字幕表 動画を再生する 英語字幕をプリント A decade ago, robots still seemed pretty limited. Now, not so much. And computers don’t just win chess any more, they can win Jeopardy. “Watson.” “What is the of the Elegance of the Hedgehog?” They can win Go. “There are about 200 possible moves for the average position in Go.” This is all happening really fast. And it’s causing some to forecast a future where humans can’t find work. “There will be fewer and fewer jobs that a robot cannot do better.” “And what are the people gonna do?” “That’s the $64,000 question.” I believe this is going to be one of the biggest challenges we face in the coming decades. “People who are not just unemployed. They are unemployable.” But if you ask economists, they tend to have a pretty different view from the futurists and Silicon Valley types. Do you worry that new technologies could cause mass unemployment? Yes. No. I have devoted my career to worrying about the labor market, particularly worrying about the living standards of low and moderate income workers. So I worry a lot about things. I am not worried about this. One of the reasons a lot of economists are skeptical about robots taking all the jobs is that we’ve heard that before. There was a spike of automation anxiety in the late 20s, early 1930s when machines were starting to take over jobs on farms and also in factories. This article from 1928 points out that there used to be guards who opened and closed the doors on new york subway trains, and people who took tickets before there were turnstiles. And I just love this quote: It says “building materials are mixed like dough in a machine and literally poured into place without the touch of a human hand.” Automation anxiety surged again in the late 1950s, early 1960s. President Kennedy ranks automation first as job challenge. “Computers and automation threaten to create vast unemployment and social unrest” “What should I do Mr. Whipple?” “Stop him!” This article from 1958 is about 17,000 longshoremen who were protesting automation on the piers. And if you don't know what longshoremen are, that’s because there aren’t many of them left. Technology destroyed a lot of those jobs. And yet, we didn’t run out of work. This chart shows the percentage of prime-age people with jobs in the US. Ever since women joined the workforce in big numbers, it’s stayed around 80%, outside of recessions. During this period, technology displaced some 8 million farmers in the US, 7 million factory workers, over a million railroad workers, hundreds of thousands of telephone operators, we’ve lost gas-pumpers, elevator attendants, travel agents. Tons of jobs have died but work persists. What you realize when you look through those old reports is that it’s really easy for us to see the jobs being replaced by machines. It’s a bit harder to visualize the jobs that come from what happens next. New technology creates jobs in a few ways. There are the direct jobs for people who design and maintain the technology, and sometimes whole new industries built on the technology. But the part we tend to forget is the indirect effect of labor-saving inventions. When companies can do more with less, they can expand, maybe add new products or open new locations, and they can lower prices to compete. And that means consumers can buy more of their product, or if we don’t want any more of it, we can use the savings to buy other things. Maybe we go to more sports events or out to dinner more often. Maybe we get more haircuts or add more day-care for the kids. This process is how our standard of living has improved over time and it’s always required workers. The key economic logic here is automation does indeed displace workers who are doing work that got automated, but it doesn't actually affect the total number of jobs in the economy because of these offsetting effects. Warnings about the “end of work” tend to focus on this part and not all of this -- like a widely cited study from 2013, “According to research conducted by Oxford University, nearly half of all current jobs in America --” “47 percent of all our jobs--” “47 percent of US jobs in the next decade or two, according to researchers at Oxford, will be replaced by robots.” That study assessed the capabilities of automation technology. It didn’t attempt to estimate the actual “extent or pace” of automation or the overall effect on employment. Now, all this doesn’t mean that the new jobs will show up right away or that they’ll be located in the same place or pay the same wage as the ones that were lost. All it means is that the overall need for human work hasn’t gone away. Technologists and futurists don’t deny that’s been true historically, but they question whether history is a good guide of what’s to come. Fundamentally the argument is that this time it’s different. That’s what I think. Imagine a form of electricity that could automate all the routine work. I mean, that’s basically what we are talking about here. And so It’s going to be across the board. And it is easy to underestimate technology these days. In a 2004 book, two economists assessed the future of automation and concluded that tasks like driving in traffic would be “enormously difficult” to teach to a computer. That same year, a review of 50 years of research concluded that “human level speech recognition has proved to be an elusive goal.” And now? “Ok Google. How many miles has google’s autonomous vehicle driven?” “According to Recode, that’s because the company announced its self-driving car project, which was created in 2009, has racked up over two million miles of driving experience.” This is the textbook chart of advancement in computer hardware — it’s the number of transistors that engineers have squeezed onto a computer chip over time. Already pretty impressive, but notice that this isn’t a typical scale: these numbers are increasing exponentially. On a typical linear scale it would look more like this. It really is hard to imagine this not being massively disruptive. And as the authors of The Second Machine Age point out, processors aren’t the only dimension of computing that has seen exponential improvement. The idea of acceleration in your daily life when do you encounter that? Maybe in a car for a few seconds? In an airplane for seconds again? The idea that something can accelerate for decades literally just continuously is just not something that we deal with. I mean, we think in straight lines. But even though there’s been all this innovation, it’s not showing up in the data. If we were seeing this big increase in automation we would see productivity growing much more rapidly now than it usually does, and we are instead seeing the opposite. Labor productivity is a measure of the goods and services we produce divided by the hours that we work. Over time it goes up - we do more with less labor. We’re more efficient. If we were starting to see a ton of labor-saving innovation you’d expect this line to get steeper, but when you look at productivity growth, you can see that it has been slowing down since the early 2000s, and not just for the US. It’s possible that new technologies are changing our lives without fundamentally changing the economy. So will this all change? Will today’s robots and AI cause mass unemployment? There’s reason to be skeptical, but nobody really knows. But one thing we do know is that the wealth that technology creates, it isn’t necessarily shared with workers. When you account for inflation, the income of most families has stayed pretty flat as the economy has grown. One of the problems we've seen over the last 40 years is that we have seen all of this rising productivity growth but actually hasn't been broadly shared, it's been captured by a thin slice of people at the top of the income distribution. Even if unemployment stays low, automation might worsen economic inequality, which is already more extreme in the US than it is in most other advanced countries. But technology isn’t destiny. Governments decide how a society weathers disruptions, and that worries people on both sides of the debate about the future of work. We’ve adopted policies that instead of really trying to counteract the trend caused by technology and globalization and other things, we’ve in many cases exacerbated them. We’ve put a wind in the back of them and made them more extreme. And that’s a big problem. We will probably always be fascinated by the prospect of robots taking our jobs. But if we focus on things we can’t really control, we risk neglecting the things we can.