Let’s say for now that the day comes when robots and artificial intelligence can outperform human beings at every conceivable job, from waxing floors to waxing eyebrows to waxing philosophical at a lectern. Will there still be work for people?
There could be, says Noah Smith, a blogging economist. “It’s very possible that regular humans will have plentiful, high-paying jobs in the age of A.I. dominance — often doing much the same kind of work that they’re doing right now,” he wrote Sunday on his Substack.
I ran Smith’s argument by several economists who think a lot about these issues, and they were skeptical. But there’s so much pessimism around the future of work these days that Smith’s take comes as a welcome ray of sunshine. That’s why I’m taking it seriously.
Smith rests his case on one of the most beautiful theories in economics: comparative advantage, which boils down to “do what you’re best at.” The theory implies that even if Martha Stewart has an absolute advantage in ironing shirts — she does it better than anyone else in the world — she should still have someone else iron her shirts for her, because her time is better spent on something she’s even better at, namely producing TV shows. (I took this example from a Marginal Revolution University video.)
The relevance to artificial intelligence is obvious. To maximize profit, the owners of A.I. will want to put it to work doing things it’s a million times as good at as people, and not bother with jobs for which it’s only twice as good as people. That leaves space for human beings to keep working for a living rather than taking handouts.
The theory of comparative advantage would hold even if A.I. got super cheap, Smith argues. As long as the computing power available for A.I. is a scarce resource, its owners will always want to put it to its highest and best use. Which is probably not going to be teaching yoga or writing this newsletter.
Problem solved! Or actually not quite. Smith acknowledged one “important and scary” limitation to his argument. That’s the fact that people compete with computers for resources. The resources consumed to employ people, such as electricity for their offices, might be more profitably used to power more artificial intelligence.
Smith referred to a 2013 piece by the economist Tyler Cowen, who pointed out that businesses didn’t keep using horses for transporting goods once trucks were invented, even though horses had a comparative (not absolute) advantage in pulling things. Horses were sent to the glue factory. Not a pleasant thought for human workers.
David Autor, an economist at the Massachusetts Institute of Technology, also used the example of unneeded horses in an email response to my questions. Human workers “have a real cost of upkeep,” he wrote. “Accordingly, humans might become a noncompetitive factor of production for any activity.”
Autor said he didn’t believe that robots and A.I. would become better than people at everything, which is the scenario we’re exploring here. If they did, though, he wrote, “it’s likely that workers would eventually become ‘too expensive to employ’ — or would earn too little to cover their own costs of upkeep.”
Autor is a faculty co-director of M.I.T.’s Shaping the Future of Work Initiative, along with fellow economists Daron Acemoglu and Simon Johnson. Acemoglu emailed me that he, like Autor, is optimistic about a continuing role for people in the labor market. “An upper bound of the fraction of jobs that would be affected by A.I. and computer vision technologies within the next 10 years is less than 10 percent,” he wrote.
Like Autor, though, he said that to the degree that A.I. does get better than people at everything, one shouldn’t expect human labor to gain market power or even hold its own.
Ethan Mollick, who teaches innovation and entrepreneurship at the University of Pennsylvania’s Wharton School, has a new book, “Co-Intelligence: Living and Working With A.I.” He writes that A.I. is really good at exactly the types of things human beings pride themselves on, such as creativity. “Research by economists Ed Felten, Manav Raj and Rob Seamans concluded that A.I. overlaps most with the most highly compensated, highly creative and highly educated work,” Mollick’s book says.
That would imply that the jobs A.I. will leave for us human beings will be the scut work. Mollick isn’t that pessimistic himself, but he did tell me that everyone who gets deep into A.I. has “at least one existential crisis,” where his or her economic value as a flesh-and-blood being is questioned.
I asked Smith by email what he thought of the comments by Autor, Acemoglu and Mollick. He wrote that the future of human work hinges on whether A.I. is or isn’t allowed to consume all the energy that’s available. If it isn’t, “then humans will have some energy to consume, and then the logic of comparative advantage is in full effect.” He added: “From this line of reasoning we can see that if we want government to protect human jobs, we don’t need a thicket of job-specific regulations. All we need is ONE regulation — a limit on the fraction of energy that can go to data centers.”
