One of the questions is whether higher productivity will make us richer or rather reduce the number of jobs.
The Potential Impact of AI on the U.S. Economy and Budget Forecasts
Will artificial intelligence (AI) transform the economy? For today, I thought it would be good to take a break from my usual topic of current crises and better look a bit at how technology might change the economic landscape in the years ahead, especially on a topic that, despite its relevance, has not received much attention: how might AI change U.S. budget forecasts?
Starting last fall, we noted that AI generated a lot of hype, both positive and negative. That hype seems to have died down to some extent, as the use of ChatGPT, the most famous application of the technology, has declined somewhat in recent months. In addition, many more observers have realized that what we have called AI (or what more thorough people call “generative AI”) is not actually intelligence. Rather, it’s an extrapolation of pattern recognition or, as some people I’ve talked to explain it, it’s essentially an enhanced auto-corrector.
Of course, that doesn’t mean it’s not important. After all, arguably much of what human workers do, even highly skilled ones, is also enhanced autocorrect. How many workers regularly devote time to creative thinking? Even among creative workers, how much time do they spend on being creative and not on pattern recognition activities?
I say this not to disrespect those who work in knowledge, but to explain why it seems to me that what we call AI could be a very significant technology for the economy, even if it doesn’t lead us to create HAL 9000 or Skynet.
Obviously, no one really knows. Some people interested in estimating that impact started their observations from the lowest hierarchical levels, in different types of work, and so formed guesses about the proportion of that work that might be replaced or augmented by AI. The most widely circulated estimates are those made by Goldman Sachs, whose base case is that AI will raise the growth rate of productivity (output per man-hour) by almost 1.5 percentage points per year for a decade, so that it will total about 15 percent over that period.
Is this possible? The truth is that it is. A parallel, for anyone who has studied the historical relationship between technology and productivity, is the boom in productivity observed between 1995 and 2005, which followed decades of low productivity growth.
By the time that increase in productivity dissipated, it was about 12 percent higher than would have been expected based on the trend of the previous two decades. Since AI is arguably an even more profound innovation than the technologies that drove the boom from 1995 to 2005, 15 percent does not seem unacceptable.
But whether this higher productivity will make us richer or rather reduce the number of jobs is another question. The fear of technological unemployment (a term invented by none other than John Maynard Keynes in 1930) goes back at least to the early 19th century. It even inspired Kurt Vonnegut’s excellent novel “The Pianola”. However, while technology has indeed eliminated some jobs, historically it has always been, as Keynes wrote, “a temporary phase of mismatch” followed by the emergence of other forms of employment to replace those lost. For example, it seems that Microsoft’s Excel jolt (the emergence of spreadsheet programs) eliminated many accounting jobs, but, later, they were replaced by more jobs, including in the area of financial analysis.
However, while there is no reason to believe that what we call AI will lead to mass unemployment, it could affect people who are displaced from their jobs and have trouble finding another job or are forced to accept a lower salary. Who might be the losers?
The answer is most likely to be that the most noticeable consequences will be seen for people in relatively exclusive managerial jobs, many of whom are currently earning high wages, but most manual jobs will not be affected.
Now, while this prediction seems accurate in the case of generative AI, there are other applications of big data that may affect manual labor. For example, with all the fuss over ChatGPT, relatively little attention has been paid to the fact that, after years of failed advertising deployments, autonomous vehicles have already begun to enter service. Either way, at this point it seems more likely that AI, in contrast to the technological advances of the past 40 years, is a more influential factor for inequality in the low-income sector than in the high-income sector.
Finally, it may be worth considering the effects of generative AI on an issue that has come to the fore again: concerns about government debt.
Until recently, many economists, including myself, claimed that government debt was less of a concern than many imagined, because the interest rates charged on debt were below the economy’s long-term growth rate, “r<g.” So the common idea that the debt would inflate tremendously because interest payments would increase the debt and thus raise interest payments even more was wrong: in reality, the ratio of debt to gross domestic product, the number that really matters, tends to melt rather than inflate.
That said, the rapid rise in interest rates has made debt a much greater cause for concern. Conventional estimates of the long-term sustainable growth rate of the economy, such as those of the Federal Reserve, usually indicate that it will be around 1.8 percent. And real interest rates on the federal debt are already above that figure.
Strangely, however, talk of debt sustainability has no bearing on the generative AI discourse. In fact, I’m pretty sure there is no shortage of people who have warned about a debt crisis and massive unemployment because of AI, although I haven’t made any effort to find them. But if the optimistic estimates about the boom produced by this technology are correct, growth will be well above 1.8 percent over the next decade and debt won’t cause as much concern after all (especially since if growth is faster, it will boost sales and reduce the budget deficit).
Of course, these are all mere speculation. No one really knows what the impact of AI will be. But, I repeat, it doesn’t have to be “real” artificial intelligence to have much influence on the economy and, if we venture a forecast, it most likely does matter a lot.