Are we betting too much on AI? | Nobel Laureate Daron Acemoglu

30:54So one problem, for example, imagine we have pro-worker AI, but the developing world is falling behind in terms of infrastructure education. that's another challenge because now pro-worker AI will drive growth in Canada and the US but not in Mexico or not in Paraguay so we have to worry about those things as well but there isn't just even two there is a continuum and a multi-dimensional continuum of futures we talked about the pro-worker AI future another one is the Chinese future where some aspects are better than the US, where they're actually much more insistent on integrating AI into the production process.

31:40So they may actually get more productivity gains faster than the US. On the other hand, they're copying a lot of the US technology, so that's not feasible in the long run. Then on the downside, AI is a powerful tool for surveillance that has really pacified the population. If we go the sort of the Elon Musk path, then I think we really have a private company's dominating AI as a centralizing technology. And the whole thing possibly leading to a two-tier society in which a lot of workers are sidelined. or it could lead to a very non-democratic society because even though we may want to keep democracy alive, either the foundations of it are shaken because of this two-tier structure or because inequality reaches such levels that the powerful parties may decide, well, we need repression a la China.

32:45So there are so many possibilities. We also have one in which, as I mentioned, we get a big slowdown in investments and some companies go bankrupt and then we have to reconstitute like a pattern where we've had these AI springs and AI winters in the past and this will be like the mother of all winters. There are just so many potentials. Some of them are under our control. That's where the democratic process comes in. but some of them are not under our control at all. And also, final point on this, our democracy is already ailing, obviously. That's why I wrote that book. So we have to be realistic about what we can achieve and do that before it's too late.

33:36So you made a few points there. The first is that let's say we don't have this fast takeoff and this really destabilizing transition where we wake up tomorrow and AI is super reliable and agents work, and we have modest productivity gains. But if we have designed it and shaped it in such a way that it is pro-worker, workers could actually benefit as they have historically. You might see rising wages, productivity might be shared, and that's a pretty good future. So we don't even need to think about new redistribution mechanisms. And I know UBI is very linear thinking, but those types of ideas don't even need to come into the picture.

34:11We could all win. But then there's the other complexity of, well, there's the surveillance aspect of the technology. There's the fact that most of it is run by the private sector, which also goes back to taxation decisions to kind of offload investment in R &D and all sorts of things to the private sector. But here we are. So then there's the surveillance complexity. And then you have, but potentially there is a scenario B where it does take off. I mean, you can't predict a breakthrough and it is possible that Transformers isn't where it ends. And something happens in the next two to three years where we do see a faster takeoff and we get that wobbly transition for a decade or more, is there a different economic structure?

34:51I know we hear about UBI. We've talked a lot about that in this podcast, but is there something different we should be doing to how income tax, all of that system works right now? Because it seems to keep going like this. You're asking, again, fantastic questions, and I wish I had better answers, but I'm going to preface it with the same thing. there's just so much uncertainty. Even pro-worker AI. I've been advocating it. I think it's our best chance. But there is a chance that it may not work. Like it may actually not be feasible. If indeed we transition very quickly to something like AGI, and it's a comprehensive AGI where AI models are better than us in everything, then you can't really have meaningful pro-worker things.

35:35Second, imagine we transition to something like pro-worker AI, we still need a lot of public support for workers. First, the labor market will be almost certainly different. Many people today are still today, many more were in the 1980s, but still today are in routine jobs. Those jobs will be done by AI. So when I talk about pro-worker AI, I don't mean that AI should not be used for automation. It will be used for automation. It should be used for automation. We welcome automation if it's coupled with other things that creates jobs at the same time. So we need different skills. Where will people get those skills?

