About this transcript: This is a full AI-generated transcript of Federal Reserve Chair Kevin Warsh Hosts A Panel On SI At Fed Task Force Session In California from Forbes Breaking News, published November 17, 2026. The transcript contains 34,663 words with timestamps and was generated using Whisper AI.
"This feels faster, more transformative, and less familiar than anything I've seen in my lived experience. And probably the lived experience of most of you. Maybe it's like electrification. Maybe it's like the industrial revolution. Maybe it's much, much better. The one thing we know for sure is..."
[00:00:00] Speaker 1: This feels faster, more transformative, and less familiar than anything I've seen in my lived experience. And probably the lived experience of most of you. Maybe it's like electrification. Maybe it's like the industrial revolution. Maybe it's much, much better. The one thing we know for sure is it's definitely different. So for policy makers, these possibilities, these promises, they're important, but they're only part of our equation. We need to know what will happen, not just in the future, not in the thing we can all agree with will be different. But only across the timeline, throughout the change, throughout the transformation. We need to know what happens to growth, productivity, inflation, and the labor market. And this is where insights from the task force and from this conference and the dialogue we have here are so critical. They help us understand and distinguish between promises and reality. So today, we'll hear from technologists, academics, businesses, and other leading thinkers talking about what AI could be, what it is now, and how it's likely to evolve over the coming years. I personally expect to have a very lively conversation. You have the liveliest conversations when no one's really sure and everybody's evaluating and investigating together. So I'd like to get started. And here I have literally the distinct pleasure and privilege and honor of welcoming Kevin Warsh, chairman of the Federal Reserve, force behind this task force and the conference to the stage. Chairman Warsh has long been an important voice in the economic possibilities that AI creates, and he's brought that same thinking front and center to the Federal Reserve in his role as chairman. Chairman Warsh will be providing some opening remarks to set the stage for our day. Mr. Chairman, Kevin, the podium is yours and welcome to the San Francisco Fed.
[00:02:12] Kevin Warsh: Thank you all very much, Mary, thank you for the most kind introduction. I'm grateful to her, I'm grateful to the staff at the San Francisco Fed and the board staff for pulling this off on relatively short notice. Now, this is my first visit to a reserve bank as chairman. I wouldn't call it a snap inspection, but I'm certainly impressed by what I've seen in the first couple of hours. What I really see are high expectations, from Mary, from her colleagues, a reform orientation, a view that we need a forward-looking inquiring spirit because the world is changing every bit as fast as Mary said at the Outlook. It's also great to be joined by plenty of my Fed colleagues from the board and from around the system. As you guys know, central banking, well, it's a team sport, and I'm pleased that you're all here. This conference is an expression of the mission we all have at the Fed. One Fed, fit for purpose, focused on the future. I also want to thank the task force leaders that are here, Mark Andreessen, Chad Jones, Asha Sharma, for leading this task force. You're giving the scarcest of resources, your time, and your commitment. It's a service to the Fed and our nation, and I am eternally grateful. The task force is one of the five I've convened as chairman. These groups are a way of broadening our thinking, of inviting robust discussions that take us far deeper, far more involved than fleeting data points or daily market moves. Our task forces are an intellectual investment to better equip us for the next policy conjuncture and beyond. We need to get this right. The common attribute of those leaders I mentioned, Asha, Mark, Chad, the reason I asked them to lead is not just their exceptional insight and experience, it's their battlefield bravery in the world of ideas, their intellect, their willingness to think hard, and at least as important, think anew. They need to help us think about what might happen next. They're the ones who asked me to open up the task force discussions to all of you, to broaden the discussion among academics, businesses, leaders of all sorts, because none of us have a monopoly on the truth. As for me, I feel particularly engaged to be back here in California. I arrived here in this part of the world in 1988 as a Stanford undergrad. I remember sitting in Green Library over Thanksgiving. Didn't go back home. I did a little bit of extra work my freshman year, reading and wrestling with the great works from a canon that I had scarcely known existed. That was, in my own little way, an initiation to the world of ideas. It was a special time to be in the Bay Area, neither boom nor bust, a quieter period between tech surges. And today, just as then, I'm keen to listen and learn from the greatest minds and the people that are thinking anew about this new era. The current era may be the most transformational period in my adult lifetime, marked by the arrival of this new general-purpose technology, SI. It's an entirely new force in economic life and society, and it demands our special attention. Even a few years ago, the idea that we could find in our pockets instant access to genius-level intelligence on almost any subject would have seemed far-fetched. Today, with the advances of SI, it is happening. The marginal cost of satisfying limitless curiosity on questions large and small is falling fast, perhaps even trending towards zero, a subject that financial markets are trying to figure out every day. So now we're going to turn to the wonkier part of the discussion, but frankly, the more consequential. In the tradition of endogenous growth theory and economics, ideas are the ultimate resource. If living standards are to advance, conditions must be favorable to the generation of new ideas. It should be no surprise, then, that the United States is leading. Societies that bring forth the best ideas, well, they tend to prosper the most. The implications for economic policy can hardly be overstated. While an idea can be costly to discover, it's free to use. And because an idea is non-rivaled, that is, one person's use does nothing to diminish that of anyone else. Its value grows with the number of people who use it. The Nobel laureate, Paul Romer, among others, including my friend Chad, has shown that a doubling of ideas and other key inputs can more than double output. The key word there is more, allowing increasing returns to scale. Even so, in the U.S., over the last several decades, breakthrough ideas had been harder to find. You'll recall at conferences like this, this is when there would be discussions about secular stagnation. By some efforts, research efforts have increased more than 20 times since the 1930s. Yet, Nick Bloom, another friend who's here, Chad Jones and other co-authors, have pointed out that research productivity has fallen by an even larger amount. The result? Productivity growth has fluctuated. And while in recent quarters, it has been reasonably strong, it hasn't yet broken out definitively, structurally, reliably to the upside. At least not yet. Long-run growth, I should note, is also a function of the rate of increase in the number of people producing these ideas, setting them in motion. The conventional view holds that as population growth slows, there may be fewer new ideas to drive prosperity. The increasing challenge of generating breakthrough ideas, combined with slower population growth, helps explain why many in the economics profession, especially in the post-crisis environment, return to the old theme from the 1930s of secular stagnation. So the question for all of us, the one that I want to explore with our colleagues today, is does superintelligence dramatically change this picture? Well, there's certainly some indications that it does. SI might be somewhat different from general-purpose technologies, the kind that Mary referenced in the outset, the internal combustion engine, or electricity, or the transistor. Those inventions made for decades a follow-on innovation, so I don't mean to dismiss them. But they did not themselves automate the creation of new ideas. What might set SI apart, as an economic influence, is that it could supercharge idea generation. It appears to be changing the method, and at least as important, the speed of innovation. If idea generation broadens, growth could be sustained at a level substantially higher than the long-run average. But for all of that, however, let me make one essential point. Humans should not be underestimated. There is no replacing the minds and hands of men and women. Machines that can amass or even exceed human knowledge, well, they're still machines. They do not take the place of human judgment, morality, dignity, or responsibility. Humans are the designers and the builders, and humans must be the deciders. So, for our country in this era, consider how much is on the line. Consider the reason why, five and a half months ago, we put this task force in the front of the line. I remember Bob Mendel, another Nobel recipient, saying years ago that the most important fact of the 20th century was the rise of the United States as a superpower. For us today, that's more than a fact of history. It's our charge to keep. And so we must ask, can you be a superpower without superintelligence? My Fed colleagues and I are not at this conference as futurists to speculate on each and every turn in technology, and we're certainly not here to make predictions. We can't know the future, and as you've heard from me before in other settings, we ought not to pretend otherwise. Our job is to prepare for whatever comes next. That means contingency planning, a subject that I will return to in other task forces. Contingency planning becomes only more important when the world is changing at this fast speed. So our aim today is a humble one, to ask the right questions. As policymakers, we have to be mindful that great opportunities rarely come risk-free. And the risks, in this case, if overlooked, can be highly disruptive and economically quite costly. SI empire empowers all, including bad actors. So we're thinking especially of critical infrastructures and their vulnerabilities. And of course, I think of economic policy in this new era of SI. At the Fed, our role is neither as advocates or opponents. Our responsibility is to recognize that the widespread adoption of this technology will have a very significant effect on the economy, with hugely important implications for our day jobs on monetary policy. So let me say a few words about some of the big questions. They're mostly about magnitude, scope, and timing of this development. First, will this new technology deliver a rise in total factor productivity? Will the democratization of specialized knowledge significantly raise standards of living? When will this surge start, if it hasn't already? And when will it taper? Will it happen sooner or faster here in the U.S. than in other countries? And will adversaries race to deploy SI in whatever its form happens to be, intending to do harm to our economy? These answers will matter greatly. And that is why we're thinking about them today in a sober, serious way. And I expect more clarity with the passage of time. Still, we've got a job to do in monetary policy, so it's worth considering the range of economic outcomes we could see over the next several years, depending on how SI progresses. Board staff has begun to evaluate various contingencies. There's that word again. Applying stylized model-based frameworks to generate scenarios. They're illustrative scenarios. They're not forecast. They're not predictions. Moreover, many other outcomes, probably including the one that will actually happen, could be drawn. Consider them a starting point for the very useful discussions we've got lined up today. First, at one end, SI leads to an increase in productivity and economic gains that are shared broadly across labor and the owners of capital. In that event, real GDP grows persistently above its long-term average, and the U.S. stays close to full employment. Other outcomes, however, could be less positive. An SI-driven expansion could bring substantial labor market disruption, as some jobs are displaced rather than complemented by technology. GDP growth would still rise above its long-term average in this contingency, but widespread automation combined with labor market frictions could push the jobless rate above levels consistent with full employment. Let me reference a third. Or, SI could disappoint. After significant capital investment in stock market gains, expected returns could turn down. Equity valuations could fall, leading to an economic slowdown and a more notable rise in unemployment. In time, the economy would recover, but without the productivity gains. I stress that while I'm not here to offer predictions, I'm not exactly impartial about American leadership. In a time so filled with both challenge and possibility. I'll end where I began, thinking back to my first visits out to this part of the world, to my days at the Hoover Institution, and to a collaboration with my fine mentor and colleague, John Kogan. As we wrote some years ago, without the knowledge that SI might be upon us, but I must say as optimists and believers in the U.S., and I'll quote a paper that he and I wrote together. If we meet the challenges before us, we said, America has a bright future with strong, durable economic growth, abundant economic opportunities, vibrant and secure communities, and a healthy environment. Future generations will inherit a dynamic economy capable of delivering technological advances that will extend lives and promote human flourishing. That's the goal. That's the ultimate scenario. That's what all of us here hope to see. I can just end with a final note. It is a privilege, privilege of a lifetime, to be a policymaker here at a time of such consequence. And it's the most encouraging sign to be joined by some of the most serious, sober, responsible thinkers, great men and women who are willing to join the Fed, even as an aside, to try to put our SI goals on the right path. Thank you all very much. Let me turn now to our first panel, and I'll ask my panelists to join me up here on stage. Thank you. Thank you. So let's turn to this first panel. It's an honor to be joined by these three gentlemen. I've known them all for decades. Let me do, let me introduce this like we would a proper, beautiful old Russian novel. Introduce the characters, and then we'll see where the plot goes. I have the wrong one in front of me. Hold on a second. That's the wrong characters I was going to introduce. So this panel is called At the Frontier. The idea here is to ask questions from people that are users of the technology, who are some of the most innovative thinkers that I know. We didn't call these guys here because they're at the labs and they're designing this great stuff. We called them here because we wanted to know what's really happening inside of companies. That they happen to run technology companies might be an accident of history. Is that because these technology companies themselves are more innovative, are willing to cast aside how they've done business, or is it just three guys that happen to work down the street? Let me do the brief introductions. Nikesh is immediately to my left. He's the chairman and CEO of Palo Alto Networks, which he's led since 2018. Before that, he was the president at SoftBank, where I came to know him. And even before that, he was at Google, where he was the chief business officer. Next to him, Patrick Collison. I've known him and his brothers since they came to these shores as young men from overseas, poor immigrants, and they've done pretty well in the United States. Millions of businesses use Stripe, the business that they started and founded, to accept payments and payouts, and manage other complex online enterprises. Max Levchin, who's at the end, has been a friend also for, I don't know, 20 years. He also came to the United States and is one of the smartest people that I know. He's the founder and CEO of a firm, the co-founder and chairman of other companies. And he created an innovation lab called HVF, which stands for Hard, Valuable, and Fun. I don't know if I was supposed to disclose that. Max was the chief technology officer at PayPal back when we were all much younger than this, and before at least some of us had gotten mixed up in public policy. And he was the chairman of Yelp, not bad for someone new to our shores. So those are the three people I want to sort of probe on a few areas of questions. One is adoption. What's really happening? And I want to try to understand if what's happening is so special and different in their companies, what about the rest? What about the other 450 companies in the S&P? The other 2,000 companies in the Russell 2000? Because I want to know whether this is just a technology that's at the frontier and going to stay there, or is going to find its way more broadly. When I start at the end, Max, with you, tell us about how your business is different, how you're using this new technology, and exactly what you think that means, not just for your business, but for the broader economy.
[00:20:15] Speaker ?: Sure.
[00:20:15] Speaker 3: Thank you. Can everyone hear me? Yes. Thank you, chairman, for inviting me to participate. So a firm is a lender, a fairly sizable one, and everything we do is done by machines. And so our underwriting models are entirely ML, superintelligence-based, et cetera, et cetera. And so we've been an early adopter of the earliest models. We've been using them primarily to increase our engineering productivity, which for scale is about five times more productive. So the cost per feature is one-fifth, or what it was. We expect it to be one-tenth. So these are pretty extraordinary gains. It's a great story. That's sort of what everybody will tell you. So none of it is new. So to contextualize what that means, as a lender, our job is to sort risk. So people come in. They ask to borrow money because they're trying to buy something. We need to pick the ones that are most likely to pay us back on time every time versus the ones that don't. And it's all done by machines, but humans are the ones deciding what makes an interesting predictor. So the fancy term is feature. The kind of colloquial term is a signal. As you try to decide what might be a great signal that predicts your ability to repay, the ideas greatly outstrip the available time to train and the engineers to put those things into motion. So there's lots and lots of really good ideas that end up on a cutting room floor. So what does this efficiency or this productivity gain really do for us? We're able to pick up a tremendous number of these ideas off the floor and put them into production. So in a kind of almost too meta a story, we used LLMs to design our most recent model, which in itself is actually built on top of a transformer model. So not exactly an LLM, but a token representation of repayment events that allowed us to do almost twice as well as the next non-LLM model would have done. So we ran the two side by side and we saw an 80% increase over the baseline, which again for context translates to about two more billion dollars per year of loans for us. And so this is a direct consequence of just being able to free up engineering talent, all human, zero layoffs due to AI at a firm, and just deploy them for what you might think of mass customization, which I think is what's going to happen. Maybe I'm too Pollyannish, but as a techno-utopian, I'm prone to love our machine overlords. What I think will happen is we will see more and more of these long tail ideas that no one would ever have the time or the productivity or the staff to lift off the floor come to fore. Some of them will fail, but that's what we're supposed to see. But we get to find out if the thing that you need, and only you, which now can be produced at a much lower cost, is going to be vitally created. And you might in fact find that there is the intersection between supply and demand that makes it work.
[00:23:16] Kevin Warsh: So Patrick, let me turn to you. I presume you're going to also tell us a story that Stripe is at this frontier and you're making great progress. We want to hear about that. We want to hear also from you about how long will the delay be between those at the frontier and the rest? What do you think is holding down? What's the bottleneck? When will the rest see it? And what happens if they don't?
[00:23:43] Speaker 4: Well, again, great to be here. When you said it's like a Russian novel or something, I thought it was because we're all unhappy in our own ways as CEOs. But so, yeah, like Max, we've seen a very significant acceleration. And it's been changing quarter by quarter where 12 months ago, maybe the narrow work of engineers was being somewhat accelerated in code production. Whereas today at Stripe, and I'm sure at both Paleta Networks and at Affirm, it's more about these longer horizon loops where the agents are identifying tasks autonomously themselves. They're going and actually implementing the feature or fixing the bug or whatever end to end and deploying it to production with substantially reduced human oversight nonetheless, but significantly reduced human oversight as compared to, say, six months ago. And in terms of what that means for Stripe, the organization, I mean, it's interesting. There's always kind of this question of, you know, to what extent AI is kind of a substitute versus a complement. At least in the near term, we're seeing substantially reduced hiring for our GNA functions as compared to some, you know, prior counterfactual. And, you know, GNA as a share of our revenue is definitely diminishing quickly. But for engineering, that is kind of R&D, you know, our kind of product production, and for sales, thus far we are seeing acceleration because, you know, effectively, as Max said, the cost of feature production is diminishing. So, you know, in the face of that, there's a sale on for feature production. And in the face of the sale, we want to buy much more than we did previously. And similarly, you know, our sellers are much more productive than they were previously. And so, therefore, on an, you know, the ROI math is quite straightforward, right? It's mathematically, arithmetically, makes sense to hire more of them. In terms of what we're seeing across the economy generally, you know, I remember early in Stripe's history, the CEO of a very large mega cap company came to visit. And, you know, we were all, we were using all these new SaaS collaboration tools, and everyone in the company could see what was going on and all the things. And, you know, he was shocked, I mean, the Stripe was a 30-person company at this juncture. He was shocked that we were, you know, that we would operate this way. And, you know, here's the governance and the SOC2 and, you know, all the kind of enterprise considerations that, you know, that looms large for him. And I'm now sensing some of this same disparity and kind of mental model impedance mismatch between incumbent organizations and new companies, where large companies, and I put Stripe to some extent on the incumbent side of this, we have all the same questions that you have in terms of, well, where's the human supervision and where's the governance and what's the certification and the compliance and, you know, all the things that a sort of prudentially oriented infrastructure company, I think, is wise to consider. New startups are new startups. And they're happy, you know, we would not autonomously plug the agents into sort of just, you know, feature development loops and, you know, dust off our hands and walk away and say, hey, you guys do whatever you want to do. The new startups are happy to do this. And the 2026 cohort of companies started on Stripe is massive, so new companies, is massively outperforming any prior year's cohort by, you know, a dramatic magnitude. The number of new businesses being launched on Stripe every week is up approximately twofold year over year. And, you know, we've had more than one decade-long time series here. You know, Stripe has been in operation for around 15 years. And so a 2x year-over-year increase is not something with historical precedent, even, for example, during the pandemic. And so I guess I'm reminded of, you know, the, well, many people here in this room can correct me. There's the Paul David 1990 paper, I think, about how with electricity, by and large, the benefits accrued to new firms rather than existing incumbent firms adapting. And, you know, obviously, existing firms are trying to adapt. And, you know, there's certain, you know, AI is incredibly salient for CEOs across the industry. But I will say that new firms that we observe and whose business performance we directly see on Stripe are proceeding at an unprecedented pace. And I'm very curious to see where it goes.
