Skip to Content

From rust belt to robot belt: Turning AI into jobs in the US heartland

Artificial intelligence is offering an amazing opportunity to increase prosperity, but whether or not ­we will seize it is our choice.
June 18, 2018
Rob Larson

The vast vacant lot along the Monongahela River has been a scar from Pittsburgh’s industrial past for decades. It was once the site of the Jones and Laughlin steelworks, one of the largest such facilities in the city back when steel was the dominant industry there. Most of the massive structures are long gone, leaving behind empty fields pocked with occasional remnants of steelmaking and a few odd buildings. It all stares down the river at downtown Pittsburgh.

Next to the sprawling site is one of Pittsburgh’s poorer neighborhoods, Hazelwood, where a house can go for less than $50,000. As with many of the towns that stretch south along the river toward West Virginia, like McKeesport and Duquesne, the economic reasons for its existence—steel and coal—are a fading memory.

These days the old steel site, called Hazelwood Green by its developers, is coming back to life. At one edge, fenced off from prying eyes, is a test area for Uber’s self-driving cars. A new road, still closed to the public, traverses the 178 acres of the site, complete with parking signs, fire hydrants, a paved bike path, and a sidewalk. It doesn’t take much imagination to picture it bustling with visitors to the planned park along the riverfront.

The gem of the redevelopment effort is Mill 19, the former coke works. A structure more than a quarter-­mile long, sitting amid the empty fields, it has been stripped clean to a three-story metal skeleton. Crews of workers are clearing away remaining debris and preparing the building for its reincarnation. By next spring, if all goes according to plan, its first occupant will move in: the Advanced Robotics for Manufacturing Institute.

The symbolism of robots moving into a former steelworks is lost on few people in the city. Pittsburgh is reinventing itself, using the advances in automation, robots, and artificial intelligence coming out of its schools—particularly Carnegie Mellon University (CMU)—to try to create a high-tech economy. Lawrenceville, five miles from Hazelwood, has become a center for US development of self-driving cars. Uber Advanced Technologies occupies a handful of industrial buildings; self-driving startups Argo AI and Aurora Innovation are nearby. Even Caterpillar has set up shop, working on autonomous backhoes and other heavy machines that could one day operate themselves.

Can AI, advanced robotics, self-driving cars, and other recent breakthroughs spread prosperity to the population at large?

This has drawn billions of dollars from Silicon Valley and elsewhere, a welcome development in a city whose economy has been moribund for decades. And the effects are visible. Self-driving cars out for a test ride are a common sight, as are lines outside the trendy restaurants in what civic boosters call “Robotics Row.” While many longtime residents complain of skyrocketing home prices near the tech firms’ headquarters and test facilities, they’ll also tell you these are the best days the city has seen in their lifetimes.

But despite all this activity, Pittsburgh’s economy is struggling by many measures. Though the city’s population is no longer hemorrhaging away—between 1970 and 1980 it fell by roughly a fifth—it isn’t growing, either, and is aging quickly. During the last half-decade, almost 70,000 people aged 35 to 54 have left the region. And not far from the city and its elite universities, in areas where the main hope for prosperity lies in coal and natural gas from fracking rather than self-driving cars, well-­paying jobs are scarce and towns are being devastated by opioid addiction.

This makes Pittsburgh not only a microcosm of the US industrial heartland but a test case for the question facing every city and country with access to new digital technologies: Can AI, advanced robotics, self-driving cars, and other recent breakthroughs spread prosperity to the population at large, or will they just concentrate the wealth among entrepreneurs, investors, and some highly skilled tech workers?

To prosper, says Scott Andes at the National League of Cities, Pittsburgh “can’t just be a producer of brilliant talent and ideas that then don’t turn into job generation.” He adds, “Pittsburgh is a great case study for the 21st-century economy, because it is beginning to leverage research strengths into economic value.”

Changing jobs

There is no sillier—or more disingenuous—debate in the tech community than the one over whether robots and AI will destroy jobs or, conversely, create a great abundance of new ones. In fact, the outcome depends on various economic factors. And how it will play out as the pace of AI intensifies, no one knows.

Automation and robots have certainly wiped out many jobs over the last few decades, especially in manufacturing. In one of the first attempts to quantify the impact of industrial robots, research by Daron Acemoglu at MIT and his colleagues, based on data from 1990 to 2007, found that for every robot on the factory floor, some six jobs are lost. That means as many as 670,000 jobs for the years that they looked at, and as many as 1.5 million jobs at 2016 levels of robot usage in the US.

Automation is changing work

Gauging the net gain or loss of jobs due to robotics and AI is a tricky business. But it’s clear that the kinds of jobs in demand are changing as the need for manual labor declines and that for digital and human skills soars.

