Inside a hip hotel room in Cupertino, California—just a short walk from Apple’s headquarters—the phone rings. I answer. A recorded voice tells me that my delivery has arrived. Just outside the door, I find a squat, white pillar of plastic and metal waiting expectantly.
A gray hatch on the top springs open, revealing a small, contoured compartment. I grab the courtesy toothbrush. Two beady white eyes blink at me from the touchscreen below. “Did you get everything you requested?” a prompt asks me. “All set,” I press, and it asks me what I thought of the service. Amused, I rate it five stars. The little robot hoots and twists a few times in a poor imitation of Chubby Checker before motoring off down the hall.
As Botlr nears the door to the lobby, Tessa Lau takes control of the robot, halting the demonstration. PlayStation controller in hand, she guides the little guy back to his resting spot. Lau, an enthusiastic woman with short, cropped hair, is “chief robot whisperer” and co-founder of Savioke, a company started by alumni of the famed Willow Garage research institute to bring robots to the service industry. Her ward, Botlr, is the first autonomous robot to be deployed in a hotel. But it certainly won’t be the last.
Watching Botlr go about his rounds is revealing. Unlike a human bellhop, he would rather be tweeted, not tipped. He knows where all of the furniture is located in the hotel, he has memorized every room number, and he chats with the elevators to get up- and downstairs. He navigates in smooth arcs, pauses when people get in his way, and, on occasion, still needs a little assistance. He proceeds confidently at times, cautiously at others, as though the future of robotics depends on him. Which, in a way, it does.
Botlr isn’t the first mass market robot, but it is among the first to perform a service industry job that was once exclusively done by humans. Work as a bellhop may not be the most appealing job, nor does it pay particularly well. But usually, it’s a person’s job. In this Aloft Hotel, at least, a person isn’t needed anymore.
What I just witnessed will repeat itself time and again in the coming decades. A robot will take a small step into a new industry, one that we had assumed would be unaffected by automation, at least for a little bit longer. The robot may not be particularly good at first, but it will get better. And better. And better, to the point where its performance is nearly indistinguishable from a human’s. Then that human will be out of a job.
Tensions will emerge. People will grow to distrust the robots working alongside them, fearing that one day their job will be next. Eventually, some of them will be robot-replaced. Their pay will tumble, their remaining skills will be marginalized. If history is any guide, they’ll struggle. Meanwhile, others will rise, assisted by the myriad robots that will build, serve, drive, and deliver. Those people will grow more skilled and more valuable. One hour of their work will accomplish far more than it did just a few years earlier.
The impact on society could be substantial. “I think that there are bigger technological changes likely in the next ten years than there were in the last ten years,” says Erik Brynjolfsson, director of the Initiative on the Digital Economy at MIT and co-author of The Second Machine Age. “And we know that those were pretty disruptive for the economy.”
We don’t have to be robot-replaced, though. We can be robot-enabled. How we do that—well, no one knows for sure. But there are ideas, ones that could reshape society to be more inclusive. Whether we implement them or not, that’s up to us. “It’s a choice that we have to make,” Brynjolfsson says, “and right now I don’t think we’re taking it seriously enough.”
Technology has been upending the status quo since the early days of civilization. Take, for example, plows and draft animals, which reduced the need for hoeing, an inefficient and labor-intensive task. Compared with a group of people hoeing, one person guiding a team of oxen could prepare a field for planting in a fraction of the time. Farms could expand without needing to employ more workers. Economists call this substitution, where old jobs are lost due to some technological advancement.
The Luddites may be the most famous substituted workers. In 19th century England, as automation was sweeping the textile industry, hand spinners and weavers smashed mechanized spinning frames and looms. Needless to say, their sabotage didn’t accomplish much. In 1820, there were about 240,000 cotton hand weavers in Great Britain. Just 20 years later, their numbers had been cut in half, and by 1863, just 3,000 people were employed as cotton hand weavers.
Other transitions in the 19th and 20th centuries were similarly fraught. In the early 1890s, Andrew Carnegie—whose U.S. Steel was a technological leader—decided to increase automation at the Homestead Steel Works in Pennsylvania to counteract a decline in steel prices. The plan worked, and productivity per worker ticked upwards. Carnegie and his management team decided that skilled workers were no longer as valuable, and in 1892, they demanded that their employees accept lower wages. The workers went on strike. The Pinkertons, a private security force, were called in to protect the scabs and violence erupted. Thirteen people were killed.
