WEI Chenyang: Professor Spence, it's great to welcome you to the 2023 Tsinghua PBC School Finance Global Finance Forum. It's great that we are able to have you as our keynote speaker, again. And thank you so much for your support in the past few years. Welcome!

Spence: Thank you so much. I appreciate the opportunity, and I'm hoping that a year from now will be doing this in the same room together.

WEI Chenyang: Absolutely. And we have a short slot because this year we're trying to convene a lot of speakers from important domestic institutions, foreign think tanks and distinguished institutions. So today, I still want to start with the key word “uncertainty”. I remember in last two years when we had you at this event, we talk about at each point of time, the specific events and policy uncertainties and all the challenges brought by the pandemic and etc.

Now the challenges from the pandemic seems to be largely behind us. But we are still in the world in which we see no less uncertainties. I hope you don’t disagree with me. So for my first question, I want to come back to one of the key factors that people are still really worried about - and I want to start with the monetary policy global-wise. We see good news and less certain situations in the US economy and in the rest of the world. I just saw the number of the first quarter GDP growth and job numbers in the US, which beat expectation. And people are talking about maybe that changes the perception about interest rate path going forward.

But at the same time, we just had Silicon Valley banks’ failure along with First Republic, for example, which were quite eventful from the financial systemic risk perspective. With this all still makes together, I don't think equity market is having a clear reading of all of this. So I want to come back to you with the first question, how do you see the dynamics of the monetary policy global-wise going forward? Of course the most important one would be the Fed policies.

Spence: Well, first of all, let me agree with you. Despite the fact that we've put the pandemic largely either in the rearview mirror and just made it an endemic part of our lives, there doesn't seem to be any drop in uncertainty in the capital markets and in the global economy. So focusing on a good deal of it, is centered right in the middle of the inflation fight and monetary policy. So I think you know the narrative very well, Jason. The bad on the part of the major central banks especially in America and Europe was that the inflationary forces were transitory that turned out to be either totally incorrect or transitory had a much longer time horizon than they can live with. So we've had an extremely rapid period of interest ratehe type we haven't seen in three decades. We have to go back to Paul Walker and the early part of the regulative administration to see anything like this. This is just causing accidents, that are unanticipated, especially by a group of people in the financial sector who lived in a low inflation environment or deflationary environment for at least two decades, probably more.

So the uncertainty is great. First of all, capital markets are still being driven by trying to guess where monetary policy is going. I don't think the central banks know where monetary policy is going beyond a certain point, except that they’re being firm about the inflation fight. And I think the markets are hoping or betting, still, to some extent, that once the inflation fight is over - whenever that is, let's call it 18 months from now or whatever - we'll go back to an interest rate environment like the one that we had before.

But there is considerable uncertainty about that. And there are a lot of people, including me, who think that's very unlikely that we're likely to head into an inflation under control, higher interest rate environment. And the reason for thinking that, we don't have to spend a ton of time on it, is the global supply conditions have changed in multiple dimensions – aging, differences in labor market behavior, declining productivity, a whole host of things that at least give you some reason to think that this isn't the world that we lived in in before. So there's a lot of uncertainty around that, and that doesn't seem to be possible to just resolve it in a simple way.

Even take the recent banking - we had Silicon Valley and two other banks fail. And then a recent takeover First Republic Bank by JP Morgan Chase. Jamie Diamond, highly respected CEO of J.P. Morgan Chase, basically said that's probably the end of the turbulence. But there are people who are saying the short sellers are going to move on to a bunch of other regional banks. So even in that dimension, we don't know whether we put that kind of turbulence behind. The one thing people seem to agree on is that the stress in the banking sector, particularly among the regional banks is producing or contributing to the typing of credit conditions.

So we had a growth slowdown in the first quarter. And the remaining uncertainty is how big of the slowdown is going to have to be, even to the point of having a recession before we get inflation under control. And even there, people are saying it may be very difficult to get inflation from 3.5 down to 2, and maybe they should quit, on the way through this.

The central banks will never say that because they'll lose credibility if they say it. But there's another set of issues which is - what's the right place to stop and declare victory? The bank cases are clear, rapidly rising interest rate environment makes the long duration assets lose value, and simultaneously increases their funding cost because if they don't respond to the rising short term rates, then they lose depositors and start developing balance sheet problems. And that's not going to go away. There's lots of other places where people are worried in private debt provision markets and so on.

If you had to summarize it, you're stepping back from the details when you have regime change of this magnitude, whether it's monetary policy or the fundamentals in the way the economies put together, I don’t prefer term “accidents”, is probably “unanticipated events”, the things that people don't see coming. And we haven't got through the end of that yet. So that’s the picture I think we are in. The current debate, I just put it extremely concretely, is: is the Fed going to raise interest rates once or two more times and then start lowering them? One side says, yes. And another side says no. They're not going to be able to lower them because the inflation will start to come back, if they do that. This is the question about what is the long term equilibrium interest rates in the US economy.

