In every recession since the Second World War, companies that laid off workers during the downturn tended to rehire as soon as the economy bounced back. Until 1990 this process never took more than one year. However, it has taken longer and longer for workers to find jobs after the most recent
recessions of 1991, 2004, and 2009. In fact, the Bureau of Labor Statistics (BLS) announced that not until April 2014 did the private sector in the US regain the jobs lost from its 2008 pre-crisis peak. The so-called “jobless recovery” had lasted six painful years, or longer than the previous two recoveries combined.
According to economist Michael Evangelist, as quoted in the New York Times, the new jobs were quite different from the ones that were lost six years earlier: more than 1 million high-wage jobs were never replaced, and, instead, more than 1.8 million low-wage jobs were created. Most of those new jobs, by far, were in the food service industry, which is the lowest wage category in the BLS report.
One reason the US employment index responded so sluggishly after the 2009 crisis was that many of the workers who were laid off were not rehired. And they won’t be. Ever. Those workers were replaced by machines.
In a paper entitled “Jobless Recoveries” published by the centrist think tank Third Way,
economists Henry Siu and Nir Jaimovich observed that in each of the three most recent recessions employment recovered significantly slower than gross domestic product (GDP). This trend is not an economy-wide phenomenon, however. According to Siu and Jaimovich, “It can be traced to a lack of recovery in a subset of occupations; those that focus on “routine” or repetitive tasks that are
increasingly being performed by machines.” The researchers point out that occupations focused on these routine tasks tend to be middle-waged, which may explain why the new jobs post-recovery tend to pay less than those that were destroyed. It looks like the robots really are stealing our jobs.
Economists also tell us that destroying jobs is a normal part of a healthy industrial economy. For example, automation has been a part of manufacturing since the inventor James Watt added a governor to his steam engine in 1788, and concerns about machines displacing workers have been with us just as long. As early as 1930, economist John Maynard Keynes coined the term “technological
unemployment” to describe this phenomenon. So is it really a problem? On the one hand, alarmists suggest that machines might cause a permanent rise in unemployment levels. On the other hand, economists tell us that, historically, the labor market has always generated enough new and better jobs to more than offset the ones eliminated by technological progress.
They point to the cycle of displaced workers devising new crafts and gaining new skills over several months after their layoff and then trading up to more interesting and rewarding occupations.
The long-term result has consistently been the expansion of economy, increased productivity, and much better and more diverse jobs. And everyone benefits because prices for machine-made goods are significantly lower, and generally the new jobs pay higher wages than the old ones. For economic progress to occur, a certain amount of job destruction is simply necessary.
This kind of healthy job destruction has been a reliable—and vitally necessary—feature of the economy since the Industrial Revolution. What’s changed is the number of recent improvements in robotics. Suddenly, robots are everywhere, especially if we look beyond the humanoid clichés to recognize a robot for what it is: automation. Many observers fret that we are automating old jobs faster than we can create new ones. For the purposes of this argument, any automated non-biological substitute for human labor is a robot.
Mechanized robots were initially a replacement for muscle power; they perform the physical work that would have been done by human beings or animals. Forty years ago, repetitive tasks on an
assembly line or food processing plant were deemed suitable for robots. More recently, so were tasks that require stamina and sustained focus in a predefined area, such as a factory floor or a
warehouse. Robots can now handle pattern-detection jobs, like pulling weeds in a field of crops.
Sorting, stacking, and moving barcoded items for transport is a robot-ready task.
Thanks to rapid advances in technology, the machines are smarter than ever. Robots are now versatile and capable enough to be used in a much broader range of occupations than the factory robots of the 1990s. Today robots perform their work silently in every corner of the urban landscape:
as automated teller machines (ATMs) at the bank, self-service check-in counters at the airport, self- service checkout stations at the grocery store, pay-at-the-pump machines at the gas station, and automated payment machines at parking lots and tollbooths.
As they gain dexterity, limbs, and more intelligence and situational awareness, robots look less like stationary kiosks and more like the humanoids we recall from cheesy 1950s science fiction movies. Lowe’s hardware stores are experimenting with robot greeters who guide customers to the correct aisle in the store. And Rethink Robotics is selling Baxter, a $22,000 robot with a friendly face, suitable for small businesses and light manufacturing. Add a dash of personality and these anthropomorphic machines are capable of dealing directly with people.
Humanoid automatons are showing up in customer-service roles. In Berlin, there’s a robot bartender; in Tokyo’s Shinjuku district, a robot cabaret show; and at Germany’s giant CeBIT trade show in 2012 there were robot booth babes, gyrating and pole dancing for bemused attendees. Oh, the horror. More and more, robots are taking on roles in some unusual places:
> Robot hotel staff: The Henn-na Hotel in Nagasaki, Japan, is staffed by ten humanoid robots. They speak fluent Japanese, Chinese, Korean, and English, and can check visitors into the hotel, carry bags to the rooms, change the sheets, clean the rooms, and deliver laundry. The president of the company expects robot minions to perform 90 percent of hotel services.
> Robot warehouse staff: Kiva robots, part of an automated order-fulfillment system, dance an
intricate ballet inside an entirely automated Amazon warehouse. They move entire shelving units gracefully to workers who select, sort, and package individual orders. Think of the smart
warehouse as a giant computer that routes packages just like packets on a digital network, moving, sorting, stacking, and tracking items.
