A few weeks ago, MIT News published an article on “What do we know about the economics of AI?” The article is based on the research of MIT Nobel laureate professor Daron Acemoglu, who in October was awarded the 2024 Nobel Memorial Prize in Economic Sciences along with his MIT colleague Simon Johnson and University of Chicago economist James Robinson. “For all the talk about artificial intelligence upending the world, its economic effects remain uncertain,” said the article. “There is massive investment in AI but little clarity about what it will produce.”
Professor Acemoglu has long been conducting research on the impact of technology on the economy. “Artificial intelligence (AI) has captured imaginations. Promises of rapid, even unparalleled, productivity growth as well as new pathways for complementing humans have become commonplace,” he wrote in “The Simple Macroeconomics of AI,” a research paper published in May of 2024.
“AI will have implications for the macroeconomy, productivity, wages and inequality, but all of them are very hard to predict. This has not stopped a series of forecasts over the last year, often centering on the productivity gains that AI will trigger. Some experts believe that truly transformative implications, including artificial general intelligence (AGI) enabling AI to perform essentially all human tasks, could be around the corner. Other forecasters are more grounded, but still predict big effects on output.”
Let me discuss some of the key topics covered in the MIT News article.
What are the measurable effects of AI?
According to Acemoglu, AI Advances Aren’t Likely to Occur Nearly as Quickly as Many Believe. AI will contribute only modest improvements to worker productivity and will add no more than 1 percent to U.S. economic output over the next decade. Based on his research, he is skeptical of the significantly higher estimates made by AI boosters. US GDP growth has averaged 3% annually since 1947, while productivity growth has average 2% annually. Some of the more bullish predictions have claimed that AI will double GDP and productivity growth over the next decade.
A lot of that growth is predicted to come from the deployment of new AI applications across economies. Where will the AI-based GDP and productivity growth come from?, he asked. “I don’t think we know those yet, and that’s what the issue is. What are the apps that are really going to change how we do things?” Some recent studies have found that only about 20% of the tasks in US jobs might be exposed to AI. Other studies found that about 23% of computer vision tasks could be profitably automated within the next 10 years, and that the average cost savings from AI is about 27%. “I don’t think we should belittle 0.5 percent in 10 years. That’s better than zero,” said Acemoglu. “But it’s just disappointing relative to the promises that people in the industry and in tech journalism are making.”
How different is the U.S. economy going to be in 2030 because of AI?
Many forecasts of AI describe its impact as revolutionary. “You could be a complete AI optimist and think that millions of people would have lost their jobs because of chatbots, or perhaps that some people have become super-productive workers because with AI they can do 10 times as many things as they’ve done before. I don’t think so. I think most companies are going to be doing more or less the same things. A few occupations will be impacted, but we’re still going to have journalists, we’re still going to have financial analysts, we’re still going to have HR employees.”
By 2030 AI is likely to mostly impact tasks in white-collar jobs that deal with large amounts of data, where AI can analyze a lot of inputs significantly faster than humans. But those jobs constitute about 5 percent of the economy.
Increase worker productivity or worker replacement?
Acemoglu argues that we currently have the wrong direction for AI. We’re using it too much for automation aimed at replacing workers instead of using AI to provide expertise and information to workers in order to make them more productive. “It is the difference between, say, providing new information to a biotechnologist versus replacing a customer service worker with automated call-center technology. So far, he believes, firms have been focused on the latter type of case.”
In a 2019 paper, Acemoglu and Boston University professor Pascual Restrepo warned about the rise of so-so technologies, which they defined as automation technologies that are just good enough to be adopted but not more productive than the workers they are replacing. Their paper warned that “as we go deeper and deeper into AI-based automation, we are moving into areas in which human labour is quite good, and machine productivity, at least to start with, is not always impressive, to say the least. Automation technologies aimed at substituting machines for humans in these tasks are thus likely to be of the so-so kind. As a result, we cannot even count on powerful productivity gains to increase our living standards and contribute to labour demand.”
What’s the best speed for innovation?
One may generally assume that if a technology helps generate economic growth, then fast-paced innovation might seem ideal, by delivering growth more quickly. Should this be the case for AI?
In a recent paper, Acemoglu and his MIT doctoral student Todd Lensman developed a framework for analyzing the optimal adoption strategy for a major transformative technology like GenAI which promises to accelerate productivity growth across just about all sectors of the economy, but which also presents major new risks to society from its potential misuse.
Their analysis concluded that “If a disaster occurs, some of the sectors using the new technology may be unable to switch back to the old, safe technology. Whether a disaster will occur is unknown, and society gradually learns about it over time. Consequently, adoption should be gradual, … initially growing slowly before accelerating later. Most surprisingly, a faster growth rate of the new technology should lead to slower adoption when potential damages are large.”
Their model of innovation adoption is “a response to a trend of the last decade-plus, in which many technologies are hyped as inevitable and celebrated because of their disruption. By contrast, Acemoglu and Lensman are suggesting we can reasonably judge the tradeoffs involved in particular technologies and aim to spur additional discussion about that.”
How can we reach the right speed for AI adoption?
Acemoglu offers some suggestions in the MIT News article. One possible role is government regulation of AI. “However, it is not clear what kinds of long-term guidelines for AI might be adopted in the U.S. or around the world.”
But, if the hype around AI diminishes, then the rush to deploy it will naturally slow down. This possibility may well be more likely than regulation because companies and financial markets are not seeing the returns that justify their large investments in AI, as is often the case in the early years of a major new technology, — remember the 1990s dot-com bubble.
A number of financial experts share these concerns. For example, a recent NYT article, “Will A.I. Be a Bust?,” wrote about the views of Jim Covello, Goldman Sachs’ Head of Global Equity Research. Covello jolted markets based on an interview published in a June, 2024 Goldman Sachs Research report, “Gen AI: Too Much Spend, Too Little Benefit?.”
In the interview, Covello challenged whether business and investors would see a sufficient return on what by some estimates would be over $1 trillion in AI spending over the next few years. He argued that to earn a reasonable return on investments (ROI), AI applications must solve highly complex and important enterprise problems given the capital investments required in specialized chips, large data centers, and electric utilities. The crucial question is: “What $1tn problem will AI solve?,” he asked. “Replacing low-wage jobs with tremendously costly technology is basically the polar opposite of the prior technology transitions I’ve witnessed in my thirty years of closely following the tech industry.”
Covello added that truly life-changing inventions like the internet enabled low-cost solutions to disrupt high-cost solutions even in their infancy, unlike costly AI tech today. While we can debate whether AI will deliver on the promise many people are excited about, “the less debatable point is that AI technology is exceptionally expensive, and to justify those costs, the technology must be able to solve complex problems, which it isn’t designed to do.”
Hype is a tangible aspect of the economics of AI, said Acemoglu in his concluding MIT News observations, “since it drives investment in a particular vision of AI, which influences the AI tools we may encounter. The faster you go, and the more hype you have, that course correction becomes less likely. It’s very difficult, if you’re driving 200 miles an hour, to make a 180-degree turn.”
Comments