Fed's Daly on the Economic Impact of AI, Labor Markets, Monetary Policy

Well, like most things in the world, neither of the extremes are usually useful as good descriptors of where we're likely to end up. So I think ultimately the future that we create is our decision. Technology doesn't make choices. It's a tool and is a powerful tool. But we need to make the choice of where we want to be, and I think they touched on it in the last panel. Daniela talked about that, about how you have to think about what do you want this to do for you and what do you want in life that comes from it? And that's where we are. So I'm enthusiastic about what the models and tools can do. But I do believe strongly that we all need to appreciate what we need to do with it, and how we need to help everyone come along. Otherwise, the world of the doomsayers versus the enthusiasts gets determined by things outside of us, and really, we want to harness it. I always tell people, technology can harness you or you can harness technology. I think we should be in the driver's seat and not the passenger seat here.

I'm really turning to businesses and asking them, what are you doing? And not the technology businesses, the businesses that are going to determine whether this is as transformative as many think it will be, or whether it's just a cost-effective play that makes you better at doing the jobs, but you're not going to be creating new opportunities. And I'm already seeing very clear signs that businesses are asking not the question of how can I use it to simply do things faster, better, cheaper, but how can I do things differently? What did I never think I could do in my business that now I have the capacity to consider? And one of the CEOs we talked to said, we thought we were going to be thinking about costs, and now we're thinking about revenue. And I said, well, tell me more of what that means. He said, we thought we were a business that did this, and now we know we can be a business that does this and we can bring our design. We don't have to go and get designs to do our machine tooling. We can build our designs because we can use model-assisted work, and we can change how we think about agriculture to make it have better yields and less disease, and importantly, bring prices down for consumers should all of that come through. So it's one of these things where I see people tackling the technology that they thought they needed to learn to keep up, and now they're thinking about how do I use it to grow.

You know, the Federal Reserve is taking a system approach because we are part of a Federal Reserve system. And so we really worked hard all the way back when ChatGPT first came out. The first thing we have to do is say, you're not allowed to use these technologies to do your work because we're in a confidential environment and we have ring fences. I know all the private sector companies started thinking about that, but then we quickly realized that it's an important tool. For the tool time I've worked at the Fed, which is longer than I now talk about anymore, but really my whole career it's really been about being not the first out there taking risks, but the next out there being thoughtful about how technologies could be used elsewhere because we have to study the economy, we have to know how financial institutions could use them. But we're also really committed to being good fiduciary stewards of public funds. So we want to be using technologies that can make work more efficient, more effective, more resilient. But we're recognizing constantly, even though you're right, we are regulated, but we are also self-regulating in this way. We are fiduciary stewards of public funds, but also fiduciary stewards of public trust. So we're always balancing how do we continue to modernize while we're making sure we're safe and we're sound and we're doing our work well with a human in the loop. That's what we always think about. Humans got to be in the loop.

You know, well, we've been on this journey for a while. We, all the way back, starting in 2023 and thinking about this in San Francisco, we've got lots of ambassadors for this. But on any growth curve or maturity curve, there's the early adopters, the enthusiasts. I'm probably one of those. I'm using all the models I can get on my private device thinking about how I can do practice, you know, can I run my equations that I've written down much faster? Can I estimate models better, can I code faster, all of this type of thing. But the important thing is we quickly built a sandbox in the Fed where we can do these things safely and people can practice well. That means that it's not just the leaders who are using it and saying you must use it or you should use it. It's actually ground up as well. And so there's the enthusiasts, the early adopters, and then you bring the others along. And I would say that we have, and I feel proud of this at the system level, we have taken this as it's the individual's responsibility to invest in him or herself, but it's also the organization's responsibility to make sure we're training the workplace that's ready for tomorrow, not just the workplace that's ready for yesterday. And so thinking about how we enterprise train people on these tools in a safe and sound way, making good judgment. It's also, you don't need generative AI for every single task. We've long been using machine learning and robotic process automation. You could use older technologies like those which seemed new and cool before. You can use those older technologies to do your work well. And so I think it's just a matter of helping people in our organization. We've got widespread adoption at this point, helping people understand that when you get down to it, it's not just about learning a tool, it's about seeing how it can benefit you. And one of the ways we think that we can change business processes is have people who are actually in the work innovate within that space, but do it in a safe and sound way by having a group of people take a look at that and say, do we want to take this to scale across the enterprise, or is this just a one-off thing? That was nice to experiment, but it's not going to give ROI.

Well, the very first thing to know is that there's this famous phrase that productivity growth is everywhere except in the data. The data itself, at the aggregate level that we collect. That's a true thing that was said by Robert Solow. It's like this is important because it always happens where the productivity growth is starting to take place, the productivity gains. But in order for that to get to the aggregate level, it has to be across a wide group of individuals and firms. So we are seeing evidence in particular firms in particular sectors where you're already seeing those gains. But we haven't seen it at scale yet, in part because if you think back to—I'm going to use any sort of historical example, and they're not perfect guides, but I think it's useful. Think of electrification. We had electricity for a long time before we got the rapid productivity growth that came from electrification. The change wasn't that we knew how to use electricity. We had that. The change was that instead of just putting the electric motor at the end of a factory line that was once powered by steam, where you saved costs for energy, but the line looks identical, the unit drive came in, and they could make machines with specific motors for themselves, and they could rearrange the factory anytime they wanted. And then transformative things happen. What's the key? It's business process change that generates sustained productivity gains. And firms are just at the early stages of interrogating, learning the technology, using the technology and then thinking, how do I change my business so it doesn't actually look the way it once did? For that, you need workforce. You need investment. You need learning. And then, of course, as Daniela mentioned, the models are changing so quickly, the capabilities that most of the firms we talked to, you don't want to just take one thing and say, that's our new, because they know six months, one week from now, things could be totally different.