> Robot animals: Boston Dynamics creates impressive robots inspired by biological designs:
BigDog, an autonomous packhorse, can carry 340 pounds over hills and through rough terrain for 20 miles; Cheetah, the fastest legged robot on Earth, runs faster than Usain Bolt, the world’s fastest human; Atlas, a humanoid robot, walks upright and uses its hands like a human; Sand Flea, a rolling robot, can leap over buildings; and RiSE, a six-legged robot, can climb vertical walls and trees using micro-claws.
> Robot freighters: Eight European companies have joined forces with the Fraunhofer Institute to launch the Maritime Unmanned Navigation through Intelligence in Networks (MUNIN) project to design an automated freight ship. Robot ships will operate slowly, at speeds human crews would not tolerate, but they will also save 50 percent on fuel costs, lower greenhouse gas emissions, and, of course, save the money that would have been paid in the form of crew salaries. Eventually, the 100,000 merchant ships on the seas today could be operated by robot crews.
So, should human workers be concerned? Yes and no. Robots are going places that humans can’t or
don’t want to go, including hot war zones, toxic waste dumps, and the stars above us. The US Navy has developed a robot drone that swims like a shark for underwater reconnaissance and may be phasing out manned aircraft in favor of drones. Speaking at the Sea-Air-Space 2015 conference, US
Secretary of the Navy Ray Mabus said that the new F-35 Lightning II “should be, and almost certainly will be, the last manned strike fighter aircraft the Department of the Navy will ever buy or fly.”
At the same time, artificial intelligence (AI), a type of robot that doesn’t need a body, is allowing computers to do more tasks that were previously the domain of human intelligence. Current-
generation AIs are like “cloud robots,” consisting of software running across multiple servers. They can recognize speech, translate between languages, and make decisions. So it’s not a stretch to
contend that the global economy recently entered a transitional phase in which far more occupational tasks can be handled by a machine. Factory labor is just the beginning. Next in the queue: the
knowledge-processing and analytic skills of white-collar workers. Some jobs that require mastery of a defined body of knowledge, such as law, accounting, journalism, and medicine, can now be
partially handled by software robots.
THE BOOMING SECOND ECONOMY
During the past fifteen years, the automated software-driven control systems that operate constantly in the background of every business, every transaction, and every communication have begun to migrate to the cloud along with everything else in the software-defined economy. Now computer systems, payment systems, communication networks, and business process management tools rely on cloud- based automation to function. Soon our autonomous vehicles and smart cities will communicate with the cloud to do their jobs too. Automated systems have merged with the Internet. This, too, is no coincidence. The Internet consists of a vast network of automated systems exchanging billions of messages every day. From inception, it was designed to operate without human intervention, relying on countless software-defined robots that serve, observe, and anticipate what human users will do.
For example:
> Servers and routers direct every packet, web page, email message, music stream, and frame of video automatically to the correct destination.
> Automated filtering algorithms and recommendation engines determine which stories appear on individual newsfeeds, which friends are recommended on social networks, which offers show up on e-commerce sites, which fulfillment center will handle a particular order.
> Automated systems handle the real-time bidding, buying, and placement of advertising targeted to individual preferences and behavior.
> Spambots fill online discussion boards with junk messages, create fake accounts to send email, and crack passwords.
> Robotic journalist tools like Narrative Science’s natural-language generation platform, Quill, crank out news stories.
More and more of what we experience and consume consists of digital products and digital
services. Most of the work to generate and deliver these experiences is handled by automation. As the vaporized digital domain spills out of the computer screen and blends and merges with the real world via the Internet of Things (IoT), more and more automated software systems will permeate our daily
lives. According to economist W. Brian Arthur, an external professor at the Santa Fe Institute, this automation comprises a vast, unseen economy. As he explains in a 2011 article in the McKinsey Quarterly, “Another economy—a second economy—of all of these digitized business processes conversing, executing, and triggering further actions is silently forming alongside the physical
economy.” In fact, these everyday processes in the physical economy are now tightly interwoven with the digital world too: as soon as contact is initiated between them, elaborate communication and data processing occurs in the digital domain, unnoticed by humans.
Routine events, such as checking in for a flight at the airport or shipping freight through Rotterdam, involve conversations between servers, satellites, and distant computers. In the past, Arthur points out, human clerks were obliged to manage a sheaf of paperwork, supervise logistics procedures, and physically handle packages. Today the entire process is digitally scanned and automatically
dispatched, to be tracked with barcodes, radio-frequency identification (RFID) chips, and other machine-readable interfaces. Arthur says, “What used to be done by humans is now executed as a series of conversations among remotely located servers.” Welcome to vaporized labor, expressed purely as digital information.
Arthur calculates that over the long term the second economy will be responsible for a 2.4 percent annual increase in productivity in the overall economy. By 2025 he estimates that the second economy will be as large as the entire physical economy was in 1995. Even if these back-of-the-envelope
figures are subject to quibbling and fine-tuning, “What’s important is that the second economy is not a small add-on to the physical economy. In two or three decades, it will surpass the physical economy in size.”
What’s different about this new, networked kind of automation-based economy is awareness.
Unlike the robotic brutes on auto factory floors that can neither see nor hear, the invisible automated world of software has eyes and ears. Just like the smart devices of IoT, the automated systems track and respond to events that transpire in the physical world. And detecting changes in the outside world and reacting to them is a form of primitive intelligence. Arthur argues that the emergent intelligence of automated systems is the biggest change since the Industrial Revolution. “With the coming of the
Industrial Revolution,” Arthur writes, “the economy developed a muscular system in the form of machine power. Now it is developing a neural system.”
Thanks to Moore’s law, Metcalfe’s law, and the growth of really big data sets, we are about to experience a step function increase in the power of automation and robotics. This combination of factors means that the age of artificial intelligence may finally be at hand.