Note to Moderator: This is from an email message, not from either a printed or an email issue of Technology Review, the MIT Alumni magazine. It is included in an issue, however, with considerable additonal material.
There’s a growing fear among many American workers that they’re about to be replaced by technology, whether that’s a robot, a more efficient computing system, or a self-driving truck. It’s an anxiety palpable in pop culture, national polls, and news stories; on presidential debate stages; even in the halls of the Institute. “People are coming to MIT and literally saying, ‘I need MIT to tell me—is a robot going to take my job?’” says David Mindell, PhD ’96, a professor of aero-astro and the Dibner Professor of the History of Engineering and Manufacturing.
To address the anxiety, a new MIT report, “The Work of the Future: Shaping Technology and Institutions,” explores the legacy of technology in the American workplace to help us understand what lies ahead. No, robots aren’t about to take over all the work, but technological advances have devastated many middle-class jobs, contributing to record levels of inequality and significantly reduced social mobility. And if we don’t do anything, the future will be even worse.
The report is the first step in “finding a more coherent way to talk about these topics,” says Mindell, one of the co-chairs of the task force behind the report. “If MIT doesn’t have a voice on the subject, then who will?”
MIT president L. Rafael Reif first convened the Task Force on the Work of the Future in 2018. He wanted to help address what he describes as “the sense of powerlessness—the worry that automation is ‘coming to get us’ automatically,” and that no recourse is available, he said in September at a briefing announcing the report in Washington, DC. “It is up to those of us advancing new technologies to help make certain that they do not wind up damaging the society we intend them to serve.” The task force was made up of 18 faculty members from nearly a dozen departments, supported by an advisory board, a research board, and a number of staffers and graduate students.
Men without college degrees working full time made less per week on average in 2018 than they did in 1980.
President Reif charged the group with investigating the relationship between work and technology. Through surveys, case studies, historical analysis, and other research projects, task force members aim to restore some of that lost power to American workers—and offer guidance to the governments that serve them and the enterprises that depend on them.
The task force’s interim report makes clear from the outset that current American fears about how technology will affect the workforce are “neither misguided nor uninformed.” Indeed, they’re based firmly in real economic trends.
Over the last four decades, “ordinary software, computers, and information technology have transformed everyone’s work,” says Mindell. Partly as a result, US productivity—the measure of how efficiently the economy as a whole is using its resources—grew by 75% between 1973 and 2016. The report calls that “healthy,” although it is smaller than the increase during the first three decades after World War II, and the pace of growth has slowed considerably in the past decade and a half.
Major productivity gains didn't boost paychecks for blue-collar workers
Percent changes in labor productivity and compensation, 1948-2016
But the benefits have been unequally distributed. As productivity rose over those 44 years, the median wage increase was a mere 11%. “Productivity growth did not translate into shared prosperity,” the report explains, “but rather into employment polarization and rising inequality.” Indeed, income inequality has soared: in 1973, the top 1% of earners in the US accounted for 11.0% of national income while the bottom 50% had 20.3%. In 2014, the most recent year for which data is available, the top 1% earned 20.2%, while the bottom half had just 12.6%.
A major factor in whether automation helps or harms a particular person professionally is education, Mindell says—and those who lack a four-year college degree typically lose out.
Technology tends to make workers in high-skill jobs—the term economists use to refer to jobs that require a lot of education, experience, and abstract thinking—even more productive. For example, an architect can use a design program to create blueprints more quickly, thus completing commissions faster and earning more money. But at the other end of the spectrum, technology currently isn’t much help when it comes to performing many so-called low-skill jobs, such as those of retail clerks and cleaners. These jobs require physical dexterity, face-to-face communication, visual recognition, and situational adaptability—things that today’s computers are generally terrible at but humans can do with relatively little training. So although most low-skill jobs remain safe from the threat of automation, few low-skill workers are reaping any of the benefits of technological innovation. In other words, as the report points out, “digitalization has had the smallest impact on the tasks of workers in low-paid manual and service jobs.”
