In April 2018, Elon Musk announced that he planned to use fewer robots on the factory floor at Tesla, the electric vehicle company he founded. He blamed overreliance on automation for Tesla’s persistent failure to meet production volume targets for the Model 3 sedan.
“Excessive automation was a mistake,” Musk tweeted in response to a reporter’s comment about this shift. “To be precise, my mistake. Humans are underrated.”
Was Musk right? More broadly, what is the appropriate relationship between people and machines in the workplace? What level of automation will maximize business productivity and economic prosperity? Should we view automation mainly as a threat to jobs, or rather as a set of tools that can free people from mundane tasks and allow them to focus on more creative and meaningful work?
Worries about the impact of automation on employment date back at least to the fourth century B.C., when Aristotle suggested there would be less demand for musicians and weavers if lyres played themselves, and looms wove independently. Aristotle’s concerns about automation were mostly theoretical, but seem prescient in a world of self-driving cars, factory robots, and customer service chatbots, not to mention computer-generated music and automated textile mills.
Predictions that machines will supplant human workers seem both exaggerated and historically obtuse.
Automation generally means using machines to perform tasks once handled by humans. Increasingly, this technology is enabled by software that “thinks” and “learns” on its own. Aristotle may have been the first to express anxiety about the impact of automation on labor markets, but he was certainly not the last. Today, 82% of Americans expect that by 2050, robots and computers will either probably or definitely do most of the work currently performed by humans, according to the Pew Research Center.
Most of us are not bullish on this trend. The Pew survey found that three-quarters (76%) of Americans think inequality between rich and poor will increase if most jobs are automated. Only a third (33%) say it’s likely that widespread automation will create many new, better-paying jobs for humans.
These numbers help explain the unlikely rise of Andrew Yang, a political neophyte who is running for president on a platform that predicts mass job losses from automation and proposes a taxpayer-funded, universal basic income to cushion the impact on American workers. As of late November, Yang ranked 6th out of the 10 Democratic candidates still in the race, with a national polling average of 3% and $15.1 million in individual contributions, according to the New York Times.
Off the campaign trail, research on automation suggests its impact will vary widely by job category, industry and geography. A 2016 study by the McKinsey Global Institute (MGI) found that less than five percent of U.S. occupations can be automated in their entirety. However, many existing job categories involve repetitive tasks, which are likely to be automated.
The MGI study found that within 60 percent of jobs, at least 30 percent of activities could be automated by adapting currently demonstrated technologies. In order to stay employed, workers in these fields will need to acquire new skills, notably the ability to work productively with intelligent machines.
A more recent MGI report on the future of work in America found that 14.7 million young Americans have highly automatable jobs in fields like manufacturing, food service and back-office government functions. By contrast, the study predicted strong job growth in healthcare and STEM occupations, along with creative and arts management. Many of these latter industries are concentrated in big cities, which as a result will see fewer workers displaced by automation.
MGI found that 25 major urban areas have generated more than two thirds of all new jobs since the Great Recession. These cities have younger, better educated workforces. By contrast, 54 trailing cities and around 2,000 rural counties have older and shrinking workforces, higher unemployment, and lower educational attainment.
Education makes a huge difference: The report found that workers with a high school degree or less are four times as likely as those with a bachelor’s degree to be displaced by automation. And rising automation could widen all these disparities at a time when workforce mobility is at an all-time low. Fewer Americans are willing to relocate, even for a better-paying job. While there are many reasons for this trend, an important one is that living costs are much higher in the big cities that are generating most of the good-paying new jobs.
While automation tends to disrupt labor markets in the short term, it generally boosts economic growth and living standards over the longer term. Automation usually increases the productivity of labor, which lowers the cost of goods. In addition, automation has always created new job categories that eventually replaced jobs eliminated by technological disruption.
The Industrial Revolution destroyed millions of farm jobs but replaced them with new factory and office jobs. More recently, the Digital Revolution has disrupted many industries and eliminated many jobs, especially those that consisted largely of repetitive tasks. It has also created a wide array of new job categories, from app developers and user experience designers to Instagram and YouTube influencers.
“Automation’s upside is that it is the primary way that economies grow and standards of living advance,” writes Brookings economist Robert Litan. “This happens because automation is responsible for cheaper and often better goods and services. Automation’s downside is that it is disruptive, and however essential ‘creative destruction’ is to economic advance[s], when it is your job or that of a close relative or friend that is ‘destroyed,’ the ‘creative’ advantages don’t count for much.”
Supply and demand
Automation isn’t the only variable that impacts employment. Warnings about job-killing robots generally assume that automation reduces demand for human labor by replacing people with machines. They rarely address the supply side of the labor market equation. In theory, reducing the supply of labor will tend to increase demand for labor. And in fact, demographic trends suggest the U.S. labor force will shrink in coming years as the Baby Boom generation retires.
How will this impact jobs and wages? Google chief economist Hal Varian has compared the most aggressive estimates of how much automation will reduce labor demand with demographic trends that point to a shrinking workforce. In a recent panel discussion at the Stanford Institute for Human-Centered AI, Varian reported that the demographic effect on the labor market was 53% larger than the automation effect. (His model excludes unpredictable factors such as immigration policy and natural disasters.)
Varian’s conclusion: When you consider both supply and demand trends, real wages are more likely to increase than decrease.
Like Tesla, most companies are still trying to determine how best to structure the relationship between workers and machines. To that end, they are spending billions on digital transformation initiatives that seek to boost productivity, customer satisfaction and employee morale.
Based on previous chapters in the history of automation, we can expect this race to create winners and losers. Companies and governments would do well to mitigate the impact of technological disruption on workers via reskilling programs and similar initiatives. However, predictions that machines will supplant human workers seem both exaggerated and historically obtuse.
Musk said it best: Humans are underrated.