The Future of Work

Technology has always created as many jobs as it destroyed. That may soon change

Loading the podcast player...

Hi, this is Steven Cherry for IEEE Spectrum’s “Techwise Conversations.” This is show number 79.

Sixty years ago, there were about 350 000 switchboard operators working for AT&T. Today, there are fewer than 20 000. Nowadays, automation is moving up the skills ladder in just about every profession. It’s not just toll collectors and supermarket checkers—with continual improvements to machine intelligence and robotic dexterity, it’s lawyers and nurse practitioners and engineers and maybe even taxicab drivers as well.

People have been alternatingly hopeful and worried about machines taking over their jobs almost since the beginning of the Industrial Revolution. In an 1870 novel by Edward Bulwer-Lytton, a mysterious technology called Vril makes possible a world without war, poverty, or hard work, and in Edward Bellamy’s 1887 utopian novel, what little human labor is still needed is shared by all, and everyone retires by age 45. On the other hand, the original Luddites rioted and destroyed the textile looms that had spelled the end of their jobs.

The future probably lies somewhere in between extreme utopia and dystopia, but it makes a big difference exactly where on the continuum things end up. And if your job gets automated, it’s cold comfort to know that everyone else will eventually be made obsolete too.

My guest today is Andrew McAfee. He’s a principal research scientist at the Center for Digital Business, which is part of the MIT Sloan School of Management, and is the coauthor, with Erik Brynjolfsson, of a new book, Race Against the Machine. Andrew, welcome to the podcast.

Andrew McAfee: Hi. Thanks very much for having me.

Steven Cherry: Andrew, the “machine” part of your book’s title refers to the increasing sophistication of a bunch of technologies, and we’ll turn to those in a minute, but I want to start with the “race” part. Is the race whether more jobs are created by technology than are lost?

Andrew McAfee: That’s one way to think about the race metaphor that we keep using in the book. Another one is the fact that technology itself—and when we talk about technology in the book, we’re talking about digital technology. So “the machine” and technology is our shorthand for the bundle of hardware and software and networking equipment that makes up modern information and communication technologies. And we keep on presenting an image of this technology racing ahead and doing things that were the stuff of science fiction even a short time ago. So some of the examples that we give include the Google autonomous car, the Watson supercomputer that beat the best human “Jeopardy” players, and the software that now exists, some of it free and over the Web, that can translate with decent quality from one human language to another. All of these have been longstanding goals of computer scientists and artificial-intelligence researchers. Our progress has been really frustratingly slow in a lot of areas up until quite recently, and we really get the impression that technology is racing ahead in its ability to do stuff that used to be the domain of human beings alone.

Steven Cherry: Yeah, economists used to talk about something called the “productivity gap”—there was all this spending on information technologies and economists weren’t finding a corresponding rise in productivity. First of all, did I explain that right? And then, what’s your take on the productivity gap?

Andrew McAfee: Yeah, the label we usually use for that is the “productivity paradox.” It was first articulated in 1987 by Robert Solow, who was a Nobel Prize–winning economist that year, so people tended to listen to what he said. And he had a great quote—he said, “We see evidence of the computer age everywhere except in the productivity statistics.” My coauthor on the book, Erik Brynjolfsson, was one of the first careful economists to reexamine the productivity paradox. And we started to come to the conclusion as long ago as the mid-90s that the paradox was being resolved, that if we took another look at the productivity statistics we did see the influence of computers. And as time has elapsed, we see more and more clearly the productivity paradox as a thing of the past. No one really disputes the link between computerization and productivity anymore.

Steven Cherry: So productivity is higher than it’s ever been, and businesses love productivity gains, right? It’s like free money to them almost. So sure enough, corporate profits are pretty darn healthy, but official unemployment is stuck above 9 percent, and real unemployment is probably twice that. It seems the workers aren’t sharing in these gains—in fact, just the opposite.

Andrew McAfee: Yeah, and you bring up this central tension that runs throughout our book. And I want to go back to the statement you just made that businesses love productivity increases—that’s absolutely correct. It’s also correct that societies and economies as a whole love productivity increases for the simple reason that because we can produce more, we can also consume more. And there’s kind of a flip way to understand that which is, oh, great, we’re just going to consume fast food and Happy Meals and other flavors of junk out there; it’ll just become an increasingly shallow and materialistic society. But I want to challenge that, because one of the other ways to think about increased consumption is that we can increase our consumption of things like health care and literature and leisure time, and a productive society is a fast-growing society and it’s a society that takes better care of its people. The problem you bring up, though, is one we decided to focus on in our book, which is that it is possible for some people to be left behind, even as productivity surges ahead and even as our society as a whole benefits. There is no economic law that says that everyone has to share in those benefits equally, and when we looked around we started to see a lot of people—and particularly a lot of workers in the economy, people who wanted to offer their labor to the economy—we saw a lot of people being left behind as the machine races ahead.

Steven Cherry: Doesn’t it seem like it’s just a matter of time before we can automate large chunks of what doctors and nurses do, and what lawyers do, and a whole bunch of other white-collar professions?

