The labor market is about to be transformed by machine intelligence, the combination of ubiquitous data and the algorithms that make sense of them. That’s according to economist Tyler Cowen, in an argument spelled out in his recent book Average is Over. As Cowen sees it, your job prospects are directly tied to your ability to successfully augment machine intelligence. He writes:
Workers more and more will come to be classified into two categories. The key questions will be: Are you good at working with intelligent machines or not? Are your skills a complement to the skills of the computer, or is the computer doing better without you? … If you and your skills are a complement to the computer, your wage and labor market prospects are likely to be cheery. If your skills do not complement the computer, you may want to address that mismatch. Ever more people are starting to fall on one side of the divide or the other. That’s why average is over.
But what about management? I interviewed Cowen last month about his vision of the future, and where he sees managers fitting into it. An edited version of our conversation follows.
WF: What do you see as the main career lessons of the book?
TC: One thing the book suggests is that only being technically skilled may not be that useful, because those jobs can be outsourced or even turned over to smart machines. But people who can bridge that gap between technical skills and knowing some sector in a way that’s more creative or more intuitive, that’s where the large payoffs will come.
A classic example is Mark Zuckerberg with Facebook. Obviously a great programmer, but had he just gone out to be paid as a programmer he wouldn’t be that well off. He was a psychology major — he understood how to appeal to users, to get them to come back to the site. So he had that integrative knowledge.
For people who are not technically skilled, marketing, persuasion, cooperation, management, and setting expectations are all things that computers are very far from being good at. It comes down to just communicating with other human beings.
WF: Where does management fit into this?
TC: In any company, you need someone to manage the others, and management is a very hard skill. Relative returns to managers have been rising steadily; good managers are hard to find. And, again, computers are not close to being able to do that. So I think the age of the marketer, the age of the manager are actually our immediate future.
WF: I was talking recently with Andrew McAfee of MIT, whom you reference the book, and he mentioned the way software might even replace some pieces of management, though not managers themselves. Do you see algorithms and software encroaching majorly into that area?
TC: If it’s just measuring how hard people work, how long they’re at their desks, how good a job they do, how good a doctor or a salesman is adjusting for quality of customer, there’s a huge role there for software. But to actually replace managers, for the most part I don’t see that. At least not for the time horizon I’m writing about, 10 to 20 years out.
WF: One thing that McAfee talks about is the idea that people want to get their information from a human; they can be very distrustful of a computer just spitting out a recommendation.
Do you see that being true in a management context as well? That part of my job, if I’m running a company or division, is being able to understand machine intelligence and deliver it personally?
TC: That’s right. You will translate what the machine says and try to motivate people to do it. Professors and teachers will be more like coaches or tutors, rather than carriers of information. They’ll steer you to the program, tell you which classes to take, and be a kind of role model to get you excited about doing the work.
It’s a very important skill, and hard to learn. But I think you’ll see this kind of pattern again and again.
WF: How else do you see management changing?
TC: Management — for all the change we like to talk about — has actually been pretty static for a while. But smart machines and smart software are going to change management drastically, and in general we’re not ready for this. We will need truly new managerial thinking, not just new in the cliched sense of repackaging with new rhetoric and new categories.
WF: I wanted to ask you a little more about the machine-human teams. Freestyle chess — where human and computer teams play together, and outperform either on their own — is the example throughout the book.
What skills does the freestyle chess master have that the grandmaster doesn’t?
TC: The program, of course, does most of the calculations. The one skill the human needs when playing freestyle is how to ask the program good questions. Knowing what questions to ask is how you beat a solo program playing against you, and you don’t even have to be very good at chess. You need to understand chess at some core level, and you need to understand what different programs can and cannot understand.
It’s a kind of meta rationality. Knowing not to overrule the programs very much, but also knowing they’re not perfect, and knowing when to probe. And I think that, in management, those will be the important human skills.
WF: How do you determine where the line is between when the employee is able to add value above and beyond what computer intelligence is giving them, and when the software itself can replace them?
For example, I’m imagining some analytics software. Someone is very skilled at using it, they’re drawing some conclusions, they’re presenting them to people. But the next version of software may have built-in the ability to make those inferences. What skills keep that human from being replaced?
TC: People who can judge that there’s more to the matter than the software can grab; people who can judge the fact that there’s a need for a different kind of software for the problem; people who know when to leave the software alone and get out of its way.
Those are difficult to acquire and often quite intangible skills, but I think they’re increasingly valuable. You can think of other professional areas, like law or soma medicine, where you let the software do a lot of the work but you can’t uncritically defer to it. Software is bad at common sense in a lot of ways and it misses a lot of context. It’s people who can provide context.
WF: You have an interesting section in the book with respect to specialization in science. Do you see the path towards greater specialization as the path to career security? Or is there still a role for generalists?
TC: There’s a role for both, but you need to ask whether these terms lose some of their meaning. If someone says to you “I specialize in being a generalist,” it’s not actually a crazy claim. Most people cannot be a generalist, and you have to work really hard at a bunch of particular things to be good at it. You’re specializing in doing that. These are people who integrate and understand the contributions of others — that’s a lot like managing. So what you call generalists — I would not oppose them to specialists — there’s a big and growing role for them.
WF: You can’t go a day without seeing a story about who should learn to code or not learn to code. The same thing with respect to statistics. Given the way you see the labor market breaking out, are there specific things you advise people go out and learn?
TC: Statistics will be an increasingly big area. And even knowing a little can have a pretty high return. Coding’s tricky. If you can learn it, great. But if you can’t do it right, you really shouldn’t bother. There’s no half way.
But if you’re a manager or you work in health care, you might not ever be doing statistics, but if you can grasp some basic stuff that you can teach yourself, there’s a very high return. And it’s really quite feasible, unlike coding where it’s a major undertaking. If you’re a doctor trying to figure out which parts of the hospital are bringing in the money and someone hands you statistics — if you’re helpless, that’s really bad.
WF: One area that strikes me as one of the more difficult for machine intelligence is strategy. Where to position your business in a marketplace seems like on the far end of what machines can tell us.
TC: Yeah, that’s all humans, though you might consult machines for background information. But in no sense are machines close to being able to do that. That’s a very long way away.
All of our sectors are all on different paths, and the differential timing actually will be useful because we won’t have to figure it out all at once. We’ll get lessons from different areas sequentially and adjust. Humans will switch into the sectors they’re still good at in a rolling way. And that will make this socially more stable and better for most people.
If you woke up one morning and the machines were better than you at everything, that would be pretty disconcerting. That’s the science fiction scenario but it’s not that realistic.