The American middle class is shrinking and it’s technology that’s causing it. It’s not all bad. The gains in efficiency begotten by automation have been great for productivity. And productivity means progress. It always has. Since the Industrial Revolution began around 1760, new technologies have been stealing jobs, and since 1760, people have responded by finding or inventing new jobs that contemporary technologies couldn’t do.
It’s a good system — in the long term, everyone benefits from technological progress, and while the workers losing their jobs in the interim might feel a bit miffed, people have always found a way to bounce back into an ever-adapting economy. Besides, if machines can do something better than people can, it would be senseless to ignore such utility and hold back progress for fear of a few temporarily lost jobs.
Unfortunately for today’s average worker, finding or inventing a new job is harder than it once was. When economists look back, they see that it was around 1999 when something changed. Productivity kept going up, but where in the past median household income and employment per capita would have also hitched along, they instead diverged. Median household income is on a steep decline, employment isn’t bouncing back strongly after the Great Recession, and a greater percentage of Americans now identify themselves as “lower class” than at any point in history.
Andrew McAfee and Erik Brynjolfsson, director and professor, respectively, at the MIT Sloan School of Management, named this divergence of productivity and employment “the great decoupling.” Technology is a broad term: It can be equally applied to a stick being used by a chimpanzee to extract insects from the Earth and to a rocket launching a chimpanzee into suborbital flight above the Earth. Today’s technologies are more scalable and complex than the machines people needed to outsmart in the past, which is a big reason for the decoupling.
Surveying a county road in 1998 meant that a team of workers needed to get into a pickup truck with a government seal on the side of it, drive, take photos and measurements of the area, return to the office and assess the information that was gathered. Researchers at the Michigan Tech Research Institute found in 2013 that it’s much easier to just send a drone. They also developed software that takes the sensory data gathered by the drone and generates a fully characterized 3-D model. People are still needed in the process to make high-level decisions and babysit the technology when things go wrong, but not as many people are needed. And even fewer will be necessary if the researchers’ concept is honed and commercialized.
The trouble is that the guy who once rode along in the pickup truck is now unemployed and he doesn’t know how to design drones or code 3-D modeling software. The average American is looking more and more like that guy. A study by two researchers at the Oxford Martin School concludes that within the next 20 years or so, approximately 47 percent of all jobs could be replaced by automation.
“Technology is racing ahead, but our skills, our organizations, our institutions aren’t keeping up,” Brynjolfsson said. “As they adjust, we will see more of the benefit show up in the economics, but right now there are a lot of technologies with more potential than has been fully realized.” This trend is just getting started.
A lot of economists, technologists and policymakers agree with McAfee and Brynjolfsson, but some say today’s technologies aren’t special — people will find new jobs just as they always have and no intervention is needed. It’s just the recession, they say. To understand why that’s not the case, people need to look closely at today’s technology, Brynjolfsson said.
For example, there are prototypes of autonomous vehicles on the roads. Legislators in Nevada, California, Florida and Michigan have penned laws allowing the vehicles limited public use, and some estimate that almost all vehicular traffic will be autonomous by 2050.
In a recent span of two months, Google purchased eight robotics and machine learning companies, not just because the technology will make cars drive themselves, but because smart robots can improve productivity across almost all of the company’s businesses. The problem with this extremely fast progress is that people are relatively slow. Although it won’t happen overnight, today’s 233,000 taxi drivers and 1.7 million truck drivers aren’t ready to become an anachronism.
In addition, decoupling means the upper 1 percent gets a bigger piece of a growing pie, Brynjolfsson said, which also accounts for the shrinking middle class. “A lot of these digital technologies have winner-take-all or winner-take-most economics, where you can get a small group of people producing a better piece of software or insight, and once they’ve digitized that, they can replicate it 10 times or a hundred million times, and dominate the market for that,” he said. “You can see it in checkout counter software, you can see it in tax preparation software — there are 17 percent fewer tax preparers than there were a few years ago — you can see it in airline reservations. In more and more categories, software is eating the world.”
Software can give legal advice, analyze data and automate data entry, and robots like IBM’s Watson can even diagnose and recommend accurate cancer treatments much better than humans can. One study showed that Watson can diagnose lung cancer accurately 90 percent of the time, compared to a measly 50 percent rate for human doctors. “There’s no economic law that says everyone is going to benefit from technological progress, even if it does make the pie a lot bigger,” Brynjolfsson said. “So both in terms of theory and evidence, I think there’s a potential to be concerned, and I am concerned.”
Amazon spent $775 million buying out Kiva Systems in 2012, the maker of a disc-shaped robot used in warehousing. Today Amazon uses the robots to fetch pallets of goods, saving workers the time and energy of running around to find products themselves. The purchase came soon after a report showing that some Amazon warehouse workers were walking 10 to 15 miles per shift. Most workers are probably grateful that their job is now to pack goods and do various administrative tasks rather than play fetch all day, but on the other hand, Amazon doesn’t need as many human employees now. If technology gets automatic and simple enough, it will eventually just be one guy in the Bahamas running the company from his smartphone. That might seem far-fetched, but it was in living memory that the idea of autonomous vehicles and talking robots were science fiction. Now those technologies border on passé.
Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory are developing a theoretical model and algorithms that would let robots like the ones used by Amazon communicate more intelligently with one another so they can solve logistical problems on the fly and also communicate with other sensory agents in the environment, human and otherwise. Once robots can talk to each other and solve problems on their own, even fewer people will be needed in those warehouses. When robots like Baxter, a $25,000 production line worker with a calm demeanor, get cheap enough, even sweatshop workers will be out of a job.
What the Industrial Revolution did for muscle power, Brynjolfsson said, the second machine age is doing for brain power, and especially so in government. “As you automate and augment a lot of mental tasks, it’s a little less clear whether those technologies will be complements or maybe substitutes for human labor, and that’s one of the things we’re working through now as a society,” Brynjolfsson said. “Government jobs on average tend to include more information processing, and on average the workers in that sector are more educated and doing more knowledge work than in a lot of other parts of the economy. So they, in some ways, are likely to be more affected.”
Today’s robots are very poor at doing some things, such as folding laundry (as one YouTube video confirms), and even worse at others, like offering compassion to a young student, or making a sound moral decision as a police officer. Those traditionally human skills are expected to gradually improve in robots, but in the meantime, data analytics and the Internet of Things provide a waypoint for progress. Facilities like the Domain Awareness Center now being constructed in Oakland, Calif., will make monitoring cities for crime and dispatching help a more efficient process. Likewise, predictive crime software is a rising trend in law enforcement in many cities. Making better use of data and distributed sensor networks means that fewer officers will be needed to cover a given geographic area because officers are better informed and more efficient. Even if it’s not the extreme scenario of Robocop taking over for human police officers, technology finds gains in efficiency everywhere and the cost is usually displacing human jobs.
Some are optimistic about what all this means for the world, and others less so, but to take either position is to accept that technological progress has a foregone conclusion, Brynjolfsson said, and it doesn’t. “It’s been said that the best idea America ever had was mass public education,” he said. “That helped us make the transition from an agricultural economy to one based on industry and services. It didn’t happen by accident; it happened through public policy. We’re going to have to reinvent what education is and focus more on creativity and interpersonal skills — things that machines are not very good at — and less on having people sit quietly in rows, listen to instructions and carry out those instructions.”
Nicco Mele agreed that education and government policy need to be revolutionized and that’s why he teaches future policymakers at the Harvard Kennedy School of Government. Mele is a consultant to the Fortune 1000 and was named by Esquire as one of America’s “best and brightest.”
Modern technology, Mele said, compels us to rethink the assumptions of every discipline. “Look at our health-care policy, look at our retirement policy,” he said. “Those policies are built on this assumption that people have 9-to-5 jobs and stay with one employer their whole lives. That’s profoundly not true for the American workforce and hasn’t been true for well over a decade. A third of American workers are self-employed and another third are contingently employed, which means only about a third of the workforce has a traditional 9-to-5 job. Yet our policymakers and our politicians are building all the policy on the assumption that this is a good way to do it.”
American policymakers are on average older and richer, but they also tend to know less about technology, care less about technology and delegate technological responsibilities to others, Mele said, and this is unacceptable in a world where tech influences everything.
“Imagine a state legislator saying, ‘I don’t think about money. I don’t worry about money, I don’t worry about the tax rate, I don’t worry about the budget. I just let accountants deal with it.’ We’d kick them out of office, right? That’s not an acceptable answer, and we need that same kind of expectation technologically,” he said.
The idea that tech is separate from the rest of the world is rooted in a past era where digital technology was less prevalent. Today, there are remnants everywhere of that old way of thinking that reinforce technological illiteracy in some of government’s most critical positions, Mele said.
“One of the things I hate most in the world is the Genius Bar at Apple because it sets up this dichotomy,” he said. “It sets up this world where they’re geniuses and we have to just do what they tell us.”
People must demand that their leaders be technologically literate, Mele said, because “it’s profoundly dangerous to have elected officials or policymakers who don’t have any technical literacy to evaluate what’s going on.” A recent Gartner report identified that 60 percent of CEOs dismiss the idea that automated and smart technologies could displace a huge percentage of jobs in the next 15 years.
Michael Armstrong, CIO of Corpus Christi, Texas, is not a denier of the second machine age’s power and influence. As a CIO, Armstrong said, the best one can hope for is to influence policymakers.
The coming years will bring incredible new changes, he said. Advances in 3-D printing have applications in medicine and other fields that haven’t even been realized yet. Advances in prosthesis are allowing people to live longer, and there are even philosophical questions being opened up about what it means to be human. The Shadow Robot Co., based in London, spent $1 million building the Bionic Man, a robot composed entirely of artificial body parts and internal organs. Life is changing fast for everyone.
“I think we’re hitting the knee of the curve and things are getting exponential,” Armstrong said. “Make sure that you understand and your leadership understands what is happening in these areas and what the implications are because that’s going to drive social policy and government policy to a huge degree. A lot of this stuff is happening very quietly.
“Government has been shrinking for so long, that’s been an accepted way of doing business. I think this is not going to leave anyone alone. One way or another, it’s going to affect us all,” he said. “In any kind of revolution, we always lose jobs, but there’s always been something to replace all those jobs, and that may not be the case this time.”