[00:28:07] Kevin Warsh: We're going to return to the Paul David product. Did I get that right? You did 1989, not 1990, productivity paradox paper with The Economist. I wasn't born then. I'm going to keep you focused. So, Nikesh, you run a big company. You've got a history of big companies. I know you have one foot in the venture world and startup world. Tell me two things. One, about how much is this new technology changing your own business at the frontier. But when I met you, what you were is the best salesman I ever knew. You would sell products to enterprises that weren't sure they needed us. See if you can tell us. See if you can tell us.
[00:28:50] Speaker 5: Is that a compliment? It is. I'm not quite sure where that was going.
[00:28:53] Kevin Warsh: They bought them. And Google seems like it's done okay. But what is the reason why there would be a big distinguishing feature between the frontier and the rest? What's the reason why, if we've got this hypothesis right, which is the Fortune 450 aren't coming along so quickly, what's the reason? Have they not changed their organization, their mental model, their way of thinking, their process, their governance? Help us understand on the diffusion question, not just your own internal workings.
[00:29:24] Speaker 5: Got it. Well, first of all, thank you for having me here. I think it's fascinating. We seem to be focused on talking about AI from an enterprise and enterprise productivity lens. If you look historically, since, you know, there was a time about 25 years ago, technology started getting built for consumers and not for enterprises. Historically, we had built technology for enterprises, mainframe, et cetera. When mobility came about and when laptops came about, we started building for the consumer. And I want to go back to something you said. The idea that we will carry genius in our pocket and have access to intelligence at a marginal cost of zero is extremely compelling. And I think that's the optimistic view we have to start with because there's 6 billion people in the world who do not have access to expertise or people who are their physiotherapists and nutritionists. All that intelligence is going to be carried in the back pocket. So I think that's going to be hugely amazing for society and humanity at some level. Now, if you transpose that and say, why are we so obsessed with the enterprise version of this new technology? It's because it's expensive. We need to find a lot of money to go build these large data centers required to train these models to get really good intelligence. That's why the focus is inadvertently on how much money can we make out of this technology and back to your point about how can we get enterprises to adopt it quickly and embrace it. I think there are certain things where this technology is really good at. Processing large amounts of data, retaining large amounts of information. If you have 3,000 people doing customer support, you can have one brain that understands all these interactions and it can be the complement of the human being that's sitting at the edge solving the problem of the customer. So yes, it already has immense applicability in terms of how enterprises operate. But what you're noticing is, as Patrick said, it's really good for repetitive tasks where a lot of analytics are involved, a lot of understanding is involved, but it's still early to give it arms and legs. What I mean by that is giving it agency or giving it independent authority to go act on our behalf. And I think that's where the fear is because that's where the job displacement begins to happen when you can say, oh, I don't need you. I will take care of it. Now, as you mentioned, I worked at Google and not just in sales. I was there in 2009 when I sat in the first driverless car. It was a Lexus with many, and I still wonder what the hell was I doing, thinking of letting it drive me in 2009 or in 2026, which is now called the Waymo. Just think about it for a second. It's taken 17 years since then to take something and replace one human job, one human job. It's cost more than $30 billion. It's not the mainstay that we worry about. It's the edge cases. I work in cybersecurity. Our products are all about the edge case. It's about that one time, somebody will come, find a vulnerability, get into your organization, and exfiltrate data. So it's not the 80% use case we worry about. It's the other 20% that we have to worry about. I don't think this technology is fully there yet in terms of being able to understand edge cases, being able to plan for those. That is why we still need that human judgment, the human in the loop. I know there are certain companies which are out there to the races because they found use cases where we need this superior intelligence. I think there are use cases, and you'll be able to find distinctions. Cybersecurity is one use case. Superior intelligence allows bad actors to go find vulnerabilities and come attack companies because that requires real smarts and lots of compute. I'm pretty sure it's going to work in cancer research. I'm pretty sure it's going to work in hard science problems, which we need to sort of further humanity. But I suspect the benefit of this technology will be over time. It will become the new normal. We will all have access to amazing amounts of intelligence. And I'll end with this. In 2004, when I was at Google, we used to say, search democratized information. Everybody had access to the same information wherever they wanted. I think AI has a power of democratizing intelligence.
[00:33:21] Kevin Warsh: So let me start back with you. We're going to go down the line. I'm going to have a couple of individual questions. So, Nikesh, when I see you, I think cybersecurity, your business is focused on that. How worried should we all be? Well, we're relatively early in applying these technologies, both you on the business and revenue side, for many of us for protection. How worried are you? What are the vulnerabilities to payment systems, infrastructures around the world, and what, if anything, can we do about it?
[00:33:53] Speaker 5: I tend not to worry, but it could be hard otherwise. But this is a harder time than normal because what we've done is this capability that AI has, has the ability to find vulnerabilities and bad code and bad things that were written by humans for the last 50 years. And the problem is, we, cybersecurity is an asymmetric battle. The bad actors would be right once, we have to buy it 100% of the time. It's very hard when you bring this technology to bear where things which used to take days now can be done with pretty much, you know, what we've noticed in the last six months is there are no people available in the industry who you can call and say, can you help me? I'm having a problem. Literally. Everybody's backed up right now because every company is going through some version of bad actors trying to get into their infrastructure and understanding all vulnerabilities. So the technology which came out six months ago called Mythos has now become more commonplace and is available to most bad actors in open source models. Unfortunately, it's not that the hackers have become smarter. The technology has become so good even bad hackers have become okay. So that's what we're dealing with. So right now, I think we're at a bit of an asymmetric situation. We have a disbalance in terms of what we need to get done. There are things which are getting exposed. We will see more cyber events. I think you already see there are more cyber events happening every week than we've seen in the past. I don't think this is going to stop. I think it's going to persist. I think it's particularly important to make sure that interconnected systems understand who the weakest link is. And I think the banking system is one such place where we have to be very careful because lots of people are connected to a lot of different people. And the weakest link is going to determine how safe our system is over time. I think the next two, three years is going to be a bit of a technology rebuild moment. We're going to have to go back and see every mistake, every thing we kind of did not pay attention to how to rebuild. Now, the good news is AI or SI has a possibility of not just finding vulnerabilities. It also has a possibility of allowing us to refactor code and rewrite code and do a bunch of that stuff. So I think if we take the positive outlook, we have to go spend more time investing in upgrading our technical infrastructure. We saw, for a totally different reason, the airlines were down this past week. We saw Microsoft had an outreach. So you are seeing examples of how, if you don't pay attention to upgrading old technology, we're going to end up in a slightly worse position than we are.
[00:36:17] Kevin Warsh: Patrick, since I met you and your brother, you always majored in technology but you minored in economics. So let me give you a question that straddles in between. The question for financial markets and maybe for some people in this room is, what's the market structure on the other side of this? You were once a little startup company and by your own description, you sort of feel like a legacy incumbent after a decade or more. Who gets to make the money? Will the money be made by the big labs? Will the money be ultimately made by the users of these technologies which are those things that could drive that productivity which would be the biggest fundamental change in economics in generations? So what's your sense of what the world looks like in market structure terms?
[00:37:02] Speaker 4: I mean, I think that, well, actually first, let me just echo the last part of what you said and maybe not the first part which is you said that the next couple of years is going to be a technology rebuild moment. I think that's exactly right. It's what we're seeing across the industry and even the most scrutinized parts of technology. Like the Linux TCP stack, there was tremendous attention devoted to making sure that it was as secure and as bug-free as possible. A huge incentive for people all around the world to find every little minute error or exploitable codicil. We thought it was pretty good. like, you know, the Linux TCP stack is about as scrutinized as software gets and, you know, turns out it is riven, absolutely festooned with bugs and problems and exploitable issues. And so I agree with that part of what you said and you're, I mean, this is maybe why you're more successful and just, hey, but I'm worried. I'm worried. I think it's going to be... We make more money
[00:38:05] Speaker 5: when people are worried, so... Yeah, so...
[00:38:07] Speaker 4: I could be thinking my... I think it's going to be a mad scramble. And so, that's my Irish sensibility here. But on your question about the market structure, look, I think this... I think this reduces to the question of what does the... As in the market structure question, reduces the question of just, well, what's the world configuration? You know, 10 years hence. Maybe we can make near-term projections about, you know, to whom different surpluses might accrue and so forth, you know, a year out. But, you know, many of you have, I assume, read the essay Machines of Loving Grace. No? Yes? Okay, seeing some nods. Right, so this is an essay published by Dario Amode from Anthropic. And kind of laying out the positive vision for where AI goes. And it is a reference to a poem from Richard Braudigan in 1967. And it's a very short poem. You can read it. And the last verse of the poem says, you know, he says that he dreams of a cybernetic ecology where we are freed of our labors. So, at least in Braudigan's conception, you know, this is not just a compliment to help us, you know, be more productive. In his vision, we're freed of our labors and returned to nature with our mammal brothers and sisters. And all watched over by Machines of Loving Grace. And this is a vision, Braudigan's vision, for where things could go. Elon Musk has said that, you know, humans are a biological bootloader for superintelligence. And I guess I bring these up to say that, you know, there's some pretty far-out stuff being posited as, you know, potential destinations or at least, you know, intermediate points in terms of where these trajectories point. And so, given all of that, I find it pretty hard to, you know, then, you know, work backwards. Well, what's the market structure of the cybernetic ecology? Look, I think today and for the foreseeable future, the vast majority of the surplus is being captured and will continue to be captured by the users of these technologies. And there is immense productivity gain and new creative possibilities being unfurled. But, you know, what the general equilibrium looks like, I will defer to people here.
[00:40:42] Kevin Warsh: Max, I won't ask you to define what festoon means, but maybe in the old country, it has particular meaning.
[00:40:49] Speaker 3: My socks are festooned with my logos.
[00:40:51] Kevin Warsh: Max, let me try this with you. I've been trying to frame this enterprises at the frontier versus the rest. I might have gotten this from you some time ago. Is SI really driven by individuals and not businesses? Are there people resident inside of your company where there doesn't have to be a corporate strategy on SI? They're just playing and excited by these new tools. So is the transformation less about is the company small or large? Who are the individuals that are practicing these new ideas?
[00:41:25] Speaker 3: So I think, like most profound innovations, it begins with individuals and it's up to us as leaders and frankly, all of you as leaders to institutionalize the gains. I think the, in the industry, for a little while, there's been this narrative of there's about 500 people in the world that are good enough to think up the next advance in LLM structures and sort of the math that comes with it and haven't heard that one in a little while because I think that number is expanding and I think today I can name the handful of geniuses on my MLAI team and I know what they're working on. I'm excited to sit with them because they teach me things that I just had no idea were even possible. I think this is true for most companies and all of us are hanging on to our researchers unless the frontier labs come and get them and overpay them but over time like everything else our job is to take kind of the patterns of genius that we're seeing and institutionalize them and I think that's a, that's reasonable and a good thing and there's some degree and I'm reaching into places like national security and kind of a short term concerns around leakage of IP where we must know who they are and really care about where they go and what they work on I think over time this becomes diffused and maybe we care a little bit less and since we're all on the topic of security predictions I'll stick with my, I am the optimist apparently on this panel or maybe the, the most optimist anyway I actually think that we're going to go through a period of upheaval for sure there's quite a lot of discoveries of these latent bugs that are being unearthed very quickly by sort of the likes of mythos but I think that motion has always been there the entire security industry is all about lots of very brainy people going wait a second that's a buffer overflow if you don't know what that is I encourage you to look it up it's one of the more fun parts of computer science and then you're like oh here's there's an if statement missing we'll fix that right now and what we're going through is an extremely compressed period in time where we're finding all the missing if statements and we're fixing them with SI as our propellant both in the finding and the fixing the fact that the bad guys have access to open weight models and they're chasing the one thing they need to be right and then all of us have to be right most of the time or hopefully all the time that's a temporary effect the long-term effect is that we'll build more robust software the thing that we're getting from this revolution which is probably the best word for it there's a tremendous amount of creativity that's being unleashed in all of us the thing that LMs have not gotten to and I will take the bear case they have yet to come up with a single brilliant idea that was not originated with a human the job changes we're going to witness is going to be all about pivoting all of us towards more creativity and the real question I think all of us have to ask how creative
[00:44:37] Speaker 6: am I really
[00:44:37] Speaker 3: and what am I doing to unleash all of that and if you're finding yourself devoid of good ideas that's the moment to worry but retailers will become better at marketing more creative at marketing because the fulfillment and price comparisons are all going to accrete towards just efficiencies because this productivity gains will apply to all those areas and so creativity is going to be the currency of the moment for security researchers too
[00:45:03] Kevin Warsh: Max, let me follow up on an important thing we'll come to this with C.C. Rouse and other economists in that panel what's happening to your workers what do you expect there to be you use the word upheaval algorithms that follow the Fed just traded several basis points on upheaval coming up at the Fed conference what does upheaval mean to your company and to the enterprises you know in terms of your own labor force
[00:45:30] Speaker 3: so I think in companies I suspect like all of ours to a varying degree but probably not too different upheaval is mostly a positive motion we are primarily trying to figure out where to apply the gushers of productivity some degree of budgeting some degree of capital allocation we have to engage in even if productivity is just running off the charts because tokens aren't free we're not electricity isn't free etc and so in my world prioritization has become ever more important what is the appropriate risk to take Black Friday, Cyber Monday is upon us that's the sort of hunker down part of the year for the two of us you have to have your five nines or six nines or seven nines because retailers expected and there's a small army of people saying I have this great idea let me ship it and it's a human it's not AI that says I thought it through and that's the right thing this is the risk I want to take and the risk is particularly high in the moment before Black Friday so on the flip side if you ship it and it works that's another five basis points of growth that you wouldn't have had so I think upheaval is mostly judgment and attempts to decide what is the right bet to place right now and that's always been our job that's the thing that brings us to work every day but the pace of decision making is increasing and I think it's being propagated down where every single human that's sitting in front of this rapidly increasing capability doesn't decide well do I write five more lines of code or is this time to call it you know Friday night for you know it's more like well do I put out this feature it'll take me and my agents several hours and then we'll ship it and hope it doesn't break anything and so I think the the tension we're all feeling is this productivity needs to be directed and well managed lest it causes actual upheaval in our code bases and introduces more bugs the one thing that AI unfortunately does right now it writes code that is not entirely perfect yet and therefore all the bad guys probing for holes or finding holes that were introduced of course by the very same AI they're using
[00:47:37] Kevin Warsh: so Nikesh I'm going to start with you with a question go down the line so we've limited most of this discussion to thinking about our day job at the Fed so we're thinking about implications for the U.S. economy you three are leading important U.S. companies what about the diffusion of the rest what about the U.S. versus the rest of the world how should we think economics isn't zero sum but how should we think about what our enterprises are doing what our individuals are doing in the U.S. and whether that's being replicated among big trading partners around the world and what this does for U.S. economic growth relative to our trading partners
[00:48:15] Speaker 5: well let's start with the comment you asked Max about upheaval or how we're deploying and how other companies are going to be deploying it I think what you're noticing and seeing is a gradual reskilling of our workforce it's not explicit but it's gradual every employee at Palo Alto is now experimenting with AI we're watching them this is a technology nobody knew you can't go to school and learn how to use it you literally have to be naturally curious you have to be naturally creative and learn it now you're seeing that across the board it's not as uniform as you'd like it to be but you're seeing that across the board I think by virtue of the U.S. being in the center of the AI storm in a positive way you think there's a lot more ability to pivot change, embrace every CEO is interested in seeing how their organization can embrace this new technology so I think you're seeing a bit of a groundswell desire I'm not suggesting that everybody will make it across the other side but I think the desire to embrace the desire to understand the desire to absorb it and leverage it for economic gain or organizational efficiency is the highest in the U.S. I think your question around U.S. versus the rest of the world I think we have to separate two parts I think we get over indexed on who's producing the fastest smartest model versus who's diffusing it most effectively and where are we seeing real economic gains in enterprises which have competitive advantage as a consequence of using it effectively I think yes the U.S. is the center of where a lot of these models are being produced or where this intelligence is being produced and the mass of that is still to be seen whether that results we'll find out in the topic but it's still to be seen in terms of is $50 billion every time the right number to go build these models most countries in the world can't afford to create this intelligence of their own so what you're going to discover there'll be a few centers around the world which will build the intelligence and the question becomes let's assume that let's take it for the positive side and we can debate the you know how this intelligence is in the hands of people who could use it for bad things but let's take the positive scenario which is 98% of the scenario there I think we're more than likely to see the diffusion in the United States faster than anywhere in the world I think that will result into efficiency gains that will result into margin gains that will result in global competitiveness so I think you know this is ours to lose from that perspective I just think we get too hung up on who's building the next fastest model who has it I think the diffusion here is the most in anywhere else in the world Patrick
[00:50:35] Kevin Warsh: let me follow up with you on the same question so we care in the United States about our absolute level of growth our potential productivity potential growth is Nikesh right that the delta between American growth and the rest of the world could gap out unlike any period that we've seen or is that unlikely
