AI is going to remake every aspect of our lives, and not just our lives. The lives of 6 billion people who are outside of the United States and China, they have no voice, and we have a voice. We still live in a democracy. It is our responsibility to have a say about how our future is being remade, even if it's being remade by two or three companies. It doesn't mean that their CEOs have the right to make every decision. I think the democratic input is a very important part of that.
Welcome to Radio Davos. This is the podcast that looks at the biggest challenges and how we might solve them. I'm Linda Lacina, stepping in for Robin Pomeroy this week, as we had the chance to snag a very special guest in our World Economic Forum's New York City offices. Our guest is a renowned economist, an MIT professor, and a 2024 Nobel Prize winner. He is also the world's foremost expert on AI's impact on the economy, which is one of the most critical issues of our time. He has a new book out. It is What Happened to Liberal Democracy, and a lot of very smart people are calling it a "tour de force," a "must read," and even "the most optimistic thing that they've read in a very long time." He is going to explain the crisis as he sees it, a crisis that is widening the economic divide, what's needed to fix it, and also the role that AI, especially pro-worker AI, could play in solving this issue. Welcome to Daron Acemoglu.
Thank you, Linda. It's my pleasure to be with you.
We are very, very honored to have you. Why don't we get started, dig right in. What is this crisis? What is the biggest crisis that is facing the economies and humanity right now, as you see it?
I believe that liberal democracy is one of humanity's greatest achievements. If you look at our past, the million-year revolution, the existence in fighting bands, or very unfree societies, people at that point would have had a very hard time imagining that we could build a complex society with very different people, but all cohabiting, participating in politics mostly peacefully, and using state institutions for creating also shared prosperity, meaning economic growth from which many people, most demographic groups, are also taking their fair share. That is just an incredible achievement. But the more immediate challenge is that I think that achievement is now in danger. And we need to understand why liberal democracy is in crisis today and how we can rebuild it, because liberal ideas and liberal democratic institutions, I am convinced, have a special role to play in our future as well. And we can make them better so that they can play that role.
And just to get our listeners and our watchers sort of all on the same page, can you define liberal democracy as you see it and as you wrote about it in the book?
Liberalism, at its heart, is about individual freedoms, freedom of expression, association, religion, but also about turning those into an actual behavior in which people can realize their aspirations. And that requires state institutions that provide opportunities, education, infrastructure, insurance against very tough things happening to people such as disability or inability to earn a living for their families, but all the time consistent with liberal principles. And liberal democracy is the most successful institutionalization of those liberal ideas because it brings elements of liberal thought and principles together with democratic participation.
If we're not able to get liberal democracy out of crisis, what happens?
Today, liberal democracy is viewed as the establishment. And anti-establishment feelings are everywhere. And not only those movements that challenge liberal democracy are illiberal, they're not democratic, they don't want to provide voice and ability to self-government to people, but they are often ideologies that separate people, create more divisions, and also they are stances that have not been able to deliver any kind of welfare or improvement to people. So we really would be looking at both bad ideas, sometimes hateful ideas, but would also have to put up with worse corruption, more arbitrary rule, no equality before the law, and ultimately nothing like shared prosperity, which we are still aspiring for.
The fault lines started around 50 years ago, you started to see them. Let's talk a little bit about that to understand what's shaping this, what's really putting shared prosperity into great risk. Can you take us through what changed at the end of the industrial era to the post-industrial era? What was this change that we saw?
I think if you want to understand what's going on with liberal democracy, you have to start with diagnosis. I would say the roots of our crisis go back at least 50 years. Liberal democracy's problems are rooted in the transition from industrial society to a post-industrial society. By post-industrial, I mean an economy in which digital technologies play an important role. Globalization combines supply chains and flow of goods much more seamlessly around the world. And these processes fundamentally change the nature of the economy, in particular making it much harder to reach shared prosperity. Not impossible, and there's nothing inevitable about either globalization or digital technologies. But the way that things have transpired, in the United States and many other industrialized economies, was that one of the major channels for creating shared prosperity, which went from businesses innovating, expanding, producing more, to workers also benefiting, was through businesses' desire to hire workers. And in a democratic framework with strong labor unions, that meant they had to pay higher wages, as well as create jobs. And that was something that created a powerful drive towards shared prosperity, because it meant that greater profits and greater revenues for businesses came hand-in-hand with higher wages.
