Harnessing AI for Economic Growth While Ensuring Equality with Julian Jacobs: Episode Rerun

Welcome to AI Government and the Future, a podcast by Corner Alliance. We explore the intersection of artificial intelligence, government and the future. We work with government to create results. We ignite your agency's mission by helping you to design and implement high impact and innovative federal programs in AI broadband, cybersecurity, public safety, and more. Being a government ally is at the core of all we do. All right, Joining us today on the podcast, Julian Jacobs. He is a doctoral student at Oxford over at the UK on AI and economy issues. Deals a lot with inequality, policy implications, retraining, things like that. So it's pretty great to get him here on the podcast today and talk a little bit about the implications on society and governments for us. So Julian, welcome to the podcast. Thanks for having me, Alan. So first, just kind of just give us a sense of the work you're doing and what you think is foremost in mind. And I'm sure you're getting input from tons of people about questions and things to look at. So kind of give us a sense of like, what are the top topics these days? I think one of the first things I'll say is the research space as a whole has exploded since ChatGPT has started. You can imagine it used to be that you had a handful of economists and social scientists that were taking AI seriously in terms of its economic social effects. That's very different. Now where seems like pretty much everyone is in some way thinking about what AI's effects will be on economies and socio political order. Now in terms of my particular focus, I've been for a while very oriented around the quote, disequalizing effects of technological change, the capacity of AI to, to widen inequalities between people that are in some ways complemented by the technology or in a position of ownership of the technology versus people who are in a job that could potentially get automated. That strikes me as a particularly important question that maybe doesn't get discussed quite enough in policy circles. Now in terms of the substantive research questions, very focused on the financial cycle, very focused on the gap between America's rich and poor, and then very focused on the process by which we might work to build a middle class amid AI. And that deals with things like worker retraining, welfare and workforce adjustment policy, things like that. So I guess the idea is AI technology provides a potential opportunity to retrain maybe middle skill workers who might not be getting paid that much into jobs that are higher paying. Is that the idea? That's right, exactly. So you're obviously researching that. So is that possible? Is there or you know, and, and what does that look like? That's, you know, you always hear like, oh, you're going to get a green job or you're going to get an AI job. It sounds like very amorphous. What is specific? Well, it's a very difficult thing to study for many reasons, but probably the biggest reason is with worker retraining in general. In the data, we don't know what the counterfactual is. So, so you could look at federal workforce retraining programs and you could look at the outcomes, you could look at their wages, you could look at whether they were employed one year, two years out of the program, but we don't really know. You know, that's a very self selecting group. We don't necessarily know what the counterfactual would have been. We don't know would their outcome have been better or worse by how much if they had not participated in the worker retraining program. So I'd say that in general, measuring it is very difficult. What good evidence we do have of worker retraining programs is quite mixed. A lot of federal programs are underfunded, underutilized, and their effectiveness is pretty mixed and as a vehicle to bring people into good middle class jobs. So there's a dissonance because you read these AI policy papers and a lot of them are very much articulating some vision of workforce retraining. And then you go look at America's record in particular on worker retraining and it's not a particularly auspicious record. The best evidence or the best types of worker retraining programs that we can see in data tend to be sectoral programs. So you're retraining people specifically to turn them into coders or it's worker retraining with, you know, one end goal and one sector that they're focused on. But of course, you know, there's selection bias there. But we can talk about, I guess this is like anticipating the issue. I mean, I don't know. In the States here today, we had another banner jobs, you know, numbers, 275,000 jobs added and unemployment's been under 4% for, you know, a couple, you know, a year and a half, whatever now. So is this more like anticipating the problem or is it also, I guess there's an aspect where as AI creates more wealth, you would want some of that wealth to come down to the workers. Right? As well. So I think that this is one of the most important things that we could talk about on this topic, which is there is I think a misconception that the tendency and the bend of technological change has been to essentially bump up the unemployment rate or lead to aggregate job loss. And that's not really what previous technological shocks did, didn't really boost unemployment, but what it did lead to was this very substantive reallocation of where labor was in the economy. So you saw in the aftermath of post 1980s computerization, you see significant growth at the head and the tail of the wage distribution. So there's a lot of high wage jobs that are getting created that are in effect working with computers and then there's growth in low wage jobs and you see a kind of hollowing out of middle wage, middle, middle