[00:50:56] Speaker 4: not only do I think it's likely but it's happening and we see it directly in our data AI usage intensity is higher in the US than elsewhere obviously on the supply side an overwhelming majority of the revenue and the firms and the creativity and the innovation is occurring in the US generally and indeed in the Bay Area specifically and you know as you mentioned in the introduction I grew up in Europe and you know the European economy is when I left Europe the European economy in aggregate was you know not wholly dissimilar in magnitude to that of the US and you know we don't need to be mean here but suffice it to say that there's been a considerable divergence since then and again every indicator and sort of leading datum that we have access to would suggest that that divergence is going to increase and then generally speaking you know I guess in econ you always want to avoid corner solutions and you know on the one hand there's sort of this sense with AI that maybe like the international at least in its early days it's sort of showing up everywhere but in the productivity statistics now you mentioned that well maybe we're seeing some kind of wiggles of hope and so forth but it's certainly not a dramatic trend break or not you know the peaks that we saw you know in the aftermath of the second world war and so forth so in that sense some degree of you know humility or caution or something would seem to be warranted and then at the other extreme and I mean something I would hope perhaps to convey to all of you is the magnitude of that which is contemplated in San Francisco doesn't really have you know smoky back rooms it's more in kind of desultory and drab conference rooms but that which is spoken of in these drab and desultory San Francisco conference rooms is two people I know have a and you know these are both people in positions of I would say considerable influence over how AI unfolds from here they have a bet between them as to whether in some year of the next 10 years global energy use doubles in a single month now you know you don't need me to point out that that is not something with historical precedent or that you know has occurred hitherto but that is at least what one of these individuals feels sufficiently confident in that this vet has been placed and so you sort of have these two extremes of what we can't exactly see in the productivity stats and global energy use doubling in a single month in some month in the next 10 years and I would say stripe sits somewhere in between where we do see it again we are observing this dramatic arguably exponential acceleration and on the relative framing of your question U.S. versus rest of world I would say the U.S. is in a definitively preeminent position today
[00:54:13] Kevin Warsh: Max join in on that question the U.S. versus the rest of the world and give us your sense about whether that gap is growing
[00:54:23] Speaker 3: certainly feels that it does I agree with you I think that it's a little bit earlier than the numbers would allow us to sort of declare but just looking at the sort of external factors perhaps the ones we're very used to I think European Union is still debating the correct framework for debating about AI they're a couple of meta levels away from accepting that this is happening whether they like it or not
[00:54:54] Speaker 4: excuse me we're the regulatory superpower
[00:54:56] Speaker 3: as someone who also grew up in eastern Europe you know I take you've seen regulatory superpowers I take a lot of pride in being an American by choice and no no maybe maybe one of those strands is the my love of bureaucracy being cut I think the present company being excluded I think I think we are seeing extraordinary gains that are going to add up to some pretty staggering I'm not sure about the doubling of energy use but it's very clear there's a tremendous number of tokens that would get consumed if they were cheaper and so it's incumbent on us to build access to those tokens therefore access to energy compute and markets are really really good at figuring out the intersection of supply and demand so I'm not even a little bit worried that we're going to overproduce or underproduce the number of GPUs TPUs you know your favorite chip that'll sort itself out in time but I think the time to make those investments however perfect is right now and the fact that U.S. generally speaking independent of your political spectrum is embracing that and it doesn't seem to be a tremendous party divide around the question of should we be the world leader in this all important technology I think gives me a tremendous amount of hope I think if you sort of look at Europe the conversations about you know we have to stop this thing it's going to take us ten years to figure it out like ten years from now maybe we'll be doubling energies I think our obvious adversaries slash trading partners in China have exactly the same idea and I think there's an enormous amount of future tension that we're going to have to navigate hopefully in peaceful and primarily trade ways and I think that's sort of it which maybe is the sort of the world that we were always going to have but it's coming sooner than we think
[00:56:59] Kevin Warsh: so we've got two minutes I'm going to ask one question I'll just add a statistic which is imperfect I think you guys will concur token prices are falling about 50% per quarter and are falling at an accelerating rate so on the question about who gets the gains boy it sure looks like markets might work there let me end with a final question for you I began mistakenly by beginning with a great Russian novelist so let's end with a depressing conclusion this has all been way too optimistic we central bankers are two-handed and I think that I've lost control of the three of you because you think this is easy and we're going to win so I'll start Nikesh with you what could go wrong
[00:57:44] Speaker 5: I don't know Kevin you're not going to drag me down that path I know you want to be unhappy at the end of this but we don't want to be unhappy look we've powered through a pandemic as a society we all thought it was the end of the world as we knew it and here we are sitting in the room and no masks and life is wonderful so I don't think there's anything we'll not be able to get through I think this is going to be a phenomenal technology you have to look at the positive sides of it all the different tasks that do I want my 20 year old kid to go become a customer service agent and say how can I help you your internet is broken that's not a job I want in the future in my society so I think the fact that this technology will allow us to unleash the creativity as Max said and get people to focus on more important things we're seeing a gradual reskilling this is not going to happen as a shock to the system this will happen as every other revolution has happened as over time people will reskill and you suddenly discover 10 years now when you look back and say holy shit that used to be a category of jobs that doesn't exist anymore and we seem to be fine so it's going to happen are we going to have some little bumps in the world for sure are we going to have a cyber event that's going to be there that's going to impale as something for a while possibly are we going to have some system that doesn't work for a weekend for a week possibly but they're not existential
[00:58:56] Kevin Warsh: to you when we've got to do quick answers to quick questions what could go wrong
[00:59:01] Speaker 4: well being an immigrant I felt like I should read democracy in America so I just finished it and at the very end during the last chapter he describes how how he you know the sort of failure mode that he can imagine for democracy is a kind of soft despotism where life becomes too comfortable the state provides sucker and comfort and a kind of suffocating tranquility that inhibits and enervates and does not support the cultivation of a great society and ambition and achievement and great works by its individuals many people
[00:59:47] Speaker ?: many people
[00:59:47] Speaker 4: in the AI sector expect drop-in remote workers within the next 12 months so you can be a little critter it's on slack join the zoom meetings you can do anything that somebody with access to a monitor and keyboard can do and I think there's a question for us as to what the I mean there's all those doom scenarios and cyber hacks and bio risks and all the rest but there's also the countervailing question or set of questions about the positive scenario let's just say the catastrophes do not unfold we manage to tread the perilous path I think there's a question about what the utopia looks like and how it is what the shape of the society in the future that we want to inhabit is what the nature of work is and so forth and how we avoid de Tocqueville's soft despotism
[01:00:37] Kevin Warsh: with the final word with you Max
[01:00:39] Speaker 3: I'll try to be less positive than the although for the record I'm extremely positive on this topic and many others so I think the thing about AI revolution is unlike most other at least in my memory revolutions the incumbents in particular incumbents flush with extreme access to capital are very aware of it aware of it from the very beginning and so the disruption by the plucky startup is not actually given enough time to go reconfigure the industry and so maybe if there is a concern to be had it's that the incumbents correctly invest in the opportunities further solidifying what sometimes can be a monopolistic or oligopolistic grip on the industry and so what could go wrong we could end up with a hundred greatest American corporations completely dominating our economy and not a whole lot in the middle tier I do believe that the long tail the creativity being the solopreneur the bedroom creator of a unique product that one person in this room will buy but that's okay because it's a profitable purchase will endure and better than endure and so there could be a weird economy where we have this incredible number of businesses starting on Stripe that never really get to Fortune 100 and Fortune 100 just obliterates the Fortune 10,000 because they can't because that's a productive purchase for them I'm not sure that's a terrible thing but it's at least a weird thing that the market may or may not be able to quickly turn around that said I think every person in that infinitely long long tail the thing that makes us human is agency we look at things and say hey let's fix that that looks stupid like I don't want to live in a world that looks like that so the little guys will start repopulating the hollowed out middle should my scenario unfold I think but in the meanwhile we may see sort of a boom of M&A that may or may not be well advised for things like consumer pricing
[01:02:37] Kevin Warsh: so on behalf on behalf of the federal reserve to the three of you we're grateful thank you and we'll turn immediately to the next panel to figure out the implications thank you thank you thank you
[01:03:17] Speaker ?: thank you thank you thank you thank you
[01:05:17] Kevin Warsh: this panel is entitled reading the signals how will SI reshape the economy so we just talked to some practitioners inside of firms now I'm looking at my fellow economists to try to say what are the implications for policy makers for workers for enterprises I'll introduce each of these four economists they need no introduction to Fed personnel because they play such an important role in public policy I hesitate to say they need no introduction I was on a panel many years ago with Dr. Henry Kissinger and they had introduced me at this long introduction no one knew who I was but of course everyone knew who Dr. Kissinger was and so after a long introduction from me they said Henry Kissinger the man who needs no introduction and he said while I need no introduction I expect one and so let me try for each of these four gentlemen Nick Bloom immediately to my left is the Eberly Professor of Economics at Stanford and he co-directs the NBR's program on productivity innovation entrepreneurship he also helps pull together surveys that we at the Fed are consummate users of so we're grateful to his immediate left is Dan Huttenlacher he's the inaugural dean of the MIT Schwarzman College of Computing before that where we met is he ran Cornell Tech when Cornell took over the city took over Roosevelt Island and I was blown away then when I met him that a computer scientist had thought so much about the implications of what was happening in technology to the economy he co-authored The Age of AI with Dr. Kissinger and Eric Schmidt and he brings a computer scientist sense to the economic discussion we're going to have next to him John Levin full disclosure John was my boss at Stanford until five months ago so if you see me playing favorites that will be the reason he's a distinguished economist in addition to being an administrator has to think what are the implications of this for the teaching of undergraduates and graduate students the recipient of the John Bates Clark Medal his work spans industrial organization market design digital asset markets we're lucky to have him join us C.C. Rouse is an old friend and colleague she was the chairman of the Council of Economic Advisers in the Biden administration a labor economist by training she's taught for many years at Princeton and was the dean of its School of Public International Affairs her research focused on a lot of things but importantly for our purposes on education and community colleges and maybe most important how do workers adapt what you heard from that prior discussion was overwhelming with optimism for the United States its role in its world the benefits of productivity I don't mean to dismiss any of that but we central bankers we have a day job and the day job involves trying to stay true to the remit that Congress assigned to us so I'll be pushing some questions to C.C. to help us navigate that let me begin if I could and I'll start Nick with you and we'll go down for this series of questions let me begin by see if I could frame this we think about workers and the implications of this rather optimistic scenario you heard over the last 45 minutes what are the implications for knowledge workers versus the rest what we had come to know until this generation of SI is knowledge workers well they were just going to be fine and the shocks maybe the China shock and others well that was going to be bad news but knowledge workers would be fine is this the opposite
[01:09:01] Speaker 7: I was going to start by saying you've obviously gone back to the Russian theme gone back to something more depressive with the number of economists here you know return to this so why don't I start to say why I'm here to give backdrop so I've been running with the Atlanta Fed for many years this survey of CFOs several thousand CFOs across the US large small firms different regions different industries turns out once a month if you look at it it aggregates up very well to GDP very well to employment so it tracks the economy extremely well and as we were talking earlier it feels like now AI is everything so the last year in particular the last six months we've been pointing the survey to ask very precisely what's the impact of AI for these executives on their businesses and we ask about the last three years and the next three years the next three years kind of the window I think they can answer about so the last panel was fantastic and if you want to look five, ten years out you want to think about technology as was discussed the next three years really is much more about bottlenecks and implementation so what did they tell us two things if you look at employment they said looking back three years they've basically seen nothing so 90% of firms said employment has not been impacted by AI to the extent the other 10% said anything they were almost perfectly balanced small negative small positive so as of now there's no effect on employment these firms tell us looking forward three years the story is slightly different so they said on average no effect but there was a lot of mass out to the right and the left so I think what it's telling us and again this was discussed in the last session there's a lot of churn so there's going to be a large number of people going to lose jobs and a large number of people going to gain jobs the other thing
[01:10:44] Speaker ?: the other thing
[01:10:44] Speaker 7: we asked them about is productivity and Kevin staying I noticed you were very precise on the date of the Paul David things I'll be very precise we asked about productivity with sales per employee which is something that execs can talk about but it's not exactly the same as TFP they said over the last three years on average we're seeing numbers about plus 0.1 or 0.2% a year that is small but it's not zero so looking back over the last three years these firms are saying look we haven't seen much we've seen something it's primarily in the tech sector out of that not much but looking forward for the next three years we see really very large numbers so on average they're talking about 0.8% a year and to be clear that is just a huge number for the US economy and it would you know when I step back with Chad Jones and others we mentioned we have this paper about how productivity growth has been declining in America for the last 40-50 years these numbers feel like maybe we'll add a point of inflection it hasn't happened yet but you asked on timing the data looks like 27 maybe 28 is when things start to really turn around and it affects many parts of these firms so Nick
[01:11:51] Kevin Warsh: I can say this to you as a former Stanford colleague I think President Levin says you did not answer the exam question CC will you answer the exam question the exam question is the China shock had a big effect except on these knowledge workers it had a big effect on our economy is this shock coming for them
[01:12:12] Speaker 8: very possibly you know the thing that is different about the China shock is that the China shock was about the manual workers the trade workers that the work that could be done less expensively abroad and then we re-import the goods here this is about knowledge workers and you know I think we don't fully know what the impact is going to be you know every time I talk to folks that are you know even you know folks at Anthropic or open up like people who know don't know because jobs are tasks they're made up of individual tasks and we see that currently AI can do some tasks well but you kind of need some human judgment still in the mix so firms aren't letting go workers because it's catastrophic to have a mistake by an AI it's one thing when a human makes a mistake there's some accountability but you still need some human judgment etc. in the mix so in your last panel they were talking about when do you let the AI make some of those judgment calls so we're not yet seeing it I don't know if it's fully in the next three years could be the challenge that we learn with the China shock though I think this is what you're alluding to Kevin is that workers don't move so our economic model suggests that yes their jobs could be created on net we may see new jobs being created but if they're not in the same place where the jobs were being displaced then they're going to be workers who are we have large places where there's large unemployment and that did not go well some of my colleagues at Brookings you know this is guesswork at this point because we don't fully know who's going to be impacted but if we look at workers who are you know do largely administrative work a lot of these are women that are more you know middle-aged and that live in monopsony towns like college towns or where there's a state government that's largely administrative those workers may see you know may face more displacement and have a harder time transitioning but we can talk more about that
[01:14:14] Kevin Warsh: so so John let me turn to you you and Dan are educators here so we're going to turn in just a while to you to what are you seeing what are you doing on campus but with your old economics hat on will the gains from SI flow to workers to consumers through lower prices the owners of capital help us understand what are the implications of this
[01:14:35] Speaker 9: do you want to start Dan I think he was asking you're John
[01:14:41] Kevin Warsh: John can answer any question that he wants
[01:14:45] Speaker 10: I don't think anyone has a crystal ball on this particular question I think some of it's going to depend on competition because the question of are there going to be very large rents that accrue to the model firms or to the AI firms is going to depend and to significant extent on how differentiated are the models how big are the returns to scale how much to open source and models catch up and how long the bottlenecks and chip production memory chips energy and so forth last if some of those supply constraints are relaxed there'll be more competition and prices that'll be deflationary on many prices and a lot of gains will accrue to consumers from having more products more innovation if that doesn't happen then that'll justify the very high stock market valuations of a lot of the technology firms that we're seeing right now and more gains will be accruing to the firms and to capital I think the question about you know what happens to the labor share and I think Nick could speak to that better than I could but I think it is possible I think there is a scenario where many of the gains accrue to capital and less to labor but a lot depends on the exact shape of the way the technology advances and how much it ends up augmenting human productivity versus replacing it and I think it's actually just too early to know with any degree of certainty on that question
[01:16:37] Speaker 9: Dan over to you So I think this is sort of the $64,000 or you could make it millions or billion dollar question which is to what degree are we going to see human augmentation and I think that the challenge we're seeing right now is that this intelligent set of technologies are so different than any other kind of technology that we've seen that the deployment lag is something we don't understand at all and that's what makes it such a fundamental question in my mind so what I would say is that you know we're not in Kansas anymore in the following sense which is that these technologies are really a different form of intelligence which means that when we think about how people apply and apply their intelligence in their work and knowledge kinds of work the ways that these technologies get integrated into the workforce is going to have a huge effect on the degree to which this is complementary versus replacement and I think that to me part of the reason this lag is going to be quite a bit slower than what we've seen before unless we change our focus and we can come back to that a little bit more people are interested is that we don't have the right mental models you sort of heard on the panel before the word mental models came up we don't have the right mental models for what it means to actually deploy this kind of technology so what we're seeing is you know certain individual employees who sort of figure out how to make it work there may be sort of blessed with the flexibility to have a different kind of mental model or maybe they're at firms like like the CEO's we saw before and they're starting to take advantage of this and I think that those employees are the kinds of people who will do very well in this world but we don't understand whether that's some small minority or whether that's a big majority and I think we also just don't have the organizational expertise to really start to try to scale this deployment right now so I think this is less about the technology at the moment and much more about what kinds of organizational changes do we need to put in place what kind of new medical models do we need in order to really take advantage of this technology as quickly as possible but if we do that then I think we're going to see a huge amount of the value going to the users because that's where that those those changes to where the users are but in the three-year time frame right now I'm skeptical about that because of these sort of gaps in understanding
[01:19:18] Kevin Warsh: so Nick I'm going to try you again so we had we had capital and labor in our basic economic theory producing output this time the K and the L produced these T tokens which we're all trying to figure out about your surveys look at another T time how important is the time dimension here if the technology continues to the scaling laws keep working we're at a hyper moore's law and there's nothing that society either chooses or can do about the time at which this technology finds its way into everybody's pocket everyone's enterprise what's your judgment about the implications in that scenario for businesses and for labor
