New research from MIT economics professor Daron Acemoglu finds that AI will increase economic productivity by just 0.71% over the next decade. Joining him on Market Domination Overtime, Acemoglu discusses how he quantifies AI's impact on economic growth and the distributional effects of the technology.
“This is a big achievement… but when we look at the data, we see that most of the things that humans do, these models still can't do,” Acemoglu emphasized. About 4.5% of the tasks done by U.S. workers could be directly affected by AI, and when combined with the productivity gains these models can currently achieve, “it amounts to less than 1% of GDP,” Acemoglu explained.
The distributional effects of AI will “not be very large,” he noted, adding that society is unlikely to experience large-scale job losses over the next decade because jobs affected by AI are more evenly distributed across regions and economic groups than those “affected by robots.” Still, Acemoglu said AI will likely widen the gap between capital and labor, “helping the owners and managers of capital more than workers.”
For more expert insights and the latest market trends, click here to watch this entire episode of Market Domination Overtime.
This article was written by Gabriel Roy
Video Transcript
Unless you live in isolation, you've probably heard how AI will transform our lives, drive economic growth and increase productivity.
But new research from our next guest suggests there's no need to rush. We're joined now by Doron Amou, professor of economics at MIT. Thanks for being here.
Um, so basically your new study, like your other studies, tried to quantify how this affects economic growth?
The difference is that you came up with a much lower number.
So firstly, tell us how you got there.
Yeah.
Guilty.
So, there's a lot of excitement around AI, and I think there's no question that these models are doing things that were thought impossible 10 years ago.
So there's been a big accomplishment here, and it's impressive in many ways.
But actually, if you look at the data, you'll see that most things that humans do still can't be done by these models even 50 years later.
There's no telling what will be possible.
Nothing is impossible with AI, but we more or less know what the technology will be able to achieve in the next 10 years because we already have prototypes, they are evolving at a certain rate, and we more or less know how it will turn out.
So while there's a lot of uncertainty, we can at least put some numbers together, and based on estimates from others, we've come up with a figure that suggests that about 4.5% of the jobs performed by American workers could be directly affected by AI, and many more could be indirectly affected.
It's hard to know, but about 4.5% of the economy will be directly affected.
And if you combine that with what productivity gains and cost savings these models currently bring about through tax automation and in some cases worker assistance, it comes to around less than 1% of GDP.
Exactly as you summarized it.
Well, Jerome, that's interesting, because I know you've also studied economic inequality, how these changes are distributed.
So, when you look at that increase, how is it going to be distributed?
Who will benefit most?
Well, the good news is that if the figure is that 50% of all employment will be affected by AI, the distributional effects will be large and millions of people could lose their jobs.
But if it amounts to only about 4.5% of the economy, the distributional effects will not be very large.
There will be no mass unemployment within the next decade.
But the data shows that the types of jobs affected by AI are much more evenly distributed, both geographically and demographically, than those affected by robots.
For example, when you look at robots, what do you think they did?
You know, they took over the jobs associated with heavy industry, like welding and painting, and that impacted the blue-collar workers.
Well, it was a specific group of men with low to moderate education, many of whom were in areas like the Midwest.
Well, the effects were very concentrated, but the AI is much more distributed.
Therefore, it is unlikely to have a significant negative impact on inequality.
But on the downside, some people dream that AI might suddenly make us more equal by helping less productive workers.
Not really. I don't see much evidence of that.
which one.
The bad news is that AI, like other automation technologies, is likely to widen the gap between capital and labor.
That would help capitalists and managers more than workers.
You know.
Well, when I listen to people in the industry, I think of Jensen Wong talking about the AI revolution and a lot of other people using that kind of language.
Of course, you've looked at the history of other kinds of economic revolutions, right?
Industrial Revolution, etc.
What can we learn from past major historical shifts in the way we work?
good?
So Jensen Wang is absolutely right.
A big revolution is happening in NVIDIA stock.
And it's a stock that's been doing better than any other stock for a while that I can remember.
We can see this from history.
New technologies often produce big winners, especially sectors and companies that provide key inputs and expertise to the new technology that benefit greatly in the early stages of the industrial revolution. Those who build the first big factories make a lot of money, and others provide GPU power, either in old formats or computing power.
They will benefit greatly, whether that be in the form of GP US or other types of chips that may be developed over the next decade or so.
But for all of us to benefit from new technologies.
A: I think it will take time and other conditions will need to be met for it to happen.
So I don't subscribe to the view that ultimately everyone will automatically benefit from new technologies. New technologies expand our capabilities, and AI certainly does that.
But depending on how you develop it and how you use it, it can create a lot of inequalities. You know, in the British Industrial Revolution, we owe comfort, health, productivity, all this industrial technology to that process.
But if you look back at the first 80 to 90 years of the Industrial Revolution, there were a lot of challenges and many people were unable to benefit.
In fact, real wages for many groups of workers, including those in the most dynamic sectors of the economy such as the textile industry, may have fallen.
To avoid that, I think we need more guardrails, institutional guardrails, and we need to be more careful about how we develop AI technology.
There are also some recent examples we would like to avoid, such as social media, its incredible technological capabilities, and improvements in communication that could have been truly transformative.
But there is a lot of evidence that shows we have created a huge amount of polarization and mental health problems.
So AI has the potential to make these things worse.
So we must be careful about how we use this very promising technology platform.
Gerone, let's take a break from talking about AI for a moment, but I'd like to hear your general thoughts on the current economic situation.
I mean, we've been talking about some of the economic data that we've had recently.
That wasn't very encouraging.
Well, I think overall, the U.S. recovery since the COVID crisis has actually been better than other developed countries, to some extent, probably even since the financial crisis.
And I think a very important point is that this is a dynamic economy, but governments have stepped in when they needed to, both after the financial crisis and after the COVID crisis. Of course, mistakes were made. You can go back and sue over whether inflation could have been contained, whether we could have done better in terms of financial regulation, etc.
But the reality is that the U.S. economy has proven stronger than nearly every other developed country.
And I think that's likely to continue in the medium term, but there will be ups and downs.
Maybe it's trending down a bit now.
Let's see how it plays out.
Daron Ajou.
Thank you very much for coming.
I am truly grateful.
Thank you for allowing me to appear on this program.
have a nice day.