36:15Well, in vocational education programs, but mostly in schools. So our schools need to adapt. So that's a public education problem. Most likely, many jobs will be much more dynamic. So workers will need to be more flexible. Again, that is a very difficult skill, by the way, to teach, especially in low-income schools and neighborhoods. But that flexibility isn't enough. We also need better social insurance programs. So we do need to adapt our institutions. And exactly in what way we're going to adapt them depends on how those uncertainties that I mentioned will be resolved. Yes, UBI is very linear thinking.

36:52I love that term. I had not heard it in this context. It's like a very simplistic solution. First of all, it won't work. I mean, can you imagine us in our current political economy to actually fund a decent UBI? I can't. The political economy of it doesn't work. Second, to me, it's like throwing the towel. It's saying AI is out of control. The only thing we can do is create our modern version of bread and circus. Third, it's actually, even if it worked and if it was generous, it would not prevent what I've called a two-tier society. Everybody would understand that 80%, 70%, however many people just live on the crumbs of the tech billionaires, and that would create a very big, very steep status hierarchy, and that would be very inconsistent with liberal democracy.

37:48Yeah, I think UBI, I know depending on who's talking about it, for some people who bring it up. They mean it with the best intention. But to me, it takes us to a state of disempowerment and then asks, now what? Versus we're still here. The future hasn't happened yet. We just need to get a lot more creative in how we're thinking. And I get the intention behind we should stop this technology, but we need to, that's actually an easy scapegoat for companies. It is easy. It's impossible. And it's also, you know, look, I've spent as much of my career studying political economy and the history of political economy as I've studied technology.

38:30And I would be the first one to tell you that almost every example of trying to stop or block technology in the past has been disastrous. So you really have to be very careful whenever you're asking people to resist technology. You need to have a positive, proactive plan. So the image that I have in my mind that I try to communicate isn't that we should try to slow or stop AI. The image is we are in a fast car. We're driving 200 miles an hour towards a steep cliff. We need to steer away from it. You can't steer a 200-mile car. You might need to hit the brake a little bit and then do it gently.

39:19So that's the kind of slowdown that perhaps we need to think about, that, yeah, let's not put another trillion dollars into this race for AGI. Let's think how we can use both the engineering talent that we have, which is very scarce, and the funding and the entrepreneurial energy to develop applications that are going to be more useful for society and world workers. That's the redirection. Slow, gently cut, do a steep one. And I think some of this is also, we don't have a proper long-term vision from leaders. And I think some of, democracy is amazing, but some of the challenges are, we think short-term, we think in election cycles.

40:03And so we don't really hear from a leader in 12 to 14 years, here's why this technology is going to be worth it if we build it this way. So we're stuck in this kind of, this presentism. We really suffer from it. That's so important. This is what I've also emphasized almost in every conversation if the opportunity arises. Our approach to regulation is completely wrong. The way we regulate tech right now is we have the most powerful corporations humanity has ever seen. First, we give them a complete carte blanche. They can do whatever they want. They have lawyers to enable them to do whatever they want.

40:39And in four or five years' time, if some of the things they have done turns out to be sufficiently disastrous, then we put some backward-looking reactive regulations. That is a complete recipe for disaster. Instead, we need to have two realizations at the same time. And you put your finger on both of them. First, we need a proactive regulation, meaning which starts with where do we want to go and how we can go there and how can we set the institutions and the governance structure to achieve that. Second, recognize that you're dealing with the most powerful corporations you can imagine. Don't kid yourself that these are like small corporations in the local economy.

41:29They are going to be really powerful. If they have their own agenda, they're going to push that agenda. You have to grapple with that fact. They're not our enemies. There's a lot of innovative talent there, but they have their own agenda, and they're very powerful, so you have to take that into account. Yeah, I think there's so much with this technology that is worth fighting for, and I mean, I wouldn't be able to do the work I do in foresight if I didn't see visions of the future that I thought worked for most of us, at least. But you do have to take people there. and not all of it even has to be regulation.