skilled jobs. And of course there's a ton of other effects, particularly in the American context of you know, policy decisions, outsourcing and what have you that are happening at the same time. But computerization, that shock did seem to have a substantive role in spurring that kind of hollowing out effect. With AI, the fear is that it's going to do something very similar. And there are papers that on a preliminary basis seem to suggest that the quote productivity effects of AI are, are being borne out most for the quote, let's put it this way, the less skilled workers within a particular firm. So this is the Klarna Diana, I assume you've seen their, their recent claims. Yeah, a good paper to look at on this would be the Erik Brynjelsen paper which is being cited like crazy right now where he looked at the kind of empirical effects of generative AI in a few industry case studies. Now the results themselves are narrow but if that is an effect that is borne out more broadly across the economy, that's interesting. We'll get that link because I'd like to look at that. Yeah, and what I was referring to is Klarna, this buy now, pay later company did an early experiment with ChatGPT where they built, you know, chatbots more based on their information, claimed that they were able to get same customer satisfaction scores for customer service, got rid of 700 people in their customer service organization, took query time answers from 11 minutes to 2 minutes but on and on and on and on. And so obviously this know where they targeted first was like, you know, and those are considered low skill call center job. So in this case we're not seeing like the middle hollowing out like before. I think in manufacturing or other technological revolutions you might have seen that middle getting hollowed out. We're really talking about lopping off the bottom skilled stuff. Well, I Think that we're not seeing, you know, the, the lowest wage workers in the world necessarily experience. I mean, there may be an automating effect there as well. There probably will be. Many of the jobs that we are talking about are actually would be called middle class jobs. And so what is true is that they may be lower skilled, but they're lower skilled within the context of a particular firm or industry, which is, I mean, in many cases right now with ChatGPT, LLMs, white collar workers. And so the million dollar question, I think with AI economic adjustment is what is the quality of the jobs that we're going to have over the next 10, 15 years? Not so much you had the aggregate total, but just what is that quality going to look like? And are we going to continue to see this bifurcation in the types of work people do where you've got a lot of very, very well paid, well remunerated jobs on the one hand, and a cohort of jobs where it's not very well paid and you're essentially providing services for the class of people that are really riding on the capital returns of these innovations. A few people really do well and then everybody else providing personal services to them kind of idea. So part of the issue you always have in this kind of technological change is the jobs of the future aren't at the table yet to advocate for themselves. And we don't know what those are. Right. I mean, I've heard some people speculate, like think about whenever you wanted to create an app in the past, like how many people have thought about that and just never done it because it was expensive. You got to go find some programmer and you need to hire them. You need to understand enough about coding to control what they're doing, blah, blah, like within a couple of years it feels like you're just gonna be able to spin up an app on the fly. Right. So how many businesses will that inspire? Just few, just a couple of people. Right. Able to go out. It's, I mean, it's gonna be an incredible generator. So it's, it does feel like that middle to high will kind of move into those roles, right? Oh, absolutely. I mean, look, I, you know, I think we're sometimes relatively good at having a sense of which jobs are gonna be displaced, but we're not typically very good at predicting which jobs are going to be created by new technology. And I think that's very true here. And the one caveat that I'll say here with AI, you know, regarding this point about labor displacement is on Some level, I think it's worth it for social scientists to just say we don't know. We don't really know what AI is capable of, what it's going to be capable of. I don't think anyone really knows. And so it may be that AI is a technology that is uniquely displacing and breaks many of the previous trends that we would have seen in the early 20th century Industrial Revolution and the kind of computerization revolution too. So it may be a totally different beast. So in the long term of economics, right, And I know in the long term we're all dead, right? But it's not like people are working in worse jobs than they were 150 years ago or something. So is it just that over time the workforce, the people were stuck kind of like in the old technology disruptions, they just died off and new people came in and they took the new jobs that were better. Is the issue that you have localized people who aren't that mobile and aren't that retrainable so they can't kind of play in the dynamic economy. And then they basically end up dying off and new people come in and then it changes or it's tough to say. So I'll say that there's probably two schools of thought or answers to that question. So one answer might be that there is a latency with technological change where it's, you have this displacing effect. You see significant growth in low wage work that's again providing services for higher wage workers. And then there's a latency in the creation of new middle