What AI means for growth, incomes and inflation

• This would parallel the late 1990s growth surge spurred by a wide range of technological advances, or (in more grandiose scenarios) the results of electrification in the first half of the 20th century.

There is plenty of evidence that AI can improve worker productivity in specific scenarios. One study of a customer support center showed an average 14% gain in workers' output per hour when assisted by an AI bot, for example.

• The open question is to what extent, and how fast, that will scale to an economy-wide boost in the ability of companies to produce more goods and services relative to the work put in.
• Moreover, argue Martin Neil Baily, Erik Brynjolfsson and Anton Korinek in a Brookings Institution paper, AI-based tools could pay dividends not just in making existing work more efficient, but in making companies better able to identify productivity-enhancing strategies.

By the numbers: If AI increases the productivity of all cognitive work — labor using words, numbers and ideas — by 30%, and cognitive work accounts for 60% of all labor, multiplying the two implies an 18% gain in overall productivity over time.

• If that gain were to take place over 10 years, it would imply an extra 1.7 percentage points of productivity growth a year — more than double current mainstream forecasts. (The Congressional Budget Office puts it at 1.5%.)
• That could translate into commensurately higher GDP growth.
• Incomes should rise similarly — on average, at least. There would be winners and losers in this scenario, as some jobs are eliminated and others become more productive and thus more lucrative.

Yes, but: New technologies do not pay productivity dividends overnight, as it takes time for companies to make effective use of them. In 1987, economist Robert Solow famously remarked that "you can see the computer age everywhere but in the productivity statistics."

• It wasn't until the mid-1990s that IT started to drive a productivity surge.

What they're saying: Brynjolfsson, an economist at Stanford, argues that the AI-driven productivity revolution is at a stage more equivalent to 1995 than to 1987.

• "It's not in the laboratory, it's not just some technical benchmark, it's people really using it and changing the way they do business," he tells Axios.
• He argues that this set of innovations is poised to pay productivity dividends in the very near future. Compared to the 1980s and '90s IT revolution, "this time I think the takeoff is going to be faster because there's not as much complementary investment that is required to get these technologies to work."

If that world were to materialize, it would likely have profound effects on interest rates and inflation dynamics, as well.

• However, JPMorgan chief U.S. economist Michael Feroli argues those effects could cut in opposite directions. Stronger trend growth should coincide with real (inflation-adjusted) interest rates — but the benefits of higher productivity could bring inflation down in the short run.

The intuition: "Imagine your real income is expected to grow at a 2% rate and then something comes along so that you now expect it will grow at a 4% rate," Feroli writes in a note.

• In this scenario, "Most people would likely want to spend more today. But society can only spend today what is produced today," he writes. "To maintain equilibrium, interest rates need to increase to reduce current desired demand."

As for inflation, when there is a supply shock — something that increases the productive capacity of the economy — it tends to exert a downward pull on inflation, at least temporarily.

• Past examples include the late 1990s productivity surge, the expansion of trade with China in the 2000s and the shale gas revolution in the 2010s.
• In effect, the economy can suddenly produce more while demand changes only slowly, keeping a lid on price increases.

The bottom line: If AI is as big a deal as many people think it could be, look for higher growth, higher incomes, higher real interest rates and lower inflation.