Then They Came for the Lawyers
Technology has already driven blue-collar workers into the underclass. Professionals may be next.
Strange though it may sound, there was a time when manufacturing work resembled professional work today. In the 18th century, on the cusp of the Industrial Revolution, life wasn’t bad for skilled tradespeople, who enjoyed a remarkable level of freedom and flexibility in performing their work. They represented a relatively well-off, aspirational class.
Then came the machines. The mechanization of manufacturing transformed industrial work. Because the new machines cost lots of money, bosses kept a close eye on workers to make sure they were doing their jobs and taking good care of the equipment. Over time, this monitoring allowed industrialists to rejigger production in efficiency-enhancing ways. Where manufacturing work was once something of an art that relied on knowledge built up through apprenticeship and experience, it became a highly scripted slog, broken down into repetitive tasks. Anyone could do it, and so the special status and relatively high wages once enjoyed by manufacturing workers disappeared. Laborers became mere extensions of the machines, handling tasks the machines could not — until, eventually, they could.
Across advanced economies, the professional class — white-collar workers in business management, technology, law, finance, and medicine — has largely escaped the ill effects of the recent changes, including rapid globalization and automation. Those changes have disproportionately hurt workers engaged in routine sorts of tasks — running machines on factory floors or carrying out back-office jobs — and those without a college (or especially a graduate) degree. Conventional wisdom long held that this immunity was likely to continue. A paper published by Carl Benedikt Frey and Michael A. Osborne in 2013, which famously estimated that 47 percent of job categories would be vulnerable to automation in coming decades, ranked positions such as manager, engineer, and lawyer among those at lowest risk of displacement.
But the forecast for highly skilled workers is starting to look less sunny. The professional world is about to be transformed by artificial intelligence. As that process unfolds, it could reshape white-collar work much as industrialization transformed manufacturing.
Anyone who has ever interacted with companies such as Amazon, Google, and Facebook is already familiar with AI, whether they know it or not. Every time we like a friend’s photograph, send an email, or search for a good nearby restaurant, we provide massive amounts of data to those firms. The tech giants use that data to train machine-learning programs to provide us a more customized experience. That process, in turn, allows the firms to sell us more stuff or to sell more advertisements to others who want to sell us more stuff.
But the same techniques that generate the ads that follow us around the web are increasingly finding their way into the workplace, as Ajay Agrawal, Joshua Gans, and Avi Goldfarb write in their new book, Prediction Machines. Many firms already rely on AI to help them assess business risks (to tell banks which borrowers are most likely to default, for instance) or anticipate consumer demand. And new applications are appearing all the time.
Employers already amass piles of human resources data on their workers: the roles they have held in the company, for instance, and the way both they and colleagues around them performed in each. Machines that are fed that information on a large scale can divine what sorts of worker characteristics are most associated with high team performance — and, therefore, which job applicants are most likely to thrive within the company. By chewing on reams of detailed career histories, machine-learning programs can predict whether a worker on a particular career arc should be promoted or directed toward the exit — or whether a valued employee is about to quit. They can be given sales data and asked to predict which accounts can be made to generate more business and which are dead ends.
Yet this is only a start. According to Agrawal, Gans, and Goldfarb, machine prediction will get steadily better and vastly cheaper. Those advances will give companies incentives to more aggressively find ways to deploy AI — and more concertedly collect information on their employees to feed into those systems. The result will likely be much closer monitoring of workers of all sorts.
From Foreign Policy, here.