[01:20:07] Speaker 7: great I was listening very carefully I don't want to get a B minus he hasn't gotten a B minus in his life yeah yeah but two questions wrong in a row well actually Stanford is probably a little bit more relaxed at this point so oh no John so I so two things one is you know just coming back to exactly as Dan was saying before it looks like in the near term a lot of the obstacles around adoption what's often been called kind of messy jobs bottlenecks etc I'll make kind of two upsetting comments so one comment is in a different piece of research with the St. Louis Fed we've been looking at job postings and technologies in a paper we published last year and are updating we looked at hundreds of technologies and looked at how quickly they diffuse out and the explicit metric is how rapidly they are mentioned in jobs postings online jobs postings and in paper before that what you see is AI is mentioned at a rate that's incredibly fast compared to other technologies so if you look back yeah actually your poor David example is perfect so if you look at electrification that's an amazing article what it talks about is factories used to be run by steam engines that have like a central pillar that was rotating around the steam engine belts coming off it was kind of mechanical power the electric motor comes along but in order to use that you've got to redesign factories make them flat lots of motors all over the place and it takes several decades computers the internet took maybe a decade or two you know cloud is maybe a decade so in terms of jobs what we're seeing is the uptake of AI is much faster than we've seen before so we're seeing you know two three four years a lot of uptake but the question is will it impact productivity what the timescale there is is much less obvious and so I think it's kind of you know Michael and I were talking in the pre-break about it's very kind of wide but shallow there's just not a lot of kind of deep use of AI because it's hard to change systems and so I think the timing question is very rapid if you look at surveys just of individuals in America fifty percent of them now report using AI but we're just not seeing it show up in productivity so the fusion is very fast but I think impact is going to take longer because it's about these organizational changes and bottlenecks
[01:22:10] Kevin Warsh: CC let me let me hit hit you on the same question which is as the speed of this technology is accelerating what are the effects in your judgment on the labor markets at the Fed and economics profession we've had rules of thumb between output employment that have been reasonably reliable guides as output moves then we're going to see gains in labor so that's still true
[01:22:36] Speaker 8: you know this is a question I have been thinking about which is with this with the with AI one can produce a lot more and maybe this goes back to the weakest link idea if it's producing a lot more but you need a human to consume things at some point we can only consume so much so is that increased output really really productivity output that one of my colleagues sent me a report that the UN publishes I don't know how many reports a year and nobody reads them I said great with AI they'll do twice as many reports and still nobody will read them and so I so I I genuinely ask myself this question I think we will see we will see some productivity out and we will see some output gains that but the question is how is it diffusing and and and and in what purposes but can I go to my nightmare trying to find nightmares the last panel wasn't so agree okay great which is and it's related to something that was said in the last panel which is that to date and these are changing so you know what we're saying today at 10:25 could be you know completely overturned by 4:30 but today the LLMs haven't found anything novel that hasn't been based on human innovation and creativity and if we think about like economic growth we know it's human you know labor capital we need innovation in order to get the gains I worry that we are like little frogs as the temperatures going up and that what we are slowly going to be destroying is the human capital in this country because our students are not learning how to use their brains our entry-level workers are not getting the human capital production as they're over relying for very simple tasks for on this LLM but they're not learning to exercise the creativity which your last panel says is actually what's going to be most valuable and so that is my that is my long term I watch Wally I put off doing it because I always thought it was a horror film and I stand by it it's a horror film because what is it that we're going for I have not gotten a clear articulation of what is utopia what there are choices that these firms can make about whether it's labor augmenting whether it's going to compliment creativity or replace us which which fork in the road are they going for so I'm sorry I just wanted to put out there my big concerns
[01:25:04] Kevin Warsh: because that's my nightmare is longer term John do you want to build either on CC's nightmare or say talk about your students undergraduates and graduates most of whom are going to be employees if not very quickly employers how's this changing your job that are
[01:25:24] Speaker 10: responsibility of the University I think AI is going to hit you know it already is hitting every aspect of you know both education and research at universities and that'll that's just going to continue no question about it and you know like you said your universities are sort of in the business of doing two things we create knowledge and we create you know we create people I think the education part I think there's some aspects of of changes in education that are basically just about trying to adapt teaching and classrooms so that you don't get the cognitive offloading and the short cutting and so forth that can undermine learning I actually don't I actually don't view those problems as you have to execute on them and do some experimentation and figure them out but I don't view those as sort of a fundamental thing that will prevent students from from learning so just to give you an example MIT put out a nice report that kind of summarized what many people are seeing in higher education which is students don't you know class with problems they don't go to discussion section anymore why because you used to go to get your TA to help you with a problem now you go to Claude the students don't work so much in study groups anymore in those classes because why go walk down the hall and talk to your the other person in your class when you know chat GPT is available at all times and knows all the answers why do you think the problem set scores when you think about the problem set scores when you know the problem set scores when you know the problem set scores go up to 100 and exam scores go down because people haven't actually learned anything students don't do the reading and they show up with a little list of here are the three clever things to say in class and okay so what do you do about that these are not insurmountable problems so what are the types of things that we're doing I'm sure MIT is doing many many universities are doing you you don't count the problem set scores in the final grade you put the the the the the the all the points on the exam you require the students to go to section but in sections that of giving them the answer to the problem you have them get up on the blackboard and solve a new problem they've never seen before you require students to work in groups in the in in our introductory core curriculum which I'm teaching in this fall we give the students a quiz every single day at the beginning of class just to test them to make sure that they've actually done the reading grading is based on blue book writing and on oral exams and on class participation so you know it took a year or two or three years or something to make those adjustments but universities will make adjustments because the faculty are clever and they want the students to learn they'll do those things so that's that's those are all important changes it'll take some time to figure those out and percolate through and so forth but I think that the fundamental goals of a college education I don't actually think they're going to change that much everyone want to learn how to use technology they'll have to be facile with it left to be ready for the workforce they're going to need to be really really good with these tools but ultimately they're they're going to need to be able to ask questions to be able to communicate effectively be able to think hard to solve problems to be adaptable that to and to have some you know good character that will sustain them in in the workplace and in life and and that's a that's been a north star for higher education program I think it'll continue to be one and we'll you know we'll take some time to navigate these changes and figure things out but I actually think we'll get through it and and the college education will still be an incredibly valuable asset for many many people in the in this country
[01:29:16] Kevin Warsh: Dan the story John tells is one of organizational change just like enterprises to get the fruits of the productivity that CC reference you can't just go do your own thing and drop in this new technology and AI is changing everything just that way we're you're changing how the organization moves maybe you can speak to that both at the university context but also more broadly in the enterprise
[01:29:39] Speaker 9: yeah so let me just start in the university context I agree with you on some of this part of it I I feel like there's maybe a piece of context here that to me is very important which is that I think in many ways we sort of industrialized education we scaled it so that you know big lectures and other ways of delivering learning compared to when you think historically about how how learning worked and I think one side effect of that is that it's maybe diminished one of the most important things I think about education which is feeding people's curiosity because in industrializing education we sort of have structured the information that we think we need to get to people and and and so I think that we're missing a huge opportunity here in sort of what these technologies are going to do in education if we don't sort of refocus on some of those goals so like to me one of the things that's most important is comes back to this if you're going to make big organizational change let's make sure we're doing it for the for the right things and the you know much of what we deliver in any kind of you know product we sell or you know company we build or education is the result of the kinds of organizations we could build to deliver it and I think in education in particular we really need to refocus on how we feed curiosity because learning is hard it's really hard to learn but without understanding we're not going to be able to play all of the roles that the previous panelists were saying were so important they have these people who are or education is the result of the kinds of organizations we could build to deliver it and i think in education in particular we really need to refocus on how we feed curiosity because learning is hard it's really hard to learn but without understanding we're not going to be able to play all of the roles that the previous panelists were saying were so important they have these people who already understand for the most part on their staff so i think we need to think about you know and i think these tools and the changes that we can make can really help us drive curiosity and and help people become motivated to learn in a world where you're carrying this genius around in your pocket so you need to be motivated to learn and understand and not just sort of consult the genius and have a sort of very surface kind of understanding of it on the broader question about organizational change i think i think that you know one of the the key issues and this actually goes back to the book that you know i wrote with kissinger and schmidt you know then we made these we knew we couldn't make statements too early so we asked lots of questions against what i would say are sort of philosophical statements and one of the philosophical statements was that this is a completely foreign kind of intelligence so that uh you know it's it's you know how do we understand the world historically human intelligence and faith and this is none of the above so suddenly we have something that and if anything i believe that statement much much more today than i did in 2020 and 2021 when we were writing the book uh and it may be the only thing about ai that i believe more now than i did five or six years ago but i think what we're starting to see is the organizational consequences of that as we're looking at how to try to roll out ai in organizations we have no mental models for what this new form of intelligence means or what it can do and so what we're starting to see in some of these you know sort of let's call them tech firms the ones that sort of have already because their businesses have moved me so quickly had much more more organizational flexibility uh and that's why we see them also leading in the rolling out of ai but uh but just sort of one more quick quick point is i think that um something i think it was uh maybe patrick said that really just i think is critically important is you know you get a straight jacket as a big company you can't afford to experiment in the same ways as small companies and so i think a very important part of this for larger companies comes back to issues of the sorts of control that we have when we have a bunch of human workers that we've developed their organizational control mechanisms so that managers and ceos and so forth and what would it mean to have analogous kinds of control for an organization that's a mix of human and ai workers or even predominantly ai workers and i think we we have no idea right now so i'm just going to pose the question but i think those are the kinds of things that we need to be focusing on and kind of really trying to address so let me
[01:34:12] Kevin Warsh: build dan i'm going to follow up with you and build on a thread that cc raised which was almost all the amazing innovations made possible by ai were ideas that started with humans yes i tried to say in the outset you know the the types of folks that have been under uh appreciating this are humans humans here make the difference so you wrote very recently that humans have to be have to remain accountable for what si systems can do humans are not outsourcing their decision making to a bunch of agentic bots and that you know as a computer scientist how to do that maybe you can help us with it
[01:34:49] Speaker 9: yeah so this this is it this is an extremely important set of questions around accountability and what accountability means and it comes back to this fact that we don't have reasonable mental models because we're used to delegating things to human colleagues all the time and then we assume that they're taking some portion of the accountability and sometimes we're very explicit about that but even we're not explicit that's that's a background assumption machines cannot take accountability for anything they're machines this is no different than you know yourself you know you're yourself riding lawn mower or something it just it can't take responsibility you're the one who's responsible if you know you you you hurt yourself or hurt somebody with that device so uh so we need to change our mindset we need to recognize that accountability rests with us but then back to this fact that we don't have reasonable mental models um so how do we think about this today we think about this as sort of humans in the loop and then everyone's like oh my gosh we can't approve all of these things that machines are doing at the right speed so then we have humans on the loop which is a little less well defined but you know we're still there that technological loop is happening and we're kind of holding on for dear life uh and and i think that that um that you know there are ways to start to think about uh things that uh things that can give us the ability to share accountability with some kind of a system but they depend on things like you know in the op and op-ed i had one of them which is just independent sources of information i don't think turning to another large language model to be your independent confirmation of what this large language model did is very independent um but there are other technologies under development which i you know uh uh maybe highlight here briefly uh there's stuff that's been going on for decades you know i've been working in this field for a long time uh called automated reasoning which uh um you know until recently has only been used for things like verifying flight control systems in aircraft when you get on an airplane you care a lot about that software and so it's formally verified there are mathematical proofs generated by machines about what that software can do those formal verification techniques have traditionally not been very scalable and it takes a lot of very specific human expertise to use them we're starting to see these glimmers of where those technologies could start to also scale and so i think that the technological answers to this are going to be new ways of us gaining confidence in the fact that we can have some more kind of accountability sharing than we've been able to to date but the warning that i had in that op-ed is you cannot do that today you really you you you need kind of uh you know oversight mechanisms that we don't quite have yet so four of the most esteemed esteemed
[01:37:53] Kevin Warsh: economists i know and we've quickly turned into questions of philosophy yes we haven't talked about oakland's law we haven't talked about the shape of the phillips curve we haven't done any of that and in some ways that's the point the point is that the answer is not to go to some mean regressing furbus model as if it's going to have the answer for what policy makers are going to need to do over the next several years so i'm glad we've done it but many of us here have jobs to do so i don't want to leave the let the four of you off the hook so i'll start nick with you and i want to go down the row what if you were policy makers would you be probing most about what long-held assumption of the economics profession of our rules of thumb would you say not so sure this time what's the question you'd be asking which is something that we at the central bank have to reflect on
[01:38:46] Speaker 7: i i believe this morning i was thinking on the way up like what will be the three pieces of policy that you do and two i repeat and one i was very unsure about so and the one that kind of goes to i think what john cc were talking about earlier so you know classically one is have flexible labor markets flexible markets why things are changing and as this and the last panel discussed who gets the gains the more this competition the more the gains go to consumers the more it's kind of things are flexible so flexible labor markets flexible markets competition seems good the second is you want some coming back to your china shop question tc's discussion you want some kind of insurance for the folks that clearly you're going to lose they're going to be out of jobs you know that's hopefully temporary but you want some kind of downside some kind of you know government protection the third i'm really less sure about so classically this goes back and then you know john mentioned this for the the liberal education classically we've always seen educated people have done better in response to new technologies and less educated and so the classic textbook is more education more training why because these folks are more flexible and adaptable and they just move i you know that still would be my take now for ai but i'm less certain to be honest because if you look at what it's doing it's you know the people that i know that are most scared are some of the most educated folks my students my phd students are really nervous because the things that ai doing is you know it's like the china shop for white colored jobs instead and so i still go with that but i'm less certain and the kind of skills you know david deming at harvard has this stuff about you should focus more on people skills rather than analytical skills and maybe that's the kind of skills i'd focus on but it's not so clear how to imbue them into individuals same question for you dan so
[01:40:23] Speaker 9: um this is easier to say as a computer scientist as a practicing economist um but but i and and and actually i think some of the stuff nick's been doing is very much along these lines but i would try to go further i think you need some more disaggregated data um you need more specificity in some of the questions that these surveys are asking so some of the things that have you know come up here that i think are super important um is looking at patterns in the you know college graduates by you know major or other way of sort of looking at where they would have normally gone in the job market and looking at is the demand for them going up or down and in what kinds of firms you have to somewhat disaggregate the firms somewhat disaggregate the majors but looking at what's happening with those college graduates at all levels undergraduate masters and and doctorate because there is incredible churn there and i think all of us i think hopefully everyone in the room is uh you know rightly are worried about the fact that a lot of you know professional training happens in the workforce in these entry-level jobs and that is the place where we're seeing the most change in companies so i would really try to get some surveys that focus in on that disaggregation and i know it's hard like i know disaggregating causes all kinds of challenges
[01:41:47] Speaker 10: but that would be my my kind of uh take on it john i have a short term and a longer term answer to that i i think in the short term if i was working at the fed right now i would i would be looking really hard at the enormous sort of enthusiasm fueled investment boom in uh in ai and data centers and and hoping that that was going to be backed up by you know growth in in revenue and earnings and and solid credit and would would would deliver uh whatever the three trillion or so planned investment is over the next couple of years that's increasing gdp and and so forth and and wasn't going to collapse in in some way but longer term to me and i mean i think this is certainly what we're seeing it at at
[01:43:01] Speaker ?: Thank you. Thank you. Thank you. Thank you.
[01:45:01] Kevin Warsh: Thank you. Thank you.
[01:45:33] Speaker ?: Thank you.
[01:45:33] Kevin Warsh: Thank you. Thank you. Thank you. For a policymaker, you can get data that's more relevant, more actionable, less likely to be revision so we can be helping do our day jobs. Is that fair?
[01:45:50] Speaker 6: That's exactly right.
[01:45:51] Kevin Warsh: Okay. Let me do one final go-around with each of you. I'm going to stick on the economics and, Dan, if you want to opt out of this and weigh in with some computer science, that's fine. Inflation tends to be something we spend a good bit of time at, perhaps now more than ever at the Fed. I'll give a proposition. I don't want to hold this too tightly. But if in some ways over the last generation or generation and a half, the China shock, for lack of a better word, lowered prices of goods in the United States, the way that I heard you talking about the philosophy of this new technology is it might do something for services. What we might be doing is productizing services. Services now can be bought the same way the widget could have been bought. And I wonder whether we should expect disinflation over the medium term from what AI could do to services. I'll let any of the economists weigh in on that and then I'll ask you, Dan, a final question.
