April 16, 2023
What New AI Might Do to Labor Markets
Artificial intelligence, or AI, is back in the news with the rapid adoption of a technology called ChatGPT. This is interesting, as I will shortly describe, but the economic effects are perhaps not that much different than earlier technologies, like the wheel, plowshare, harness, electricity, or computers. So, it is good to keep in mind what automation and new technology do to the workplace.
Human beings are unusual animals. We aren’t especially strong or fast, nor do we carry great weights. We really only have two advantages over every other animal on the planet. We are creatively intelligent, and we socialize extraordinarily differently than other creatures. I’d call these uniquely human skills.
The archeological record suggests our ancestors used stone tools as long as 3 million years ago. The earliest known arrowhead is more than 60,000 years old, the wheel a tad bit more than 5,300 years old, and stone plowshares 3,500 years old. What these technologies have in common with steam power, electrical power and computers is that they were designed to replace the labor far from the core of those uniquely human skills. That is mundane, difficult, and distracting tasks.
The steam power hammer allowed metal workers to more quickly craft horseshoes and iron bridge beams. The harnessing of electricity permitted trolley cars and factory conveyor belts. The microcomputer allowed impossibly large computations to be performed in minutes or seconds.
It takes humans to think of horseshoes that can leverage an animal to do something that humans do poorly. It takes a human to build cities that are sufficiently large to need transportation, and it takes humans to devise a problem that would require the linear algebra of inverting a 40 by 40 matrix. The new knowledge comes from creative intelligence and social organization, not the technology.
Artificial intelligence is the newest iteration of these technologies. It is developed to eliminate more mundane tasks, and it permits us to spend more time at those things that are uniquely human skills. Just to see for myself, I took ChatGPT for a little test. I asked it to explain to me a topic that I knew really well and was old enough for most of the information to now be in the public domain. So, I chose a subject that I wrote a 350-plus-page book upon more than 15 years ago.
I asked ChatGPT to review the research on this topic. The result was interesting, but hardly overwhelming. I got a very crisply written summary of a tiny fraction of the research on the subject. My 2008 book cited more than 200 hundred relevant studies 15 years ago, but ChatGPT cited only four studies. I asked the question again, in a more detailed way, and got the number of studies up to five. Hardly very helpful, but this time ChatGPT gave a ‘wrong’ answer, in the sense that it incorrectly stated the results of one of the studies. I know this, because I authored the study.
ChatGPT isn’t the only AI used for scientific research. There are other platforms that perform literature reviews and rewrite academic papers to make them more readable. Empirical researchers, including myself, use early versions of AI to optimize some math or statistical problems. Others use AI to interpret large bodies of texts that no one could ever read. My favorite example of this is a study team that examined Congressional Records of speeches on immigration over the past century to judge how much the support changed over time.
All of these technologies perform the same fundamental labor market tasks of the wheel, wagon and plow. They release humans from the drudgery and difficulty of work, allowing us to focus more fully on those tasks that yield themselves to uniquely human skills.
Recent sensational headlines have warned that AI will “kill humans” or “destroy humanity.” To be sure, AI will doubtless kill people. Automobiles, airplanes, and electricity kill people every day. My hunch is that it will save many, many more people than it kills. Indeed, one obvious application of AI is in early detection of cancer in screening images; it will save a stunning number of lives in the decades ahead. I feel certain there are many other similar ways AI can help improve lives from air traffic control to product quality assurance.
As with any technology, there are also costs. Cancer screening algorithms will eliminate the need for some skills of radiologists. As with past technologies, nearly all will shift their work to tasks that are more uniquely human. In the future, radiologists may need to create more algorithms for even more obscure diseases, and help develop new types of imaging.
If a millennium of history is any guide, the workers most likely to be displaced will be those with the least aptitude for human-specific skills. The wheel, harness and plow placed the biggest strain on the strong man with few other skills. The vacuum cleaner, indoor plumbing and gas stove eliminated whole classes of jobs for servants.
However, for every person who must adapt themselves to a work in which their greatest skill is rendered of little use due to a machine, many others are freed up to become better at these uniquely human skills. The single best evidence of this is the pace of technological change.
It took most of 30 million years to go from using raw stone tools to fashioning our own. It took another million years to craft metals, and then thousands of years to create machines. We went from the internal combustion engine to the Indy car in 31 years, and 68 years from powered flight in a canvas-and-wood airplane to walking on the Moon. It is no wonder that the technological changes we now experience are so head spinning.
The reason for this is not that we have become evolutionarily adept. Rather, it is simply because many hundreds of millions more humans can focus on tasks that use their uniquely human skills, while the mundane work is outsourced to machines. There are today countless geniuses sweating through 12-hour days in Chinese factories or walking behind oxen in Africa. The faster we unleash their uniquely human skills, the better off we all become.
Technological innovation continues to offer the most enduring opportunity for our species. All we must do is make sure our children are learning well enough to develop their uniquely human skills and not wasting education on learning those tasks that are soon to be redundant.
About the Author
Educational Attainment, the 21st Century Fund and the Future of SchoolingIndiana ranks 42nd in educational attainment.
Big Savings for Ending Prevailing WageMy statistical models show that repealing state prevailing wage laws save taxpayers money.
Re-Thinking Economic Development A large share of the most mobile families—perhaps half—no longer need to live near where they work.
Money Illusion and InflationPrice fluctuation could cause inflation to last longer, but it didn’t cause the inflation, it simply extends the pain.View archives