Andrew McAfee: One of the things I’ve learned from writing this book is, really, “Never say never about technology.” And to make that concrete, we relied on a book that was written in 2004 called The New Division of Labor, which was the best attempt at that point in time to understand the relative skills and relative strengths of computers versus people as labor. And that book said people are just inherently better. They put in a few different categories; two of them were “complex communication” and “pattern matching.” And as an example of a pattern-matching test that a person does pretty well and a computer does quite poorly, they gave driving a car in traffic. They said that’s just an area where you have to do such a huge amount of internalizing information and matching the patterns against the rules of the road, both formal and informal, and doing it in a very dynamic environment, that people do it quite easily and computers are hopelessly bad at it. You fast forward just six years and you find yourself in the world of the Google self-driving car, and I’m pretty convinced that if we replaced all of our cars right now and all the human drivers with the best autonomous cars available at this point in time, our roads would become safer. So I learned never to say never about technology. That said, there are still some skills where we haven’t seen the computers become the masters yet. So human abilities like creativity and empathy and the ability to look at a broad situation and come up with an innovative approach—computers just aren’t much good at that yet. And to give a fairly trivial example, we’ve programmed computers so they can summarize the results of a baseball game and put that into serviceable English prose automatically. Now that’s not too hard to do, because there is a template for what a baseball game is—who won, who lost, who pitched all that sort of stuff—so it’s a fairly formulaic as far as prose goes. I have never seen a computer that could write a 10-page report about this business problem, for example—I just haven’t seen them come close. So there do appear to be some areas where people have a defensible lead at least for some time to come. But two of the professions you mentioned in your list are really interesting—you mentioned both doctors and lawyers. It turns out a lot of what they do, not all, but a lot of what they do is pattern matching. The work of discovery, for example, as part of a lawsuit or another legal proceeding is an exercise in large-scale pattern matching across a bunch of documents. Computers are fantastic at that now, and we’re seeing that e-discovery software is hugely more cost efficient than deploying an army of lawyers to sit around and read documents in a big room. It also appears to be better at it than that army of humans would be. I know, for example, that the team behind Watson at IBM is also interested in applying that technology to the medical field. Now doctors do a lot of things; one of the things they do is diagnose what’s wrong with their patients, which again is just an exercise in pattern matching. And it’s, I believe, completely technically feasible right now to take the entire body of human medical knowledge, make that the database that Watson draws from, feed him the best possible description of the patient’s symptoms and all their lab results and their medical history. If we do that, I think I want Dr. Watson to diagnose my disease instead of a human doctor.

Steven Cherry: I’m struck by the fact that over the course of the 20th century, farming in the U.S. went from 38 percent of all workers to just 2 percent. And when you look at Watson and what’s going on with that and with robots and biotechnology and a few other things, you know, a few more generations of Moore’s Law, it seems like knowledge work could go from 38 percent to 2 percent.

Andrew McAfee: Yeah, and again I find my crystal ball here supercloudy. I have not seen computers do some of these human skills, but again, I’ve learned to say never say never. I’ll start to get a lot more worried when I see the first example of digital creativity, or very deep digital empathy, or the ability to communicate with another human being. I’m not saying that can never happen.

Steven Cherry: You know, your book’s title is just one single character away from the phrase “rage against the machine,” which is the name of a rock band and their first CD. Do you worry that the race could turn to rage?

Andrew McAfee: When we look around at the protest movements on both the left and the right, the Occupy movements and the Tea Party movements, we see this huge discontent with the status quo, and that’s aimed at a bunch of places, obviously. It’s aimed at Wall Street in come cases; it’s aimed at the gridlock in government in a lot of places. But I get the sense that it’s also aimed at this—I would call it a fraying of the American Dream. And what I mean by that is for the recent history, several decades worth of history in America, the social contract has been—and I’m going to oversimplify it a huge amount—but the social contract has been something like, If you are willing to work, there will be a job waiting for you. And we’ve done a fairly good job with that social contract, not a perfect job obviously, but pretty good. When I hear a lot of the protesters talk, I get the impression that they believe that social contract is fraying. When we looked at the data, we came to the same conclusion.

Steven Cherry: I read that you and your coauthor were offered a sizable advance by a major publisher for Race Against the Machine, but you went the self-publishing Amazon e-book route. Is publishing another one of those industries that’s losing the race against the machine?

Andrew McAfee: I think that the data are pretty clear that publishing is losing the race as we head deeper into the digital world, and you bring up the fact that we self-published our book. The main reason we did that is that the publisher we were planning to work with just could not meet our aggressive timeline for getting this book out there into the world. We had some specific dates that we needed to hit, and we wanted to get these ideas out as quickly as possible after we were finished writing them down. The publisher couldn’t commit to do that; their timelines were just too long, so Erik and I looked at each other and said, “I guess we’re going to experiment with self-publishing,” even though neither of us had done any of it before. And what we learned pretty quickly was, even starting from a dead stop, it was strangely easy for us to assemble all the pieces necessary to get a professional looking book, at least a digital one, out there into the world. So we found people that could take our word-processing manuscript and turn it into an EPUB file or a MOBI file. We found a proofreader who could clean up and tighten up our prose. We crowdsourced the cover for the book on 99 Designs. And we were able to kind of piece it all together ourselves in a short amount of time, even though I wouldn’t say that project management is the best skill that either Erik or I possesses. So we looked at each other at the end of this and said, “Wow, this experiment in self-publishing was remarkably smooth and easy.”

Steven Cherry: Very good. Well, I’m glad you got it out as quickly as you did, and I’m glad we were able to talk about it today.

Andrew McAfee: Fantastic. Thanks for having me.

Steven Cherry: We’ve been speaking with Andrew McAfee, of the MIT Sloan School of Management, about the opportunity and threat that technology poses for the future of jobs. For IEEE Spectrum’s “Techwise Conversations,” I’m Steven Cherry.

Announcer: “Techwise Conversations” is sponsored by National Instruments.

This interview was recorded 22 November 2011.
Segment producer: Barbara Finkelstein; audio engineer: Francesco Ferorelli
Follow us on Twitter @TechwisePodcast

NOTE: Transcripts are created for the convenience of our readers and listeners and may not perfectly match their associated interviews and narratives. The authoritative record of IEEE Spectrum’s audio programming is the audio version.

Advertisement
Advertisement