Digital technologies changed every aspect of that bargain. First of all, they started favoring more college-educated workers, who could take managerial positions, work with data, administrative positions. But something else more important was also at play. That was the use of digital technologies as a platform for automation. By automation, I mean the use of machinery and algorithms to replace tasks that were previously done by workers. And automation is a game changer because it severs the link between business expansion and shared prosperity. I want to be very clear that this is not in the nature of digital technologies that they always have to increase inequality. It's the way that we developed and used these technologies.
Equally important, post-industrial society brought very different social relations. It elevated the social status and political power of college-educated workers, especially those integrated into the global economy in major urban centers, such as in New York or Boston. But at the same time, they started segregating into their own neighborhoods, their own social networks, their own communities, and diverging from the rest in terms of values. So this created a tension in a broad coalition that was essential for liberal democracy. For example, you can see that in how center-left parties, the Labor Party in Britain, social democratic parties in Europe, or the Democratic Party in the United States, which used to receive the majority of their votes from working-class people, people without a college degree, working in manual or service jobs, now that demographic group votes for the center-right or the hard-right, and center-left parties are overwhelmingly the party of choice for people who live in large urban areas with college degrees or postgraduate degrees. But center-left parties are essential for liberal democracy because they have been the pioneers of the welfare state, of shared prosperity, supporting the rights of workers. And now they became less concerned with the rising inequality or the non-college groups in society falling behind, and then ultimately, with the cultural politics of our era, even became a part of a big divide within society that has created a major backlash.
There's a lot of very unique examples you have in the book. In fact, you can correct me if I'm remembering this wrong, but in Ohio, the Anchor Hawking Company, where all of the people, whether you worked in the factory, you worked at the office, you were a manager, everybody's kids went to the same school, everybody lived on the same block. And then within a few decades, that was very much not the case. And certainly all of us sort of understand that there's certain areas where elite people live and certain places where they don't. How important are those examples to helping drive home to people how these ideas walk and talk in their everyday life?
You summarize it so well, thank you for doing that, but they really tell the story. Not only has there been a big sea change that you don't see managers and workers living in the same communities, sending their children to the same school, going to the same churches, but you have also broken one of the major ways in which you learn about the plight of other people or learn to empathize with them. And that has accelerated in the age of social media and online communication because we are interacting with each other much less in natural social networks. And that's very important for recognizing something that really, in hindsight, is shocking. In the 2016 election, when Hillary Clinton was defeated by Donald Trump in the United States, or in the 2016 Brexit referendum, most people who were very well-educated, living in major cities like London, New York, Boston, were genuinely shocked that there was so much reaction against the establishment. But that showed the extent of the discontent. But the shocking thing is how such strong feelings went unnoticed for almost three decades, and that is because we came to live in silos. The major newspapers that we read are part of that silo. Who we talk to is part of the silo, our education system is part of that silo, and it's not just a failure of our information ecosystem, but it also makes it much harder for us to form bridges with each other, to form that broad consensus that is so important for democracy to work.
People forget or overlook the connections between political institutions and economic prosperity. What is the before and after that? How does one feed into the other?
Institutions are critical. Part of the reason why I am drawn to study and want to improve liberal democracy is because I believe liberal democratic institutions are essential for the kinds of lives that we aspire to: peaceful, based on dialogue and understanding, prosperous, and free. If we start sacrificing our institutions, we risk losing all of these things very badly. That's why the book talks both about liberal ideas, how they have evolved, how they worked, and why they were important for shared prosperity, voice, and other things.
You write in the book, you say, liberalism's crisis is intimately linked to its failure to adapt to a changing world. And you argue that we should remake liberalism. Why is a new framework needed?
First of all, clearly something's not working. There is a real crisis here. People are willing to put to power ideas, leaders that really are very questionable in terms of their corruption, in terms of their incompetence. But it shows the strength of the feelings. I think liberalism needs a new governing philosophy. You cannot remake institutions without something that brings people together around their shared moral values and an understanding of how you could build a better society. That's what I mean by a governing philosophy, which is a collection of coherent ideas that need to be part of that. In the past, for instance, social democratic ideas played a very important role in the development of the European welfare state, in the affirmation of the desirability of shared prosperity, of providing much higher quality public services to poorer parts of the country. So we do need something like that again. Social democracy was for the industrial age, so that's why I emphasize something related, but different, what I call working-class liberalism. And that is my rallying cry for this sort of new governing philosophy.