The McKinsey Global Institute estimates that about 50 percent of tasks done in our economy could be automated. But such statistics are often misinterpreted. The 50 percent merely describes the “technical feasibility” of what can be automated with existing and emerging technologies, says James Manyika, the institute’s chairman. The number of actual jobs lost will depend on the costs and benefits of replacing people with machines.

Even more uncertain is how many new jobs will be created. Many technologists, especially roboticists, assert that advances will lead to a wealth of new kinds of work. So far, though, that hasn’t happened, and few of the breakthroughs have reached the largest sectors of the US economy, such as health care.

Perhaps we just need to be patient; technology advances have always increased incomes, which then increased demand for goods and services, which then led to more jobs. But Laura Tyson, a top economic advisor to President Bill Clinton and a professor at the University of California, Berkeley, asks the question that is on everyone’s mind: What if, this time around, the goods and services that people want just don’t require much human labor to produce? “This is the first time that technology, we think, could on net reduce the demand for human workers,” she says.

Reinventing Pittsburgh

  • Mayor William Peduto, a Democrat elected in 2013, has been at the center of the city’s high-tech reincarnation. We asked how it’s going.

  • On the best use of new technologies:

    “What good is it if we develop autonomous vehicles that only create more congestion on our streets? That deny people mobility unless they have a smartphone or credit card? Are we really creating a better society? And if we aren’t, why are we investing in it?”

  • On the region around Pittsburgh:

    “Our neighbors throughout the Rust Belt are still going through a recession. They’re still not seeing any part of the new economy directly connecting to them. One of the reasons that President Trump did so well is that he offered them a false narrative of bringing back the mills and bringing back the mines. As Democrats we offered them no hope. Saying to someone who has generations’ worth of making and building things that you’re going to be retrained as a coder is an insult.”

  • On new jobs:

    “We don’t want to create a society based on PhDs; we want to include GEDs.”

  • On the city’s progress:

    “We’re right at the beginning. We’re at the beginning of the next phase of Pittsburgh. Those who lived through the ’80s and ’90s will tell you these are the best days. We went through those decades exporting people like we used to export steel. It has been a slow death of a former economy, but from it has come a new, more diverse economy. So it’s at the beginning phases.”

  • On lessons for executives in Silicon Valley:

    “What I reminded them is that we were you before Silicon Valley existed. We were where the great wealth was and where all the jobs were being created 100 years ago. And we created air that was dangerous to breathe, water that was poisonous to drink, and the greatest disparity between the haves and the have-nots in American history. Learn from us.”

  • On hearing from Amazon:

    “Nah. They made it very clear. They said they will make a decision in 2018, but they never gave us a specific timeline.”

“The naïve view among macroeconomists for several decades has been that technology will always create jobs,” says Acemoglu. “The alarmists’ is that this time is different and it will destroy jobs. The truth is it’s capable of doing both.” Though in the past the economic benefits from new technologies have always been enough to create more jobs than were lost, he says, “lately, for a variety of reasons, there has been a much more job-destroying face to technology.”

Part of what he’s describing is the so-called productivity paradox: while big data, automation, and AI should in theory be making businesses more productive, boosting the economy and creating more jobs to offset the ones being lost, this hasn’t happened. Some economists think it’s just a matter of time—though it could take many years (see “The productivity paradox”).

But the debate about how many jobs are gained or lost obscures a much more important point. The location of jobs and the kind of work they involve are changing, and that’s what’s causing real pain to people and to local economies.

In the US, demand for low-­paying work in places like warehouses and restaurants is growing; so is demand for well-paying work in occupations requiring lots of technical skills, such as programming. At the same time, many traditionally middle-class jobs in areas like manufacturing and data processing are shriveling. These trends have contributed to record levels of income inequality. “There is not a lot of disagreement that technology is changing the skills and occupations in demand,” says Tyson. “And that will continue to increase income inequality.”

This movie has, of course, played out before. In 1900, about 40 percent of US workers were on farms; today fewer than 2 percent are. In 1950, about 24 percent of the jobs were in manufacturing; today around 9 percent are. Similar shifts are occurring in other developed countries. But today’s changes are happening faster and more broadly than before, leaving little time for people to adapt.

Many are simply giving up on finding a decent job. Labor-force participation—basically, the proportion of people working or seeking work—is showing a troubling drop, especially for men aged 25 to 54. Melissa Kearney and Katharine Abraham, economists at the University of Maryland, have looked at why. They think there may be several causes, but they say robots and automation are a critical one. Many people without a college degree simply think the prospects of finding a well-­paying job are too slim to make it worth looking.

Inequality is up as growth slows

Despite advances in AI and robotics, productivity is sluggish, and fewer people are enjoying the benefits. To boost growth, especially as workforce growth slows, we will need more AI, and we’ll need to learn how to deploy it better.

Princeton economist Anne Case and her coauthor Angus Deaton have identified what’s likely a related trend. They found that mortality is rising among middle-aged white people in the US with a high school diploma or less. The culprits: high rates of suicide, drug addiction, and alcoholism, which Case and Deaton call “diseases of despair” because they don’t seem related to poverty per se, but rather to disappointment; in a reversal of expectations, people are realizing they won’t be better off than their parents.