[01:46:57] Speaker 7: Go for it. Actually, my grade is going down even lower now. I'm answering a different question. But I was just going to build on what Cece said. There is a major hole in AI data just on how much firms are spending on AI. It's kind of insane because it's not in accounts typically. It's not in administrative data and classically it hasn't been collected by census. And so, you know, talking to journalists and other people out there like how much is each firm spending? We actually have very little data and so that, you know, that's before anything else. So now from my B minus, I would say going back to your comment in the last panel, one of the things that is positive is the token cost is falling so rapidly. I mean, it's quite amazing. I was looking at -- I won't name who, but a very large tech firm's use of internally of tokens. You can see it going like that. If you look at expenditures going downwards. So that actually makes me quite positive that that stuff feeds through downstream and actually puts quite substantial downward pressure on inflation services. So on that side, I'm kind of bullish.
[01:47:56] Kevin Warsh: John, do you want to -- do you want to weigh in another former GSB dean did Nobel Prize winning work on the non-tradable sector? I wonder whether that services sector here might be affected by what's happening in AI or am I being -- am I looking for a rainbow that doesn't exist?
[01:48:16] Speaker 10: I think that's a -- I actually think that's a hard question to answer. So let me give an answer that has actually not much to do, in a sense, directly with economics, but more to do with human psychology or philosophy, which is -- just to start with -- which is there are -- I think there's one general question with AI and the labor market and product markets, just the evolution of the economy, which is -- there are certain types of goods and services in the economy that we think of as obtaining from other people. education is one of those things. We think of education as being sort of a fundamentally very human activity. We -- the teachers are human. The students are human. Small groups tend to do particularly well for -- for learning. Health care is very human, by and large. You have a relationship with the physician. You talk to the physician. They show empathy. That actually helps in many ways with health care. There's many other parts of the economy. Those tend to be the parts of the economy where the disinflationary pressures from automation have not hit. Those are the parts of the economy where the costs have continued to rise with wages as opposed to being -- because there hasn't been automation. Those are the parts of the economy. Should those relational parts of the economy be automated? Will they be automated? I think a lot just depends on the way people actually react. Do they -- do they want to be -- do they want to learn from an LLM? Or do they want to learn from another human? Do they -- do they want to be -- have their medical care provided by a very sophisticated LLM that sees them in the office? And there's also regulation and so forth that will limit a lot of the automation. But, you know, could we end up in a world where when you go to the doctor's office, you're talking to open evidence or chat GPT? Yes, there is a world where that happens. And then there would be a lot of deflationary pressure on health care provision because you could automate it and scale it in a way that you can't with human services. It would be a different world. And I don't know that in many ways that's what people want or are ready for right now. You see even just the complexity of trying to get automated drivers to replace human drivers. People are very, very anxious about that and worried about it. It's taken a long time even though probably just statistically you might be safer in a Waymo than in a car that's driven by a human driver. I think there's going to be a long -- I think this is just speculation. I don't know this, no one knows, but I think for a while people will want to be taught by their humans and they'll want to be taken care of by other humans. And so those sectors will continue to be relational sectors of the economy with a lot of good jobs, but also less deflationary pressure from automation. CC.
[01:51:25] Speaker 8: So I think what you're asking really the service sector is very labor intensive. And so, you know, one part of my, you know, question, I hadn't really ever posed it in that way, Kevin, but it's like what are humans then left to do? There's the more manual work building the data centers, but at some point, you know, the machines may be able to do even some of the fine parts of that better. I don't know. So what are humans left to do? I'm going to go back to my earlier question about what are we aiming for? Like what is left for people to do? But two, you know, I don't know. And this goes to the human brain and we don't have a good model of it. And what do humans actually need in terms of warmth and comfort? I don't think kids learn better from a, kids need a human being to interact with to learn, at least today. But maybe we evolved to be different kinds of beings. You know, we are getting into the philosophy and, you know, these metaphysical questions because I don't, there are certain things that humans actually need another human to do. A robot can do it to some extent. You know, I consult the internet. I have a Google MD, so I understand the power of the internet for my, for health information. But I also know what I don't know when I'm consulting that. And again, a human doctor can take a whole bunch more context in, you know, these LLMs are basic, their language, right? So in the end, it's based on one kind of intelligence, whereas a person we touch and we feel and we, you know, we smell. And there's just so much more that goes into those interactions. So I don't know. But I don't know what kind of a world we're sort of driving for. In the short term, I think we're probably safe. But maybe you still have services inflation. I don't know. But I just don't know what really the end game is. Can I stick with this product? Please. To the services question.
[01:53:23] Speaker 9: So actually, I think this question of what humans want in terms of from humans versus from technology is incredibly important. And this is sort of like a hallucination of mine. But but I actually think that the most likely place for us to see productivity increase in services is actually in individual and small businesses. So sole proprietor and small businesses because they're doing the human piece of this and they're spending a lot of their individual human labor on. I'll just call it back office stuff. And to the degree to which intelligent technologies can do this much more effectively probably even than they can and more cost effectively also because the token costs are dry. You know the conventional software for those kinds of things very expensive actually. So if we think about sort of intelligence systems that do this more effectively more cheaply that could actually be a large increase in labor productivity in these small businesses. And I think and this is where this is the hallucination part. So if that's true I think we could be entering a golden age for individual and small business proprietors which historically has been such a driver of growth in the economy. If you look over the you know century and a half time period as opposed to you know the sort of large corporation dominated time period that we've been in for most of our lives. If that really were true I'm very excited.
[01:55:08] Kevin Warsh: So let me just end with a closing comment here first to build on this question of how humans have been underestimated. I believe the following statistic is broadly true. The average three year old learns a language with somewhere between one tenth and one one hundredth the data of the most advanced large language model. Well that means we're pretty impressive. And you can describe ascribe that to anyone you want. And in spite of my best efforts to turn four economists three and a computer scientist into answering narrow questions to the right of the decimal point I should have you know. About what's this mean for productivity and wages and inflation. You see in each of their answers they abstract back. They go to the left of the decimal point. Because whether we are thrilled with it or despondent about it. This is a paradigm shift. It's a paradigm shift in the business of universities. In the business of teaching economics. And in the conduct of everything we do at the central bank. I am grateful for you four for your insight. And also for sharing with the broader community the big questions that we have. With that we're going to take a break of several minutes. People are free I think to go grab a cup of tea. And Mary Daly the president of the San Francisco Fed will lead the next panel. Thank you. Thank you Kevin. You were great. You gave me the comic form. Thank you.
[01:57:03] Speaker ?: Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you.
[02:11:04] Speaker 1: Next part of our panel and actually, Kevin previewed it in his question he asked to the first panel, but they didn't answer, which is past the technologists who are really using this technology for the reasons of being at the cutting edge. Where is it going in the rest of the industries across the United States? Where is it going in companies that seem less technology companies and more mainline companies in the U.S.? Now we know from history, and I know electrification is only one moment in time, but we know from that history that you can't get a productivity boom in a sustainable way until you get diffusion across industries that make up the bulk of what people in the United States do or in the world do. So today we have a perfect panel to start thinking about that. These are businesses that actually aren't in technology per se, some of them a little bit more than others, but not technology businesses, but businesses that are trying to do things. So here's who we have today, and I'm very excited to introduce them. Jim Liefer is the CEO of Ambi Robotics, which develops AI powered robotic systems for logistics operations. You can already hear it, logistics operations. Jim had to once kick me out of their tour of their company because it's so interesting what they're doing, so you'll hear a little bit more about that in a moment. Then we have Michael Senske is Chairman and CEO of Pearson Packaging System, a packaging automation equipment manufacturer. So now you can see we're getting into the meat of what we do in the United States and elsewhere. Eric Wexler is President and CEO of Providence, one of the nation's largest national health care providers or nonprofit health care providers. That's another system. John talked about this in the last panel. He said, you know, health care, are we really going to change health care? So we're going to get some of the answers here. And Zdenia Goodell is chief economist at Indeed. You've all heard about Indeed. She follows job postings and labor demand in real time. And this is going to give us some insights into, as we talked about in the first panel, especially what's happening out there in the labor market and how widespread is the adoption and use of AI. So please join me in welcoming the panel. And then we're going to move to our first question. I'm going to characterize these first questions as the state of adoption today. Where are we right now if we do a kind of a time check on where is AI and how many companies are adopting it? Let's start with Eric. Since we talked about health care last, Eric, I'm going to ask you first. Where is AI delivered measurable outcomes in health care? And my follow up would be, you know, do you see immediate additional gains on the horizon? But where are you? Where are you focused?
[02:13:58] Speaker 11: Well, you know, health care as the last panel talked about is a space where we believe we're going to see very substantial improvement. And I'll start out with three examples. Some of the earliest example of improvement is in radiology images. And so those images when checked by AI can improve accuracy substantially. But the other two places are kind of innovative and interesting. The University of Washington and Microsoft and Providence collaborated on something we call Gigapath. And what Gigapath does is it literally advances the review of tumor specimen so that we can get outcomes of that diagnosis within minutes. People with pancreatic cancer used to have to wait weeks sometimes to get those results. We can now result those in minutes. And that allows treatment to start quicker and really reduce a lot of the anxiety, the substantial anxiety that goes with this. I'll give you a third one. A third one is, and this was referenced in the previous panel, so I wanted to jump up on stage and participate. You know, how do we use artificial intelligence in the interaction with the patient or the clinician? So we created at Providence something called Empathy AI. Now, if we try to train clinicians to give bad news, it's very hard to do person to person. If I'm doing that with you, I look at you and I feel like you might be judgmental with me and my answer while you're training me. But in Empathy AI, we're actually using artificial intelligence to have that conversation and in a non-judgmental way be able to do that type of training. I listened to this myself because I was unsure that I could see AI be empathetic on the other end, but I was taken by it. And in fact, my heartstrings were pulled as I was listening to the response of the agent as it was giving its response to the clinician.
[02:16:21] Speaker 1: Did you get any feedback about whether their empathy improved or their process of talking to patients with negative news?
[02:16:30] Speaker 11: Yes, because we're using this in the space of palliative care and hospice. I see. So this is a place where you're talking to people about end of life. And the physicians and clinicians that are using it are saying, I was able to work my way through it so that when I had that next conversation, I felt like I could do it with better competence, with better compassion, and with accuracy in the discussion versus feeling like I might fail in that discussion.
[02:17:00] Speaker 1: That goes a little bit to answering the question I think John posed in his answer to Kevin's question on services and inflation. It's really not talking about inflation, but it's talking about where are the barriers where people don't want individual agents to interact with them. That sounds like a good use of that. So let's go from -- thank you, Eric -- let's go from healthcare to broader manufacturing. And, Michael, you know, your manufacturing business makes machines that help people do packaging. But you're another place where I think people imagine AI could be helpful but don't really have great examples about where it is helpful. So can you do a little more elaboration on that? And we're talking about right now the state of the world. Where are we today? Yeah, there's a couple of things.
[02:17:43] Speaker 12: I mean, there's -- I would characterize there's two different kind of use cases for us. There's actually on the machines themselves to make them more flexible and responsive to changing customer needs kind of dynamically. But inside the organization is the other -- is the other place. And a lot of it, you know, are the initial -- the initial things are, you know, might seem rather mundane compared to what the panelists earlier were talking about. Just improving response accuracy. Our customers are primarily Fortune 500 companies. They're working 24 hours a day, seven days a week to get us our products. When they have a problem with their machine, they need spare parts, they need repairs. Humans have been in the loop trying to diagnose that for quite a while. And so now a lot of those requests that come in can be handled instead of in hours, in minutes, the response, the spare parts can be identified, you know, on legacy equipment, can be identified, can be shipped, can be tracked and sent to the customer and keep, you know, their downtime at a minimum. So there's a lot of those types of things. It's not really particularly -- it might not sound impressive, but at the end of the day, it has a profound positive impact. The other thing is actually when you start leveraging it for design, you know, actually using it to stress test mechanical designs, using it to -- you know, we just went through a process where we were changing from step logic to state logic, which probably doesn't mean a lot to people here. You know, on our machines, the language of the controls of the machines, the code that controls the machines. And we were able to have a human do the first conversion, look at what that conversion essentially entailed, and then replicate it in mass across all the different product lines, which was -- which was really, you know, kind of saved a lot of time, a lot of effort, and allowed us to really focus on innovation instead of just going through those menial steps.
[02:19:34] Speaker 1: I mean, it's one of the things I think about when you do more mundane things. We all talk about these frontier models, what they're capable of. We just talked about cancer research and radiology, but also just the back office things. But it does free you up. And so then you're more productive, you would say, and you're redeploying that productivity for innovation. Yeah.
[02:19:52] Speaker 12: At the end of the day, like in that particular case on the code, we're actually able to redeploy those controls engineers that would otherwise be, you know, kind of their time would be monopolized doing that, which is not really creating any new value. It's just transitioning from one programming state to another. We're able to essentially now deploy them on things like product development, you know, kind of unique developments, you know, things like that, and actually improve the -- you know, our product lines and our -- you know, and the functionality we're able to offer customers.
[02:20:23] Speaker 1: So Xenia, I'm going to talk to you next and ask you, so these are two industries. You have the benefit of looking across the whole economy. So where do you see AI demands right now that might give us a window into where companies or industries are deploying it?
[02:20:39] Speaker 13: You know, in our research, we found that AI really touches every single job, every single skill out there, every single industry to varying degrees, right? And some impacts are very direct and some are more nuanced. So, for example, in manufacturing or even installation and maintenance services, we see the AI impact because of data center construction or construction jobs being elevated because of that demand. We -- interestingly enough, you know, I feel like when Claude kind of came on the scene, everyone was like, oh, that is the end of the software developer, right? Like, you're going to never hire another software developer ever again. It turns out that was wrong. We see in our data that actually demand for software developers is up 25% year over year. Now, that's off of the lower level compared to pre-pandemic levels because there was the right sizing that happened, but really strong demand there. So, AI is impacting the jobs quite heavily that we thought, you know, would actually go away. And they're not going away. Marketing, finance, all these types of industries are seeing that AI impact as well, just in slightly different ways. So, it is interesting to see. I will point out, though, one thing that's really, I think, really important is that it's very uneven. We're seeing a lot of demand for more senior roles. So, even within industry, within every single sector that you're looking at, it's the director of, for example, communications that will be more demanded than the PR coordinator, for example. So, entry-level jobs are much more impacted by AI in terms of not seeing as much demand there, whereas AI is really causing more demand on the more senior side. And it's, you know, also having the impact that wages pay for that as well. If you have that AI impact or knowledge, you will be able to be compensated for that.
[02:22:34] Speaker 1: Well, I'm going to return to that topic because that's something CeCe referenced about, you know, what are we doing in terms of our human capital? And you do have to wonder with some of the AI hiring and also some of the statistics we're seeing on younger workers, whether this is a type of age-biased technological change. I think that's not a phrase you hear a lot. We think about skill-biased technological change. But I'm going to return to that. I first want to ask, okay, this is where we are so far. Sounds like there's more usage in the periods today than just in the technology sector. So, what's next? And, Jim, I'm going to talk to you about that because when I think about the future, the next one to three years in particular, I always think of you and MB Robotics. Talk to us about what's next. But for those of you who don't know or maybe I'll just level set, what is physical AI? Can you start there because you work in the area of physical AI? Sure.
[02:23:26] Speaker 14: So, the difference is between digital AI and physical AI. So, digital AI really has unlocked knowledge for the world. That's how I think of it. Physical AI is really unlocking productivity potential across the world. And I think of this as like intelligence into action. It's done, physical AI is done with models that understand how the real world works and leverages hardware like a robot arm in order to accomplish that. So, let's go back to digital AI for a moment. All of us in the room for the last several decades have been entering information onto our computers. And all of those clicks are creating data that has made all of the digital AI be the incredible platform that it is today. Conversely, in the physical world, intelligence is actually disappears. As soon as the task is done, that intelligence is gone. As soon as you think about a person, all the billions of times that a person and people have picked up an item, interrogated that item, maybe looked at a label, sorted it, stacked it, put it somewhere. When they walk away, that information is gone. And for physical AI, robots in particular, you can't deploy them without high reliability because if you make mistakes in the physical world with physical AI, it's very costly. So, what we do at Amby, across our fleet, all of our robots, every time we interact with an item, about 150 million times now, we interact with an item, we retain that knowledge because all of the robots are on the platform, on the Amby OS platform. So, the big difference is digital AI, knowledge, physical AI, action.
[02:25:11] Speaker 1: So, one of the things that I learned when I visited your company is you originally trained the models though with simulated data, right? You didn't even go into the physical world because you wouldn't have had a place to start. So, you had to simulate what it would be like and then train the models.
[02:25:26] Speaker 14: Exactly. Our PhDs from Berkeley spent a long time working in sim-to-real models. And so, for our first customer contracts, we relied fully on the sim-to-real data because we are enclosed, we have safety systems and all of those. We keep the robot away from the people who are at these facilities. So, we started out just to give a couple examples here about how much it improved. In our first contracts back in 2021, we were operating in the buildings at 325 pieces per hour. We're now over 500 pieces per hour because we took the sim-to-real, then you do in the real world and then you feed that back in through our loop and then that's how you get to those numbers.
[02:26:12] Speaker 1: So, in terms of where, maybe I will do this health check with you in terms of where we are. In terms of where we are today, would you say we are at the beginning of the maturity curve in physical AI or are we as far along as the digital space? Where do you think we are?