You say AI could enable this shift. Right. You say AI, if steered correctly, can be complementary to the expansion of collective knowledge and the rebuilding of shared prosperity, and then you introduce this idea that we mentioned earlier in the introduction of pro-worker AI. Tell us, what is pro-worker AI?
Well, first let me take a step back, and again I'm sort of partly responding to your previous question because I didn't want to go on and on, but working-class liberalism at its heart has to be about shared prosperity, jobs, and better economic conditions for the working-class people who don't have a college degree, who are going to work in manual jobs, service jobs, and a variety of trades. So that's essential. The other very big, important element in working-class liberalism is community. So you have to bring communities in, give them room for self-government, to make different choices, experiment with their different values, and do that while being consistent with the founding principles of liberal individual freedom, individual rights to expression, and this drive for people to have voice.
Now, taken at its face value, AI is a threat to both of them. If AI goes in the AGI, artificial general intelligence, and an artificial superintelligence direction, it could displace a very large fraction of the workforce from their jobs in the United States, in Europe, ultimately in the emerging world. And that would make shared prosperity essentially impossible. It would make workers dispensable, not so important, second-class citizens, so to speak. It would usher in an era of inequality far surpassing our very, very unequal present. But AI would also, could also, continue the trend of social media and pacify us more, take us away from our communities, don't encourage us to get into the messy but ultimately essential business of making compromises in real-world interactions, getting to understand different parties, different segments, perspectives, form bridges, deliberate with them. All of these things are a real threat to the future of liberal democracy. But here is the catch. There's nothing in the nature of AI that it should do those things. It's how we develop AI and how we use AI. And the idea of pro-worker AI, it's about how we should conceptualize AI and how we could achieve it, and why it might actually be complementary to building shared prosperity and making a better liberal democracy.
And how does it walk and talk, pro-worker AI?
My argument is that AI is not just a very powerful generative technology that could be used for automation or become the basis of something like artificial general or super intelligence, but it's also an amazingly versatile and powerful information technology, meaning that it can provide very relevant, accurate, context-relevant information to individuals in real time. What that promises then is that all of us could become much better at solving problems in which we specialize, expanding our expertise, starting doing new things, new tasks, more innovation, provide better or different goods and services in our line of business. And those are very different from AGI or artificial super intelligence because they still have the worker in a critical part. We are making the worker's capabilities expand. We're making the workers perform new tasks. As a result, rather than being sidelined, this could increase demand for workers and re-energize that channel that I emphasize: business is doing well, but also paying higher wages and creating jobs. And that's so important for the future of shared prosperity and the future of liberal democracy.
Oh, that's not the direction in which we're going. Right now, there is more than a trillion dollars being spent on AI, and a very large fraction of that goes into building more and more sophisticated foundation models. They're very, very impressive, by the way. I will not sympathize with anybody who says these are just stochastic parrots. These are really capable models, and we can benefit a lot from them, but we also need to develop them in a way that could be complementary to humans. That's the pro-worker agenda. I give many examples in the book or in my other work, but there are so many places where you can see the promise of pro-worker AI. You can make nurses far better. They can play a much more important role in diagnosis, care, prescribing drugs, making follow-up decisions. You can make educators much more effective, essentially a single teacher can become, with the help of AI, much better at doing new things in terms of reorganizing the classroom and coming closer to a cost-effective but quite competent version of individualized education programs. You can make electricians or plumbers and all the trades much better because the problems that they are solving are becoming more complex. Now you can give them tools that expand their expertise and enable them to do many things than they could do. All of us working in knowledge occupations obviously can benefit from AI in the same way as well.
All of us have heard the promise of these wonderful things that could happen. But still, plenty of folks are equally worried, equally, like, gnawing on their fingernails. Why is it that this might not happen?
Oh, they are absolutely 100 percent, 1,000 percent right to be worried. This is not the direction we're going.
The kind of pro-worker AI I'm describing is actually not a big technological barrier. It is within our capabilities right now. There are a few companies that are working here, but they are marginal companies without the best scientists, the best models, the GPUs, and they are making progress. So we need much more of the resources to go in this direction. But it's not what's happening. It's not happening because of the economic model of Silicon Valley, but it is also because of the ideology of Silicon Valley, where all of the leaders are prioritizing building more and more human-like or human-improving models.