More recently, the automotive industry has been swept by a wave of automation that began in earnest in the 1980s. Automakers installed industrial robots in factories, eliminating many jobs in the process. In North America alone, over half a million automotive industry jobs were lost between 2000 and 2012, in part because of increasing automation.
It’s not just factories where automation has eliminated jobs. Computers have reduced the need for secretaries, and accounting software has eliminated many bookkeeping positions. “That set of tasks, which was a skilled set of activities, is no longer scarce,” says David Autor, a professor of economics at MIT. “We have machines that can do it a lot more cheaply than people.”
“Technology has always been replacing jobs, and it’s always been creating jobs,” Brynjolfsson says. This time might be different, though. “Recently, we’ve been seeing more of the replacement, more of the automation, and less of the complementing and creating of new jobs.”
Autor argues that, unlike the transition that occurred in the post-War boom, automation is hollowing out the workforce rather than fattening it. Middle-class jobs are being replaced by robots and computers. As workers are forced out of good-paying factory and clerical positions, many take what they can find, maybe working as a janitor or preparing food at McDonalds.
Those jobs never paid very much to begin with, and the oversupply of applicants depresses wages further, reinforcing the hollowing out. For now, many of those low-wage jobs appear safe. Cleaning office buildings and cooking and serving food is currently hard to automate, at least with the level of service people demand. But as robots get better, those jobs could be in jeopardy, too. Eventually, we’ll all have to find some skill that can’t be substituted by automation.
At Stanford University, just 20 minutes up the road from the Aloft Hotel where I met Botlr, I’m about to perform my first surgery. The patient’s skull is laid bare, from just in front of the ear canal to just past where the bone starts curving to form the back of the head. Fortunately, I’m not alone. Ken Salisbury, a professor, is sitting next to me. A thoughtful and soft-spoken man, Salisbury calmly instructs me to pick up the tool.
Wielding the drill like a pen, I move it closer until I feel the bit touch bone. Salisbury encourages me to start drilling. I depress my index finger. The tool whirs to life and immediately sinks into the skull. “Oh! There goes the occipital bone,” Salisbury exclaims. I’ve drilled too deep.
Luckily, this isn’t a live patient but a simulation. Salisbury and his lab are developing technology for doctors to rehearse surgeries before they enter the operating suite. As I move the stylus around, I can feel the thin sutures between plates, the holes through which blood vessels and nerves escape the cranial cavity, and the smooth, horrifically large hole I just made. Salisbury chuckles. Clearly I’m not as skilled as the surgeons he normally works with.
Salisbury’s work on virtual surgeries was inspired, in part, by his long history as a roboticist. In graduate school in the early 1980s, he helped develop the first of many robotic hands that would bear his signature, an accomplishment that required a keen understanding of haptics, the science of touch. Later, in the 1990s, he was part of the team that created and patented the first remotely-operated surgical instrument, the Black Falcon. That led to a stint at Intuitive Surgical, which had licensed the Black Falcon patent and commercialized it as the da Vinci Surgical System. Today, the da Vinci robot is used in hundreds of thousands of operations every year.
When robot surgery systems were first introduced in the late 1990s, people dreamt up scenarios where doctors would operate on patients thousands of miles away. With telesurgery, there would be no need to travel to a specialist—the specialist would see you, remotely. In 2001, that vision came true. A doctor in New York City removed the gall bladder of a 68-year-old woman some 4,100 miles away in Strasbourg, France. While successful, the procedure was more proof-of-concept than harbinger of the future. The demonstration was highly orchestrated, drawing on dozens of experts including engineers from France Telecom, who ensured sufficient bandwidth to avoid potentially dangerous buffering delays. To Salisbury and others, that operation helped put the nail in the coffin for so-called telesurgery. “There are limits in how far you can put the remote doctor because of time delays,” he says.
But it wasn’t the end for robotic surgery. Over the years, we’ve found that surgical robots work best when the doctor is sitting just across the room from the operating table instead of across the ocean. Robots have the potential to make minimally-invasive procedures even less invasive. Today, many minimally-invasive procedures use laparoscopic tools, which are inserted through a small incision in the patient and directly manipulated by the surgeon, who watches what’s happening on a screen off to the side. That disconnect between what they see and what they’re doing takes training to overcome.
Many doctors are adept at using laparoscopic equipment, but Salisbury and others think that surgical robots can help them perform even better. For one, surgical robots minimize the disconnect between a surgeon’s tools and the on-screen view. With robotic surgery, Salisbury says, “when you move up, the tool moves up.” Such systems also get doctors off their feet. Fatigue-inducing, multiple-hour operations happen surprisingly often. With robots, surgeons can sit in front of a console instead of standing over the operating table, a less taxing position.
