Automated Factory Machines

Fear of technological progress can be observed throughout history. From British textile workers rioting against labor-economizing techniques in 1811 to John Maynard Keynes warning of “technological unemployment” in 1930, many have worried that new technology would lower wages and eliminate jobs. In reality, while some people lose work temporarily, most are reallocated to new, higher-paying jobs. In the words of French philosopher Frederic Bastiat, “to curse technology is to curse the human mind,” and in the face of these Luddites’ worries, technological process has consistently led to increased productivity, higher incomes and more jobs: all three of these statistics (GDP, median income, total employment) have increased steadily through the end of the twentieth century.

There is intense debate, however, over whether this trend will continue. A recent Pew survey revealed that barely half of technology experts believe that automation will continue to create more jobs than it replaces. The other half believe it will displace “significant numbers of both blue- and white-collar workers” over the next decade, and that it will leave “masses of people who are effectively unemployable.” They think we have reached a point in technological progress where human skills are finally becoming obsolete. Former Treasury Secretary Larry Summers, for instance, compared American workers today to horses in 1910, whose jobs were permanently replaced by automobiles and tractors.

Recent trends in the aforementioned data are indicative of these worries. While GDP has steadily risen over the past sixty years, employment has been mostly stagnant since 2000. Similarly, GDP per capita has more than doubled since 1975 while median household income has barely increased. As MIT economist Erik Brynjolfsson notes, “productivity is at record levels [and] innovation has never been faster,” yet the benefits of the growing economy have not been felt by average workers. Technology, it seems, may be reducing the demand for human labor.

Historically, the victims of technological unemployment and dislocation have been low-skill, low-wage workers. From 1979 to 2010, the number of Americans working in factories declined from 19.5 million to 12 million while total U.S. population increased by almost 100 million. Similarly, the agricultural industry that employed 41 percent of the American workforce in 1900 only employed two percent in 2000.

This historical trend may be coming to an end, however, as technology now poses a huge threat to mid-skill, mid-wage jobs as well. Today, most at risk of automation are well-paid jobs requiring repetitive tasks, such as the work involved with accounting, law, clerical work, medicine and even education. The recession of 2007 likely sped up the destruction of these types of jobs. They “fell off a cliff in the recession” states Henry Siu, an economist at the University of British Columbia, “and there’s been no large rebound.” According to Siu, this type of work, ranging from white-collar jobs in sales and administration to blue-collar jobs in assembly work, makes up about 50 percent of employment in the U.S.

High levels of income inequality in the U.S. today are indicative of the loss of these middle class jobs. In what is termed the “polarization” of U.S. employment, the middle class is “hollowing out” as demand increases for not only high- but also low-level work. Most of the new employment opportunities afforded by recent technological advances have been high-wage jobs, such as software engineers and executives, which require high levels of education. This trend is evident in the growing income gap between high school- and college-educated Americans, which has more than doubled since 1965. Demand has also increased for low-skill jobs such as restaurant workers, janitors, and other service workers, whose work cannot yet be automated. The effects of these changes are exemplified by trends in inflation-adjusted incomes from 1970 to 2010. The upper class share of income has steadily increased and the lower class share has remained stagnant, while the middle class share has decreased nearly to the point of becoming equal to that of the lower class.

MIT economist David Autor argues that this trend does not seem likely to continue and that the scope of technological substitution in middle-skill work is “bounded.” To explain this, Autor cites machines’ inability to acquire tacit knowledge, which involves judgment, common sense and flexibility. He says, “there are many tasks that people understand tacitly and accomplish effortlessly but for which neither computer programmers nor anyone else can enunciate the explicit ‘rules’ or procedures.” Speaking a language, for instance, is learned at a young age without significant instruction in grammar. Yet programming a computer to understand a language (not to mention the nuances of voice inflection and body language) is extremely difficult. Autor uses this concept to argue that the increasing polarization of jobs that has been occurring in the past couple decades may be coming to a close because machines are nowhere near the level of sophistication that would be required to perform a task requiring tacit knowledge.

Recent advances in areas such as machine learning, however, could prove Autor wrong. Computers today are able to accomplish tasks that were once considered impossible to automate, even high-level tasks like writing this article. Such machines are called advanced natural language generation (NLG) platforms and are sold by companies such as Narrative Science. Their program, called Quill, is given data (such as a sports box score or company annual report) and can produce complex, engaging stories that are, based on surveys by the company, indistinguishable from human writing. While this technology is relatively new, it is gaining prominence quickly. Large content producers like Forbes and the Associated Press are already using the service, with many others planning to incorporate it in the future.

Similarly, computers can now drive cars, stock shelves, clean buildings, diagnose illness, detect fraud and prepare briefs for lawyers. Jobs that require tacit knowledge, Autor’s alleged saviors of the middle class, are becoming increasingly easy to automate, begging the question: what is the future of the middle class? In addition to the income stagnation that we have already witnessed, some experts predict widespread unemployment, a long-term displacement of as much as 50 percent of the work force.

It is impossible to accurately predict how dramatically technological progress will affect the labor market, and there are still many experts who doubt the scale of its impact. Nonetheless, evidence suggests that technological progress may permanently supplant a significant portion of the workforce in the future, and it has already replaced a substantial number of American middle class jobs and driven inflation-adjusted income downwards.

