Bloomberg Audio Studios, podcasts, radio news. Tech leaders, including Sam Altman, the head of open Ai, which created chat GPT, have been brimming with optimism about the future of artificial intelligence. A lot of the things that people are starting to experiment with now, you know, sort of super cheap energy, virtual reality, genetic editing, really great AI. You know, these things are going to transform the world in very fundamental ways.
And Meta CEO Mark Zuckerberg shares Altman's enthusiasm.
The next five to ten years AI is going to deliver so many improvements in the quality of our lives.
They've successfully sold the promise of AI its transformational power to many investors. The titans of Silicon Valley have poured billions of dollars into research and development, and the share prices of their companies have risen in kind. But the recent arrival of a Chinese competitor called deep Seek made investors question some of the prevailing narratives that had emerged
around this buzzy technology. Deepseek says it created a rival to chat GPT maker open AI's model that can perform human like reasoning at a fraction of the cost, and that's raised some new questions about where the frenzy surrounding AI is going to lead and who the winners and losers in the AI era are going to be. It's something Tom Orlick, that chief economist at Bloomberg Economics, has been wrestling with.
So if we look at the grand sweep of history, hundreds of years, thousands of years, it's really clear that the tech visionaries have it right. The plow, the windmill, the textile factory, the electric motor, the automobile, the PC, the Internet, all of these have driven increases in prosperity. And that's the claim that the AI visionaries in Silicon Valley and China shen Jen are making about the large language models that there.
But Tom says, lives are not lived over the span of hundreds of years or millennia. Lives are lived, he says, over years and decades, and with the development of AI, it seems like time is moving even faster.
Technology. Powerful technology can have positive impacts on the people who invent it and the people who own it, but also significant negative impacts on workers who find themselves displaced and unable for whatever reason to retrain, reskill, relocate, and get a foothold back in the labor market.
I'm David Gerat and this is the big take from Bloomberg News Today. On the show, Tom lays out three cases for what AI will mean for the economy, companies and investors, and for you and me. Tom Orlick says, the first scenario he and his colleagues at Bloomberg Economics considered for how AI will transform our lives has an hour that's pretty rosy.
If we think about the revolution in robotics and automation which swept the manufacturing sector in the nineteen nineties and early two thousand's. Well, the promise there was that we would have machines that could do the work of factory workers better, faster, cheaper. Of course, that was good news for the folks that owned the factories and the machines, not such great news for many of the workers who lost their jobs. Still, in the grand sweep of history,
doubtless a positive. What's the promise of AI. Well, the promise of AI is that it can do something similar for the white collar workers. Right, you're a lawyer, you're an accountant, you're an economist. Well, AI can supercharge your productivity, enable you to get your job done more quickly.
That best case is productivity goes up and a lot of people are going to benefit from that. Do I have that right?
That's right, David even podcast hosts.
God Will I. What is the second scenario that you're considering?
So the second scenario is that AI turns out to be more of a parlor trick than a paradigm shift. Yes, these chatbots look pretty impressive. It's fun that we can ask chat gpt to draft a legal document in the style of a Shakespeare tragedy and it does it in a couple of seconds. But maybe the downsides of AI turn out to be more important. Maybe AI stumbles on the path from the lab to the market and it just can't do the job, and so the booster productivity is there, but it's not a game changer.
The final scenario that you weigh is the most worrisome, and I wonder if you could lay that out for us.
The last path is kind of a dystope in path, and that's one where AI is powerful. It can do the job of accountants and lawyers and economists, and it can review X rays and it can write architectural plans. But instead of supercharging productivity for individual workers, that ends up just replacing a vast swath of the workforce, and white collar workers face the same challenge in the twenty twenties and twenty thirties that blue collar workers faced in
the nineteen nineties and the early two thousands. Massive job losses, lost income in miseration.
Leads me wondering sort of how all of this is going to shake out.
So one of the things that's happened in the last week is that the sudden appearance of deep Seek has suggested that developing leading edge AI models could just be much cheaper than we previously thought. What it also suggests is that the competition between Chinese AI champions deep Seek, Ali Baba and others and the US champions is going
to get more intense. And as we saw in the Cold War in the technology race, the space race between the US and the USSR, when you have those sharp geopolitical incentives, well that can amp up investment accelerate progress
past the technology frontier. And both of those things, cheaper AI and sharper incentives, more competition between the AI champions both suggest the moment at which we find out if AI is going to be a game changer for productivity and how that cake is going to be divided up, that moment of kind of revelation is going to come forward.
So how will we know when that moment of revelation has arrived. We'll get to that next. We've talked about this question of how AI is going to impact productivity, and I'm curious how economists measure that.