[02:26:29] Speaker 14: I think we are at the very beginning of that curve. And I believe also, as on the digital AI side, it happens quickly. When you get to that part of the hockey stick, it happens quickly. And we all feel that coming. You can just feel the excitement of it coming because we're at that tipping point and it's going faster and faster.
[02:26:51] Speaker 1: So, I'm going to ask a few. Yeah, go ahead, Svenja.
[02:26:54] Speaker 13: I was just going to say, it's really interesting to hear you say that, Jim, because I feel like overall AI is still so very much in its early days, right? If you look at, for example, mentions, overall mentions and job postings of AI, we're around 7% to 8% of all jobs out there. So, it's incredibly early and I think private or leisurely adoption of AI is much further along than even companies are along in terms of adopting AI. But we're starting to see this unfold in stages because I did want to mention that even though AI started within the technical space, two-thirds of AI mentions and job titles now are actually outside of tech completely. So, we've kind of hit that next stage, but overall, it's still incredibly early in terms of where we are with AI adoption across the board.
[02:27:44] Speaker 11: And I just want to pile on because I think the work you're doing is incredibly important. You know, robots in hospitals are cleaning the institution itself, the rooms, the halls, and that's protecting patients from infection. Or in pharmaceutical distribution, as we have a drug that needs to be distributed and ready for a patient to take, a machine, a robot, is making sure accurately, perfectly, that it's the right pharmaceutical. So, you're doing great work.
[02:28:17] Speaker 1: Thank you. I mean, one of the things that I find remarkable about listening to people who are actually running these companies and doing these things is that you think we're in the very early stages of AI, and yet we're already doing quite a lot with it. And it's very hard to match those two things unless you talk to businesses and see where they're deploying it. And really, since you're not technology businesses by and large, I find it a reassuring component of there is deployment out there. Is it everywhere it will need to be? Likely not, but there is deployment out there. Okay, so let's stick on the future a little bit. Michael, what do you think is coming next in terms of -- so you've told us where you are today, but where do you think your business or your customers' businesses are ripe for additional usage of AI?
[02:29:07] Speaker 12: A couple of areas. I mean, again, there's inside the building and then there's -- we do very similar stuff, you know, as you guys do on the machine. So I'll talk about internally. I would say it's really, you know, AI models being able to understand the mechanical capabilities of the existing machine or the robots that we deliver. And then when presented with a new or a unique novel request from a customer, the ability to actually change that, you know, kind of redesign the machine on the fly. Or to come up with different concepts very quickly and even test them, you know, so that we can, you know, just really dramatically reduce our response times. I mean, I think that's one of the things. I think, you know, debugging code, original code. You know, I know original code is being done in a lot of other industries, especially in the tech industry. But there's a lot of capability where we'll be able to identify exactly what it is we want to do. Like, okay, we need to be able to run this product using that hardware. And based on its ability to look at the code we've written previously, I think it's going in that direction very quickly where it'll be able to extrapolate very quickly and write that original code. There will still be a human in the loop checking it, making sure that it works for quite some time, I think. But I think those are the types of use cases that we see kind of internally that are emerging very quickly.
[02:30:33] Speaker 1: So now since you said human in the loop, I do have to pick up and ask, are you on the human on the loop or human in the loop? Because I hadn't really fully understood human on the loop. But that's a level of step back that I'd like to know if any of you have done yet in your business.
[02:30:50] Speaker 11: Well, you know, we're committed to human connection. We don't even like to think of it as human in the loop. It is the human. And artificial intelligence is supporting the work that we're doing to try to enhance our efforts and make sure we're delivering good, compassionate, whole person care. And so this concept of human in the loop, we kind of feel in our space as it's foreign. And we need to continue to own and control the work that we do for the people that we serve.
[02:31:22] Speaker 1: Dan would really like this from the previous panel. He talked about accountability and how do you have accountability models. So I think that's perhaps an industry where you're starting to see that. So one thing that I want to turn to is, you know, Chad Jones, who's on the task force, has a paper and research on the weak link that might slow the process of full transformation on AI because you always have a weak link. So I want to ask each of you, where do you think the weak link is? And Eric, I'll start with you since you last spoke. Where's the weak link in healthcare?
[02:31:54] Speaker 11: Well, we're part of the weak link and maybe purposely. A lot of the panelists and Kevin talked about diffusion. And sometimes I refer to that as saturation because it's a less technical term. But saturation of the technology. And so for our organization, we're being really thoughtful and not trying to implement 10 or 15 things at once. So some of the potential weak link is our commitment to deploying the right type of technology but getting that full saturation and diffusion. But I'll give you a second one. You know, in healthcare, there's a lot of regulation. There's the global regulation as well for how technology is created. But let's just think about here in the United States within our own borders. If one state creates a regulation that restricts the use of artificial intelligence or super intelligence in one way and another state doesn't, what could happen is that the improvement of care and its ability to create high reliability gets reduced. Because then you have a fracturing of the quality of care that we need within the communities that we serve. So I'm worried about that. We've seen a little bit of that regulation rolling out. We are in seven different states. So it's almost like sometimes being in seven different countries.
[02:33:16] Speaker 1: And that's a really useful point, I think, to make is that the AI is only good as the data inputs to it. And if you have a fractured delivery system or a fractured training system, then you will have less production from it. That's right. Yep. Michael, what's your weak link, you think, in your deployment? You know, one thing that I wondered about when I saw Chad give this work before is does just the speed of change, does that create itself a slowing of transformation because customers can't keep up? So I was particularly interested in whether your customers want to buy the newest machines if they think the newest machines are going to change tomorrow. If it's really going to be 4:30, like CeCe said earlier.
[02:33:57] Speaker 12: Yeah, that's an interesting question because historically when you look at industrial automation and you go way back, the way you achieved high out, high reliable levels of output was you forced your customers to provide, to present their product to you in a very homogenous, very ordered way. They had to take responsibility for that. I think that what's really going to change is that our customers still, even though we've developed much more flexible systems, we use AI-driven vision on our robots, things like that to adapt, and they're much more flexible than they were, but customers have been trained over nearly a century that, you know, that there are certain things they're not able to do, that they have to kind of move toward homogeneity. As customers' marketing departments understand that some OEMs are going to be adopting this technology and that they can create almost mass customization on their packaging floors, I think that's going to really dramatically then lead to our customers are going to want to make sure they have a platform that they can either retrofit AI to, or that they have a system that's delivered out of the box originally with it. But it's interesting because customers aren't quite catching on to it yet. I see. But I think as soon as they really figure out what's possible and start pushing the OEMs to deliver that level of flexibility, it'll really start to take off very quickly.
[02:35:26] Speaker 1: And then your future is bright. Yeah, yeah. So, Jim, what's the weak link in physical AI? What's the thing that slows this process down?
[02:35:36] Speaker 14: Well, for us, we are perpetually worried about just getting more robots out there, right? That's our mission is, like, just keep getting them out there. So, I would say adoptability. That is the weak link, meaning the customer, our customer type retailers and shippers are incredibly busy every day. Like, there is an enormous amount of volume coming into these buildings. There's billions of packages. They're moving them to the customer's home. It is very difficult for them to take just even a moment to stop and say, oh, this actually can be a better piece of technology. It will increase my throughput. It's better and more accuracy. So, what we are doing at Amby is we deploy our systems into brownfield locations mostly. And then we find that one of the most important parts to the customer is throughput per square foot. So, if our systems can fit into square footage that's relatively small in almost a million square foot building, that throughput increases and that's what causes them to take more of them. There's the one other factor is the employee satisfaction part. They care about these employees, giving them jobs that they want to take and want to keep. So, they want to retain those employees and we can give them a better job alongside of the robot.
[02:36:56] Speaker 1: So, when your machines show up and your robots show up, you don't see a workforce that gets really nervous. You see a workforce that is a little nervous. How do you think through that? I know that wouldn't be your job, but that is a question I think many firms have.
[02:37:11] Speaker 14: My sense is that people are a little nervous when they see the fencing go up and there's going to be a robot there. Right after we go live though, what we have found from these people is that from the employees is that if you think about what that job used to be, I, the employee had to pick up every parcel, look at the label, figure out where to take it, sometimes throw it there. This is very like arduous labor intensive kind of work and now the robot shows up and just imagine you roll a bin into it with a bunch of packages, hundreds of packages. The robot takes over and does all of the mundane repetitive work and now the employee proudly stands next to it and can handle exceptions and work alongside, manage the robot instead of managing the parcels. So we, we see that really resonating with our customers employees. I've heard, I've heard employees and customer sites call it my robot. Um, and I've, I've seen them take a selfie in front of the robot to perhaps share with their kids to proudly say like mom has a great job or dad has a great job. Uh, and, and that's what we see when we deploy the robots.
[02:38:25] Speaker 11: And I just want to say that what he's describing is spot on true. So where we have robots in our pharmacies that I referenced earlier, when I go into those pharmacies, one of the first things that our caregivers do, our pharmacists do is take me over to the robot. Let me show you how this works. And they feel very proud about it. And they feel very proud about it and it's taking the space of administrative burden that is involved in the care delivery process. It's reducing that down so that they can focus on what they need to given the demands that they have to do their jobs. It's very effective.
[02:38:58] Speaker 1: So one of the things, and so Eric and I'm turning to you, Sven, you so get, so get ready because I want to ask you a broader question about, about weak links. The, the thing that I've wondered about is whether, you know, you see surveys, especially after the pandemic, that healthcare workers just weren't very satisfied with their jobs because, and the burden for insurance, you know, writing down notes has gone up. Do you see an improvement in your teams when they use AI to sort of de, make their job a little less tedious and a little more patient oriented? Yes.
[02:39:31] Speaker 11: Let me start with the clinician. So, you know. You're enthusiastic about healthcare.
[02:39:35] Speaker 1: Can you tell? Yeah.
[02:39:37] Speaker 11: Sorry. I'll take it back a couple of notches.
[02:39:39] Speaker 1: No, go ahead. Please. I want to hear it. Please. I think it's a good thing to be enthusiastic about your work.
[02:39:45] Speaker 11: Well, I think we should be enthusiastic about how this, I'm glad you said that because I think that the way this technology can be used to actually improve care for all of us is really essential right now. So my enthusiasm is busting because I will tell you in, I've been in this space for 35 years, not once in all that time for all the things I've ever talked to physicians about have they said something we've given to them has been life changing. So let me give you an example. Taking a step back to where we are post COVID, physicians and clinicians in the United States are enduring and endemic of burnout. They just, we're aging, we need more care. They don't have the time to take care of patients in the way that they used to. So one of the early types of technology from an artificial intelligence perspective that help physicians is ambient technology. And maybe some of you have experienced it. You walk in to see your physician. They're going to give you an exam and they tell you that they're putting their smartphone on the table and it is going to be listening to the conversation. And what it does is it listens to that conversation. It extracts, for example, my tendency to talk about walking my two dogs. It takes the unimportant things out and it only takes the important things and it loads that into the electronic medical record. So why does that make a difference? Because traditionally what we have seen is our physicians staring at a screen while we're talking to them, pounding away on a keyboard. And what physicians have said is this is life changing. I've now been able to save two to four hours of time a day. And you know where those two to four hours of time are? At home when they should be renewing and recovering. I'll give you a second example. We've created something called Prav Aria. And what it does is every morning we all wake up. We have email in our inboxes. And we might be a little stressed about it. But can you imagine being a physician and you have a hundred of these in basket messages, as we call it. And you have to go off to see your patients. They need you. So what is in that hundred? Some of them may not be that important. But maybe in basket or email 67 is somebody saying, you know, my jaw's a little raw. My shoulder's been bothering me. That email could be somebody having a heart attack. And it shouldn't be in an email. So what Prav Aria does is it goes in, it extracts out the things that are not urgent. Like we send thank you notes to our physicians. We take those, we put it into a gratitude box. Physicians go in there and look at it anytime they want, especially on a bad day. Okay, then we take the, you know, I need a new pharmacist, I need a new prescription. That goes off to the MA or the APP. Or, hey, I have something urgent I need to talk to you about. And then we prioritize that right to the top. We've been able to remove three, excuse me, 1.3 million messages in a year and a half of this work. It needs to be many, many more. But it is significant improvement in the quality of delivering care.
[02:42:57] Speaker 1: That's really, I mean, I experienced that myself. I will say, I will tell a joke on myself here. So I, you know, my physician uses the listening device and it does all that. And I kept asking her about how that works. And how much, how many hours it saved her and what she was doing with those hours. And so it said, she has a, she talks a lot about AI and the technology. So, you know, there it is. I'm, I'm transparent. Now it's recording me. That's my disclosure for the day. Okay. Svenja, I want to ask about weak links in the sense that, you know, one of the things that people worry a lot about is workforce. And are we training, helping, lifting our workforce so that they're ready for a changing environment? You know, Patrick Collison was on earlier. He, on a panel a couple of years ago, he said to the group, I'm hiring as many people as I'm, I have a lot of churn. I'm hiring more people than I'm letting go. But the reason is because they're different people. And that struck me as one. It's good for overall job growth. But what about the people who are no longer ready for the skill sets that they need to have? What are you seeing out there? Are you seeing evidence of mismatch or employers looking for things that there's only a few of? Yeah, absolutely.
[02:44:14] Speaker 13: I think skills mismatch is already happening. I think it will become an even bigger issue in the future. I mean, we talk a lot about AI, obviously, today. But that's not the only thing that's happening right now, right? We are an aging population. We're going to see demographic impacts. And those are not felt evenly across the economy either. So some sectors have a much older workforce and you're going to see higher retirement rates and government and education roles, for example, than you will on the information side. And so as the population ages, we're going to need certain people with certain skills in one pocket. But it turns out our education system will have trained them over here for a whole bunch of different skills. And then there's AI, of course. And then we're figuring out, is AI going to be augmenting or replacing? And we probably don't fully know quite yet, especially as it's going to become better over time, right? And so skills mismatch is going to -- is already here and it's going to become even worse in the future. And I think employers have to really think about this as they think about workforce planning, right? What's -- who do I need to hire for the jobs of tomorrow in order to make all of this work? And I think the employers we talked to at Indeed certainly try to always think about this. And the number one question is who should I hire? What skills are really important to hire for? And it's interesting when we look at our data, the world isn't getting easier in some sense. We're seeing more and more skills being demanded. So if you, for example, are hiring a nurse, all the typical care and nursing skills are included, plus a bunch of technology skills that you now have to possess. Same software development, for example, you have to have all the basic -- all the basic skills. Although a lot more people are paying attention to shipping code versus actually coding and coding languages themselves. But then, you know, you used to have to learn all of these things. And also, you have to know the AI skill set on top of that. So it's getting trickier along the way. And then you have to figure out, okay, who do I hire in the, you know, for that job of tomorrow? Particularly if, you know, if I used to hire by perhaps a degree or, you know, I used to always hire from the school or I used to hire someone that had this job title, what if that person doesn't exist anymore tomorrow? Who do I hire then? So you have to really do skills-based hiring and look for that exact set of skills. And then we're finding that more human skills are becoming more and more important and more in demand. I mean, we've talked about human on the loop, in the loop, human-centric itself. Well, we heard about creativity in the previous panel that creativity was in high demand.
[02:47:03] Speaker 1: I think the first two panels we heard that.
[02:47:05] Speaker 13: Absolutely. So really paying attention to that. And, you know, I don't think there's a fully right answer that holds everything together here. But having that skills mismatch is going to be -- keeping an eye on that is really important. And then I'll say that it's, of course, really hard to actually take someone from this area and say, I don't need you working over here. I actually need you over here, right? What if you have someone who is trained in finance but you're noticing that I actually really need someone in healthcare? It's incredibly difficult to then just reskill, upskill that person over here. There's a lot of education and training and certifications that go into that. So it can be really tricky moving that -- having that very fluid labor market, which I think we need in order to push innovation. That's going to be really tricky. And we're all going to have to work towards that.
[02:48:00] Speaker 1: So I want to ask employers a little bit about that. Jim, I'll start with you because it looked like you wanted to say something. And then I also want to go to Michael on this. Like, who are you hiring? And who do you see firms you work with hiring? And importantly, one of the things we've been asking ourselves at the Federal Reserve is, are employers waiting until they know more before they hire? You know, we're in this low hiring, low firing environment. And there are probably a myriad of reasons that's going on. But one of them that we hear is, we just don't know who we need yet and we're going to wait until we do.
[02:48:31] Speaker 14: Yeah. Well, for us, it's about execution capacity. So we are hiring people who can help us with manufacturing, skilled installation people in the field, field service engineers. We hire a lot of those. And then engineers at large within our team and all of those types. And I want to say a little bit about in the loop or on the loop. Yeah, I'd love to hear more about that.
[02:48:56] Speaker 1: That's a new phrase for me.
[02:48:58] Speaker 14: For me too, because it used to mean something different for me. When other tech companies would talk about someone in the loop, that meant like a pilot that was helping a robot do something. And we don't do that, Andy. So we're not in or on the loop in that way. But we have deployed, and this year has been significant change. We have deployed our agentic robotic harness, we call it. So we have agents that are looking at interrogating real time production data and looking for areas where something could be improved. And then those agents will write code, test it, safety check it. And this is the harness part. Not release it, but provide it to one of our engineers. And then the engineering team can look at, let's do some A/B testing in the field. And so we do that. We proved it recently with a difficult package type that we couldn't figure out how to handle. That one customer had a certain kind of package. And then the agentic side of it took over and did that in days instead of months. So that was incredible. The other part is we use these agents in production for support operations and for even maintenance issues that go on. So agents, call them bots, will look at an issue that has happened. Some robot in the field has done something. And then the triage bots will come in and suggest how to solve that problem. And we don't let those agents release the robot back into production. But it will do all the steps and then it will suggest, okay, now you can do the safety release or go check this thing before you do it. It's pretty incredible. Quite different.