Furthermore, surgical robots can alter the entire surgery environment. Robots can give surgeons an extra set of hands, allowing them to work with one set of tools one minute and then easily switch to another without having to “let go” with the first. Robots could also allow surgeons to set boundaries within our bodies, telling the machine that certain parts are off limits, thus preventing the surgeon from accidentally bumping into any delicate structures. Robots can easily alter the scale at which surgeons work, too. “I actually saw a blood vessel being worked on, and it looked like a hose on the screen,” Salisbury says. The surgeon was able to easily sew around it without puncturing the vessel wall. There’s even talk of using robots to fine-tune a surgeon’s motions. Rather than sweating tiny incisions, surgeons can scribe wider arcs that the robot then scales down, effectively increasing their dexterity.
Surgeons working with robots are a perfect example of what economists call complementarity. “A surgeon’s skill is in no way eliminated by use of a surgical robot,” says Autor, the MIT economist. “What the surgical robot does is it allows the surgeon to use their knowledge of the human body more precisely.”
Complementarity isn’t limited to surgeon. “In many cases, automation complements us and makes us more valuable,” Autor adds. Nearly every worker today is complemented by some form of automation, whether they realize it or not. Trading algorithms give stock brokers an upper hand, iPads help appliance salesmen field consumer’s questions, and engine computers help mechanics diagnose a car’s problems. It’s part of the reason why productivity increased during the Great Recession, even as people were being furloughed or laid off. Those who kept their jobs were able to be more productive, thanks in part to automation.
Designing robots that will make people more productive isn’t a simple task. Pamela Hinds, an associate professor and co-director of the Center for Work, Technology and Organization at Stanford University, studies how robots fit into the workplace. In one case, she observed how a pharmaceutical delivery robot functioned in a hospital.
It didn’t do well. The robot would move from station to station, delivering medication. If all went as planned, nurses would open a door to retrieve the drugs they had requested, just like I had done with the toothbrush Botlr delivered. But unlike Botlr, the hospital robot wouldn’t wait for the nurses. Instead, it would sit at a station for a set amount of time, demand that the medications be taken out, and then move on even if they hadn’t. If the nurses were busy attending to patients, too bad. The robot was gone.
Needless to say, it wasn’t well accepted. The nurses were frustrated with it. Their first priority is to patients, not to the robot. Perhaps worse, when they were busy tending to patients, they would miss the medication delivery they had requested. The machine, Hinds recalls the nurses saying, was disruptive. “That’s common with a lot of technologies, where they’re not well designed to integrate with existing practices,” she says.
“It may be even more challenging when you’re talking about autonomous robots, because they are a physical presence in the workplace as well,” Hinds adds. In other words, while most of us have accepted email in our workplaces because it sits unobtrusively on our monitors, a physical robot may not slip into our workflow so easily—simply because it takes up space. An autonomous robot, Hinds says, is “different than a desktop that sits in an office.”
That probably won’t keep physical robots from entering our workplaces, though. To prevent the sort of friction Hinds observed in the hospital, robots need to be designed to be as unobtrusive as possible. Hopefully, the thinking goes, we won’t notice them until we need them. Adrian Canoso, design lead at Savioke, went through nearly a prototype a month before he and his team settled on the right height and form factor for Botlr. At three-feet tall, the robot is just short enough so its top-mounted bin is easily accessible, yet tall enough so you don’t trip over it in the hall.
Roboticists will also have to instill their creations with a level of awareness. Eyes are a good place to start. Botlr has small white orbs that imbue its pillar-like form with some emotion. The general purpose industrial robot Baxter also displays eyes on its LCD screen so operators know where its attention is focused. Cynthia Breazeal, a roboticist at MIT, designed her Kismet and Leonardo robots with eyelids and eyebrows so it could emote more expressively. They also track people’s faces, giving them a certain lifelike quality. Design strategies like these can help ease the robot transition, making them feel more like coworkers and less like usurpers.
Preparing for Automation
Ultimately, how we feel about robots will be affected by whether we’ve been substituted or complemented. “If you are directly substituted, and you don’t have a set of skills that are complemented, then that’s less good news for you,” Autor says.