You mentioned the algorithmic cage. Can you talk a little bit about that and how that impacts how we think and maybe also the direction we might go in?
Communities, almost by their nature, create these cages of norms. They have their own norms. They try not to let people break out of them. When communities do that, we still have to protect individual liberty, and that's the complex relationship between community and freedom that is at the heart of the book as well. What I call the "algorithmic cage" is that when the same thing is done, say, by social media, you don't get many of the benefits, but you get an even more powerful version of that cage of norms because now the algorithms are policing you, or the algorithms are finding other people who are going to very strongly police you and are going to be very, very militant if necessary in forcing you into certain modes of behavior. And all of that happens without the shared moral values, the shared understanding, the development of information exchange, etc., that community brings.
There's an interesting anecdote you had in the book where after the pandemic you went to a conference and you were all excited, you're like, oh gosh, this is a big deal, we're going to talk about wage stagnation, and this is a really important thing, people need to be digging into this, it has all these other implications, and then folks didn't talk about that, and you're like, what? What was going on? And why am I bringing that up? Because do you think that people are focused on the wrong signals?
To some extent, yes. And that's, again, it's because of the algorithmic cage. And this is because of that division between the college-educated groups that have become very dominant in politics throughout Europe and the United States and the less educated. So wage inequality affects all of us, but really the burden is on people without a college degree in the United States. It's an extreme case, but it's a very telling extreme case. While wages of college-graduate Americans have increased, those of post-graduate Americans have increased very rapidly. By the 35 years from 1980 onwards, the real inflation-adjusted earnings of non-college workers, non-college men, declined almost every year. So inequality discussions have to start there. That is just a phenomenally large change. But who are those bearing that burden? Those are the non-college workers. And if you start thinking that the non-college workers have the wrong values, they're not coming part of your coalition, you stop empathizing because you're not interacting with them as much, then that goes to the second rank of the problems that you focus on. And that's exactly what I witnessed.
You make the point in the book that business models can also influence political frameworks. Can you tell a little about that connection and how maybe that can help unlock better perspectives for AI?
I think the most important aspect of the business model that we need to debate is what managers, CEOs view labor as. If you think that labor is a cost, that is going to push you more and more towards automation, because if something is a cost, you just try to cut it. That's the right thing to do. But labor, your workers, are also your main asset. And rather than just trying to limit the costs, what you should do is ask yourself, how can I make that key asset be more productive? That's where pro-worker AI would come. That's one aspect. Another aspect is the business model of, for example, social media. The social media mess that we have today, and I say very clearly in the book, social media is not responsible for the crisis of liberal democracy, but it is exacerbating it. But that mess is very related to the business models of the social media companies, which is capture people's attention and then make money by sending them individualized digital ads. That's a particular type of business model, but it's not consistent with the promise of digital technologies and online technologies that people believed in at the end of the 1990s or early 2000s, when, if you remember, everybody was talking about new economic democracy, news democracy, more organic forms of journalism, better communication. And they were looking at what was happening in the Arab Spring and celebrating that as a gift of Facebook and Twitter. Well, these people weren't fully naive. Of course, their analysis turned out to be incorrect, but their aspirations weren't naive because digital technologies are fully capable of doing that. They can create new public spheres for us. They can provide us with better information and more effective ways of forming bridges and sharing information with each other. But it's just that social media's business model isn't that.
If I'm a business leader listening to this, and I read your book, and there's that line where you say, you know, "we have to get better at identifying pro-worker, sorry, pro-worker technologies," right? We have to keep identifying pro-worker technologies. What do I do? What is my instinct that I'm moving towards, and what's the thing I should be doing instead? Can you help them sort of right-size? Business leaders who are actually producing things are very sympathetic, but what can they do?
First of all, they need tools from the tech sector to actualize pro-worker AI, and they're not getting that. Second, we've created expectations in the stock market through their boards that unless you're going lockstep in automating work using AI, you're falling behind your competitors. So they have to demonstrate to their boards, to their shareholders, that they're doing that. So they're actually between a rock and a hard place as well. That's why the first step is changing the narrative. Davos has been at the heart of this agenda of automation, and businesses should make more profits, they should get leaner, we should build more complex networks of supply chains. I think Davos people should also ask themselves that question: Is this the right direction that we're pushing as the aspiration for our future? Nobody thinks that they're taking anything into a bad direction, right? Everybody is trying to do the right thing.