Christopher Mims reports in the Wall Street Journal:
Once you’ve solved a particular use case for an artificially intelligent bot, you’ve solved it forever. (And) the devaluation of human labor will continue up the income ladder because we have the ability to eliminate higher-paying knowledge work.The number of tasks AI can accomplish grows exponentially, while the number of humans behind it grows much more slowly, or stays flat. Now imagine that happening to every automatable task in our economy.
There is a reason we live in a golden age of dystopian science fiction: Increasingly, it feels like it is coming true. From “The Hunger Games” to “Elysium,” stories depict a world in which the trend of growing wealth and income inequality continues to its logical conclusion.
This narrative seems inevitable because it has occurred throughout history. The Luddites who attacked the automated looms that displaced them aren’t so different from the millions of truck drivers who could be displaced by self-driving vehicles.
What we’re going through now is called the fourth industrial revolution, marked by rapid innovation in automation, artificial intelligence, biotechnology, nanotechnology and other areas. Microsoft Chief Executive Satya Nadella worried it could lead to social unrest or excessive regulation. “If we don’t get it right we are going to have a vicious cycle,” Mr. Nadella said.
Economists assert that in the long run, at least, such revolutions don’t lead to mass unemployment. I’ve written previously about how automation creates more and new kinds of jobs. It is one reason America is approaching full employment, according to Federal Reserve Chairwoman Janet Yellen, despite more than two hundred years of industrialization.
Still, while there are jobs to be had, they aren’t—to put it bluntly—all that they used to be. The same economists who laud tech for increasing standards of living also note many forms of employment people are pushed into don’t pay as well or aren’t as rewarding as the old ones.
When workers lose a middle-class manufacturing or clerical job and end up in the service sector, the effect on their wages, benefits and job security contributes to what economists call polarization. In a polarized labor market, a minority of highly skilled employees—the ones who can leverage technology to be more productive—effectively replace the labor of others and are paid accordingly. Everyone else sees their fortunes dwindle.
Polarization has hit the middle class hard, but the devaluation of human labor will continue up the income ladder, says Branko Milanovic, an economist who specializes in income inequality.
That’s partly because, more than ever, we have the ability to eliminate higher-paying knowledge work. Ian Barkin, co-founder of Symphony Ventures, which helps some of the world’s largest companies automate everything from call centers to human-resource departments, says this phenomenon is known as “no-shoring.” The idea is that digitizing back-office tasks brings them back to the country in which a company operates, but without bringing back any jobs.
“One of our retail utility customers in the U.K. has about 300 robots doing 600 people’s worth of work,” said Alastair Bathgate, CEO of Blue Prism, another company that helps multinationals automate critical business functions.
“You can imagine that’s quite a big impact,” he said. “Before, you needed a building to house 600 people, but all that gets crushed down to one cabinet in the corner of a data center.”
AI could accelerate this trend, said Dennis Mortensen, CEO of x.ai, a startup that created a digital assistant smart enough to set up appointments.
In training x.ai’s artificially intelligent assistant, Mr. Mortensen’s team discovered that while they first needed to hire people with Ivy League degrees to work in the company’s Manhattan office, once the bulk of the training was done they were able to outsource the remaining work to a team outside the U.S.
Once you’ve solved a particular use case for an artificially intelligent bot, you’ve solved it forever, Mr. Mortensen said.
One way to visualize the impact of artificial intelligence is to plot the productivity of an AI and the number of people required to create and maintain it. In Mr. Mortensen’s experience, at some point the number of tasks the AI can accomplish grows exponentially, while the number of humans behind it grows much more slowly, or stays flat.
Now imagine that happening to every automatable task in our economy. A recent McKinsey study released this month concluded that 49% of the time workers spend on their jobs could be supplanted by automation, just by using technology that already exists.
It is possible this transition can be navigated more deliberately than in the past.
One solution to polarization is to address the “skills gap” in our labor market. The U.S. Bureau of Labor Statistics projects that up to one million jobs for programmers will go unfilled by 2020, for example.
Education is often touted as the answer to the skills gap, but it is generally a blunt instrument, says Muriel Clauson, an academic who hatched a solution to this problem at Singularity University.
Known as udexter, her idea is that if we study the tasks that comprise jobs, we can figure out which are automatable, and therefore which jobs are most at risk. If we also assess what skills individuals have, it should be possible for governments and companies to figure out what other jobs laid-off workers would be suited to, and get them trained in just those skills—perhaps pre-emptively.
A solution like this relies on the willingness of companies to use such a tool to retrain workers, or on laid-off workers winning a government-sponsored retraining race against increasingly adaptable machines.
The fact that so many CEOs and government leaders are talking about technological disruption of jobs, at Davos and elsewhere, means all is not lost.