Automation might be partly to blame for these social problems. But if economists like Acemoglu are right, the key to creating more good jobs is not fewer of these advances but better versions of them that are deployed faster throughout the economy. 

Pittsburgh reborn

That, in essence, is what Pittsburgh’s attempt at reinventing itself is about. So far the results are mixed. “The transformation of the city by new, young people working in AI and robotics has been spectacular,” says Andrew Moore, dean of computer science at CMU. “But it has been more of an approach of gentrification rather than an inclusion of the community.”

Economic Pain

  • Laura Tyson, a former top economic advisor to President Clinton, explains how technology is increasing inequality.

  • On inequality:

    “I think there is a consensus based on what we know from the past 30 years about what will happen as a result of ongoing technological change; it will be labor-replacing and skill-biased. And this kind of technological change will lead to the continued erosion of the labor share of national income, growing wage inequality, and growing income inequality.”

  • On what’s new:

    “The pace of automation is increasing, and it is spreading across more skills, tasks, occupations, and sectors. So it is not just the pace but also the breadth of change.”

  • On not leaving people behind:

    “If workers in their late 40s or mid 50s are replace by automation, will they actually be able to acquire the skills required for new jobs, and will they actually be hired if they acquire those skills? We have to invest a lot in skills and education and mobility, and—even though that sounds hard—it’s harder than you think.”

  • On understanding what’s happening:

    “It’s almost simple Econ 101. Technology will displace more and more humans and more and more tasks currently done by humans. That will reduce the demand for labor, which in turn will reduce the returns to labor. My concern is what this does to wages. If a significant share of human workers have the value of their skills undermined by automation, it means lower wages. It’s that simple.”

  • On productivity:

    “I continue to believe productivity gains will be substantial. The question is, how will they be shared?”

  • On tech unemployment:

    “I’m of the view that we’re not headed for sustained technological unemployment. In a market economy, wages adjust over time and people will find jobs. The question is not the number of jobs but the quality of jobs. Will they provide livelihood levels and opportunities comparable to livelihoods and opportunities of the jobs lost through automation? This worries me.”

That criticism resonates in a place that prides itself as a working-class city with strong unions and a rich history of progressive politics. Mayor William Peduto helped attract Uber to the city, but he has since soured on the San Francisco–based company. “The Silicon Valley model doesn’t [put] people in the equation. It is based on what return will be derived for VCs,” he said in a recent interview at city hall with MIT Technology Review. “In places like Detroit and Pittsburgh, when we look at the future of work, we want to know what the future of the worker is.”

According to a recent poll, more than half of Pittsburgh residents would strongly support Amazon’s building its second headquarters there. That’s far more than in many cities on Amazon’s shortlist—in Austin and Boston only around a third of the population would welcome the move. It’s hardly surprising: Amazon is pledging 50,000 jobs and $5 billion in investment, which would be transformative for Pittsburgh. It’s rumored that the city is tempting the company with the site along the Monongahela River that includes Mill 19.

But if Amazon picks Pittsburgh, that’s likely to exacerbate the anxiety over how to match residents with new high-tech jobs. “There is nowhere near enough people in the city and the region with the technical skills,” says CMU’s Moore. “We’re great in terms of the rare genius leaders, but [Pittsburgh] really needs to skill up the local population to take part in this.”

The challenge facing the city and the rest of the country, though, is not only to include more people in the high-tech workforce but to expand the supply of those well-paying jobs. Advanced robotics can modernize the factories in a city like Pittsburgh and help make manufacturing more competitive. But the factory jobs lost through the years aren’t coming back. As a country, we’re struggling to imagine how to build an economy with plenty of good jobs around AI and automation.

A person standing on the flat roof of a building in the Lawrenceville neighborhood can get a glimpse of the future. On the first floor is a large garage housing several of Aurora’s self-driving cars. Off in some weedy fields is a Caterpillar backhoe belonging to the company’s research outpost for autonomous machines. Beyond that is a fenced-in testing area next to yet another former steel facility—this one housing Carnegie Robotics, which is working on a bomb-clearing robot for the Army. In the background is the National Robotics Center, another imposing building and home—until it moves into Mill 19—of the Advanced Robotics for Manufacturing Institute.

It’s an impressive scene highlighting signs, if you know where to look, of some of the world’s leading research into robotics and automation. But it is also almost deadly quiet. There are a few cars in the parking lots—those of the engineers and programmers involved in the various robotic ventures, and probably some visitors. Beyond that, there are no signs of workers anywhere.

Keep Reading

Most Popular

Scientists are finding signals of long covid in blood. They could lead to new treatments.

Faults in a certain part of the immune system might be at the root of some long covid cases, new research suggests.

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.