen? Where does that begin?

Michael Mazarr

So absolutely, yes. And that is very often, you know, there's a phrase that's often used of AI literacy and a number of people sort of broke that down to say like, well, what does that exactly mean? But this is often one of the first few recommendations of any list of things to begin to deal with these social problems is people got to understand it a little bit better.

They've got to, I mean, I've been working on this for two and a half years and there's important aspects of the basic transformer-based large language model that I still can't explain, you know. This is one of the difficulties is, it is so intensely complex. But a fundamental level of basic understanding of what these things are, how they work, how they can be applied, has gotta be part of the solution.

A second part of the solution has to be a public-private dialogue, which has already begun with the latest executive order on this in terms of some degree of government pre-evaluation of models before they are released. Something that—and I'm talking from a political and social standpoint now—to make clear to Americans that the governments (state and federal) that reflect their common interests are part of the dialogue and are looking out for the interests of Americans in this process, rather than simply watching a process run out of control.

I think governments at various levels, including the current administration, have taken steps in that direction. There's always a trade-off here because there's a fear that if you go too far in that direction, you're going to constrain innovation, which is an issue for various reasons. But a decent balance there could theoretically be struck.

So two first steps are, yes. And how does the awareness part start? Well, that's where this broader social movement has to come in. It starts with schools at all levels, which has already begun.

China, in terms of the directives of the central government, is moving faster than us on this. How much is being implemented, we don't know. But, whether you're talking about grade school, high school, university level, conscious components and modules to get students to be more aware of it. Not necessarily to force them to use it in certain ways, but an education model in the private sector as a variety of firms have already started to do—efforts to have training programs and educate workers on this, philanthropies doing donations to schools and in other ways creating information available for people who want to learn more about it.

It's not going to happen overnight. But that sort of broad-based effort at all levels to create a basic level of awareness—so that people feel a little less afraid that they don't know what the heck is going on—that's kind of the starting point in terms of that kind of education piece that you talked about.

Deanna Lee

Right, it sounds like they're really—and I think it might be surprising for our listeners to hear—there really are a lot of kind of individual actions that can be taken. You called it a grassroots movement, and that seems to be right on in terms of how something like this would build to actually make progress.

So what would you tell our listeners in terms of what they can do at the individual level to, you know, make sure AI helps and doesn't hurt.

Michael Mazarr

Yeah, well, I would say three things to start with. One is educate yourself, right? We've been talking about that. There's a ton of information out there. There are some very good introductory books, just kind of explaining what AI is.

And so, just to get a basic sense of when someone says large language model, what are they talking about? Even when someone says artificial general intelligence, artificial super intelligence, what are these distinctions? Just some basic stuff, which is fairly straightforward and doesn't require taking a whole course in it.

Second, become a little bit familiar with these models by using them. And here it's anything, I mean, of course, millions of Americans are now using these things extensively for all kinds of purposes, and I'm not talking about the most advanced sort of agentic uses, where you have an AI going off and doing things for you, but rather just basic online access to models to ask questions, to seek advice on things, to have extended interactions and just get a feel for what they're like, what they are good at, what they're not good at, those kinds of things.

And of course, in a lot of like business settings, a lot of corporations are starting to encourage that now in various ways to get people to think of that specifically in terms of applications in the workplace.

And then the third thing I would say is educate yourself about what your public officials are saying and advocating about it. Because right now, I think there's a significant, I mean, that, you know, as a new technology and very complex, it always takes the governing institutions of a society a while to catch up in terms of knowledge and policy, right? And I think it's fair to say that it's natural that there's still a lot of members of Congress that are in this process we're talking about, still educating themselves about it, figuring out what it's all about.

But, ultimately on this issue, like on all issues, we want American citizens to hold their elected officials accountable—and appointed officials, too—accountable to the standards they would like to see. And so the knowledge becomes the foundation for a political connection between Americans and their leaders, which becomes essential to legitimacy.

That's another kind of worry of mine, is if the American people come to believe that their elected officials don't have their interests at heart, that further undermines the perceived legitimacy of our governing institutions, which is already under such threat.

So that connection from knowledge, eventually, to sort of political awareness, my senator, my representative, my governor, my county board, whatever, here's what they're saying about AI, here's the positions they're taking, here's what they're doing, do I agree with that or not? And you know, become involved, as on all issues in that, to create a chain of agency, first of all, and the more Americans that do that, the more the overall system becomes responsive to what they want, and that is the essential thing for a democracy in the long run.

Deanna Lee

And we've established that this is not a top-down process. We just talked about individual responsibility and what everyone can do. But what is one thing that you would urge policymakers to focus on right now when it comes to this problem?

Michael Mazarr

Um, you know, I think I would say, I mean, there's a couple of broad AI policy issues that are critical in terms of like safety, but I'm sort of going to put those aside because that's a separate discussion in a way from the societal impacts that we're talking about here.

In terms of that, at various levels, one I would just nominate an issue that I think is ripe for initial consideration and that's education. And because it affects all levels of governance. And again, the idea here is not, oh, sit down and figure out the right answer and just implement it.

It's how do we envision the next two to five years as a period of experimentation, where we outline a number of principles that we want to be reflected in the policy that comes out, um, we encourage different schools, different teachers, different people to experiment in different kinds of ways. We collect data on it, we study it, we see what's coming out. We engage AI in that process in many ways, as it can be a powerful research tool.

And we consciously work toward the best possible outcomes in terms of enhancing opportunity, enhancing agency. There are definitely some universities that are beginning to try to move in that direction. There's some governors, there's others that have sort of spoken to this. But we're really just putting our toe in the water.

So, in terms of an issue that I would recommend that people—like whether it's a county board member, a governor, a state legislature, a federal legislator—thinking about AI's implications and developing, being part of a conversation to develop conscious policy actions that encourage reflection on a series of experiments guided by key norms and principles that we want to promote, that's an issue that I would suggest.

Deanna Lee

So, it sounds like education is sort of at the heart of all of it, if we want to ensure that, you know, AI emboldens human agency rather than ruins it, I think, you know, to address that sort of cognitive offloading that you mentioned at the beginning. If you had to sum it up in one word, would that be it, education?

Michael Mazarr

Well, it is certainly one of the top few, yeah. I mean, and I think it's probably fair to say, yeah, that it's in a lot of ways at the hub of lots of other issues, because to the extent that it works, you get a more informed populace that then expresses its ambition and has more opportunity and is more ready for learning. So yeah, there's that.

I do think one other huge issue is the relationship between people and the major institutions in their lives, which is already a big problem in a hyper bureaucratized society where all of us feel sort of disempowered in the shadow of whether it's the IRS or a large corporation or in the legal system, whatever. And AI has the potential to tilt those relationships in favor of citizens to make them feel like these institutions are more under their influence or tilt it in the other way.

So, those kind of power relationships, I guess I would say education and power, two words that really reflect what's at stake.

Deanna Lee

All right, I think that's all the time we have for today. Mike, thank you so much for joining us.

Michael Mazarr

Sure thing. It was great to have the discussion.

Deanna Lee

And thank you to our listeners. If you want to read more about the publications we discussed today, you can find the links at rand.org/policyminded. Today's episode was recorded and edited by Emily Ashenfelter. It was produced by me, Deanna Lee.

RAND's Director of Digital Communications is Pete Wilmoth. RAND is a nonprofit institution that helps improve policy and decisionmaking through research and analysis.