Back off, robot: Why artificial intelligence may not lead to mass unemployment

This is the age of Watson: computers winning quiz shows, cars driving themselves and digital assistants knowing your whims and fancies better than you do.

Will your job be next?

Some of the world’s learning thinkers on machine learning and artificial intelligence believe otherwise, at least for now.

“Robots won’t take our jobs,” cautions Jason Furman, an economist who chairs President Obama’s council of economic advisors.

Furman spoke last week at a machine learning conference at the University of Toronto, hosted by the Rotman School of Business’s Creative Destruction Lab. The conference comes as a host of technologies?—?from Amazon Echo to Google’s new Pixel phone to robotic John Deere planters?—?upend business models left and right.

Furman argued the so-called fourth industrial revolution, rooted in computers teaching themselves routine human functions, will eliminate some jobs?—?and create others.

The earlier industrial revolutions, rooted in steam, electricity and electrons, did not lead to permanent mass unemployment. The U.S. eliminated 2 million jobs in September, Furman noted, but created 2.2 million more.

Perhaps human paranoia is again getting the better of cold logic.

Ajay Agrawal, a Rotman professor who organized the conference, argued that many jobs based on forecasts and predictions will be given over to machines. But humans will be needed to make decisions based on computer forecasts.

“Machine intelligence is a substitute for human prediction, and it’s a complement to human judgement,” Agrawal said.

AI is experiencing explosive growth. According to CB Insights, AI start-ups raised US$681 million in 2015, up from $145 million in 2011, and will hit $1.2 billion this year.

Furman argued that for every job displaced by AI, there will be news ones for programmers and maintenance workers. Moreover, he said, artificial intelligence is prohibitively expensive for most organizations to develop and maintain. Some of the most routine jobs?—?hotel housekeeping, for instance?—?would cost too much to automate.

“We can compete with them (robots) on wages,” he said.

The Rotman conference included some of the world’s top researchers, including U of T’s Geoff Hinton, who runs Google’s machine learning work; Universite de Montreal’s Yoshua Bengio; Richard Sutton of University of Alberta, whose work led to Google’s AlphaGo program; and Ruslan Salakhutdinov of Carnegie Mellon University, who was recently named head of AI research at Apple.

Hinton, a pioneer in machine learning, said AI will disrupt thousands of jobs, particularly higher-end functions that are, at their core, data analysis.

“They should stop training radiologists now,” he said.

Radiologists make their money by analyzing x-rays and medical images, looking for subtle shading and irregularities that can indicate tumors, injured and serious disease. But artificial intelligence machines, programmed with medical knowledge and fed millions of x-rays, will be able to use algorithms and patient histories to make far more accurate diagnoses?—?for much less time and money.

“If you work as a radiologist, you’re like the coyote that’s already over the edge of the cliff but hasn’t yet looked down,” Hinton said.

Until just a few years ago, artificial intelligence was a backwater in computer science. The early promise in the 1960s and a resurgence in the 1980s had run up against two big problems: a lack of data and lack of computing power. Cloud computing and storage, and massive increases in computing power, have helped change that.

Another challenge: machines have yet to learn the way children do. Unsupervised learning may be the next leap, Sutton said.

Michal Chui, a partner at the McKinsey Global Institute, said AI may also help increase labour productivity.

“Instead of being worried about AI as a driver of mass unemployment, it might be the thing that saves us in terms of accelerating productivity so we actually continue to grow the economy,” he said.

Jerry Kaplan, a computer scientist and author who has written extensively about artificial intelligence, contrasted the killer robots of Terminator 2 with the ungainly automatons whose failures are the stuff of viral videos.

Some of the world’s best brains, and billions of dollars in government, business and military research programs, still haven’t given us robots that have the coordination to climb stairs, open doors or exit vehicles.

Kaplan said the AI future is further away and more benign than some suggest, citing a line from a fellow researcher about the most common fears: “If you’re worried about the Terminator coming to get you, just keep your door closed.”