Middle class jobs are disappearing
Percent changes in occupational employment shares among working-age adults, 1970-2016
Meanwhile, middle-skill jobs—those that require a significant amount of training but involve more repetitive tasks, including bookkeeping, clerical work, and many manufacturing jobs—can be done by machines. For that reason, many of these jobs are being automated away, or in some cases moved offshore. The result has been a stunning polarization of the workforce and hollowing out of the middle class: both high-wage and low-wage jobs have risen dramatically while in-between jobs have plummeted (see “Middle-class jobs are disappearing”). In the past, you might have been able to transition from a low-skill job to a high-skill job via one of those middle positions. Now, “if you fall on the wrong side of that [education] divide, you have a very hard time getting across it,” says Mindell.
This barrier to mobility has affected black and Hispanic people more than white people, and rural workers more than urban workers (although wage disparity is particularly dire in cities). And men have felt the impact more than women because they’ve held most of the routine manufacturing jobs that have been automated away. When these demographics compound, the consequences can be both broadly unfair and personally devastating: men without college degrees working full time actually made less per week on average in 2018 than they did in 1980, when adjusted for inflation.
The increase in technology has thwarted the aspirations of many Americans: whereas 92% of workers born in 1940 earned more than their parents, only 50% of those born in 1980 can say the same. For many, the American dream is getting further out of reach.
The increased inequality and decreased opportunities for many represent a dramatic reversal from much of the 20th century. From 1945 to 1973—a stretch that saw the introduction of the bar code, the microwave oven, and UNIX—worker earnings and productivity both doubled. During that time, “we managed technological progress and shared prosperity extremely effectively,” says David Autor, the Ford Professor of Economics and a co-chair of the task force.
“If we had written this report in 1970, we would have said, ‘Actually, it works out great,” Autor says. But “the lesson of the past four decades is that technological progress doesn’t lift all boats.”
So what’s different now? One change is the type of innovation. Although productivity is still generally on the upswing, computerization—the defining technology of our era—hasn’t been as effective over the last decade at increasing it as the chief innovation types of other eras, like industrialization and mechanization.
What’s more, automation and computers are often simply replacing workers rather than complementing them or creating further opportunities. Although these inventions might benefit the companies that deploy them, their benefits don’t spill over into the labor market as a whole. Instead, they help concentrate gains in the hands of a few (see “Going beyond the so-so”).
But our own national priorities, as expressed through policy, don’t help matters. Companies in the United States have stripped away many workers’ rights and have become solely focused on increasing value for shareholders. Unions, meanwhile, have “atrophied,” the report says. The federal minimum wage has fallen over the past decade when adjusted for inflation. While new forms of worker organizing have arisen, none have been able to reverse the trend of focusing almost exclusively on shareholders at the expense of workers’ interests (see “Tech-savvy labor activism”).
As the report points out, most other market economies “broadly accept (and often enforce) the notion that employees and communities are among the legitimate stakeholders to whom a firm must be responsive.” The task force believes the United States should move in this direction, although it isn’t yet sure on the details.
Over 90% of workers born in 1940 grew up to earn more than their parents, while only 50% of those born in 1980 can say the same.
Developed countries with pro-worker policies aren’t experiencing nearly as much income polarization, even as automation advances. Sweden, for example, has more robots per capita than the US, but it also has robust unions and a strong welfare system. Wages are still high, inequality is low, and most citizens aren’t worried. They trust that “if technology benefits productivity, then most people will benefit,” says Autor.
Meanwhile, because US institutions haven’t worked to blunt the disruptive impacts of new technologies, American workers feel they’ve been left to fend for themselves in an economy that favors shareholders. “The cowboy capitalism model of the US is showing its limitations,” says Autor.
Most of the report’s suggestions are focused on fixing the imbalance between those helped and hurt by the new economy. This boils down to investing more in workers—providing them with educational opportunities and political support, and restructuring our laws, our institutions, and even our ideas of who should benefit from progress (see “Where are the robots?”). “Without taking any steps policy-wise, we will just continue to see the situation we’re in worsen,” says the task force’s executive director, Elisabeth Reynolds, PhD ’10.
