AI could spur an economic boom. The Humans are in the way.

Photo of author

By georgeskef

Artificial intelligence can drive productivity and growth but how people adapt will determine when this happens. ChatGPT was asked how soon artificial intelligence would boost the U.S. Economy. The answer was mostly non-committal, “challenging for a precise prediction.”

Many economists would answer the same way

The increasing popularity of generative artificial-intelligence tools–ones, such as OpenAI’s ChatGPT, that create new content such as text and images–by consumers and companies has some economists predicting the technology will revolutionize workplaces and spur economic growth by lifting productivity, or output per hour, out of a long lull.

This is based only on the theoretical potential. The gains could be smaller or take longer to materialize depending on a number of factors. The first is how widely technology is adopted, and the degree to which people are skilled in using it. Once the technology has been adopted, it is important to see how companies are able to translate it in higher productivity.

Goldman Sachs’ economists believe that generative AI can boost U.S. annual productivity growth by 1.5 percentage points in the first 10 years following its widespread adoption. This would be roughly twice the average rate of productivity growth since late 2007. They said that it would also result in a similar size bump in growth of the U.S. Gross Domestic Product over the same period. Federal Reserve officials, for example, put the U.S. growth rate over time at 1.8%. This could theoretically boost growth up to 3.3%.

Note the caveat, however: This is only after it has been widely adopted. Joseph Briggs is a senior economist with Goldman Sachs. He says it’s hard to say when this will happen. Briggs, a senior economist at Goldman Sachs, said that the boost AI can provide to productivity growth may be as little as 0.3 percent or as much as 2.9 percent due to uncertainties surrounding companies’ adoption timelines and AI’s final capabilities.

First, breakthroughs and then adoption

He noted that there’s usually a delay between a breakthrough in technology and its widespread adoption. AI could begin to have macroeconomic effects as early as the 2030s.

Briggs stated that “the productivity gains were significant enough, given the pace of productivity growth trend, to be a positive economic outcome” in all the scenarios they considered.

According to the Labor Department, productivity grew an average of 1.4% annually between the fourth quarterly of 2007 and 2019. This was the end of the previous business cycle. This was lower than the long-term average of 2.1% between 1947’s first quarter and this quarter.

The key to economic growth is productivity gains, which means that the same number workers can produce more goods or services and ultimately raise living standards.

Lessons from electricity and internet

Past examples show that the productivity boost from new technologies can be gradual. A 1990 paper written by economist Paul David showed that it took decades for half of the U.S. factory mechanical drive capacities to be electrified.

David used the trajectory of electricity as a comparison in history to explain why productivity was slow at the time when he wrote his paper, despite computers’ rise.

David pointed out that it was unprofitable to overhaul existing manufacturing plants which were still running on steam and water.

Anton Korinek is a professor of Economics at the University of Virginia. He said that certain factors could accelerate the pace of adoption of generative AI by industries and businesses compared to the last productivity boom, which occurred in the late 90s and early 2000s when the internet and personal computers took off. Then, it was necessary to purchase physical equipment such as routers or internet connections to benefit from the new technologies.

Korinek explained, “Now that we have all these connections in place, we only need to log on to new websites.” He estimates that generative AI could boost productivity between 10% and 20% in the next 10 to 20 years, compared to the current trend.

Economists say that it may take some time before AI effects are seen in productivity measures as workers and companies learn to use AI and integrate it into their workflows.

Indeed, in an April Wall Street Journal survey of economists, 61% said they expected artificial-intelligence tools, such as ChatGPT, to have only a small positive effect on U.S. GDP growth over the next five years. Another 29% expect AI tools to be completely ineffective over the same time period.

If you look at the history and adoption of new technologies, it is clear that they have to be combined with other complementary assets in order to make a difference. Robert Seamans is a professor at New York University who teaches management and organization. He said that you have to adopt technology along with complementary assets. “That is expensive and takes time.”

Seamans explained that companies need to hire or train workers who are able to understand the capabilities of generative AI and existing production processes in order to link them.

Internet was developed by the Defense Department back in the 1960s, but it didn’t really take off until 1990s when personal computers became widely available. Its most disruptive effects were only seen later on with smartphones and broadband access.

Some industries are particularly ready to figure out all of this. A 2023 Working Paper by economists Erik Brynjolfsson and Danielle Li found that call-center employees’ productivity increased by 14% in average when using AI assistants. The AI tool gave call-center staff real-time advice on how to best answer questions in their conversations.

Uneven adoption, unequal benefits

Goldman’s study estimates AI could automate a quarter in the U.S. with a high exposure to the legal and administrative industries. Goldman says that physically demanding jobs, such as maintenance and construction, would be less exposed.

Brad Hershbein is a senior economist with the Upjohn Institute which does employment research.

He said that it is more common for new technologies to change the tasks of existing jobs than to eliminate them.