Will Technological Innovation Steal Jobs from People?

Along with demographic changes characterized by population aging and low fertility, the impact of technological innovation is an important factor to look into in considering the future of employment in Japan. In particular, the spread of automation, robots, computers, and artificial intelligence (hereinafter collectively referred to as "new mechanization") has been the focus of attention in the business community and academia as to what impact they will have on the future of employment.

The fear of losing jobs to new technology has been reincarnated many times in history. One typical example is the Luddite riots in the early years of the Industrial Revolution, in which British textile artisans destroyed newly-introduced spinning machines and factories that threatened to leave them without jobs.

In the 1930s, John Maynard Keynes warned that technological innovation, which would bring material prosperity, might also lead to widespread technological unemployment if the pace of the resulting economization of labor is too rapid. In the 1980s, Wassily Leontief, a Nobel Prize-winning economist, said that "the role of humans as the most important factor of production is bound to diminish" just like the role of horses.

♦  ♦  ♦

However, when we look back over the past 200 years, even though technological innovation eliminated certain jobs, overall employment has expanded as new technologies led to improved productivity and the emergence of new businesses and industries, which in turn created new demand and new jobs. Will the latest wave of concern about widespread technological unemployment that may result from the new mechanization prove to be just another unfounded fear? Or will things turn out differently?

In his 2014 paper drawing on philosopher Michael Polanyi's observation that "we know more than we can tell," Professor David H. Autor of the Massachusetts Institute of Technology (MIT) coined the term "Polanyi's paradox" to describe the situation and focused on tacit knowledge (what we know) to explain the difference between humans and machines.

In what follows, I will summarize a series of studies on the new mechanization and human labor pioneered by Professor Autor. Here we classify jobs into two types: routine jobs for which rules and procedures can be described explicitly (e.g., cashiers and simple manufacturing work) and non-routine jobs in which people cannot tell but tacitly understand how to perform them. The first type, which has traditionally constituted middle-wage jobs requiring middle skills, is susceptible to the impact of the new mechanization, and the proportion of such jobs has already begun to decline in advanced economies such as the United States, Europe, and Japan.

When we classify non-routine jobs further into intellectual labor and manual labor, the former (e.g., professionals) constitutes high-wage jobs requiring high skills and the latter low-wage jobs requiring low skills (e.g., cleaning and janitorial work). With both of these types generally increasing in proportion, the polarization of jobs is occurring in advanced economies. According to such analysis, those engaged in routine jobs are the only ones who would suffer from the adverse impact of the new mechanization.

In a 2015 paper, MIT Professor Erik Brynjolfsson and Andrew McAfee, a principal research scientist at MIT, maintain that it is unlikely for the new mechanization to replace human labor within the next decade. However, they warn of the possibility that "human labor, in aggregate, decline in relevance because of technological progress" over the longer term. This is because technology is advancing at a significant pace and beginning to encroach on human territory, an area that had been considered beyond the reach of machines.

For instance, driving a car is considered a typical example of non-routine manual labor, and, up until several years ago, automated driving seemed to be far away from practical use. However, the self-driving technology being developed by Google Inc. is reaching a remarkable level of sophistication. Pattern recognition--an ability to recognize an object (e.g., chair) as such by watching the photo image of the object--is another area that has been considered beyond the reach of the new mechanization as it requires the use of tacit knowledge. However, machine learning, which is to train and enable machines to estimate tacit rules and thereby make the best prediction by inputting an enormous volume of data, is evolving at a rapid pace. As such, it is getting increasingly difficult to foresee how things will develop over the long term.

♦  ♦  ♦

Meanwhile, in a 2015 paper, Professor Autor points out that the mass media and some scholars, while overstating the extent to which the new mechanization might substitute human labor, often ignore the positive effects of strong complementarities between the new mechanization and human labor on productivity, personal income, and labor demand. As an example of complementarities between technology and employment, he cites a study on the U.S. experience with automated teller machines (ATMs). Contrary to the general assumption that the spread of ATMs would diminish the demand for bank tellers, the study found that the number of such workers actually increased over a 30-year period (1980-2010). Although the volume of routine cash-handling tasks decreased, the development of information technology enabled banks to build close relationships with individual customers, creating new tasks to provide additional services (credit cards, loans, and investment products).

Chess is a perfect example to demonstrate just how much humans and machines can complement each other. For quite some time, human chess players had been unable to defeat super computers. However, a team comprised of a human chess player and a computer can overwhelm even the strongest of such super computers.

Meanwhile, Amazon.com Inc. has introduced shelf-moving robots called Kiva at its warehouses in the United States, but not all of the tasks at the warehouses are automated. A Kiva robot lifts a movable shelf and drives it to a human picker where the person selects specific merchandise. The Kiva robot then returns the shelf back to its original position. In other words, the role of robots is limited to the routine work of moving shelves, whereas human workers handle merchandise and control overall warehouse operations with software.

Another point Professor Autor made is that the polarization of jobs is not an endless process. A closer look at middle-skill jobs reveals that each one of them is composed of multiple tasks requiring different sets of skills, and quite often such tasks are mutually complementary. Such is the case, for instance, with jobs that are a combination of routine tasks requiring a certain mastery of technique and non-routine tasks requiring skills for interpersonal interaction, flexibility, adaptability, and problem solving. Those kinds of job are unlikely to be affected by the new mechanization because it would be a recipe for gross inefficiency to discompose those jobs and mechanize only the routine tasks.

As examples of middle-skill jobs, Professor Autor cites medical support occupations, plumbers, builders, electricians, automotive technicians, and retail clerical occupations that involve coordination and decision-making functions. Meanwhile, a 2015 paper by Georgetown University Professor Harry Holzer points out that the share of new types of middle-skill jobs has increased in recent years while that of overall middle-skill jobs has declined (see Table).

♦  ♦  ♦

In order to prevent widespread technological unemployment in the future, what preparations should we make? Professor Autor points to the need to make human capital investment to develop skills that are complemented, rather than substituted for, by technological changes. Then, how can and should we foster skills that cannot be performed by machines and become more highly valued when accompanied by the new mechanization? For one thing, Professor Brynjolfsson et al. point out that people's desire to connect with others generate demand, or their willingness to pay, for human factors and skills such as artistry (theatrical plays and music), physical abilities (sports), caring (therapy services), and entertainment (restaurants).

For another, as noted in their 2014 book, The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies, what is important is to foster skills to come up with novel ideas and concepts through free and flexible thinking. Machines can answer questions, but they are not yet able to ask questions. They note that a disproportionately large number of prominent innovators in the United States have attended schools that practice Montessori education. Developed by an Italian physician in the early 20th century, the Montessori method of education emphasizes letting children learn spontaneously by following their curiosity and asking why the world is arranged as it is in free and self-organizing learning environments. This offers important insights as to how Japan should reshape its education.

>> Original text in Japanese

* Translated by RIETI.