wage jobs. Those jobs will eventually emerge. It just takes time and it's part of some kind of fundamental process of technological change. The other school of thought would be to say that actually there's nothing inherent to that process. Rather what we saw before was some government or exogenous shock to the labor market that resulted in significant middle wage jobs. You know, a good example of that would be the second World War and what that did to the growth of the American middle class and the kind of industrial innovations of place based investments of the US during that period. So it's unclear to me, and I think to many people which world we're in. It's probably a combination of the two that is most helpful. Certainly it would seem to me that there is a role of government in helping direct investment toward places where America or other countries could capture new productivity gains, spur new middle class middle wage jobs. Some good examples that people might go to are teaching or childcare or elder Care, you know, those are jobs that, you know, we often think of as maybe not being the highest paid. You know, there's nothing inherent to that. Those may be jobs where the people performing them should be remunerated and rewarded more in society. And I would anticipate that we'll see a significant uptake and increase in what those industries look like in the coming decades. I mean, you kind of referenced it earlier and sounded pretty checkered like what is the, the record on government intervention in these type of issues for training and fostering of new industries to replace. There's a role for government. What does the history tell us about how that role has played out? So since the 1970s, it's not been particularly effective. The US has seen market growth and inequality pretty much since the 1970s. And there's some evidence that that effect might be waning a little bit since the 2008 financial crisis, a little bit after. Actually where government is most helpful is essentially ensuring up some sort of social safety net for people to help with the, essentially to help navigate those periods of disruption. So if you look at the kind of great compression after the so called Gilded age in the 1920s, a period that was deeply intertwined with technology induced wealth concentration. And so, you know, people often make direct comparisons to that period and where we currently are now, the role of government in redistributing resources and directing investment to place based initiatives to spur innovation, that was deeply beneficial. That ensured that we weren't just creating new jobs, but we were actually creating new good jobs and new jobs that were connected with a strong middle class. So that's where I think government can be most helpful in terms of the worker retraining side of things. It's, as I said, it's a lot more mixed. Oftentimes employers tend to be the best place to retrain workers. And so government can be most effective if it's essentially a force subsidizing that retraining or directing workers to firms for retraining. This is sort of the German, you know, training and. Yeah, exactly. Right. There's a kind of a, there can be an apprenticeship element to it. That's interesting. So I guess what you're saying is like we don't really have a model where massive retraining in the face of technological disruption actually works. And you could argue that a lot of the attempts post Gilded Age didn't really work until the Second World War, which I think you kind of referenced. So it's this massive economic, I guess it's a dislocation Right. But it was a massive economic undertaking. And just to say on that it also extends broadly to inequality, however, its cause. There's a really great book, the Great Leveler by this historian, Walter Scheidel, where he goes through these four cases in which he basically has a thesis that there are four cases in which inequalities have been reduced across all of human history, all cultures, all contexts. So, you know, not exactly the most, you know, modest claim in the world. And those four, those four contexts are mass military mobilization, civil war, plague, or government collapse. We just had one a couple of years ago. Right, exactly. Well, so you can imagine he got his New York Times op ed around that time. And these seem to go in cycles. Right. There is some sort of massive event every generation or a couple of generations. Right. That then you see the integration of these technologies. Yeah. And I suppose the connected question too, right. Is okay, so you probably don't want to make the claim that plague is a consequence of inequality. Right. But is there some sort of self correcting almost, you know, from a social economy perspective, is there some kind of corrective mechanism when inequalities grow too high, whether it's caused by technology or not caused by technology, where it's for some kind of social unrest and that ultimately creates the leveling force? Well, hey, we're going to have a civil war next year here, so. Thought it was scheduled for a couple years. I didn't get the memo. Yeah, it's interesting to me, I don't know if you've ever heard of this website called WTF happened in 1971. And they basically just track measure after measure of increase in median wages or like, you know, low skill wages and like productivity, everything just starting to level off and go down after the early 70s. I don't think it's any coincidence that that followed the civil rights movement and all of a sudden some portion of Americans were, let's put it this way, less willing to offer welfare transfers or something like that. They were feeling less generous. That's at least, you know, one argument that social scientists often make. Right. And I guess economists would make the argument that that's