Last month IBM announced that it was selling part of its Watson Health business to a private equity company, reportedly for a fraction of the price it had invested. Watson’s artificial-intelligence—which wowed the world when it defeated two Jeopardy! game-show champions in 2011—had been touted as revolutionary for health care, and used to analyze mammograms and MRIs, communicate with patients, and analyze complex health-care data.
If Watson Health had delivered on the most optimistic expectations around its 2015 launch, it’s easy to imagine reduced demand for certain medical workers around the world, or at least significant changes to their jobs. But the sale was at least partly an acknowledgment that Watson Health hasn’t yet lived up to the initial hype.
That’s the kind of scenario that the authors of the new book The Work of the Future might point to, as they offer reassurances that technology advances generally only impact workforces over periods of decades, not years. Stemming from MIT’s task force on the future of work and written by David Autor, David A. Mindell, and Elisabeth B. Reynolds, the book argues that the pace of change might be slower than some fear.
The authors dispute the straw-man hypothesis that automation will broadly replace human workers over time and they downplay short-term concerns. “We anticipate that in the next two decades, industrialized countries will have more job openings than workers to fill them,” they write. (p. 7) While the authors forecast that “robotics and automation will play an increasingly crucial role in closing those gaps,” they acknowledge that technology “will not necessarily benefit all workers.” (p. 135) They add that policy changes are needed to ensure that individual workers have dignity, opportunity, and economic security.
At its core, The Work of the Future recounts a familiar history of US economic developments that have pressured middle-class employment, labor unions, and worker protections over the past four decades. “New technologies will evolve within a nation of empowered high-skilled work, increasing inequality, eroded worker voice, and racial disparities,” the authors write. (p. 39) Autor, Mindell, and Reynolds ultimately are on the side of more technology, not less—with their concern focused on the economic and policy context that already has reduced too many Americans to low-quality, low-wage jobs with little protection from any labor market disruption.
They make the case for a basket of policy measures, including:
- More and better worker training programs—These should focus on the lower-wage and less-educated workers to provide pathways to better jobs, subsidized through tax breaks.
- A higher federal minimum wage—They propose that it should be at least 40% of the national median wage, and rise with inflation.
- Modernization of unemployment insurance—This includes relaxing eligibility requirements and allowing the unemployed to pursue part-time rather than full-time work.
- Rebalancing taxes on capital and labor—Employers currently pay lower taxes on machinery than they do on their spending on labor, incentivizing them to replace jobs with automation.
- Targeting government investments in research toward technology applications that augment rather than replace workers—These could include human-centered AI and the science of learning and education.
“Whether rising productivity generates broadly improving living standards or instead enriches a relatively small subset of the population depends on the societal institutions that channel productivity into incomes,” write Autor, Mindell, and Reynolds. (p. 16)
They recount research compiled by the MIT task force on work that found relatively modest deployment of automation in factories and warehouses today. They point to challenges in developing generalized artificial intelligence and robotic arms with the dexterity to grasp a broad variety of objects. And they predict that headline-grabbing technologies will find broad adoption more slowly than some believe, for example forecasting that autonomous trucks and cars will have limited use by 2030.
“The ability to adapt to entirely novel situations is still an enormous challenge for AI and robotics and a key reason why companies continue to rely on human workers for a variety of tasks,” the authors write. (p. 44) Also, businesses have trouble defining processes in order to automate them.
In the long run, however, automation tends to create jobs, the authors contend, by driving productivity increases that raise total income in the economy and spurring new kinds of work.
To be sure:
- History will better tell who’s right, but the authors don’t share the urgency of writers such as New York Times columnist Kevin Roose, whose 2021 book Futureproof highlighted corporate executives’ enthusiasm for automation as a way to cut labor costs. Roose recounted the story of one AI startup founder selling “Boomer Remover” automation that would allow companies to lay off “old, overpaid middle managers.” And he cited a top tech executive who told him clients were discussing shedding 99% of their workforces through automation.
- The Work of the Future doesn’t acknowledge the recent wage gains of workers amid the current tight labor market, and focuses on the abysmally low federal minimum wage, whereas state minimum wage hikes and private-sector minimum pay increases are increasingly determinant for US workers.
- The book reads like a university task force report in places, with its didactic cataloging of various sources of worker training and contributions of federal research spending.
Memorable anecdotes and facts:
- Two-thirds of the new jobs forecast to be created between 2019 and 2029 are in occupations paying below median wage. Home health and personal care aides, fast-food workers, and restaurant cooks are the roles projected to see the greatest increases. Cashiers and secretaries are among those with the biggest net declines.
- Low-skilled US workers earn 74% as much as low-skilled Canadian workers and 57% as much as low-skilled Germans.
- One insurance company used AI to audit its legal bills, achieving 85% accuracy and gaining millions of dollars in annual savings by picking up anomalies the human auditors missed.
- Only 15% of warehouses used autonomous mobile robots, according to a 2019 survey.
- The roughly 1,100 community colleges in the US enroll about 7 million students in for-credit courses each year.
- From 1950 to 1970, the US government funded almost half of semiconductor industry research and development.
The bottom line is that The Work of the Future suggests that rather than slow automation for fear that it will destroy jobs, the US needs to improve the conditions for workers that have eroded dramatically over the past 40 years. It is not a breakthrough thesis, and arguably doesn’t acknowledge the pace at which automation is already contributing to making working conditions worse. But the book provides a valuable catalog of specific policy initiatives that could make a difference in its final chapter.