Mark P. Mills, based on his book, Work in the Age of Robots,
Encounter Books, 2018

The nation today is operating at a record low unemployment level. We are near what economists call full employment. And, as the U.S. Bureau of Labor Statistics data show, employment is growing faster in industrial domains than in health care and in “professional and technical services.” The talk in business circles these days is about the shortage of skilled labor, and about the availability—or willingness—of enough people to fill future job openings.

But the talk amongst pundits and Silicon Valley’s self-reverential futurists is quite different. They claim that the labor-savings, about to come from algorithms, artificial intelligence (AI), automation and robots, will destroy so many jobs that unemployment will radically, permanently increase.

In response, this will require, the proponents argue, the creation of a universal basic income, not just for the temporarily unemployed, but for those doomed to be never-again-employable. We’ve seen this movie more than once.

The saddest aspect of life right now is that science gathers knowledge faster than society gathers wisdom.

Isaac Asimov

In 1961, President John F. Kennedy delivered a message to Congress focused on what he called “the inevitability” of job destruction from automation. He created an Office of Automation and Manpower and proposed that Congress fund training programs and create a readjustment allowance for displaced workers.

On the morning of October 4, 2010, the manager of the Grosvenor Hotel discovered this painting – which has been attributed to Banksy, a world-famous but anonymous English street artist – on an exterior wall. The Grosvenor Hotel overlooks the seafront in Torquay, a resort town in southern England.

A few years later, President Lyndon B. Johnson convened a blue-ribbon commission on the impact of automation on work. One of its recommendations: a universal basic income.

Fast forward to May 2018 when the White House held a Summit with 40 tech companies, including the likes of Google and Amazon. The focus: hand-wringing over the inevitability of “job displacement” from automation and artificial intelligence.

So, we find ourselves at a curious point in history.

For a decade now, despite the wonders of Uber, Amazon and Apple, the U.S. has actually been in a productivity deficit. The definition of productivity is the reduction in inputs—labor and materials—per unit of output. In other words, our productivity deficit means that America is currently underinvested in automation technologies.

But now, we’re told, we are about to see a new kind of productivity that will spur economic growth, but for the first time in human history, it will also lead to the long-term net destruction of work.

Ever since the dawn of the industrial revolution, we have witnessed continual and profound advances in technologies that have improved productivity. In fact, the magnitude of labor-saving now expected to come from AI and robots won’t match what happened, for example, a century ago. In just 20 years, from 1910 to 1930, there was a 400 percent drop in labor-hours needed per car manufactured, and a 700 percent decline in labor-hours to produce a ton of steel. And those are not exceptions, but typical of what technology progress has brought to many industries over the past century and a half.

Over that time, despite this massive “labor saving” and despite population growth, which increased the total number of labor-seekers, 95 percent of job-seekers, on average, found employment. In other words, the unemployment rate has oscillated around 5 percent, despite serial technological disruptions.

Of course, there have been periods of high unemployment during that time span. But those episodes were caused not by technology but by poor policies and economic malpractice.

The essence of today’s dystopian argument about the future is, in the immortal phrase of every failed forecaster: This time it’s different.

The implicit, if not explicit, argument is that AI and robots are, well, different. Of course, the specific technologies underlying every revolution are different.

With regard to robots, in the early days of robots people said, “Oh, let’s build a robot” and what’s the first thought?  You make a robot look like a human and do human things.  That’s so 1950s.  We are so past that.

Neil deGrasse Tyson

The central question is whether the effects of new technologies are different than in the past.

A core tenet of the peak jobs thesis is anchored in offering an analogy to the historical fact every schoolchild learns: Technology eliminated nearly all farm work. But there is something fundamentally wrong with using that model as a predictor for other work.

The critical difference between work associated with producing food compared to fabricating things and producing services is found on the demand side, not the productivity side of the equation.

Growth in demand for food is bounded by the combination of two obvious variables: population growth, plus the rise in per-person consumption for those underfed.

Population growth is slow and very predictable over very long periods. And there is only a maximum of a two-fold difference between the per capita calorie intake of a subsistence diet and that of wealthy nation.

A robot may not injure a human being or, through inaction, allow a human being to come to harm.

A robot must obey orders given it by human beings except where such orders would conflict with the First Law.

A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Isaac Asimov's "Three Laws of Robotics"

Even modest gains in agricultural productivity—again, the rate of labor-reduction per unit of product—can be far faster than the rate of growth in demand. Thus, one would expect, and we’ve seen, a rapid reduction in total farm-labor-hours needed.

Meanwhile, for the things we invent and fabricate, the demand growth isn’t limited like food but is essentially unlimited. This arises from an obvious fact, but one that seems to elude economists and forecasters: Engineers and innovators continually invent new things that create new demands.

