美联储理事Waller：AI建设推高通胀或持续2-3年

Now, Christopher Waller, the Fed Reserve Governor, is currently speaking to our Selia Moshin here at Bloomberg's new economy forum about monetary policy in the age of AI. So, let's just listen in.

It's so exciting to be here with Governor Waller, we both flew in from Washington DC to have a chat with in front with you all in present. Uh, one thing I've noticed is that just like in Washington, if you walk around in Delhi, the big conversation is AI, which we're going to continue talking about. Governor Waller, you have made a lot of comments in the last couple of years about the benefits of AI, including productivity boom, uh, contri contributing to GDP growth, seeing, uh, living standards rise, and that's all talking about the United States.

I'd love to hear from you how you see the impact of AI in our economy go into the rest of the world, especially emerging markets like India. Yeah, I mean the way I think about AI is first there are those who build the I AI tools the models then there are those who supply all the inputs to build that that would be data centers chips all that and then third that's the people who actually use the models now clearly when it comes to building the models the US uh is the dominant country China's there as well in terms of uh providing the inputs.

A lot of that's Taiwan, South Korea, uh, producing the inputs. Data centers, they're all over the world. They're not just located in the US. And then, of course, the users are anybody that wants to use it for for work. Now, there's a very high correlation between GDP per capita and AI use in people's work. So that's kind of one of the things that people I think I'm a little concerned about is it's not even across the world in terms of how AI is used and that means it's going to have productivity differences if it's not actually used.

So you'll see potentially widening gaps in GDP per capita uh because of this. But I think it's an amazing technology. I really do think I'm not a pessimist. I don't believe AI is going to be the end of the human race in the next two years or anything like that, but it is a tool. And even within the Federal Reserve, that's what I tell all of our employees. We're going to train you to do this because we want you to be more productive.

If I was going to replace you with AI, wouldn't waste one cent training you. That would be a waste of money. So, it really is a tool that I expect all of our employees to use. So that's that's kind of the basic global picture of how I think the whole AI uh world works. Uh Europe, for example, doesn't produce much AI or inputs, but um they're big users of it and they're it's already kind of widely spread throughout uh the business sector and community.

Let's talk about the payoffs. You say you're an optimist. You're not worried about, you know, the human race. uh when are we going to see some of these benefits? The productivity boom, the increase in our wages, inflation being reigned in, those things haven't really turned up for the everyday person,

right? And that's it's it's often the case that when you have big technological changes, this is what economists know is you don't typically see the productivity gains for years. A big part is firms have to figure out what am I using this for? How am I going to actually increase productivity? They're not really sure. And so there's a lot of trial and error and a lot of experimentation, but then you start realizing where it really does have its impact.

I mean, I'm telling you, if anybody does any coding, this stuff is amazing. I mean, it's just shocking how good it is. So, in terms of coding, yeah, I mean, if you were a coder, your job's going to be fundamentally changed. It doesn't mean you're going to lose your job. It's just you're going to do a different job than what you were doing before. things that you couldn't do because you were too busy doing basic coding, you're going to able to now expand what you're doing in terms of the production of your output.

So, uh I I think it's just going to be a while before firms start figuring out how to actually do this. And you may look, you may not see it in terms of this is one of the measurement things I've thought about. Suppose I can do the same amount of work faster. So I instead of five days, I get it done in three days. Maybe I take all that time and just go have more leisure. My output's not any higher per a given year. It's just I've got more time to to consume my leisure and enjoy my life.

That still means it's still good for welfare even if you don't see a big increase in uh uh in GDP per capita. So you either have to rethink is it do we measure GDP or productivity per unit of time as opposed to you know quarterly it's now down to like days. So you have to kind of think about that. And then just in terms of productivity, there's a lot of things this technology is just going to make your life better even if it doesn't show up in your work productivity.

I always like to point out Google Maps is an amazing invent in uh invention. I don't use it to increase my productivity. It helps me get around in cities where I'm and I'm visiting and I don't get lost. That doesn't show up in the GDP numbers, but that makes my life a lot better. So I always try to tell people technology can just make your life better or it can actually increase productivity and how much output you produce.

One thing that's clear to measure is inflation. And right now the narrative in the United States and in other parts of the world is an affordability crisis. There is definitely uh an impulse that's coming from AI related impacts on the economy. In the short term, when do you see the AI hit to inflation dissipating in the US?

Yeah, I mean the whole buildout obviously if you're building out several trillion dollars of investment in AI and you have to buy all that stuff and your demand is perfectly inelastic. You'll pay whatever price. If you're one of the hyperscalers, you're going to bid up the price of all semiconductors of all type and that's going to spill over to anybody else who uses those products. And that's what we're starting to see in the US.

We're starting to see AI having a spreading out of inflation because it's just increasing demand for everything else. Now down the road when this stuff is done and they've kind of built it out, that kind of demand pressure goes away and then you hopefully will start seeing all the supply effects which means we can produce more output faster and at a lower price. So it's one of these two-stage things. Demand is going to be especially the size of the investment.

I mean, this this is shocking how much is being spent on this compared to any other point in history. And so, you're going to see these price pressures early on. How long are they going to last? It could last anywhere from two to three years. Given how much they've committed already to spend out in the future. Now, if all that falls off, if suddenly there's a pullback or you can't meet this buildout, you can't do it.

That'll put a check on prices. But the benefits on inflation down the road are they're down the road as far as I'm concerned. They're still a few years out.

A few years off. Is there any chance that the inflation that's stemming from the AI buildout could be embedded in the economy for a longer time?

Yeah. I mean, like I said, it's very specific what it is. It's not like a broad-based effect. It spills over into a smaller set of goods than uh than other things. So, and we know that at some point this huge buildout and huge demand is going to I won't say completely come to an end, but it's going to taper off a lot. And when that happens, we'll see the demand pressure come down. But from what all the plans that I'm hearing from the hyperscalers, they're still planning on spending trillions of dollars over the next three years.

So, this this demand pressure may still be there and we have to just decide how w