[02:50:34] Speaker 1: Yeah, that's quite different. When you hire field engineers and things, do they have specific skills that are AI based? Or are they, I think if you said field engineer to most people, they'd say, oh, I kind of know what that means. But it might mean something different in your work.
[02:50:48] Speaker 14: In our work, they are a blend of mechanical engineers, electrical engineers, problem solvers. And then finally, very good at interacting directly with the customer. We care about that human interaction. So we want to make sure that if our field support engineers are out there, they're interacting with the employees and the managers and the maintenance crews out there at the customer site.
[02:51:16] Speaker 13: It sounds like a super easy job to fill that up. Oh, yeah. Yeah.
[02:51:20] Speaker 14: But it sounds like what you were saying. There's a posting online right now you can go.
[02:51:22] Speaker 1: But you have to have electrical engineering, you know, mechanical engineering. Mechanical. Then you have to be good with people. Yeah. And you have to be a good problem solver and probably a little creative. So, you know, I think this is making, as Senia pointed out, jobs just more complicated because people can need a variety of skills at all to go to that. And I think for the questions that CeCe and others asked during that last panel, you know, how do we train our students to have that array of skills? And if you're even an employer, how do you train your entry level team to do the things you eventually want them to do? And it's creativity, right? Creativity. A lot of creativity. So, Michael, who are you hiring?
[02:51:59] Speaker 12: You know, so it's interesting. I would say our hiring practices are very similar to what they were, you know, over the last, you know, decade. I will say that there are certain positions that aren't nearly as affected by AI. And I would say it's our electromechanical assemblers, the people who build the machines on the floor, the remote service technicians who are essentially like the electromechanical assemblers. They're just not building things. They're maintaining them for our customers out in the field. That really hasn't changed dramatically. I would say that we're still continuing to hire engineers, but we're asking the question before we hire, is there a way we could, like, avoid that hire by leveraging some of the AI-related tools that are out there? Or if we do use AI, can we actually still hire that engineer but, again, focus them, let's say, on product development or refining an existing product, you know, really trying to be creative, you know, with original thought of moving the product line forward and creating more value for customers? So, I wouldn't say it's really changing now, but we are being more deliberate about asking that question before just, you know, rubber stamping, like, I need more capacity. It's like, well, can we address that with AI and then maybe redeploy people to do other things?
[02:53:15] Speaker 1: Yeah, that's a really important point, I think, is that capacity can be met a couple of ways. You can put a capitalist solution to it or a technology solution or a person solution, and you're being much more picky about when you go to the person solution. Yep. Eric, anything to add on terms of who you might be, how you're shaping your hiring by using AI?
[02:53:33] Speaker 11: Yeah, well, first I want to step back to a statistic that probably you and I look at, but differently, and that is the birth rate in this country.
[02:53:40] Speaker 1: We do look at it differently, I'm sure. Sure.
[02:53:43] Speaker 11: But we do both look at it. Yeah, we both look at it. So, but if you go back to 2007, I think that's when the highest rate of births we had in 20 years. It's now about 23% lower than it was then. In healthcare, we're seeing significant growth in demand. But we can't, we have to bend the curve down. We've got to create a deflationary measure that allows this country to afford healthcare. And it's not just healthcare, but it's also health insurance. Both of those factors require us to be thoughtful about the number of people that we have doing the work that we're doing. So, what we're trying to do is we deploy AI. An example of that is we're using AI for accounting work now. It's doing journal entries and things that, you know, we need our accountants not to be worried about. We need them to do the higher licensed work and then be able to close the books quicker. So, we've been able to reduce that demand and deploy those folks to do more important work. As we continue to hire, what we're trying to really do is scale people up from a skill perspective. And allow them to do jobs that are more, you know, required of human intelligence. And let the more menial work be relied on in the space of where artificial intelligence can jump in. You know, I'll just, I'll add one more quick example. You know, we've been advocates of returning our workforce to in-person work. We have over 100,000 people that come to our hospitals and clinics every day. We have about 10,000 people that work virtually. Those are the people that we need to recreate those human connection with that are working out in the field. So that they don't wind up being replaced by artificial intelligence for menial tasks. Not all of which, by the way, are being done out in the field virtually. Those are really important jobs. But the human connection piece of this is really critical to their future.
[02:55:43] Speaker 13: You know, it's interesting because we see that senior level jobs are up roughly 20% since the beginning of 2025. During the same time period, mid and entry level jobs are down 4%. To me, that's a real pipeline problem, right? So as employers think about who they're going to hire next, if you don't train the incoming generation, we're going to run out of people that have the, you know, three to five years worth of experience somewhere down the line. And so it brings up a really interesting problem around how do we translate schooling on the work training, on the job training, and how do we fix the pipeline issue that we have? And then ideally, you figure out who to actually hire and make more efficient in the job, right?
[02:56:31] Speaker 1: Well, I think that pipeline issue is something that, you know, needs to be elevated as something to think about. Because you can't be three to five years experience if you never had the first three years experience. Yeah. And that ultimately, you know, right now it doesn't feel pressing. But it will be pressing once we get past this original transition and we go to the next. So we have limited time left on the panel, so I have one more question. I have three last questions, so I'm trying to think about which one to ask. But I'm going to ask this one. So the question that I think, you know, as Kevin said earlier, the talking about the -- I actually love talking about philosophy because I have a philosophy and an econ degree. And so I was very pleased to see my fellow economists talk about philosophy. But the real question for us, because we do have day jobs, price stability, full employment, and we have to know how much productivity we're going to get, whether we're going to start to see this boom that we've been wanting to see and talking about that we think is possible eventually. We're going to see it in the next one to three years. Or if your optimism goes -- is high, but it goes to ten years. So I'm going to go down the line. Jim, I'm going to start with you. Do you think we're going to see the type of transformation that the first panel talked about that's possible in the next one to three years? Or do you think it pushes out to something that looks more like five to ten?
[02:57:50] Speaker 14: I'll split the difference, and to be quick, I'll say three to five. Three to five, okay. I think three to five is reasonable. Three to five. And, again, things will change quickly in the hockey stick.
[02:58:00] Speaker 12: You know, I would say, you know, if I look at the different functions, I would say in several GNA functions, you know, that were referenced earlier, I think we're going to see a dramatic increase. We're starting to see increases that are in the 50, 60, 70% increase, you know, sort of numbers. But I would say across the board, including the highly technical personnel we have, it's probably closer to that three to five range before we fully realize the benefit.
[02:58:26] Speaker 11: Okay, I'm going to stick with my characteristic of optimism. We're seeing it now. So, you know, I talked about physicians with ambient technology and in-basket reduction. And so our goal is to really revitalize the practice of medicine and the care delivery. And we're starting to see that. Our accountants are happier because they're not having to do those journal entries. So we're starting to see that now. We're not seeing it specifically in clinical delivery of care. Those are still going to be, you know, person-to-person interactions, which I'm very pleased about. Great.
[02:58:59] Speaker 13: I mean, it depends, right? Any good economist.
[02:59:03] Speaker 1: You're truly an economist.
[02:59:05] Speaker 13: I'll save the meteorologist jokes on forecasting the future here. But I think it really depends on the sector, right? Some sectors you'll see you already are seeing partial impacts today. Not so much on the productivity side quite yet, but we're probably almost there. In some other areas, it will take years to fully get there. So I'm more cautious, and I would put my money on several, several years here.
[02:59:32] Speaker 1: Well, I don't know how Kevin will score us on your answers, but I'll give you an A. So thanks to our panelists here. We're going to have a quick five-minute break, and then I think we'll have our last panel for the day. Thank you very much.
[02:59:45] Speaker ?: Appreciate it. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you.
[03:09:46] Kevin Warsh: Welcome back. This is session four of the Productivity and Jobs Task Force Conference. We're going to end with two old friends, distinguished business builders, executives.
[03:10:27] Speaker 15: Though we are not old.
[03:10:30] Kevin Warsh: We met long before the secretary was Secretary Raimondo, she was just Gina. Mark and I met long before that when we were school children. Let me just do a brief introduction. What we're going to do in the chat is to ask ourselves the question, where do we go from here? We've heard a bit from inside of companies large and small, some at the frontier of putting this new technology to work, some finding their way. We also heard from economists with a breadth of views about timing, scope, magnitude of this change. Now we're going to hear from two old friends and colleagues. Somewhat different perspectives, not as big a difference, I think, as many would suggest. Let me do a couple of brief introductions. First, Secretary Raimondo, she was Secretary of Commerce in the Biden administration, overseeing the CHIPS program investments in America's semiconductor manufacturing and led many of the initiatives in around technology. Prior to that, Gina was the 75th governor of Rhode Island and prior to that, the treasurer of Rhode Island, which is where I first came to meet her for her fine work there. She's now the CEO of Raise U.S., a nonpartisan nonprofit that works with governors, employers, and trading providers to help American workers successfully make this transition. And Marc Andreessen is one of her supporters in her new effort. Marc is a co-founder and general partner of what I would describe as perhaps the most influential venture firm in the world. He was a student at the University of Illinois. He co-creator of Mosaic, went on to co-found Netscape, created a company before the cloud was anything at all, created a company in and around the cloud. I think he might say he got there a little too soon. He serves on the board of META, is an old friend, and more important than all of those designations, he is one of the co-leads of the Federal Reserve's Jobs and Productivity Task Force, which in my mind goes to the top of his resume. So Gina, Marc, welcome. It's an honor to have you with us. Before I go into some detailed questions I have for either, for both of you, let me start with you, Gina. Where do we go from here? You know what a transformative moment this is, I think you've taken this moment seriously with concerns about what it means for the US labor force. Where do you think we go from here?
[03:12:58] Speaker 15: Yeah. Well, first of all, thank you for having me and I think you're doing a terrific job. You guys can hear me. So we had an election recently, I'm sure many of you saw that. But in the election, apart from the democratic landslide, you saw many of the data center, flavors of data center moratorium. The vast majority passed. And I actually think that's a very troubling sign. So where I think we need to go from here, so why did they pass? I think they passed because people are afraid. When the average American hears AI, they hear, I'll lose my job, my kids will lose their job, my children's mental health, national security, some version of that. And so when people are afraid, they say stop. Except that would be, I think, very bad for America, and for them, actually. So where we, I think, have to go from here is to figure out how do we address people's legitimate concerns and go further and innovate. I gave a speech the other day, I said, Americans, our companies are some of the very best in the world. The most innovative, the most dynamic. We lead the world in AI. That's where we want to be. Where do we go from here? That's where we want to be. But in order to do that, we have to figure out how to bring folks along with us. Because I think a very bad outcome would be a set of stifling regulations that stifles growth. And so somehow we have to get ourselves to a place of growing the pot, growth. And, but, you know, dividing it fairly and making sure Americans can see themselves in this AI future.
[03:14:58] Kevin Warsh: So Mark, let me turn it to you. I've heard you say that you think in and around this new technology wave, both the utopians and dystopians are going to be gravely disappointed. So where do we go from here?
[03:15:12] Speaker 16: Yeah, so let me start with, it's great to be here. It's a privilege to serve on the task force. And I do want to briefly brag on my colleague, Chad, who said, he said, he currently works for a company that's in the process of going public. And so he's in the quiet period. And he said, whatever you do, do not call on me from the stage. So I just wanted to highlight that he's sitting right over there. And then our third member, Asha, who's a fantastic operating executive at Microsoft, who couldn't be here today for personal reasons. But it's great. And then I also just wanted to note, and I guess I should be happy about this, that none of the panels today, including ours so far, have had an AI bot as a panelist. Is -- was that considered and dismissed? Or do we have another year's worth of panel appearances in front of us before we too are cruelly --
[03:16:01] Kevin Warsh: Taking it as a rhetorical question. I get to ask the questions here, Mark.
[03:16:04] Speaker 16: Okay, sorry. Okay, sorry. Okay, sorry. Yeah, so look, I mean, this is -- you know, these have been fantastic discussions. And I think a lot of the -- a lot of the, you know, twists and turns, I think, of the different -- of the debate around, you know, economic change and so forth, and social change that flow from AI have, you know, have been discussed by the very smart people who preceded us. Yeah, so -- yeah, my line has been both utopians and the doomers are going to be -- are going to be disappointed, specifically disappointed by the scope and magnitude of the real world impact, and specifically the economic impact. And I think the reason for that -- the reason for that is basically as follows, which is the world is a very complicated and messy place. And just because there's a new technology and just because it looks like there's obvious ways to deploy it and things that should happen does not mean that any of that actually happens. And in fact, there are historical precedents for new technologies that have simply been frozen in their tracks. And, you know, the big one in our lifetimes has been civilian nuclear power, you know, which at one point was, you know, was the panacea for U.S. energy, you know, global energy growth and, by the way, was a, you know, zero-emission carbon-free technology that in retrospect would have been, you know, tremendously beneficial, you know, even from the environmental standpoint. And, you know, you know, it was basically killed. And by the way, you know, the remnants of it are being killed both in the U.S. and Europe right now. So, you know, so these things don't have to get deployed. Sometimes they're simply stopped. And then if they do get deployed, they can take a very long time. And the reason is because they involve really radically changing how things operate, how organizations operate. A reference was made earlier to the now-famous Dynamo paper in the economics literature, which I think is 100% correct. And so I just think, like, the world is really complicated and messy. And so, you know, you've already seen this morning. You're going to have a set of companies that are going to race ahead on this. And, you know, I think get significant competitive advantage versus their competitors. And I think, frankly, you'll have a lot of existing incumbent, you know, both companies and institutions that, you know, it'll just take a long time. And then, look, you've got other parts of the world. You know, Patrick, you know, Patrick this morning from his vantage point as having been born in Europe, you know, observed it. But you have other parts of the world that are absolutely determined to not have any of this happen. And if anything, to shut it down and kill it. So, I think it's actually going to be quite a messy period.
[03:18:14] Kevin Warsh: So, time seems to be the big dimension. In some sense, Gina, central bankers, we're in the business of buying and selling time. That's what we do with interest rates. My sense, but I want you to disabuse me of this, part of the reason of your concern is it looks like time is accelerating. It looks like we might be having a long period, a period that would normally be creative destruction that would take a decade or two and might be happening faster. Is it that time element that's made you say, we need to find a way to get through this so that we both win with the new technologies and we haven't lost the society along the way?
[03:18:50] Speaker 15: Part of it is time, but part of it is honestly less about the technology and more about America's workforce system. So, much of what we're trying to do, you know, the chair said we're working with governors. I mean, we have, we have a workforce system that needs to be fixed, period. So, if we can use the occasion of AI and everybody's focus on it to do the big changes that need to be made, that's what I'd like to do. So, what am I talking about? You know, by the way, I agree with Mark. About a year, a year and a half ago, or a year ago, when I started to talk about Raise Us, I spoke with one of the founders of one of the big labs, who at that time was saying there's not going to be any jobs for humans in the next couple of years. And I said, I don't agree with that. And I said, don't forget, just because you can do something in a lab doesn't mean it will be automatically proliferated through the economy like that. There is friction to diffusion, and it takes time. But in any event, yes, it's the time, but it's also, look, we spent hundreds of billions of dollars last year subsidizing higher education. 35-plus percent of people who go to four-year college in this country don't graduate. Like, that is not a system that's working by and large. Those are people who wind up with no degree and a lot of debt. If you want to do something that's not four-year college, there aren't that many fantastic options, relatively. I was a governor. The way governors invest in workforce or community college or higher ed, it's just attendance. We incentivize attendance. You show up, they get paid with their Pell Grant. You show up, they get paid with et cetera. Not, you know, outcomes. Unemployment insurance created 100 years ago in the Great Depression and literally hasn't been reformed since. So there's so many examples. So what I'm worried about is this change is coming. I personally view the transition to AI just as that, a transit. And I think there will be jobs that go away. You know, there's -- you would know better than me. 10, 12, 13 million, like, administrative jobs. Benefits verification people, back office folks. Those jobs aren't -- are either going to go away or be different. That is innovation. So I think there's a transition. We are here now. We will hopefully get to an agentic economy. And there will be new jobs and new businesses and hopefully a lot of new startups. What I want to avoid is a spike in unemployment from here because some jobs go away before the new ones are created. And what I'm saying is the workforce system, support system, safety net for unemployed workers, incentives for companies to redeploy, it just doesn't exist. So I'd like to get it ready. By the way, so that we can flourish in AI. Because if we don't do it and we have double-digit unemployment, that's when you -- that's when policymakers do things that aren't necessarily long-term the right move.
[03:22:20] Kevin Warsh: Mark, you -- let me build on Gina's comment. You've said, I think publicly, that the labor displacement story is not true. That the employment shifts won't come anywhere near the people that some of the doomers would have us believe. Is that because we society won't let it? Is that because the politics are changing around it? Is it because companies can't adopt quickly enough? Help us understand why you're seemingly less concerned about labor market disruption in the near term?