So that's a really good question. And adding to the sort of the complexity and the confusion here is the fact that it's actually rather hard to measure productivity. So if we think about productivity gains at the economy wide level, or if we think about what drives growth at the economy wide level, well, it's how many workers you've got, it's how much capital you've got, and it's how smart you are at combining those workers in that capital. And
that's the kind of productivity piece. How do we measure that, Well, we observe where growth is, we subtract what we know about the labor force, we subtract what we know about the capital stock, and productivity is the residual. Right. So productivity is already kind of a bit mysterious, right, It's measured based on what we can't explain from anything else.
Add to that the fact that GDP numbers growth numbers are very odd and significantly revised, and what you've got is a situation where measuring productivity gains, especially in real time, is pretty hard to do.
Are there any unique challenges to trying to measure productivity in the context of AI. I think just perhaps given the kind of speed of uptake that we're seeing here, doesn't make the job of calculating productivity harder.
So, first of all, it's not a surprise that we don't see the AI productivity gains in the GDP data. Yet, if you think about technology and its impact on the economy, the eureka moment for the inventor is a necessary, but not a sufficient condition for the positive economic impact. You need that eureka moment, but you also need time for the new innovation to be diffused through the economy. You need time for all the factories to go from steam
power to electric power. You need time for all the companies to work out how to use PCs and how to integrate them into their workflow. These things take time. So the fact that AI is not present is not showing up in the productivity data yet, isn't a huge surprise.
What can we learn from the impact of past technological innovations. So you can go back to the cotton gin if you want, or to stee empowered locomotives. But what if we just look at, say the impact that computers had or the Internet had.
There's a few things to point to, right. So the first thing is it takes time for new technologies to show up in higher productivity. Solo a Nobel Prize winning economist, he said, indeed, not hand Solo, the the Jed.
The Jedi.
The Jedi famously said in nineteen eighty seven, we can see the computer age everywhere apart from in the productivity data. And it wasn't till a decade later that Alan Greenspan, then the FED Chair, led a kind of statistical effort to find the evidence of productivity gains from the computer. So it takes time for new technologies to show up. The second thing to say is if you allowed decades
to pass, new technologies raise prosperity for everybody. We're all better off because of electrification, we're all better off because of the internal combustion engine. We will all be better off because of computers and the Internet. But in the kind of more short period of time. In the years and decades after a new technology is introduced, the gains
very often are not broadly shared. And the reason for that is that workers who are displaced by new technologies, well, for them, the losses often outweigh the gains.
As we go forward, what are you going to be watching for? What are other economists going to be watching for, is they try to assess the impact that AI is going to have on productivity.
We're going to be looking at the technology and the advances in capability for chat, GPT, LAMA, deep seek and the other models. We're going to be looking at the case studies, the early evidence of how AI boost productivity or doesn't boost productivity, and how those gains are allocated at a micro level, at a company level. Now, where can we see evidence of a productivity boost from AI? Well, not so much in the macro numbers, not so much in the GDP numbers, but if we look at case studies,
we do see some pretty striking results. It's been a bunch of case studies thinking about whether using AI can make coding faster, for example, or help people in call centers deal with calls faster and get better results, and those case studies they're kind of micro, right, they're looking at a tiny slice of the labor market, but they are pretty encouraging to answer the big question, is there
an economy wide productivity boost? Well, I think that's a question which is still going to take years, maybe decades to answer.
The answer to that question is going to be incredibly consequent whenever we get it. If AI helps everybody, or if the technology's benefits are not evenly distributed, and we see the disappearance of rafts of white collar jobs that Tom says would have a huge effect on our society and on the balance of political power.
If we do see the cake being divided up in such an unequal way, that's going to raise some important political questions. We've just seen Donald Trump get elected for a second time as US president. Why has he been elected a second time as US president? Well, people talk about China and Mexico and trade and what that did to US jobs. But guess what. US jobs didn't get replaced just by Chinese workers and Mexican workers. They also got replaced by machines. Well, if that's what happened when
blue collar jobs get replaced by machines. I wonder what would happen if white collar jobs are replaced by machines. I'm not advocating for my fellow economists to print out their Excel spreadsheets, mold them into papier mache pitchforks, and start marching on the data centers of Arlington. But in a dystopian scenario, that's a possibility.
Tom was a pleasure. Thank you very much, my pleasure, David. This is the Big Take from Bloomberg News. I'm David Gura. This episode was produced by David Fox. It was edited by Patty Hirsch and Rachel Metz. It was fact checked by Adrian A. Tapia and mixed and sound design by Alex Sagura. Our senior producer is Naomi Shaven. Our senior editor is Elizabeth Ponso. Our executive producer is Nicole Beemster Boor.
Sage Bauman is Bloomberg's head of Podcasts. If you liked this episode, make sure to subscribe and review The Big Take wherever you listen to podcasts. It helps people find the show. Thanks for listening. We'll be back tomorrow.