when massive regulation started. Like we got things like nepa, you know, we stopped being able to build anything in the United States anymore. And like since the 70s, like we haven't built like a refinery in the United States or you know, we don't build nuclear power plants anymore. And so like I think that both sides of the political aisle would say, oh, there was like this break in societal cohesion around those things. Right? Well, yeah, so to say a few things on kind of post 1970s. Right. And I think it is probably helpful to lay out, you know, what the causes of inequality and just generally the malaise and the American economy have been. And it has been in malaise. I mean, productivity growth has been pretty stagnant since the 1980s. Sort of the Tyler Cowan age of stagnation. Yeah. And there's a lot of economists that I think are, you know, some would say that we're measuring productivity wrong. But you know, even if we're just proceeding under the assumption that there hasn't been a whole lot of productivity growth, I would point to a couple things. And some of them go outside of technology itself, which is we again post civil rights, really begun this process of substituting direct welfare transfers, access to credit and cheap credit. So if you look at private credit markets in The United States pre1970, you know, the US is pretty much in line with where Europe is. You know, and I say this with the context of, you know, today, I mean, we're known for our access to credit. You know, Americans, we like our credit cards. You know, we have a very robust but mildly debt market. And that whole process of financialization, I would argue, and I think a lot of other people would argue was inherently disequalizing and really turned the US into more of a investment driven economy, a less robust and dynamic economy as well. To give an example of what it looks like, I mean, imagine you have a real estate market in New York City. Instead of housing being just affordable or a kind of attainable thing for ordinary middle class people, you extend credit and it's cheap credit. And on the one hand that looks like a quite a good thing because they could use a mortgage to buy homes. But you have a bunch of people competing effectively driving up the asset price of that home and they're going into more debt. The person who owns that asset is wealthier. And so essentially what you have is a credit fueled spike in wealth inequalities across the US And I would say that that's certainly been one of the reasons for the malaise. Theoretically, if you'd use the credit to buy assets that you held and you could hold them, you would have benefited. Right, but we see, I mean, I don't know if you've looked at Michael Pettis and Matt Klein's book Trade wars or Class Wars. I mean, I think they kind of cite this time period as when basically the cold war policy of the US on trade stopped working. Right. We let our financial markets open, let people import to the United States to shore them up against the particularly Germany, Japan, South Korea, to shore them up against the Soviet Union. That stopped working in the late 60s, early 70s. And it wasn't until we really let it get out of control. And then we kind of clamped down on Japan and Plaza accords in the 80s. And then, then we let China in, which blew it another hole. And now we're kind of backtracking on that. So I do think they would agree that there's just been this financialization that benefits Wall street that's kind of screwed the middle class, de industrialized our manufacturing sector. And it feels like a lot of that is kind of now the effects of that have, over 40 years, have now done what they've done, and now we're kind of moving back the other way. I mean. Absolutely. And we all pay the price for this. I mean, if you're head of Goldman Sachs, you're fine. If you're head of Goldman Sachs, you're probably fine. But, you know, but even in the context of a recession where the government is stepping in, I mean, essentially what we're doing is we're taking private debt and we're turning it into public debt. So this whole cycle of growing American leverage, I think is deeply intertwined with American inequalities and some of the policy decision making that followed the 1970s. So I think WTF happened after 1971 is a pretty apt title. That's interesting. So we got to go back to AI though. We kind of got a little down the econ thing. Okay, so then you're looking at potentially there needs to be massive government transfers to people to help them. Well, first of all, number one, I guess we haven't seen massive job dislocation as a result of AI technology. Right. So we're talking about future potential. Where are you on ideas like, oh, the government's got to come in and give everybody UBI or Universal Basic Income or something like that. So I'm not a big fan of ubi. I think that it's a pretty broad, probably highly expensive policy. And there's a saying that I think is kind of apt here, which is a tolerable UBI would be unaffordable. Unaffordable UBI would be not tolerable. And I think that is kind of the point. Right. If you are using a UBI as a supplement for pretty much the entire existing welfare state, that's not in most versions I've seen of the way UBIs have been proposed. That's not exactly at all a workable way of doing economic adjustment. If you try to run a UBI in addition to the existing welfare state, well, it's not necessarily the most efficient program and there are some legitimate questions about how expensive it would be to roll out something like that. Now the good news is I actually think there are some quite good policies in the US that really just need to be expanded. The earned income tax credit being one example of that. I think a negative income tax in a vehicle that is more effective