That made me optimistic about jobs again — until I interviewed Pascual Restrepo, an associate professor of economics at Boston University. Autor recommended him because he recently gave a presentation about humans in the age of A.I. titled, “You Won’t Be Missed.”
Restrepo likened the conquest of jobs by A.I. to a rising tide that covers one rock at a time until eventually they’re all underwater. “You can still put in your time and call it work,” he said. “But its value will be a shrinking share of all the value that’s created.”
Comparative advantage is real, but it won’t stop A.I. from taking over everything if it’s better at everything and computing keeps getting cheaper, Restrepo said. We can still work, but by working, we are at best saving the A.I. some computational resources. When we apply our puny brains to a problem for an hour, we’re sparing the A.I. from having to spend a split second on it, enabling the A.I. — like Martha Stewart — to spend that split second on what it does best, which is generating enormous amounts of value from things we can only distantly perceive today. The owners of A.I. will happily pay us for that service.
The payment we get in terms of goods and services could look quite large by today’s standards. But it would be tiny compared with the overall wealth in society, most of which would be created by A.I., Restrepo said. So, people wouldn’t need jobs. Work would cease to be central to our lives. The bright side? If the gains are shared widely (not a certainty), we’re talking about a future of unimaginable luxury for all.
The Readers Write
Dear Readers: Thank you for telling me about the coincidences in your lives. Many of your stories were amazing. Then again, as some skeptics pointed out, even extremely unlikely things are bound to happen occasionally. I sense from your mail that perceiving coincidences as meaningful often comes from a yearning for connection — to a lost loved one, to an old friend, to a younger self.
You wrote that rolling 10 sixes is no less probable than rolling 6-2-4-5-1-3-2-5-4-6. I disagree, based on a statistical method called chi-square goodness of fit. The series of sixes has such low probability that we can infer a loaded die, while the other pattern is highly consistent with chance (97.7 percent likelihood that the die is fair). While human pattern bias is certainly real, sometimes something that smells funny truly is.
Shannon Walker
Indianapolis, Ind.
I am in the “just a coincidence” camp. To explain, I recommend the hilarious Spurious Correlations website, which is full of meaningless coincidences. For example, Cheddar cheese consumption in the United States is highly correlated with solar power generation in Haiti, and petroleum consumption in Azerbaijan is highly correlated with the distance between Neptune and Mercury.
Brian Brady
Seal Beach, Calif.
I specialize in twin studies. Identical twins display more “coincidences” than fraternal twins. The rarer the behavior, the more likely it is to signify a meaningful connection that is likely to partly reflect genetic underpinnings. For example, one pair of identical twins raised apart met at age 25. Both twins used a rare Swedish toothpaste called Vademecum.
Nancy L. Segal
Fullerton, Calif.
The writer is a psychology professor at California State University, Fullerton.
Regarding your piece on the Realtors settlement: As a recent seller of a property in Australia, I paid 1.8 percent to our selling agent. When we want to buy a property, we just usually go online to look for listings of properties ourselves, or buy the Saturday newspapers, where the properties for sale are usually published. The operative word being “ourselves.” We don’t pay someone else to look for us or represent us to buy our property for us, and therefore we don’t pay a buyer agent’s commission.
Khoon Tan
Sydney, Australia
Why is it that the conversation about retirement never addresses personal responsibility? At age 73 I am living very comfortably, with an in-city condo and a small summer farm in the country. I circumnavigated the world when I was 28, sleeping in the jungle in Thailand and in caves in Crete when I had to, and wheeling and dealing on the streets of Hong Kong when I needed cash. Today, I receive Social Security payments well below the national average, have no pension plan and minimal savings in the bank, but have no debt. I dress in designer clothes, which I buy at thrift stores. I make my own coffee in the morning. I am a member of a reciprocal museum association, which allows me to view the finest art for the cost of a single museum membership. I don’t belong to a gym but have exercise resistance bands that cost me $6. Sit-ups and push-ups don’t cost me anything. Walking doesn’t cost me anything. No “crisis” here.
Matthew Cassis
Damariscotta, Maine
Quote of the Day
“A gun is not an argument.”
— Ayn Rand, “Capitalism: The Unknown Ideal” (1966)