WEI Chenyang: Purely from macro environment the policy path dynamics, I totally agree with you. I think everybody is still stumbling with what would be the clear path going forward. But I think adding to that, I want to come to something clear and less uncertain which is on the technology side. I'm sure you have been watching this very, very closely, Mike. We just experienced in this AI world. Something, people say it's like the iPhone moment in the cellphone world. This Chat GPT technology and whole lot of debates in almost every sector, people are so excited and talking about new ideas, starting new experiments. I know you have been watching Fintech technology global-wise. So I am really enthusiastic to hear your view about this recent development, and what you see as these implications for the economy going forward?

Spence: So we are at the early stages. These folks, the people in the AI world - AI development world saw this coming. But I think for the rest of us, it's been sprung on us and we're in the early stages of this. These are shockingly surprisingly powerful models. It’s hard to know where to begin. There's a significant group of leaders in the tech sector who have argued publicly, stated that we should slow this down. Precisely because we don't actually know all the used cases including the ones we probably don't want to see. Having said that the CEO of Alphabet and Google was specifically questioned where is this going to have the biggest impact and his answer, which I think is exactly right is the large language models are going to be huge impact right across on the entire span of what people call a knowledge economy, so that includes pretty much all in finance, media, software development, scientific research, various business functions, in fact, hundreds of them. And I think on the positive side, once people learn how to use this, we had the potential for just an enormous productivity surge. By itself that's a good thing in a resource constraint declining productivity world where those supply side constraints are actually the operating constraints on growth. This is entirely positive.

Let me say a word or two about the worries people have, I think they come in three buckets: one has to do with jobs. So if you talk to the young people, including pretty knowledgeable ones. They say, Gee, if it's that powerful, are we going to have a job? If you get a productivity surge, of enormous magnitude. Let's pick a specific example, the large language models can write computer code. That's all we've already been demonstrated. They're pretty good at it. They're going to be powerful digital assistance to software engineers who write computer code. That's a potential to be a very big productivity increase, but that people are in the business of writing computer code, which is a growing industry for sure, legitimately worry whether there's going to be jobs, especially for the young people. So there's a question of having to do with jobs, whether they're available and the transitions that people have to make in skills and what not in order to use this technology.

A second set of concerns has to do with the bad uses of this. So these models generate content in seconds. And well some of it's better than others. It's pretty impressive. And so I think the people are saying slow down here. To some extent say, look, we don't really know what the impact of this is and nobody has an iron clad argument that the good stuff outweighs the bad stuff. So let's slow down. And that's a voice that is sufficiently powerful. It may be heard.

I'm in Italy where we live at the moment. Italy banned Chat GPT. So if you try to use Chat GPT from here, you won't get through. Not because they block it - they don't have a firewall. Instead they told open AI and Microsoft that we are not running here. And the comply the issues of privacy, which you would expect from the European Union.

So going back to the knowledge economy part, the this does not directly affect part of the economy where people don't have to be available in-person work. If you think of people in restaurants, in nursing homes, people who clean the office buildings at night, construction. This is not yet powerful enough technology. So one set of issues that aren't being talked about as much, but has to do with the distribution aspects of this. If it has a big impact on the knowledge economy, and that's eventually shows up in incomes, that tends to be the higher income part of the economy.

We may have trends or tendencies in the evolution of the economy technologically that have adverse distributional impacts. I don't think those are permanent because I don’t think that the pace of technological advance in digital area is such that we can probably look forward to some of additional advances in things like robotics that will expand the digital footprint even more and into what the technology just called the atoms or physical part of the economy as opposed to the knowledge part. But that's all speculation. But it's certainly to go back to your theme adds to the uncertainty.

I guess I would describe it as a very exciting set of developments with lots of dimensions of things that one can either speculate or worry about, but these are stunningly powerful. Let me mention one other thing. I think one of the reasons that the major players in this are the players who have the computing power of relatively few people do, or entities have the computing powered to build these things as opposed to use them. There's a difference between building them and using them.

But I think the reason they're holding back the most power and they are holding back the most powerful versions of this, is that there's unpredictability in them. They do things that even the developers of them don't fully completely understand. So in a sixty minutes broadcast that focused on google and Bard. Folks from the Google side reported that had Bard been exposed to a few prom scenes in Bengali, the language of in Bangladesh. It has been exposed for two good minutes, as far as they knew in any part of the training process focused on Bengali. They discovered and not too long after that Bard had learned Bengali. And they don't really know why or how. So they're holding back and probably for good reason until we have a deeper understanding of it.

Forgive me for one other anecdote, if you deny these large language models’ access to search and fact checking capability. No make stuffs up. They’re whole hallucinations. And apparently Bard was asked to write out a little essay on inflation our earlier topic, which it did, which is moderately sensible. And at the end of the little essay and said: “for further reading, you might want to look at the following five things” and it a listed five either books or papers or essays or articles, none of them existed. (Wow) Just made them up. ‘Cause that's what you do at the end of an article, Bard had figured out and provided some further reading for people who are interested. So it made up five of things you can read. Now if you give access to Google search, we could go out and say they don't exist, maybe we should put in something that does exactly.

This is an example of so until we have a clear sense of what is it actually being produced here, whether it's just accidental or from malicious or counter purposes that are counter to the social norms. I think there's going to be a certain amount of caution in deploying these things.