than a UBI essentially achieves what a UBI aims to achieve, which is just ensuring some basic standard of living across the board. I think that there's also a role for targeted workforce adjustment. So having a more dynamic economy where there are, let's say, employer led workers, retraining opportunities, there's industrial investment by the government in regions that may be getting left behind, that's I think an important element too, not just for those communities, but for the economy as a whole. I mean, when we have high levels of geographic concentration in AI innovation, there are productivity effects that we're losing out on. And some of those might be productivity gains in agriculture or in healthcare clusters. Healthcare manufacturing is a good example of that. A huge industry going forward. I think, you know, that people don't think about that much. Exactly. And I think that those are some of the responses that I would look to see. So you're really looking at it as like I'm not out here saying we should go try to retrain a bunch of people to be AI to get AI jobs or something. It's like, no, I mean, and I think people know that, right. Like we've all got people back in our hometowns who are working, you know, been in the same job for like 30, 40 years, kind of close to retirement or something. And you know, we all know they're not going to be retrained to do AI overnight or to, you know, they've worked very hard their whole life. They happen to be in a job where the skill set that that occupation requires is increasingly not in demand. But I don't think that they should be punished in terms of just having a poor livelihood for the rest of their life because of that change. You're kind of looking at a industrial policy combined with some, hey, maybe we get some like workforce policy where we're partnering with companies on apprenticeships, job trainings that they're doing and then some basic, however you call it, like using an earned income tax credit or something else. To do income supplementation. Yeah. And I think there is a role here for government to help direct where, you know, where we should be creating jobs. Right. If more resources in our hospitals is an important thing for us to do, there are ways that the government can subsidize that same thing with elder care and childcare and teachers. And that's to say nothing of the fact that there are shortages in many legacy industries. Manufacturing is a good example where there are shortages of workers in certain good middle wage jobs that we can't fill just because there is a lagging effect of people actually being able to learn those necessary skills to perform that work. So there is, I think, a place for government to help direct folks to the places where their work is most useful. So then I guess the obvious objection would be going to cost too much. Right. You're talking about massive government funding programs and we're already trillions in debt and getting worse every day. So how are economists kind of are people looking at this thinking about that? Plus we have obviously a huge generation retiring, are about to hit the entitlement programs pretty hard at the same time. Yeah, I'm not convinced that this would be an unaffordable program. I think that this is also a program that's deeply intertwined with growing American productivity and growth in general. I mean, and that's ultimately how I think about it. Right. If you're going into debt or you know, increasing government spending, but it's intertwined with increasing economic growth, I don't really view that as an issue. I don't think that there's any reason to be worried about that. We've done it before, we can do it again. And we had some of the best years of this country in terms of how dynamic the economy was immediately following those kinds of government interventions. And I guess we have an N of what, like one or two of those that we can point to. What do you mean exactly? Like where government directed massive resources toward different parts of society and then we had this prosperity boom afterwards, you know, so like, do we have is like literally World War II, the only thing we got to really look at? Well, American history, We don't have a particularly long history. So I think that that would be the context that I would look at in terms of a modern case. You know, you could look at other countries that did similar things. You know, I guess we could look at like the research university, land grants, you know, the railroads stuff and sort of that post Civil War era where we kind of built out a lot of national infrastructure. Yeah, you could do that too. I mean, I think that a good way of studying that kind of question is to take a comparative approach. Look at other countries, look at countries that didn't do those kinds of initiate those kinds of programs. You know, I'm not talking about having government involvement across industry or nationalizing things or spending money frivolously. I'm not one of these people that thinks a national debt doesn't matter at all. Of course it does. But this is not about just spending money for the sake of spending money. This is a program. This is an investment in American economic dynamism. What kind of grades would you give the current administration on their multiple investments in the Iraq. Right, the CHIPS Act. Are these the beginnings of what you're talking about? Well, you know, especially talking about AI, you know, the AI context. There are some promising things that the Biden administration has done. The National Science foundation has been funding AI institutes where those are regionally diffuse, essentially funding clusters that are in many cases working to complement legacy industries with AI. So there's a case of Iowa farming where they're deploying AI and