In future, there may be people who – despite being fit to work – have no economic value
On August 29 1997, Skynet – a computer system controlling the US nuclear arsenal – became self-aware. Panicking operators tried to deactivate it. Skynet, perceiving the threat, launched its arsenal, killed most of humanity, and ushered in a world in which the robots ruled. So went the backstory of the 1984 movie The Terminator . But computers did not become self-aware in 1997 – the closest they managed was when Deep Blue, a B-list supercomputer, beat Garry Kasparov, the world chess champion. Despite decades of hand-wringing about robots taking over, the robots never quite seem to rise.
But perhaps 2014 will be different. Google certainly seems to think so: early in December it purchased Boston Dynamics, a producer of military prototype robots – with names such as “BigDog”, “WildCat” or “Petman” – that wouldn’t look out of place in the Terminator films. These nightmarish machines will now be brought to you by the folks who host all your email, know what your internet searches are and are tracking your phone’s location.
But while the Skynet-esque combination of Google and Boston Dynamics is unsettling, it is not in itself a reason to expect that robot technologies really will change the world. Yet the talk in the economics profession is increasingly taking that possibility seriously.
The primary cause has been with us a long time: the familiar Moore’s law, which in various guises describes growth in computing power as swift and exponential. We have got used to swift growth, but we can never quite get used to the implications of exponential growth – meaning that whatever has just happened will be eclipsed by whatever is just about to happen.
Moore’s law, loosely applied, is that computers today are twice as powerful as the computers of two years ago, perhaps just 18 months ago. Today’s mobile phone is a match for what was once a cutting-edge gaming console; that gaming console, in turn, outperforms the kind of old-timey supercomputer that the Terminator franchise once imagined taking over the world.
Software is also becoming more efficient. We tend to miss this because the bloated copies of Microsoft Word we use do not seem faster than 20 years ago. But a mobile phone running Pocket Fritz 4, a chess program, can now beat grandmasters, despite the phone running far more slowly than Deep Blue did. A chess-playing phone is not about to lead a robot uprising, so why should we care? A growing number of economists – including Massachusetts Institute of Technology’s Erik Brynjolfsson and Andrew McAfee in a new book The Second Machine Age – argue that robots and algorithms are poised to make inroads into labour markets.
Computing power is starting to solve everyday problems – which turn out to be the hardest ones. Computers were laughable drivers in 2004, when a computer-driving competition was “won” by a car that crashed after completing seven miles of a 150-mile course. Now computers drive cars safely.
In 2008, robots still struggled with a problem known as “Slam” – simultaneous localisation and mapping, the process of mentally building up a map of a new location, including hazards, as you move through it. In 2011, Slam was convincingly addressed by computer scientists using Microsoft’s “Kinect” gaming hub, an array of sensors and processors that until recently would have been impossibly costly but is suddenly compact and cheap.
Problems such as language recognition and Slam have so far prevented robots working alongside humans; or on tasks that are not precisely defined, such as taping up parcels of different sizes or cleaning a kitchen. Perhaps the robots really are now on the rise.
Consider “Baxter”. Traditional industrial robots are major capital investments: vast machines kept apart from human workers for safety reasons. Baxter, by contrast, claims to be able to do much of the same work, is cheaper, safely works with humans, and is – its manufacturers claim – intuitive to reprogramme. And if Baxter fails to live up to the hype, Moore’s law means that the robot’s successors – with a computer eight to 10 times more powerful for the price in five years’ time – will not.
What is sobering is that we have already seen convincing evidence of the impact of technology on the job market. Alan Manning of the London School of Economics coined the term “job polarisation” a decade ago, when he discovered that employment in the UK had been rising for people at the top and the bottom of the income scale. There was more demand for lawyers and burger flippers. It was middle-skill jobs that were disappearing. The same trend is true in the US, and is having the predictable effect on wages: strong gains at the top, some gains at the bottom, stagnation in the middle.
The leading explanation is that technological change has favoured certain skills and displaced others. Typists, clerks, travel agents and bank tellers find their skills less valued. Mechanisation now dominates agriculture, large-scale construction and manufacturing. We tend to imagine that manufacturing jobs have disappeared to China; in fact, manufacturing employment in China has been falling. Even the Chinese must fear the robots.
Of course cheap, ubiquitous computing power has brought many good things – and will bring more. The question is whether we are equipped to deal with the possibility that in future, there will be people who – despite being willing and fit to work – have no economic value as employees. By the time today’s 10- year-olds have their degrees, computers could be a hundred times cheaper and smarter than they are today. A future full of robot servants could be a bright future indeed, but only if we can adapt our institutions quickly enough.
Also published at ft.com.