So we haven't seen those particular things. Certainly that's a concern if you go to particular places where data centers are being built, you can get information about how it's harder to get construction workers because they're all being moved there, or there's a demand for construction workers that might outstrip. But for the construction workers, it's a boom, right? This is a good thing. And of course, we need more pipefitters and welders and all the things that we thought for a long time, we didn't need. Now we know we do need them. And community colleges in particular are training a lot more people to do those types of jobs. But we haven't seen that drive the inflation numbers that everyone's worried about. And the number one concern when we talk to people in communities, you can see it in the Gallup surveys, etc., is inflation. But inflation is really being driven by just the ongoing—we're trying to get inflation down to 2%, of course. But then we have the tariffs and those are rolling through and hopefully rolling off by the end of the year. But then we have the oil prices which are pushing up overall energy costs and of course food prices as well. So those are the things people seem much more worried about than the data centers. It certainly has caused some supply gaps between the things that power electric plants, the chips, etc. but I don't think that's really the main driver in any way of inflation right now.

You know, I don't think of it as affecting our mandate. Our mandates are given to us by Congress. And so let me just say, they're full employment and price stability, and those are always affected by how the economy grows, how fast you can grow, what the underlying aspects of it. And so what many people are wondering right now in that particular mandate, AI goals is in the next year, will AI itself affect our specific decisions that we're making as we navigate this? And I say you always are thinking about what the potential of the economy is. But we also, if you're a policymaker, you have to think about what's really happening today and there is oil prices and other things. And then there could be this idea that maybe the job market won't be as robust as it's been in the past, because AI will do so much. So far, we have seen mostly generative AI being used to augment workforce rather than replace workforce. Now, it is absolutely true that if you're going to take—if coding is easier, you won't need as many coders. But we're hearing from firms that use coders as they're hiring new coders, new types of coders. So net hiring is going up or hasn't fallen very much. It's just the people, the skills you need are different. And so the responsibility for all of us is to think, how do we provide training for people so that we—and it's not going to be like any one component. You're going to have to have individuals saying, I want to invest in myself and learn these new skills. Companies saying, well, let's—how do we mobilize our workforce by providing training? And the public sector asking, what do we need for our nation? And I think those three things together—individual, private and public together—thinking about what's our journey forward as we make sure this workforce is ready and that anyone who used to do this job and now has to do these kinds of things, can prepare themselves to do those.

Yeah. I think, you know, one of the things that's true about technologies is when you invest in them, the investment comes first, right? You have to invest in the capital, the infrastructure, the technology itself. And then you invest in the people so that they're ready. And investments often compete for resources. That is the early part. Then the next stage is where do you start to get the gains. But if it is as transformative as some of the other technologies have been that we call transformative in our history, then you absolutely start seeing productivity gains. And that means the pie grows faster and things can fall in value. And falling prices and inflation—it can be deflationary. And the timing is really what matters. So you talked about the 5 and 10 year landscape. We're talking about the 12 month landscape. When we're thinking about policy decisions we have to think about today. And then we're thinking about the 5 and 10 of where we're heading. And if you go back to the periods in history where the Fed has had to grapple with these technological changes, I started at the Fed in the mid-90s when we were grappling with how much would computerization, the internet, create opportunities to hold prices back. And so those are the kinds of things that we are doing right now. But I don't think that's the pressing issue today. Today, you know, inflation's above target for different reasons. But going forward we definitely have to think about this in the news.

You know, we certainly—the Federal Reserve System, you can see that we released twice a year our quantitative surveillance kind of summary. But essentially we're always watching financial stability issues. But we are asking these questions and this is a question I ask: what is the value of the work being done behind that? And I think it's really hard to trace back and say that these technologies that people are investing in aren't valuable. We see them in your personal life. I'm sure all of you, if you're in a tech conference, are using AI for your personal life and your business, etc. just thinking about it. And the more you use it and the rate of change you see it making, the more you recognize there's a possibility. So I think there's some there there. Will there—is there potentially some—there's two words I don't use, I'm not saying them, but one of them is that word. Whether it starts with an H. But seriously we don't talk about these things. Now, in all seriousness, I think that it's, you know, let's think about the conversations we're having. And a lot of times, you hear conversations about AI is going to save everything. Nothing else will ever be bad again because we have AI. And on the other side of it, you hear AI is going to destroy everything. You also hear that, oh my gosh, it's a B word. It could cause this where nobody will work. And I just would caution all of us from staying at these extremes because the hard work, the work that is going to determine what our future looks like with this technology is all in the middle. It is nowhere in our extremes. It's about thinking, okay, we use this and these people don't do these jobs anymore. What do they do? And if you look back on history and you wonder why did technologies have a negative outcome on society when over time they had a positive outcome—electrification—it's often because the people who were being displaced didn't have an opportunity yet to grab another rung. And I think it's really incumbent on all of us to think, how do you enable opportunities so people can grab another rung? And how do you harness that technology so it can do work that we haven't been able to do or thinking we haven't been able to do? And every week you open the news and you see another thing that AI has done for us that has made our lives potentially better. So let's not lose sight of that while we also don't lose sight of what has to be done by all of us in order to make sure this is a positive rather than a negative experience.