MIT economics professor David Autor wants everyone to calm down about the robot jobs takeover: It’s definitely not happening now, and not likely to happen anytime soon.
In a paper released for the Kansas City Fed’s Jackson Hole economics conference, Autor says that while some occupations are capable of being automated — and have already been — the vast majority of jobs will remain free from machine’s claws.
One telling anecdote, he says, can be found in a recent New York Times article about Google trying to train computers to pick out accurately pick out an image of a cat from thousands of YouTube videos. It eventually was successful, but it wasn’t easy.
“The article’s headline ruefully poses the question, “How Many Computers to Identify a Cat? 16,000.”
Times reporter John Markoff wrote that the Google researchers admitted the program they’d used to complete the task remained wholly rudimentary compared to the brain.
“It’d be fantastic if it turns out that all we need to do is take current algorithms and run them bigger, but my gut feeling is that we still don’t quite have the right algorithm yet,” Stanford computer scientist and Google researcher Andrew Ng told him.
The main jumping off point in Autor’s paper is a maxim attributed to Michael Polanyi, a Hungarian philosopher: “We can know more than we can tell.” For Polanyi this meant, for instance, you replace a driver’s skill by teaching him about how cars work. Autor says we can apply it to mean that a machine’s proficiency will only ever be as strong as our ability to accurately program it.
“At a practical level, Polanyi’s paradox means that many familiar tasks, ranging from the quotidian to the sublime, cannot currently be computerized because we don’t know ‘the rules,'” Autor says. “At an economic level, Polanyi’s paradox means something more. The fact that a task cannot be computerized does not imply that computerization has no effect on that task. On the contrary: tasks that cannot be substituted by computerization are generally complemented by it. This point is as fundamental as it is overlooked.”
The first piece of evidence Autor deploys is a straightforward chart showing investment in machines has returned to pre-DotCom bubble levels of growth:
“One would expect that a surge of new automation opportunities in highly paid work would catalyze a surge of corporate investment in computer hardware and software. Instead, the opposite occurred,” he says. As of the first quarter of 2014, information processing equipment and software investment as a share of GDP was at 3.5%, a level last seen in 1995 at the outset of the “dot-com” era.
What happened, he argues, is a temporary “dislocation” of investment during the DotCom era that has since corrected to its historical rate.
“The end of the ‘tech bubble’ in the year 2000 is of course widely recognised, as the NASDAQ stock index erased three-quarters of its value between 2000 and 2003,” he writes. “Less appreciated, I believe, are the economic consequences beyond the technology sector: a huge falloff in IT investment, which may plausibly have dampened innovative activity and demand for high skilled workers more broadly.”
Besides Google’s cats, Autor cites other shortcomings of widely cited examples that allegedly prove the machine takeover. While IBM’s Watson won jeopardy, he blew a soft-toss question about Chicago airports. Google Translate and Netflix are still nowhere near the level of accuracy one could achieve from asking a friend to interpret or recommend a movie.
Meanwhile, automation that has come online has often proven as useful and value-added as it has destructive.
“…Tasks that cannot be substituted by computerization are generally complemented by it,” he writes. “This
point is as fundamental as it is overlooked. Most work processes draw upon a multifaceted set of inputs: labour and capital; brains and brawn; creativity and rote repetition; technical mastery and intuitive judgment; perspiration and inspiration; adherence to rules and judicious application of discretion. Typically, these inputs each play essential roles; that is, improvements in one do not obviate the need for the other. If so, productivity improvements in one set of tasks almost necessarily increase the economic value of the remaining tasks.”
Autor even believes that the hollowing out of middle-skilled jobs will soon come to an end.
“While many middle skill tasks are susceptible to automation, many middle skill jobs demand a mixture of tasks from across the skill spectrum,” he says.
Medical technicians are one of the fastest growing occupations in the country. They are relatively well-remunerated, and often don’t require more than demand two years of post-secondary vocational training.
“Significantly, mastery of ‘middle skill’ mathematics, life sciences, and analytical reasoning is indispensable for success in this training.”
Robots can grasp a lot, but not that.
Business Insider Emails & Alerts
Site highlights each day to your inbox.