People who have been substituted face two options—go back to school to train for a higher-skill job or find a job that hasn’t been substituted by automation. The former tends to serve people better in the long run. “The great human comparative advantage—way, way above any machinery we’ve yet produced—is flexibility, common sense, and making your way around in novel situations,” Autor notes. “The thing that enables us to do that is the combination of our intrinsic adaptability complemented by education that gives us analytic skills, formal reasoning skills, problem solving skills, as well as communication skills. In general, we continue to make ourselves very valuable by building on those strengths.”
It’s something our society has acknowledged over the years. “Historically, the way people have responded to automation is to educate themselves,” Autor notes. One example is the U.S. at the turn of the 20th century. The country was facing a demographic dilemma. New tractors and harvesters allowed farmers to plow and harvest larger fields more quickly and with fewer people. “People at that time understood that agriculture was not going to be the future,” Autor adds. So they sent their kids to school. High schools, which had been around for decades, suddenly grew in demand. Others quickly saw the benefits of additional education, and soon high school was mandatory. In 1900, just 650,000 students were enrolled in high schools. By 1930, there were over 5 million. “U.S. economic preeminence in 20th century had a lot to do with ours being the most skilled and flexible workforce in the world,” he says.
To see how high school shaped the 20th century, take a look at the construction industry. “If you think about a 21st century construction worker, they’re practically a cyborg by historical standards,” Autor says, listing off equipment common on work sites, from cranes to diggers and nail guns, all of which are forms of automation. But none of them replaced construction workers. If anything, automation has made them more valuable. “Nothing happens on a construction site without workers,” he adds. “They play a crucial role in orchestrating how all that stuff is used.” Construction workers today are far more skilled than they used to be, in part because of free and compulsory high school. “If we took the workforce of the turn of 20th century and put them in 21st century America, many of them would not be employable because most of them would be innumerate and a substantial chunk would be illiterate,” Autor says.
In the coming decades, we’ll all have to be cyborgs in the workplace. “I think the real issue there is not what automation is going to do, but what we are going to do with the tools,” Brynjolfsson says. “It’s not a matter of slowing down the technology, it’s a matter of speeding up our response to it.”
Another way to ease the transition would be to support those who are substituted. A societal minimum income is one option, Autor adds, though he admits that idea isn’t very popular. (Nor would it be very satisfying for those on the receiving end—people prefer their lives to have purpose, and jobs provide that.) “In the long run, we want people to do work that pays them enough that it’s a good deal—they have a decent standard of living, they can educate their children adequately, they have opportunities going forward, and they’re not dependent on societal largess,” Autor says. One way to accomplish that without a direct handout would be to expand the earned-income tax credit, which encourages people to keep working but helps them maintain a reasonable standard of living, he says.
Brynjolfsson agrees that expanding the earned-income tax credit is a viable option. He also suggests that finding ways to encourage entrepreneurship in the public and private sectors could help keep people independent while they navigate the transition to a more automated economy. “I think that creativity and exceptions and entrepreneurial activities continue to be something that’s very hard to automate.”
Back at Aloft Hotel, Tessa Lau is shepherding Botlr back to his docking station to recharge. Like other forms of automation, Botlr is both a blessing and a curse. For the hotel’s employees, Botlr allows them to do their jobs more efficiently. Rather than leave the front desk to drop off a forgotten toothbrush—a fairly menial task—they can stay and assist guests who have more complicated needs. In this way, automation helps people do the jobs they’re best at.
But for workers who are displaced, it may not be easy to see it that way. Tension in the workplace seems inevitable. There are several potential sources of friction, Hinds says. “One is the extent to which people are fearful that the robots are going to take over their jobs. Then I think there’s a natural distrust of whoever is purchasing these robots.” The people who are complemented by using robots will be just as responsible for smoothing the transition to the robot economy as those who are substituted and have to find new work.
“Ultimately, that wealth accrues to people, not to robots,” Autor says. Those who are adept at using robots will undoubtedly succeed, just like those who have mastered computers today. The question is, who will be the winners in an automated economy? “I’m quite convinced that technology could make income inequality worse,” Brynjolfsson says, “but it could also be used to make shared prosperity.”
If that latter possibility sounds a bit too perfect, remember that the word “utopia” is based on the Greek for “no place.” Eliminating poverty isn’t as simple giving the low-paying jobs to robots and expecting everyone to reap in the profits. It may require a wholesale rethinking of our economic, political, and social systems. “I wonder if we are we ready to make the organizational, institutional changes and updates in our skills that are needed to take full advantage of those technologies,” he adds. “Because if we don’t, a lot of people may be left behind.”