42:04I mean, what we've done with accounting is so interesting where governments essentially outsource the regulation to private companies that can make money from ensuring people stay in check, right? We could have regulatory markets. I know economist Jillian Hadfield talks about this a lot. We can get innovative and people can make money steering this in the right direction. This isn't a charity, but we just have to expand how we're thinking and our design space. And the details really matter. The details really matter. Like, am I in favor of the military-industrial complex? Hell no. I mean, I think we definitely spend too much on defense, and there are so many things that go wrong there.

42:45But go to the first two decades after World War II, and DARPA, which came out of the defense sector during those two decades, was a very forward-looking organization that not only had a clear vision of technology and how it should be steered, it was extremely open-minded. It ventured into areas far beyond defense and killing machines and really laid the foundations of many of the technologies that came later. So the details of how you actually support different types of technologies, even in the defense area, could have very different consequences. Yeah, I think a lot of the post-World War II innovation and investment birthed the kind of modern progress that we all take for granted now, whether you're thinking about the internet, computers, solar, if we can finally get on board with that.

43:37The list goes on and on and on. And even what came out of NASA, you also had international agreements and countries that work together that would have never even wanted to speak to one another from that science. So it is possible, right? This isn't just fantasy thinking. And I have two questions. Just on that point, it's a very important point, Shania. AI is a global technology. It's spread, its application is going to be global. So it's insane to think that we can just content ourselves with national policy. And the framing of race towards AGI has had another very pernicious effect. It has, rather than enable us to work with China, it has created this oppression of a zero-sum race with China and completely closed off all collaboration with China, not just on AI but on other things as well.

44:27And look, I am a very vocal critic of the Chinese system. It has a lot of problems. But when it comes to AI, when it comes to climate change, when it comes to nuclear non-proliferation and pandemics, we have to have communication and collaboration with China. and the AI race has also ruined that. Yeah, you can't just think national in this moment. It's not feasible. AI doesn't have a passport. It's clearly already diffusing globally. And if we're going to be honest, I mean, America's focus on AGI, China's playing an entirely different game now. America's playing chess. China's playing checkers with diffusion.

45:09And so now they're not even on that same landscape. They don't necessarily see it as zero sum. They see an infrastructure play. So I think what people are measuring is also perhaps not even helping their own cause or their own race. It may actually be holding that same argument back if we're going to actually measure where this technology is and who's adopting what where. And so what do you think we should be measuring in this moment? So if we looked at the late 1800s, early 1900s, we started to measure employment, unemployment, product safety. As we start to move into this strange to New Era, what are some of the things you think institutions, people should start measuring?

45:48Because I think the metrics matter. That's a great question. I think some of the labor market outcomes we are measuring are still very central, what wages people receive, their options in terms of employment, mobility, but also probably we need to understand job satisfaction, where they find dignity and things like that a little bit better. But there's so much more that we need to measure in terms of AI models. Like we need an auditing framework for AI models so that we know what they can be used that's damaging for society. So sometimes some of these things are a little exaggerated, but I don't doubt that Mythos, for example, did have some very powerful capabilities that could have been used for ill.

46:53So how can regulatory agencies or government understand those things and start building the right guardrails ahead of time? I think that's part of the measurement problem. Right. I would agree. And I think even with Mythos or whether people think it helped a pending IPO valuation, we're probably better off not. You're much more cynical than I am. Oh, no. I know. Who would have thought? I think we're probably better off not finding out if Mythos could accurately hack the bank and just preparing for the worst, steering towards the best, but preparing against the worst case scenario. And we do have a really diverse set of listeners to this show.

47:39We have key decision makers in governments, in global alliances, as well as civil society activists, the whole thing. So if you were to give two key takeaways from your book, the first would, what would be the message that you want policymakers policymakers to walk away with, because I know a lot of them listen. And then the second is from the general public. How can we think non-linearly and see how much power we do have in this moment? Well, so actually, it's easier for me to give, I mean, some of it is canned, I apologize in advance, but to give advice to CEOs and managers and the civil society and policymakers, and I'll explain why.