WEI Chenyang: So basically this predictability and the difficulties of putting some boundary to govern the whole evolution and utilization of the thing. I guess that makes the most capable of people a little hesitating.

Spence: I think they are. I don't think it's a permanent obstacle to the benign uses of this, but the impacts are so profound that I think they want to make sure there's at least a reasonable degree of understanding of how they work, and what the potential is for a misuse. So I think that's right. The most governments, regulators, responsible parties or potential users and the developers themselves are probably going to converge on a set of principles that have to deal with how to deploy something. The other thing is anything that's very powerful that has deployed at high speed is more disruptive than it is if it's deployed at low speed. So there's another ordinary economic transition argument in favor of not exposing people, economy, jobs and so on to the full blown storm that this technological and economic storm could create.

WEI Chenyang: And I want to come back to a big picture issue that you have been thinking about for a long time. The global growth and also the productivity issue, especially given what we just discussed in the first two questions, the uncertainty and also the policy dynamics going forward and also this technological advances. Some exciting, some unpredictable components of it. So has all those kind of uttered your view or thinking about the productivity for global economy going forward I would like to hear maybe an updated version of your thought, so to speak.

Spence: My frame of reference, you and I’ve talked about before, is that we have significant supply side constraints, in the global economy or forces that are damping down for productivity growth relative to the not just the recent past, but the impressive growth numbers that have been posted pretty much not everywhere, but across the global economy. China's growth is a perfect example of this. India’s potential growth as high. And I think it’s still high. But those constraints I think are going to operate. They'll affect the inflation and monetary policy, though. That's like what I call a short run picture.

I was cautiously optimistic about the longer run picture for two reasons. One I thought the technology, even without a glimpse into the AI world, had the potential to produce significance. Just the applications of AI pre-large language models. Now I'm more optimistic. I think the this really is a powerful set of tools that if we use them in the right way. What I’ve been saying to friends in the investment world is even if you just focus on software development. The potential is very large and that's a small slice of the overall impact.

Again, we have the issues we just talked about, speed, disruptiveness and so on, and managing transitions in a way that people don't get side swiped and so on. That's something that China, and its policy makers and leaders have experienced with over for long period of time. So it may be easier for China to get around to thinking about those then. Then it is for people who have lived in a system in which there was relatively little ideological propensity, do it intervene? So that's one thing.

The other thing is that I think although there's complications associated with this, there's a significant kindling of interest on a broad front in what you might call supply side policy or sometimes industrial policy. It's still controversial in the west. But we pretty much rounded the corner in terms of deciding both for economic performance reasons, let alone reasons having to do with resilience self-sufficiency and so on. We just have to have these policies.

So we're on a track that similar to the track that China is on. In fact, pretty much every economy. Parts of this serve is a little worrying. I think that geopolitical tensions could get out of hand and produce more fragmentation in the global economy than anybody would ideally like or desire into the system, because it's being done in these decentralized semi-non-cooperative way. But that's the direction we're going. And there's some good parts of that-investing heavily in technologies that have the potential to produce inclusive growth patterns or even and especially sustainable inclusive growth patterns. Everybody realizes just to take that last example that you can't solve the sustainability problem, if you delegate everything with the private sector. It's just not going to happen. And these are extra global externalities there. We need a ton of technology to solve this problem and incentives to adopt it. And we're in that world now and that's the world in which the government is in the public sector is a more major player. Then in the United States and it has been. It never disappeared and always had a big upstream research and technology development set of public sector investments. So it’s a change at the margin, but this one's more interventionists.

There be mistakes made, once you get into that business, and the people who were naysayers about this say well we'll just make too many mistakes of waste resources. But my view is we don't have a choice. We have to get in there. And so the answer is going to be. So I think the way the future looks is in that dimension is somewhat different than in the way it looked even a year or two ago. You don't hear as much about it, although you hear a fair amount about it now. But the focus has been so much on inflation and central banks and so on, that dominated the discussion.

The problem was going into the world in which the government is more active player. The challenge is that it makes urging everything into a more multi-lateral global economy, a more complicated one that we then we used to have. It's just much harder. If the American government is subsidizing all kinds of activity designed to move the sustainability agenda forward, the Europeans look at it and say well we have a different approach, there's competitive issues associated with this. We have prices on carbon. You tend not have high prices on it. It materially complicates the problem of reconstructing a workable global order for trade and technology transfer and investment. So I think it's inevitable, but we just have to deal with the increased complexity.

WEI Chenyang: I want to thank you again for this wonderful sharing. As we have before each time we run out of time quickly. And this time I really hope, as you said next year we will be able to discuss with you all those issues for as much longer duration in the same room here at the PBC School of Finance. That being said, thank you so much for your time, and your support in the past years for all the activities that we have been having PBC School of Finance. I really looking forward to catching up with you soon and seeing you in person next year. Thank you so much, Mike.

Spence: Thank you, Jason. It's a pure pleasure and you have an important institution there. So it's an honor for me to have a supporting role even if a small one.

WEI Chenyang: It’s definitely a significant one.