The Data Are Clear: Robots Do Not Create Unemployment
This year marks the 70th anniversary of the first electronic thinking machine that displaced knowledge workers. In February 1944, Britain's Colossus, the world's first computer, started to ‘think' and successfully automate human mental labor. By the summer of 1944 there were three such machines hard at work. The rest, as they say, is history.
But it's relevant history considering today's debate about computer-automation creating, ostensibly for the first time, the wholesale displacement of jobs requiring brains instead of brawn. That was precisely the point of the world's first -- and all subsequent - computers: to facilitate, accelerate, and replace knowledge tasks.
The task in 1944 was code cracking. According to British mathematician Alan Turing (arguably the father of computers), it would have taken "100 Britons working eight hours a day on desk calculators 100 years to discover the secret factor" needed to unlock the German encryption. Britain would quickly build ten such colossal number-crunching thinking machines, followed a year later by America's ENIAC. Turing wasn't worried about full employment for mathematicians, but amplifying their skills. In fact, Leibniz (the father of calculus), invented a calculating machine in 1673 for similar reasons: the "process of calculating is nothing but drudgery, and I would like to save people from that tedious chore." One man's drudgery is another man's livelihood.
In today's debate such automation of whole new sectors of cognitive "chores" that derive from tedious calculating excites almost as much agitation in the white collar class as did the automated machines of 1814 that gave birth to the hand weavers' Luddite movement. Start with the lawyers. The legal profession is facing stagnant income and job growth, and applications to law schools are down 40 percent compared to a decade ago. Attorney John McGinnis eloquently summed it up when he said that the Great Recession couldn't explain this state of affairs and that "the legal profession faces a great disruption" from automation.
Rote legal work - whether reading, scanning, fact-finding, or even creating documents whole cloth -- is giving way to the crushing advantage the algorithms. All this used to be done by attorneys and paralegals working like automatons in white-collar sweat-shops. Such tasks, which traditionally required a university degree, are succumbing to automation. While shocking to many, we forget that in days of yore becoming a guild tradesman required years of schooling through apprenticeship. But that was then and this is now.
We've been here before. Few lament the advent of computerized accounting or spread sheets, word processors, or graphics programs, etc., which nevertheless destroyed employment for armies of professional accountants, typists, draftsmen... the list goes on. One who did was President John F. Kennedy who, in 1961, created an Office of Automation and Manpower in the Department of Labor because of the need to address "the major domestic challenge of the Sixties: to maintain full employment at a time when automation, of course, is replacing men."
Were it the case that permanent escalation in unemployment is an inevitable result, we would see it in the unemployment statistics. But history is unequivocal. The data show no serial escalation in the overall unemployment since 1961 with the ascent of knowledge automation, nor in the 120 years of ascending machinery automation.
U.S. Unemployment Rate 1890 to 2014
Source: 1890-1940 data from Romer (1986), 1940-2014 data from Bureau of Labor Statistics
What about today? A July 2014 Harris Poll of professional hiring managers found 20 percent of all and 30 percent of large companies report replacing employees with automation, while 35 percent say more jobs were available after automation.
But neither recent evidence nor history silences the belief that "this time is different." Even the sagacious Bill Gates joined the dystopian chorus early this year, saying "technology over time will reduce demand for jobs, particularly those at the lower end of skill set." A computer professor writing in The Atlantic recently summarized the trope: "In the 19th century machines competed with human brawn. Now machines are competing with human brain."
There is some truth in this. With enough computing horsepower and the right algorithms, computers can do amazing things. Turing realized this; becoming the first to use a computer to play music, write letters, and to create a chess-playing program. (For those interested in more about Turing and the origins of computing, I recommend John Gribbin's new book, Computing With Quantum Cats: From Colossus to Qubits.) From inception, computers aimed to automate knowledge work.
But still it took 45 years for Turing's computer algorithm to trump a grandmaster in chess -- May 11, 1997, Gary Kasparov lost to IBM's Big Blue. Today, iPad algorithms regularly beat the masters. But Kasparov deserves credit for a critical insight.
He wrote in the New York Review of Books several years ago: "What if instead of human versus machine we played as partners?" Kasparov explored this idea, fostering a unique kind of intellectual chess cage fight: a 2005 free-for-all tournament. People teamed up with computers to compete against other teams, or against supercomputers. The winner? A team of two amateurs using three average computers.
Kasparov postulated -- and proved -- the power of the complementarity of algorithms plus humans. Thus what is different about this virtual robot revolution is not so much that robots will replace humans in certain tasks (they will) but that virtual robots - algorithms - are getting good enough at cognitive tasks to team up with humans, amplifying what people do.
There are no longer rooms full of draftsmen, accountants or typists - all these tasks are still accomplished and more effectively. But there are more jobs overall. Complementarity at work. We can expect much more of this now.
Algorithms also revolutionized the oil & gas industry. Shale resources are not new; nor are horizontal drilling and hydraulic fracturing. But it is information technology that has been transformative. Petroleum engineers now spend more time on computers than drilling platforms. Algorithms have lead to a huge boom in overall employment not just in oil or tech fields. In fact Houston's overall employment has grown nearly three times as much as all employment around Silicon Valley since 2006.
Algorithms, teaming up with doctors and nurses, will transform health care too. For example, rolling robot dispensaries can ensure patients get the right medications at the right times, delivered autonomously, using information from patient-embedded organic sensors, and smart pills that know when they're needed and when they've been swallowed. All parties -- nurses, doctors, patients and relatives -- are seamlessly kept in the loop. The drudgery is reduced; the place for human talent increased and amplified. Complimentarity.
Security provides another example. Here, algorithms team up with sensor-laden rolling bots that help ‘dispense' safety and security on city streets and business. Check out the recently unveiled K5 from Knightscope that automates, synthesizes, and shares with cops and citizens both, information associated with the inevitable drudgery of daily security. K5 amplifies the efficacy of police and security personnel. Complimentarity
MIT's David Autor encapsulates all of this in a new seminal paper: "Journalists and expert commentators overstate the extent of machine substitution for human labor and ignore the strong complementarities. The challenges to substituting machines for workers in tasks requiring adaptability, common sense, and creativity remain immense." Autor joins with Kasparov.
To be sure, as Autor emphasizes, "the adjustment process, as with technological adaptation, is frequently slow, costly, and disruptive." But maybe this time history won't repeat. What is new today is the accelerated pace and scale of the emergent technological transformation, which will accelerate adaption and job creation. But government, with its capacity to stifle innovation, is the wild card. Not that there isn't a role for government: it should focus on helping the minority whose jobs are disrupted in the "adjustment process," but not at the expense of overall job creation. It's a tough needle to thread. Maybe an algorithm can help.
Alan Turing died tragically 60 years ago this summer, on June 7, 1954. While Turing looked to amplify the work of mathematicians, the thinking machines descendant from Colossus will amplify the work for everyone else.