researching AI use cases in farming, and that's in theory boosting productivity in the farming cluster there. Build back better regional challenge. Many of the grants that have gone out there are again these kinds of regionally diffuse AI enabled or intertwined funding packages or innovation plans. So those are things. Two examples of things I would look at. I would say that the potential of the Biden administration to fund place based investment, a greater role for industrial investment, was much better than it turned out in practice. I think that within their policies there's the skeleton of some quite good initiatives. What do we learn from it? Like, where do we go next? Now that we've seen this? I mean, I would like to see continued growth in those place based investments. And I'd like to see a investment in the earned income tax credit, funding for the earned income tax credit, funding for trade adjustment assistance, the widening of trade adjustment assistance to include technological disruption as well, so that workers who are displaced not just by trade but also by AI can benefit from that kind of program. Those are some of the things that I would point to. And you know, I think that there's some element of broader US economic policy that maybe is isn't even focused on AI per se, that is a little bit more bent on having an economy with a strong middle class that will help with AI economic adjustment, even if it's not tailored for AI. Okay, so that's on the sort of income transfer side, but I guess on the investment into new industrial tech and that side you're kind of seeing things like the CHIPS act or ira. That's just the beginning of what we need to do. I think that's right. And are you seeing any areas or looking at some of the studies like where is it we need to put this money to? Where do we need to be investing going forward that we aren't currently. Well, I'll give one example which is I think the health care sector. I mean there's incredible potential for AI to make the health care sector much, much more efficient, cut down on costs and really bolster one of the largest industries in the United States. That's an area where I would love to see future administrations invest. Yeah, I was actually at this summit where this woman was talking about all the, just the genetic therapies that are coming and she had sort of all these concentric circles and in the middle was what already existed. There were some names there that you'd seen, but just two years out around that circle were just dozens of names and it was all broken up by sector. Right. And it was just there were names everywhere. I mean, there's so much stuff coming. And so she was interviewed after giving her presentation and they asked her what the problem's going to be. And she said, oh, manufacturing this stuff, like we can't make it. We can create like just like people are trying to get GLP1 drugs right now for weight loss. Right. Like they can't make this stuff fast enough. In a lot of cases we've lost some of the capabilities and skills, but others need to be developed. So does strike me that's a massive thing that we can be investing. And I'm sure parts of what we're doing in advanced manufacturing and things like that are hitting that, but it's an exact thing like that's a growing part of the economy, right. Aging population makes huge productivity. If you hit it right, it's going to be the world leading industry. And that's exactly where we should be investing for the future. And in theory something where you would imagine bipartisan support would be possible. And I'd say that, you know, this is not so much a prediction as more of a question, which is AI has significant national security implications. And as deeply there's, you know, a lot of the research coming out of the federal government is very cognizant of the national security, military implications of AI. And so I do wonder whether that actually makes it easier to deploy government spending in the spirit of building a strong middle class and generating bipartisan support. Again, taking kind of these historical examples, the US has tended to pair those kinds of big spending programs with some sort of military, military national security context. And so I guess overall, sort of the message I take is like, look, this is coming. We're not going to stop. I mean, we could try to stop it, but it feels like we're killing the jobs of the future and trying to do so. So it's, it's coming and like, and basically the government's role is going to be invest in, you know, some amount of income transfer assistance and investing in the areas of future growth and kind of let the AI thing play out. But don't try to go turn everybody into an AI prompt engineer. Don't go holding like, prompt engineering classes in like, rural America to take over your like, call center job or whatever, you know. No, I mean, I think in the AI age, one of the most important skills a person can have is the skill to be human. And that's something that I think from a government policy perspective, we should be rewarding as much as possible. Well, I think that's where we're going to leave it. I love that line at the end. The most important skill in the AI age is being human. Well, Julian, this has been awesome. Thanks so much for coming on. We appreciate it. Thanks very much for having me. AI Government and the Future is brought to you by Corner Alliance. To find out more about Corner alliance and how we work with government to create results, visit our website at corneralliance. Com and then make sure to search for AI Government future in Apple Podcasts, Spotify and Google Podcasts or anywhere else podcasts are found and click subscribe so you don't miss any future episodes. On behalf of the team here at Corner alliance, thanks for listening.