Rising consumption of fabricated things—and the services they enable, such as tourism enabled by aircraft or entertainment enabled by computers—comes not just from increased wealth enabling more people to own what only a few had previously been able to afford. It also comes from the continual invention of new products and services that create new demands, which in turn create new requirements for jobs in both manufacturing and services.

These stamps were issued by Sweden in 1984 to commemorate available, all-electric, anthropomorphic, microprocessor-controlled industrial robot. Since then, the design and functionality of industrial robots has remained largely the same.

Evidence of this reality is clear over the last 50 years. Consumption of agricultural goods in the U.S. has risen only slightly more than the population. But consumption of things from industries has grown 300 percent more than population. And the consumption of healthcare services has grown triple that ratio—at nine times the rate of population growth.

Of course, automation has eliminated many specific kinds of jobs. It’s easy to identify jobs that will disappear. It’s harder to identify the jobs that will appear due to innovations. Some 60 percent of the kinds of jobs that existed in the 1960s don’t exist today, such as draftsmen and typists. And today we are at full employment.

There was no demand for cars or computers before their invention. The computing industry employed about 100,000 people circa 1960, compared to more than a million workers today.

The fact is global demand for manufactured goods—and for the services the new kinds of goods enable—is on the cusp of the greatest expansion in history. We are already seeing evidence of that and the collateral requirement for productivity from robots and AI to meet the scale of demand.

We’re long overdue for improving labor productivity in nearly every part of the service sector, especially healthcare. Over the past couple of decades, healthcare productivity—value added per labor-hour—has been stagnant. Adding knowledge automation won’t destroy work in healthcare, it will make it more affordable, better and, yes, expand employment.

We can look to recent history for some obvious examples. We’ve seen word processors replace typing pools, spreadsheets replace rooms full of accountants doing “ciphering,” software replace draftsmen and many more similar examples. But over these decades, both employment and the economy have grown.

Data from the Bureau of Economic Analysis shows that, on average, the industries that spent more on software increased rather than decreased employment. The data also shows that most industries are still underinvested in software. We should hope for software investment to accelerate.

The central challenge in the current transition era is as it has always been. It is not the prospect of the end of work. It is the moral and political imperative to deal with the inevitable loss of specific types of jobs and the related—and challenging—task to help a minority of workers find new jobs.

But this time, for the first time, the technology causing the disruption also constitutes part of the solution.

One can easily imagine developing social media algorithms similar to the kind that enable platforms such as Uber’s sharing and Netflix’s preference engines. AI will make it both easier and more efficient to manage the complex and inherently social process of helping displaced employees find new work, or requalify or retrain people for new kinds of work.

A central goal in engineering has always been to make technologies easier to operate by non-experts. AI in the future will become increasingly easy for everyone to use. That will democratize artificial intelligence. And that will inevitably help far more people become productive and greatly expand the base of so-called “knowledge workers” too.

And that might be the single feature of our new age that is in fact different from previous machine disruptions. 

The Robotification of Healthcare

The da Vinci Surgical System is a robotic system designed to facilitate complex, yet minimally invasive, surgery. The system is controlled by a surgeon from a console and replicates the movements of the surgeon’s hands with the tips of micro-instruments. Currently, long-distance surgery is in the early stages of development via the da Vinci Surgical System.

Healthcare is an enormous, rapidly growing and self-evidently critical service industry. It is also the sector in most need of robotification to improve productivity.

Productivity—that is, value added per labor-hour—in healthcare has been stagnant for 15 years. That single fact explains why the costs of medical services have increased by as much as four times the rate of inflation. We should hope and pray that artificial intelligence (AI) and robots can advance quickly to become commercially viable for medical applications. Automation is needed in every corner of healthcare from diagnostics and therapeutics to cleaning hospitals and moving patients.

There are signs of progress, such as using Intelligent Virtual Assistants for physician diagnoses, deploying robots to help surgeons and to assist with ambulatory care and rehab. This will precipitate decreased costs and improved care, and also—as discussed above—the creation of more jobs.

Automation, both physical and in the cloud, will democratize healthcare and create new classes of services. It will also enable many of these services to migrate from expensive, and often hazardous, institutional settings into people’s homes.

Longer term, there’s the intriguing potential to use the new AI infrastructure to create a “digital twin,” which is a supercomputer’s digital model of a machine, a system or even a person. The model can be compared in real time to the real world thing or person, using highly granular real-time data gathered by sensors embedded in the physical-world counterpart. In principle, and still aspirational, a digital twin could predict the effect of a particular medication or procedure down to the cellular level. Who wouldn’t want that? ±