This article was first published in the November 2015 issue of WIRED magazine. Be the first to read WIRED's articles in print before they're posted online, and get your hands on loads of additional content by subscribing online.
In May 2015, a team of researchers from the University of Science and Technology of China and Microsoft's Beijing office announced a new system, based on rapidly advancing "deep learning" technology, that's capable of outperforming the average person on the kind of verbal-reasoning puzzles found in IQ tests.
Deep learning uses artificial neural networks, which operate according to the same essential principles as the biological neurons in the brain. Although neural networks have been used to perform basic pattern-recognition tasks for decades, the past few years have heralded dramatic breakthroughs that now allow deep-learning neural networks of unprecedented power and complexity to be employed in areas such as image recognition, language translation and problem solving.
A team at the University of California, Berkeley used deep learning techniques to build a robot that figures out how to perform tasks such as screwing the top on a bottle, with the same speed and precision as a person.
Deep learning offers just one of the more vivid illustrations of how smart software algorithms are becoming ever more capable at solving problems, making decisions, and -- most importantly -- learning to do new things. This advancing ability to learn and adapt has dramatic implications for the job market because, for the first time, machines are encroaching on the intellectual capability and adaptability that has, so far, ensured that most workers are able to remain relevant as technology progresses.
The impact on more routine, repetitive jobs is already evident. A recent analysis by an American consulting firm found that the number of people employed in the finance departments of large corporations, for a given level of revenue, has collapsed by about 40 percent since 2004. Routine work areas, such as bookkeeping and accounts payable and receivable, are increasingly being taken on by smart software, and the people who used to have secure white-collar jobs performing these tasks are finding themselves in the unemployment queue.
Workers who have higher education levels may be tempted to take comfort in the belief that their jobs are non-routine and, therefore, beyond the reach of the machines. However, the frontier is advancing rapidly, and software has already progressed far beyond simply automating rote repetitive jobs. Smart systems are writing news articles, performing sophisticated legal-document reviews and evaluating medical images. In fact, "routine" may not be the best word to describe the jobs most likely to be automated.
A better term might be "predictable". Could another person learn to do your job by studying a detailed record of what you've done in the past? Or could someone become proficient by repeating the tasks you've already completed, in the way a student might take practice tests to prepare for an exam? If so, then there's a good chance that a machine-learning algorithm may eventually be able to figure out how to do much, or all, of your job.
The conventional solution to technology-driven job losses has always been ever more education and vocational training. As machines and smart software eat away at low-skill jobs, workers are urged to retool themselves and continually seek higher-skill positions. It turns out, however, that workers are not the only ones who can climb the skills ladder: computer technology is proving remarkably adept at the same feat. Evidence for this is in the job market.
In the UK, average starting salaries for graduates with degrees fell by about 11 percent over the five years from 2007 to 2012, from £24,293 to £21,701. And in the US, any recent graduate can tell you that we have entered the age of the degree-bearing barista: half of new graduates end up taking jobs that don't use their education.
The upshot is that smart machines may eventually create an economy where a large number of workers cannot remain relevant. That will force us to confront the fact that we have entered a new age -- in which the relationship between workers and machines has changed. Adapting to that reality could well be one of the seminal challenges facing us in the coming decades.
This article was originally published by WIRED UK