48:19So let me start with CEOs. And my experience is that actually, when I talk to CEOs who know their business, this is advice that they sympathize with. You shouldn't think of labor as a cost to be cut. I think especially in our age where we need more innovation, more new goods and services in a changing world, labor is your most important resource. And if you start with that mindset, you can articulate a demand for technology that's much more in line with increasing productivity and using your workers more productively than just trying to automate, which is neither that easy nor often that profitable.

49:01For civil society, I think we just all need to get informed. I think that's the first important thing. And second, recognize that actually civil society as a whole has much, much more power than I could have even myself imagined 10, 15 years ago. If there is today no regulations, no guardrails, no effort to steer AI, that's because the narrative has supported a vision of AI where everybody's going to benefit, AI is inevitable, and geniuses are leading the charge. So if we want a better future from AI, we need to form a different narrative, and that starts with civil society. For policymakers, and here is the difficulty, it would be an exaggeration, but it wouldn't be a massive exaggeration to say that right now AI has two beating hearts.

49:58One is in China, one is in the US. So I would have very clear advice to US policymakers. But what about Canada? I think the issue here is that you can do things locally. You can encourage your companies to use AI the right way. You can encourage your talent to go into the right field of AI, but that's small. I think a bigger thing that policymakers can do is work towards an international alliance for building policy. At the end of the day, we need international policy. Countries like Canada, the UK have a lot more clout than first meet the eye. And they have expertise. they have more power if they can leverage a large number of countries coming together along well-defined aims for steering AI the right way.

50:53Yeah, and I loved your point. I mean, the first, especially about companies, if you're only thinking of your bottom line and AI as an automation technology, that's an existential move because you're assuming your business model stays the same versus in the future, that's how you compete versus thinking of AI as a top line technology. What can you do differently? And that's just the basic fundamentals of business. So I hope people can catch that finally. And I think civil society, it's an important one. I think we forget how much power we do have. And I know in your book, you talk about it was easier for labor to organize when we were in manufacturing companies because we were in a central location.

51:31Everyone was going to the same place, nine to five doing the exact same thing. But the irony of social media, which I think has given us all a massive headache, is that we're also all there. So maybe some of these technologies, and it doesn't have to be the platforms you see today. There's one other thing, Jeanette. Yes. So it's not just proximity enables you to coordinate. It's not just proximity can give you more power. Solidarity. so workplaces were repositories of resistance organization because they built solidarity so the question is can we find online spaces in which we build solidarity i don't think the answer is no but right now i know of no example and social media kills solidarity, doesn't build solidarity.

52:24So we have to find online interaction modes in which we can build solidarity. And I think at the end of the day, for solidarity, we need community. That's why a very large part of my book is about community and the role of community in freedom. And I think that there actually is one small example of how we could do the online, of online spaces that work and then lead to people not talking to their neighbors for six speaks afterwards. If you look at Tokyo in their most recent election, there was a candidate, the first candidate under 40 to gain a meaningful share in the race. I think he only got 2.5 % of the vote, but came out of nowhere.

53:02And he followed Audrey Tang, which was Taiwan's minister of digital innovation, I believe, or minister of the digital component. And Audrey had spoken about in their book, Broad Listening. So can we actually use these technologies to listen to what people are saying, what they need, the nuance of it? And he used technology to do that and actually played a significant role in the election. He was fifth in line. And so there are these small examples. It's not going to work. That's a wonderful example. I didn't know about the Tokyo thing, but - I'll send you the paper after. It's really interesting.

53:36I would love that. I would love that. But yeah, I mean, power and progress, I give Audrey and her achievements in Taiwan is an example of how you could have used digital technologies more generally in AI in a more pro-democracy way. And absolutely, there's a lot more that can be done. It's just that we're not doing it. Right. Right. We're just not doing it. Great example. Thank you. Okay. So we have, I think, 15 minutes and then you're out. So there are five questions, maybe six. We'll see how many we can get to that we had from our audience. They're really great questions. So I would love to hear your take on them.