Despite the threat it poses to the American labor force, however, the truth remains that technological progress makes work easier and more efficient and that its continuation is a reality of our modern economy. This is best explained by Milton Friedman. When visiting the construction site of a canal in which workers were using shovels instead of modern earth moving equipment, Friedman was told that the project was a jobs program. His response was: “Oh, I thought you were trying to build a canal. If it’s jobs you want, then you should give these workers spoons, not shovels.”

Driverless cars, data-analyzing computers and robots are all machines that can perform typical “human” tasks. Machines seem to be getting better at doing our job, and the fear of technological substitution of labor is returning to the stage of public debate.

Classical economic theory states that new technologies should be neutral with respect to labor demand. Gains from cheaper production should be transmitted to consumers in the form of expanded supply, causing labor demand to increase and offset the loss of jobs due to substitution for workers with machines.

Yet recent research by Miguel Barroso Morin of Cambridge University tells a different story. He analyzed the impact of the spread of cheap electricity in the ‘30s – spurred by technological progress — on the cement industry. The results were quite surprising: the drastic drop in electricity prices was not met by an expansion in production, and it correlated with a significant decrease in employment. Additionally, the share of income going to workers fell by 11 percent as labor productivity shot up by 36 percent. In this case, technology was not neutral with respect to labor – rather, technology crowded out workers.

One can imagine a similar pattern occurring with automation, especially if we consider the transformative nature of artificial intelligence. As the gap between what humans and robots can do narrows, technology will threaten not only manual jobs but also several high-skill professions.

Strikingly enough, according to an index compiled by Deloitte and the University of Oxford, accountants faces a 95 percent probability of being automated over the next two decades.

Although humans still significantly outperform computers in pattern recognition tasks, recent developments in machine learning seem to indicate our silicon-wired competitors are catching up quickly. Fast forward to 2050, and the most pessimistic can already picture a dismal landscape of wasted human capital as workers sit idle and machines do all the work.

But even if this does happen, not all is lost. According to Erik Brynjolfsson and Andrew McAfee, authors of “The Second Machine Age,” it’s highly possible to attain fabulous levels of wealth through technology.

The real risk is moving toward a labor-light society wherein machine capital ownership would be the only pathway to such wealth. Thus, the biggest challenge countries will face in the future is the tendency of capitalism to generate stark inequalities. Some claim that we are already experiencing this trend today, but its effects may be exacerbated in a labor-light economy.

The same old story

Yet Morin’s view has its notable critics. According to Martin Wolf, there is nothing special with automation when it comes to the labor market: it can be seen as a simple positive production shock that won’t alter the fundamentals of the relationship between technology and the labor market.

Wolf suggests economically significant technological change has, in fact, slowed down in the recent past. Evidence for this can be found in the slow growth of output per worker, stuck at slightly above one percent in the past decade. If countries were in the midst of an earth-shattering technological revolution, then their economies should have seen higher growth in output per worker in recent years.

Furthermore, even considering measures of innovation that escape crude GDP metrics, we fail to see major shifts in the way technology progresses. Take consumer surplus: when a new, economically significant invention is introduced, its price is effectively taken from an infinite value to a finite one, hence increasing consumer surplus immensely. Yet, we have not seen many large gains to consumer surplus being brought forth by today’s seemingly uncanny technology.

In addition, although recent developments, such as quantum computing, may seem impressive, the quick technological growth we’re experiencing today is not equivalent to the ground-breaking advancements of the 18th century. Although today’s high-tech gizmos appear to be changing and improving rapidly, they all pale in comparison to major breakthroughs, such as the development of steam power, electricity or even access to clean water.

The effects of automation would not be apocalyptic, but rather in accordance to a recurring pattern: as new technologies are introduced, some people will find themselves out of work because they have neither the education nor the technological skills to remain competitive within the labor pool.  As discussed in The Race Between Education and Technology by economists Claudia Goldin and Lawrence Katz, technological progress increases the demand for skilled workers relative to unskilled workers, which inevitably leads to disparities in wages and employment. As routine intellectual and manual jobs fall prey to technological substitution, we will not witness the complete disappearance of human labor but further rearrangement of returns to education and access to technology.

The upshot, however, is not all too different from what Brynjolfsson and McAfee worry about: a massive wedge in the economy between the lucky few, highly-skilled laborers (such as programmers), owners of robot capital, and essentially anyone else.

Another revolution

But is this the whole picture? The answer to the age-old conundrum of technological substitution largely depends on the theoretical framework we adopt. It is however clear that, at the present state, most jobs are safe from a total wipe out and there seems to be no impending risk of labor-market catastrophes.

But there is a different, perhaps subtler technological revolution that might be more concerning than automation – the increasing dominance of information technology.

The commercial implementation of information technology has rapidly increased in the past few years. This is favouring a shift away from an economy based on physical capital towards one that is reliant on software, intellectual property and human capital. A quick comparison between Sony and WhatsApp, two firms boasting a similar market value, evinces how there is an information technology revolution. The former employs more than 130,000 workers — the latter, only 46.

Returns on capital are skyrocketing and the traditional concept of employment may already be evolving. Soon enough we might have to say goodbye to lifelong jobs as they give way to “portfolio careers.” Workers may constantly need to stand ready to take up temporary jobs, retrain and relocate in order to act on the best opportunity available.

This trend has been apparent for quite some time, especially in highly dynamic markets. One can argue that it is largely due to market regulation and underlying macroeconomic fundamentals, but it does highlight an indirect and somewhat overlooked channel through which technology affects the labor market.

Something is indeed disrupting the way we work: it just turns out it is not a dystopian horde of robots but rather a large-scale rearrangement of means of production ultimately caused by good old human ingenuity.