[03:22:55] Speaker 16: Yeah, and by the way, look, I agree with Gina's objectives completely and I'm, you know, very proud to be, you know, to help with her efforts. So I don't mean in any way to disagree with what she said. I would just say, look, it takes 900 hours of professional training to become certified as a hairdresser in the state of California. It's illegal to be a hairdresser in the state of California if you're not professionally certified as a hairdresser. Through the 900 hours of training. The machine cannot become a doctor. You cannot get admitted. You cannot get an M.D. You cannot prescribe medication. The machine cannot get admitted to the bar. The machine cannot get hired by the government. The machine cannot get civil service protections. Cannot belong to a union, private sector union, public sector union. Like our, like, to the extent our economy is a free market economy, then maybe you could see, you know, the sort of, let's say extremely rapid displacement scenarios people are talking about. But like we've mostly outlawed the kind of job transition that I think the, again, sort of both the utopians and the doomers talk about. Just like large, large swaths of our economy are basically controlled. I mean, they're either flat out monopoly. K through 12 education. AI cannot become a public school teacher. I mean, ask Randy Weingarten, right? Like, you know, there's no chance. So we've effectively, like, outlawed that kind of economic dynamism across much of our economy. And then mechanically, and I'll just give you a preview of the work that we're doing on our task force. So my contribution to the task force is going to be to make, and the report we're going to do is going to make the following case, which is if you disaggregate sectors of the economy, you classify them roughly into two categories, which what you might call the sort of very fast sectors of the economy, that are economies that can have and do have rapid technological change, rapid economic change, rapid job turnover. You know, those are sectors. Consumer electronics, you know, media and entertainment are examples of that. You know, many sort of, you know, commodity goods. You know, those sectors can have that, but what those sectors all have in common is they have rapidly falling prices. And then you have these other sectors of the economy that I would call the slow sectors, which is health care, education, housing, and then sort of the combination of law and government. You know, sort of legal and governmental administration. Those sectors of the economy, like, are not legally allowed to have rapid change. Like, literally not. And in fact, those sectors of the economy are typically characterized by extensive subsidization, restriction of supply. So cartels, right, where you literally can't light up new supply. You can't build new housing. You can't start new universities. You can't create new hospitals. You can't, you know, whatever. You can't, you know, you can't create lots of new doctors. They have extensive subsidization of demand, which is the government that, you know, the voters get outraged that these things cost so much. And so the government steps in and does federal student loan programs and K through 12 education funding and housing subsidy. You know, I guess we're up for another big new housing subsidy in California, you know, in the face of continued restricted supply. So, of course, prices can only go up. And then extensive government involvement, if not outright government monopoly control of those sectors. And then in those sectors, prices are rapidly rising, right? And, of course, every American experiences this because, you know, this is kind of, I think, the core fundamental maybe political dynamic. You know, right now is just basically every American is basically being told, it's like, good news, your iPhone is getting cheaper, you know. The flat screen TV on your wall is getting cheaper. The bad news is, you know, the college education is going to cost a million bucks. You know, healthcare is becoming unaffordable. And, you know, good luck buying a house, like it's not, you know, it's no longer possible. And then if you think mechanically what happens is, as the fast sectors, as their prices crash, they basically exit out of GDP, right? They're, you know, they're not built. By the way, this has happened before. Agriculture used to be 97% of the American economy. It's now 3%. Food prices crashed dramatically. Food is not, you know, it's up until recently, food was not a significant part of the average American's budget relative to, you know, healthcare, education, and housing. But the slow sectors are growing. Healthcare is already a fifth of U.S. GDP. And I don't see why it doesn't go to a fourth, a third, and then half, and then ultimately 80%. And healthcare, like you literally legally cannot have technological transformation in most of healthcare. And so I think we have a political economy problem and a, I don't know, like a governance, legal, governmental political problem that is going to prevent the kind of radical transformation that people see coming. And again, if I would word it as a concern, you know, you might interpret this as good news because it means the AI revolution is not going to, you know, destroy all the jobs or kill everybody. You could also view this as very negative, which is productivity growth just simply may not grow very much. Economic growth just simply may not grow very much. We may just continue to stagnate.
[03:27:24] Kevin Warsh: So, Gina, let me come back to you. So, Mark is showing, is describing a divergence between two parts of the economy. Call it regulated versus less regulated or unregulated. In economics, as you know, we think about inflation products versus services. We look at the trends of product prices which had gone in one direction in the last couple of generations versus services. Let me ask you about another divergence which you reference, the United States versus the rest of the world. How well do you think we are positioned? What do you think the future holds? Where do we go from here with respect to U.S. economic growth and potential growth versus the trading partners that you know well from your commerce days?
[03:28:09] Speaker 15: Yeah. Is this better? Can you hear me? Yeah. So, so much in there. Let me say this. I think that A.I. provides a fantastic experience for America, opportunity for America to draw the world towards us. So, if you believe, as I do, that we are in a strategic competition with China. China is not our friend. So, whatever words you want to use, they are not our friend. Then, they are too big, my view. They are too big, too determined, too rich for us to compete and win on our own. We must have allies. And, so then you, this is what I think, Korea, Japan, much of Southeast Asia, India, Europe. Okay, what do we have that these countries want? Many things. But one thing is compute. Another thing is our models. So, we could say to the rest of the world, no, you can't have this stuff. Or, you can have a little bit of it. Or, we could say, we could show global leadership and set the rules of the road for what we think of as, you know, national security safety, cyber standards, et cetera. And, draw these countries towards us, providing them with the compute and the technology and the AI, by the way, which will make us a better country, a stronger country, our company stronger, hopefully more productive. So, we can talk about what Mark said, and also help those countries and draw those countries closer to us. And, this is something that I try to advance in the Biden administration with, candidly, very, you know, mixed success. Because there's a different view, which is, you know, we've got to keep a lid on this at all cost. And, which I think is a mistake. And, the reason I think it's a mistake, the, the, the analogy that was often used is, well, you know, AI is like nuclear weapons. But that isn't really true. I understand the dangers. I do. I really do. And, I worry a lot about cyber. And, I worry a lot about bioterrorism. And, I worry a lot about non-state actors, you know, doing bad things and criminals. So, I'm not diminishing that one bit. Quite the opposite. I'm saying, if we have the similar standards, we can all work together. But, AI is the economy, right? It isn't. It is how healthcare will be done. How banking will be done. How shopping will be done. How retail will be done. So, if you tell another country, sorry, you can't ever chip, you can't ever compute. What do you think they're going to do? They will go to China. They won't say, okay, no problem, we'll have an old-fashioned banking system and not have it be AI enabled. So, that's why I think we just have to be much more sophisticated in terms of using this as an opportunity to make friends, not enemies. In terms of, like, will it make us as productive as possible? I mean, I think that's a, this is a complicated thing. I'll just, I'll tell you one really quick anecdote. When I was secretary, I was in charge of administering the CHIPS Act, semiconductor. And we, you know, I killed ourselves, we got this thing passed, historic legislation, bipartisan, the money came to my department. I negotiated this, very difficult to negotiate deals with TSMC and Samsung. And then I was told it was going to take seven years, because of permitting, to get a shovel in the ground for the first factory. It, these guys are laughing. It's not, it's a true fact, because you have layers and layers and layers and layers. You talk about housing, same thing. We cannot build if we strangle ourselves with red tape. You want the prices of housing and manufacturing to go down? Then you have to shrink the time. Time is money. Anyway, so I worked with bipartisan group of people in Congress. We changed it, and now we're making CHIPS in America, in Arizona, in Texas. But my point is, these regulations are there for good purpose. But somewhere along the way, we got over torqued, because if it takes seven years to put a shovel in the ground, that's a problem, and that's expensive. And that's bad for America and stands the way of productivity.
[03:32:56] Kevin Warsh: Mark, so some years ago, I think you wrote something called, "It's Time to Build." The moderator here has become completely irrelevant. So, is it time to build, and how are we doing? It is, it is.
[03:33:10] Speaker 16: It came at the peak of COVID when I hit sort of maximum frustration, and everybody hit their own point of maximum frustration during COVID. Mine was when the city of New York got put on a call for people to drop off their used rain ponchos at the local hospitals to use as protective equipment for the first responders dealing with COVID patients, because they were out of surgical grounds. And so, yeah, I mean, it's just, you know, we've all talked about this at length. Again, I would go back to this, which is, this is a political economy, and ultimately, you know, an ultimately political kind of question is, you know, what do we want? What do we want as a country? You know, what kind of pace of change do we want? How much productivity growth do we actually want? I mean, you know, I just, you know, I guess I've stated my prayers before, which is, I think ignorance is bad, and intelligence is good. I think low productivity growth is bad, high productivity growth is good. I think low economic growth is bad, high economic growth is good. You know, AI gives us, I mean, AI gives us a magic formula. I mean, I did not think I would live to see what's happening in AI. Like, I didn't think it would ever work. I mean, these are ideas in AI. The things that are working in AI are ideas that were first formulated in the 1940s. I got a computer science degree in the 1980s, and these ideas were completely discredited, and the people who worked on them were, like, in the back of the building, and this is, like, the stuff that, like, everybody knew. It was just, like, never going to work. And in the last decade, it's all started to work, and it's just, like, this absolutely amazing miracle. And so, it's, like, we've been handed this elixir to boost productivity growth, to boost economic growth that can be applied a thousand different ways. And for sure, it won't be 100% positive. For sure that, you know, there will be negatives, and for sure that there will be issues. But we've been handed this. And we do, we have very, very important decisions to make about what we actually want out of this. And I think those decisions and the policy implications of those decisions are going to matter, actually, way more than what the technology is capable of. Because without those decisions being made in the right way, the technology will simply not get rolled out.
[03:35:07] Kevin Warsh: So, Gina, let me come back to you. Chad can't defend himself, but he's written some incredible academic work. And what he said about prior, my own take of his work, not his words in a quiet period, is that what he said is these productivity revolutions are amazing, but there's always a weak link. So, the question, Gina, for you is, is the weak link electricity, data centers, chips, or is the weak link us?
[03:35:32] Speaker 15: D, all of the above. So, here's what I think. I think us in the following respect. In America right now, we have a situation of political instability, divisiveness, extremes on both sides, and a level, a deep economic insecurity that most Americans feel, many Americans feel. We have 10-plus million people working full-time, working full-time, and poor enough to be on food stamps and Medicaid. And then on the other side, you know, extreme wealth. And so, and now AI, now you're putting AI on top of that. And it has the potential to be unbelievable, as Mark's been saying, like, all the things we know. But also, if unleashed, without safeguards, et cetera, to just put a match in this powder keg of inequality. And so, like, my view, it's the way I've governed. It's the way I fixed pension systems we were talking about earlier. It's always like, growth is the answer. But we do have to find our way towards a form of regulation and governance, which is growing and distributing more equally or fairly. And so, if we can do that, I'm just -- what I said earlier, there's an opportunity for America to really emerge as a global leader with this technology. And to have it make things better in our country, making us all more productive. But if we fail at that, it's because of us. Right? That's why the politics of it is, like, if we could -- and I criticize my own party as much as anything. It's like, you know, the permitting example. When I went to Congress and did that in a bipartisan way, when I opened school -- I was a Democratic governor. I'm the only Democratic governor that opened schools, public schools, in the fall of 2020. Pissed off a lot of people in my own party, but it worked. So, it's like, on both sides, if we could come together and have serious policymaking, this could be amazing. If we don't, it's by far the weakest link -- humans.
[03:38:05] Kevin Warsh: So, Marc, let me come back to you with a thread that Jean had brought up before, which was China. The distinction between where the United States is, where our G2 rival is, China. There's a couple of meetings that are involving the two presidents over the course of the next week or so. What's your mental model about this new technology and the rivalry between these two great powers? Yeah.
[03:38:29] Speaker 16: So, to start with, I mean, there's no question that, like, it's a straight-up -- it's a straight-up geopolitical rivalry. I mean, this is the 21st century Cold War. You know, both countries have, you know, very expansive visions of the way that the world should be configured and the way that things should work. And, you know, both are aggressively trying to, you know, kind of proselytize those and expand those. You know, it's a very big deal. But, you know, the technologies of the time are going to be, you know, critically important for the same reasons they were during the first Cold War with the Soviet Union. And so, you know, it's, you know, obviously, you know, completely serious. Very big deal. The way I would describe the dynamic, the puts and takes, and, you know, I'd say, like, many American investors over the course of the last 40 years were actually encouraged to go invest in China. My firm is actually late, late enough to the arrival that we never did. So, as they were kind of on the other side of this. And so, you know, my day job is investing in American companies that, in many cases, are going to have to compete with Chinese companies. So, you know, my kind of take is that in America, we have all of the pros and cons of a decentralized, you know, liberal democracy, free market system, notwithstanding all my complaints about bureaucracy, you know, to the extent that we, you know, still have both democratic and free market system in this country. Like, we, you know, we have a highly dynamic and, well, we have a highly dynamic private sector. I'll brag on the private sector, but the private sector in the US has recalibrated, and the financial system recalibrated $7 trillion or something of money in the last, what, two years to AI? Like, absolutely staggering, like, you know, way, way beyond, you know, any sort of, you know, Apollo program or, you know, anything like that in history. And primarily through public sector or private sector action, you know, supported by a financial system, you know, that, you know, that generally speaking is the best in the world. And so, we have all of the advantages of a decentralized dynamic environment. We have all the disadvantages of a decentralized dynamic environment, which is like, you know, man, you read the newspaper, it seems like everything's a total mess. You know, the government always seems like it's in some form of meltdown, like, you know, like, my God, why can't these people get together and get along? You know, like, where's our plan? Like, where's the national plan, right? China has all of, I think, the advantages of a centralized, you know, command and control system, and that they absolutely 100% do have a plan, and it is dictated top down, and it's like written down on paper, and it's very clear, and the orders fan out, they're executed. And then they have all the disadvantages of a centralized command and control system, which is, you know, they routinely go through, you know, periods of time where they just like start arresting all their entrepreneurs and throwing them in jail, and they, you know, do all kinds of crazy stuff, you know, as dictatorships, you know, tend to do. On balance, through technological revolution after technological revolution, the decentralized dynamic system has outperformed. And by the way, I think it is outperforming today. You know, the U.S. is the global leader in AI. But they have a plan, and they're definitely executing on the plan. And to the extent that we foot fault, or that the disadvantages that we have from having a decentralized dynamic system, you know, become too intense, you know, they are 100%, you know, racing to catch up. And now characterize it as, you know, there may be a year behind. You know, and it's basically, it's a foot race. And I, you know, I would hope in three years, we're still at least a year ahead, and I would hope that maybe even we're three or five years ahead. But, you know, we're going to have to make that happen.
[03:41:45] Kevin Warsh: So, let me try to distill, if I could, the last 45 minutes, maybe this morning's discussion, simply, and I'll turn to Jean and Mark for a final comment. And they should feel free to disabuse their moderator with this point. I think if there's a general view, at least, of the panelists today, and I certainly talked about it in my speech this morning, this is a hugely consequential moment. And it's easy to think that this moment, everyone would think that their moment is the moment that will determine the fate of our economy, but this time it does seem different. We are at a hinge point for the U.S. economy, the United States role in the world. And I said this morning, I don't think any of us are impartial to how we want that to come out. Secondly, I would say this, for all the considerable problems that we have in and around transitions of this new technology, the nature and structure of the economy that Mark talked about, the challenges that we central bankers have in navigating this moment. As investment hits both supply and demand, it hits demand first. But I'd rather have our problems than anyone else's in the world. So, with that cautiously optimistic statement, Gina, let me turn to you and Mark and give a final comment. I certainly agree with that.
[03:43:04] Speaker 15: I would rather be an American and an American business person than running a business, starting a business, anywhere else in the world. I really would. That being said, you know, democracy, what we're doing, we have to be continuously self-improving. And we have to be honest about what isn't working and just get to the business of fixing it. And so that's what I think. Yes, I'm optimistic. I keep doing this because I'm optimistic. I see the promise. I see the potential. But I just wish we could just kind of be practical. And there's nothing we can't fix if we don't get together and, you know, be serious and try to solve these problems. And, you know, that's why you're here. That's what you're doing. And we just have to stay at it.
[03:43:55] Kevin Warsh: Mark, a final comment. Is this human agency thing, does it matter? Yes.
[03:44:00] Speaker 16: Look, I'm extremely optimistic. I'm extremely positive. I think this is going to be absolutely fantastic. And Kevin knows I have an 11-year-old at home and he's all AI all the time. And it is one of the most amazing things I've ever seen in my life. He's able to advance in all kinds of fields at a pace I would not have thought was possible. And so I think there's just like an incredible opportunity here, incredible payoff. Maybe I could just close by saying, you know, maybe confession or something, which is like I think a big part of this is the responsibility of the technology industry and the industry to really like step up and provide leadership here. And I think that so far we collectively in the tech industry have not done a great job on that. And, you know, in some cases, I think, you know, at least some folks on our side have been, you know, maybe even, you know, not destructive in this process by, you know, maybe spreading, you know, very high levels of fear and so forth. And so I guess maybe close with like I think our industry needs to aspire to, you know, to step up and be to be more part of these solutions. But part of that is in the sense of like how we actually explain and articulate what we're doing. Part of that is the policies that we advocate for. And then a big part of that is the actual design. Somebody mentioned earlier, the products can be designed in different ways. And I think the product should be designed for human potential and human enablement and to put the technology in everybody's hands so that everybody has an opportunity to benefit. So I guess I may be close by saying I would hope and aspire and encourage my industry to try to contribute more to this.
[03:45:25] Kevin Warsh: Well, this concludes the public portion of the field hearing of the Productivity and Jobs Task Force. On behalf of Mary Daly, the team at the San Francisco Fed, all of us at the board, we're grateful for your participation. And we'll just end with applause for Mark and Gina for leading us out.
[03:46:08] Speaker ?: Thank you. Thank you. Thank you. Thank you. Thank you.