54:09So the first is from, and I really apologize in true Canadian fashion in advance if I get these names wrong, but there's Girish Meghalani. Their question, will AI enable us to improve wealth distribution and overcome this extreme concentration at the top? No. No. I think, again, the future is very hard to know. and it may well be that all these AI investments go bust so badly that some trillionaires become millionaires. But here is the problem. If you get so much investment, right now, this year probably about a trillion dollars, that's going to have returns. And if it has anything like the returns that people are expecting, that's even more returns to capital and less to labor.

55:04So already we're on a path to expand inequality. The pro-worker direction would ameliorate that. Pro-worker models would be more domain-specific, less centralized. But I think the centralizing tendencies of AI would persist even if we steered it in a somewhat more pro-worker direction. So we may need more other regulations, and that's where antitrust is really important to really deal with this. and I think that wealth inequality question should really be bundled with how centralized we want our economy to be. There is a tendency among some people to think, our centralized economy, that was the Soviet Union.

55:49But if one company plays a very, very important role, if one AI model plays a very, very important role, that's also a centralizing model. And it will be associated with inequality. It will be associated with imbalances of power. But can't we adjust the capital, the returns to capital and how much goes to labor? I know that - That's why we need policy. Yes, absolutely. I mean, I answer this by saying like, if we don't take major policy decisions. So one of the policies that I have advocated for quite a while, both on fairness, but also economic efficiency grounds is get rid of the distortions we have in the tax system, whereby we don't tax capital.

56:26We tax labor, we don't tax capital. So we may need to tax capital returns much more. I think actually a feasible, effective, fair, and efficient model would be to tax all income the same, regardless of whether we label it capital income or not, and make sure that capital income cannot go to the Cayman Islands or hidden this way or that way. That would really change things. And I suppose maybe that feeds into the next question from Natasia Bowman. What is the best measure for reducing wealth inequality? I would start with three levels of inequality. that we have to bear in mind. And we have to reduce all three of them.

57:09Labor income inequality, income inequality, and wealth inequality. Labor income inequality is what really drove the huge increase in inequality in the United States in the 80s, 90s, 2000s. Income inequality became even more as capital returns also increased later. And then because we did not tax capital and we also did not exercise antitrust, wealth inequality then ballooned. Although I should also preface this by saying that we have very good measures of labor income inequality and income inequality. So when I say inequality increase in the United States, I'm 99 % not sure of that. Wealth inequality is really badly measured.

57:54So when I say wealth inequality increase, there's much more debate on that. But I think it seems pretty obvious that it did.

58:03So I think it would be far easier to deal with labor income inequality and income inequality by redirecting AI, creating more jobs, more opportunities, more wage growth. Once wealth inequality has reached a huge level as it has today, you cannot rectify that. Elon Musk's wealth will remain at$1 trillion unless he loses it or his companies go bust. So that's where taxation and other anti-monopoly measures will have to come in. The next question, John Polo, what career should high school and college graduates enter into? Okay, that intersects with the discussion we had about, you know, the public education system needs to adapt.

58:50And I would say three principles that we should emphasize to students. One is learn AI. There's no way you're going to do okay in the future if you have never worked with AI, if you don't understand AI. Second, develop new norms that are adapted to the AI age, and there are several aspects to that. One is we need to emphasize our civic responsibilities in the age of AI. That's about using AI responsibly. That's about participating in the debates and democratic politics. After all, democracy doesn't exist independent of people's participation, and we've neglected that. I think those are actually very important.