We’ve succeeded in this type of challenge before: as the report points out, the US responded to industrialization over the first half of the 20th century by introducing free public high school, in order to outfit young people with the skills they would need to prosper as agrarian employment shrank.
The task force recommends a similar investment in education today, especially for workers without college degrees. Community colleges, vocational training programs, apprenticeships, and affordable online education can all help bridge the opportunity gap. So could offering employers tax incentives to invest in skill training for their workers.
The report also encourages investment in the right kinds of automation—those that will complement workers rather than replace them, boost overall productivity rather than concentrate gains within companies, and keep the US at the forefront of technological trends like AI and machine learning
Perhaps the saddest thing about this is that the whole 'other side' of AI and computerized 'process automation' -- the idea of pervasive computing under Negroponte at the Media Lab -- appears either to have disappeared entirely or been deprecated as 'not business related enough'.
Much of the promised gains from automation come not in the line programming of relatively primitive equipment, or the clever learning and use of computer-manufacturing software, but in the promise that AI/ES could allow 'normal' people in industry to take better advantage of a mix of increasingly-sophisticated new and 'legacy cheap' older equipment to accomplish things either impossible or not cost-effective enough in the past. Whether or not they know how to code like a boss.
The great 'revolutions' that took home computing (and thence most of the actual, meaningful early tech development in personal tech above the consumer toy level) involved things like easy enhancement and upgrade (e.g. the IBM Boca Raton PC architecture and support), significant developer buy-in for packaged and assured software development (Microsoft), and the development of a practical UI for operating a personal computer coherently and familiarly (the Macintosh Manhattan Project, as very distinct from the development work in most of the component technologies used in the 'Bandley pirate' version of Macintosh, or indeed in some of the systems and approaches that followed it). There have been some interesting versions of Linux that don't involve obligate tweaking and arcane config and builds -- but they're far more 'few and far between' than they ought to be; the very great progress that I thought would come out of GNOME (and its fork GENOME) appear to have been dissipated almost without meaningful trace.
Not to say the instantiations don't exist -- they just get privatized and used for some personal exploitation or other. Perhaps this is the natural fate of any open-system self-optimizing personal 'assistance' system ... but it it still an embarrassment, and a shame, that since the early '80s when it first became possible to build a general pervasive assistance system, no one has apparently succeeded in making one that actually 'matters' (without infuriating users more than it helps). Be interesting to see if 'Amelia' -- now showing up in my sidebar ads -- will prove to be more useful than the previous rounds.
It'll be interesting to see what I can develop as an app to prevent flash commuter buildup, for example as an ordered response to the sort of problem that occurred at Penn Station during the tunnel outage a few days ago. Much of that framework has near-immediate synergy with various kinds of pervasive access and control; it would certainly seem to be a shame to waste it... or corrupt it a la Facebook or the 'new' Kalmbach by manipulating 'relationships' with its users...
All I know for sure is the number of illegals (or undocumented-whatever) in my local area has significantly increased ever since the biggest municipality's council made it an unofficial sanctuary community. However, except for the chinese resturants, they are replacing local citizens in construction and other areas. They work for little more than half the pay, don't ask for time and a half for overtime, don't ask for medical, and don't seek unemployment compensation or file for Workers Compensation. In construction, they have their own work gangs and boss. They rarely are as good as who they replace but that doesn't matter because they are so cheap. Since they aren't skilled labor, local citizens still handle the difficult jobs, especially in construction. I know various locals who work along with them and they all complain you can't leave any tool unattended, powered or hand, for even a brief time if they are around or its gone for good. (That was extremely rare before they appeared on the worksites.)
In summary, I wonder what the MIT academics have to offer as a solution for those jobs now lost to the locals. There are a lot more than many citizens realize and it has impacted the local minority community harder than the whites.
So how does this affect commuting patterns? Most of the Downtown Office jobs require a college degree
Our community is FREE to join. To participate you must either login or register for an account.