Car buffs looking for information on the new BMW i3 electric car are being told to text the company’s U.K. sales department. What they may not realize is that they are kicking off a conversation with a computer. The artificial intelligence system, created by London Brand Management, is capable of making sense of natural language cues and learning as it goes along, allowing it to carry on a realistic back-and-forth conversation—at least to the extent that texting can be considered conversation.
It’s PR without people, and it’s the latest example of how machines are continuing to infiltrate the workplace. Three decades ago, robots took over the factory floor, displacing millions of blue-collar workers in the process. Now, they’re being tapped to do office work, sell insurance and feed hungry fast-food patrons. San Francisco’s Momentum Machines, for example, has built a burger-flipping robot that can dispense 360 sizzling burgers per hour, potentially saving restaurants up to $135,000 in annual labour costs. Other companies are turning to so-called Big Data to direct their marketing and advertising, relying on powerful computers to comb through huge databases of information about customers’ shopping and web-surfing habits.
A recent study by researchers at the University of Oxford suggested that automation—the use of machines to do tasks formerly done by people—could threaten as many as half of all occupations within the next two decades, including several occupations that were once considered “safe” because they involved making judgment calls or interacting with others. “Computerization will no longer be confined to these rule-based queries, where a certain task is repeated over and over again,” says Carl Frey, one of the Oxford study’s co-authors. “Computers can now make all sorts of subtle decisions.” One recent well-known example is IBM’s computer, Watson, which competed on Jeopardy two years ago and beat two previous champions.
What this all means for human workers is not yet clear. Some economists say the displaced will simply find new occupations, as has historically been the case when a new technology comes along. Plus, there are all those engineering and programming jobs needed to design and build automated systems. But a growing number are convinced that technological change now moves with such speed that jobs are being automated away in greater numbers than new ones can be realized. The phenomenon is even being used to explain the unique characteristics of today’s economy—everything from stubborn unemployment to the ballooning class of low-paid, low-skilled workers. “Technology will make us wealthier, on average,” Frey says. “But many will also be left behind.”
Out of 702 job categories examined in the Oxford study, 47 per cent were deemed possible candidates for some degree of future automation. The shift is being driven by technological advancements in key areas where computers have typically lagged humans. One is perception. Thanks to improved sensor technology, engineers are now able to build robots that are capable of functioning outside of strictly controlled environments, such as factories. Google’s driverless car, for example, relies on a combination of GPS, lasers, radar and sophisticated software to help it plot its course and avoid obstacles along the way. In fact, Frey and co-author Michael Osborne predict that transportation and logistics will be one of the first sectors to be hit by the next wave of automation, affecting everyone from taxi drivers to long-haul truckers.
Meanwhile, advancements in machine learning mean computers can think more like humans, allowing them to analyze unstructured data and learn from experience. Improvements in natural-language recognition have also made it possible for millions of iPhone users to turn to Apple’s digital assistant, Siri, when in need of weather updates or restaurant recommendations. But the very same technology also promises to automate a host of other occupations, ranging from office administrators to sales clerks and telemarketers. While such jobs generally require interpersonal skills, Frey and Osborne argue, somewhat insensitively, that “although these occupations involve interactive tasks, they do not necessarily require a high degree of social intelligence.” Further down the road, computers may even be sophisticated enough to substitute for some scientific research and engineering work, they say.
The trend promises to exacerbate the existing polarization of the job market. Over the years, a loss of middle-class jobs—many of them in manufacturing—has resulted in a concentration of well-paid, highly skilled positions at one end of the spectrum, and a ballooning number of low-paid service-industry jobs on the other. While this “hollowing out” has sometimes been blamed on companies shipping jobs overseas to China and India, Henry Siu, an associate economics professor at the University of British Columbia, says the research shows that offshoring only accounts for about 10 per cent of job losses in recent decades. The rest is due to technology. “It’s much easier to point a finger at someone than to blame the computer sitting on your desk,” he says.
The extent to which automation will affect overall employment remains controversial. Robert Atkinson, the president of the Information Technology & Innovation Foundation, recently called predictions of an increasingly jobless future the ramblings of neo-Luddites, a reference to British artisans who famously protested the mechanization of the textile industry in the early 1800s. “These ideas are essentially misguided speculation,” Atkinson wrote in InformationWeek. “They fly in the face of years of economic data, as well as current trends.” He argued that productivity has historically led to economic growth, creating more jobs and more wealth. “We did not see massive unemployment as agriculture mechanized in the early 20th century—the workforce shifted to other professions.”
But a turning point may have been reached about 15 years ago, according to Erik Brynjolfsson and Andrew McAfee, two normally pro-technology researchers at MIT’s Sloan School of Management. In their book Race Against the Machine, they point out that productivity and corporate profits have largely continued to improve since the year 2000, but there has been a notable lack of corresponding job growth. The conclusion? As machines become more like humans, our competitive advantage in the job market is falling by the wayside.
Frey falls somewhere in the middle of the debate. “It’s not that there is a fixed amount of work that needs to be divided up,” he says. “The issue is more about whether workers can keep up with the fast pace of technological change.” A key problem is that newly displaced workers generally aren’t qualified to fill the new jobs being created, whether in robotics or artificial intelligence fields. Most are instead forced into the low-paying service industry, which, if Frey and Osborne are correct, could soon be at risk, too. That could explain why Canadian companies are always complaining about a skills shortage in Canada, while overall unemployment is stuck at 7.1 per cent. It may also explain increasing levels of wealth disparity, as those with a high level of education and training find it easier to adapt to new opportunities.
There are always winners and losers during periods of rapid technological change, Siu says. “The difference this time around is the ones getting hurt are in the middle, and that’s the bulk of us.”