59:48Those are things that people learn in school. I mean, I think part of the reason why people have been taking democracy for granted is we haven't actually engaged in the right kind of civic education. either we've tried to impose values top-down in schools or we've completely ignored the civic values. So that's another very important thing. It becomes even more important in the age of AI. And then finally, more directly to your question,

1:00:17I think flexibility skills are going to be very important because no specific area can be said to be completely immune to AI, but there are going to be many parts of skills that a particular occupation requires that will still need human input. But what those are is often not so clear. Yes, it won't be the most routine parts, but would it be the ones that require more experience, more expertise, more social interactions, more technical understanding? So you need to have a holistic enough understanding with a good grounding that you have that flexibility to shift across tasks and find where your contribution can be greatest.

1:01:07Yeah, I would totally agree. I think one thing, and I talk a lot about this on my channels, is we have to move on from the idea of static job titles and preparing for one lane. That was a chapter of history that worked for a certain economy. That economy is gone. What are the underlying skills in that job you aspire to hold or in the job that you hold today? And think of yourself as this kind of mini institution that in your skills evolve over time and you kind of can move and work in different places. I think even the idea, I know we're all worried about the end of the career ladder. And until we get to whatever comes next, there's a lot to be concerned about.

1:01:43But the career ladder also accompanied a certain economy. And so we're going somewhere different. And education also needs to also get with the time. Education will have to look as different as the economy we're heading into. So everyone's job, everyone's institution, everyone's plan needs to change if we're going to get this right. I love the way you put it. That's 100%. What are the plans for dealing with a growing human population? And this is from Rukob always. Actually, that's one place where, depending on your perspective, the news is good or bad. world population will stop growing in about a decade or two, and it's already stopped growing in most of the developed world.

1:02:30Some people are very worried about that because they think that population declines would create macroeconomic problems, they would create innovation problems, they would create lack of young workers, would create lots of issues. Some people celebrate that. I think it creates risks, but my research also shows that labor scarcity that comes when entering cohorts are small are actually quite good for wage growth and they're actually good for how we use technology. the reason being that when labor is scarce companies really need to economize on labor and that makes the automation more productive but also it gives them incentives to use existing labor in the right way that's the reason why for example german companies didn't lay off their blue-collar workers when they introduced robots because labor was so valuable and they had trained it.

1:03:27And so they retrain them to make them technicians. So we really need a different mindset, but it can work very well. And in history, it's actually worked reasonably well. Right. Yeah. I think the data does show we're going down in population over the medium to long term versus up. And you're totally right, depending on who you talk to, that's either Godson for the planet or it's a nightmare for markets. So it totally depends on where your savings account is. And then Sharnett Holmes says, will the$40 trillion debt crisis, high energy cost, high inflation and stagnant job market cause a crash? Possible.

1:04:08I don't have a crystal ball. I think stock markets are very forward looking. So if there was a general consensus that there will be a crash in the next year's time. There will be a crash today. So I think the way I would characterize it is that there is certainly some hype in AI and the investments are amazingly large, but so far there's also a lot of private wealth. There's a lot of money in sovereign wealth funds, SoftBank, UAE, Saudi Arabia, U.S. trillionaires. So they can keep on financing hundreds of billions of dollars of losses for quite a while. So it's hard to know whether there will be a slowdown or a hard stop.

1:05:03But the risk is there. If you're squeamish about risks, take that into account in your portfolio. The final question, and I know we chatted about it a little bit. This is from, I think her name is Nikki. What can everyday people do to enact change and to shape the future of how we use this technology, of how we use AI? Become informed. Turn up to discussions. Be part of your community. And try to improve the quality of the conversation. I am perhaps naive in my thinking. but I think if in late 2010s, if instead of this unproductive false dichotomy of AI is going to be our savior led by geniuses versus killer robots are going to come for us, if we had started having the conversations that we are having today, we would be in a much better place today.

1:06:08and it's the responsibility of all of us not just all the journalists to elevate the conversation to that level i agree i think the future is much more nuanced than jobs no jobs ai is amazing air is terrible that's just not how society history has ever worked so if we want to elevate and reach a different future we have to meet the moment with the nuance that it requires. You put it always much better than I do. Thank you.