The future of work might not be so bleak
What’s the future of work? Will gigs replace salaried employment, and will robots eventually leave humans with nothing to do? I see reason for scepticism, but also for concern.
Technology, of course, is already making independent work a lot easier. It puts workers into contact with customers and helps them run a back office. More importantly, it allows individuals to build and promote their reputations at low cost. Customers used to rely on a taxi company’s reputation, or choose a washing machine by the manufacturer’s brand. Now, each worker has a brand: On Uber, customers can reject drivers based on their personal ratings. A firm’s collective reputation, with the concomitant control of its employees’ behaviour, is becoming gradually less important.
That said, technology can also favour standard salaried employment. The economists George Baker and Thomas Hubbard, for example, have noted how onboard computers could change US trucking. By monitoring behaviour, they would solve a moral hazard problem: Drivers have little incentive to be as careful with company trucks as they would with their own. As a result, more drivers could become employees of companies that buy and maintain fleets, rather than going it alone. They wouldn’t have to invest in their own vehicles; and they wouldn’t be out of pocket and out of work when their trucks broke down.
More generally, conventional jobs have a lot of advantages. First, a single worker or group of workers might lack the capital needed to set up a business, or prefer to avoid the risk of running one. Second, business owners might not want their employees to have other bosses—particularly if the work involves confidential information or team projects that require undivided attention. Third, reputations based on ratings might not be reliable: The economist Diane Coyle has shown that the quality of individual consultants can be hard to monitor, at least immediately, whereas a traditional consultancy may be more efficient at “guaranteeing” quality. In short, I believe that salaried employment will not disappear, although it might become less prevalent over time.
But what about Artificial Intelligence? Many jobs involving routine (and thus codifiable) tasks have been eliminated: Banking transactions are digitized, cheques are processed by optical readers, call centres use software to shorten the conversations between customer and employee, or even replace humans with bots.
These changes have global repercussions. They threaten the low-salary, outsourced jobs that emerging and underdeveloped countries have counted on to escape poverty. In developed countries, as the economist David Autor and his co-authors have demonstrated, they tend to benefit those employees whose skills complement the new digital tools. This “hollows out” the distribution of jobs into either high-paying skilled positions or low-paying basic service positions. In the US, the difference in salary between those who hold university degrees and those who left right after high school has grown enormously in the past 30 years.
It’s still not clear, however, which human tasks computers will be able to replace, and what the effects will be. Deductive problems, in which the particular is deduced from the general rule in a logical way, are the easiest. An ATM verifies a card number, the PIN code, and the bank account balance before issuing money and debiting the account. Nonetheless, total employment in banking rose even as the ATM network spread, because demand grew and teller jobs were replaced by new tasks.
Computers have also made great advances in induction, which starts with specific facts and works towards a general law. For example, algorithms are capable of predicting the US Supreme Court’s decisions about patents as well as any legal experts. Similar techniques are enabling automated facial recognition, voice recognition, medical diagnosis, and other tasks that previously only humans could perform.
The most difficult tasks for computers involve unforeseen problems that do not match any programmed routine. Frank Levy and Richard Murnane offer the example of a driverless car that sees a little ball pass in front of it. This ball poses no danger to the car, which therefore has no reason to slam on the brakes. A human being, on the other hand, will probably foresee that the ball may be followed by a young child, and will therefore have a different reaction. The driverless car will not have enough experience to react appropriately. Although machine learning might be able to solve this problem, it illustrates the obstacles that computers still encounter.
So humans and computers face different challenges. Thanks to machine learning, computers can increasingly cope with unforeseen situations, provided they have enough data to recognize the structure of the problem. On the other hand, the human brain is more flexible: A five-year-old child can handle some problems better than any computer. So the people best equipped to succeed in the new world will be those who have acquired abstract knowledge that helps them adapt to their environment, while those with only simple knowledge preparing them for routine tasks are most in danger of being replaced.
This is why education is crucial. If we don’t have a system that gives everyone a chance to gain the necessary skills, differences in education and family background will lead to even greater inequality. Bloomberg View
Jean Tirole is a Bloomberg View columnist.
This article is adapted from Economics For The Common Good, to be published next month by Princeton University Press.
Comments are welcome at email@example.com