WELCOME TO THE MACHINE
Why white-collar jobs are safe from the artificial intelligence surge — for now
Somen Mondal had two problems while running the company he launched 10 years ago, only one of which was a good one for an entrepreneur to have.
The good problem was that his safety compliance management software business was growing and he needed to hire a couple of new salespeople a week, which attracted hundreds of applications. That need also created a related and less exciting problem: Someone had to actually read those applications and decide who to call back.
“When we were making decisions about who to bring in for an interview, it was really based on gut feel,” Mondal said. “We’re very data-driven as business founders. We thought, ‘There’s got to be a better way.’ ”
Today, there is a better way, one that is powered by artificial intelligence. After selling that business, Mondal founded Ideal, a Toronto-based software company that uses AI to identify promising candidates from stacks of résumés, thereby leaving hiring managers with more time to meet prospective employees in person.
AI systems are now much better than humans at identifying patterns in large amounts of data and that has white-collar workers worrying about their jobs just like assembly line workers have since the 1970s. They’re likely worrying for naught — at least, for now.
The machines’ inability to use common sense or generalize means they can’t do much beyond the scope of one narrow task. As a result, AI is more likely to be a helpful assistant that takes care of boring, repetitive tasks than a job-taking threat.
“If you ask any hiring manager, they hate going through résumés. That’s the one thing they hate to do,” Mondal said. “You know what they love doing? They love talking to candidates. They love talking to people, meeting people. That’s something a computer can’t replace.”
The artificial intelligence field has exploded during the past five years due to the development of a technique called deep learning that uses code to mimic the human brain’s neural networks.
Deep learning is already powering everything from Uber’s suggestions for pickup and drop-off points to Netflix’s recommendations for what to watch next, with computers getting better at recognizing patterns as they process more and more data.
The introduction of machines capable of learning was bound to make some people queasy, given our long cultural history of stories about a future human-robot war. Add anxiety about the global economy and the decline of the North American manufacturing sector and it’s easy to understand how some people are in a full-blown panic about the future of work.
Prominent voices in the tech sector have also added to the impression that a wide range of jobs is in imminent danger of being automated.
Prestigious Silicon Valley accelerator Y Combinator is studying whether a universal basic income could alleviate some of the social impacts of mass unemployment caused by technology. Microsoft Corp. founder Bill Gates has suggested taxing robots to recover some of the lost revenue from displaced workers.
But artificial intelligence has a long way to go before it can replace most human employees, according to a November report released by a team of prominent researchers in the field.
The Artificial Intelligence Index report found that AI systems can now match or outperform humans in doing specific tasks such as document translation and speech recognition, but little progress has been made in replicating general intelligence.
The report gave the example of translating a document from Chinese to English. An AI system might be just as good at translating the document, but a human can probably also perform related tasks such as understanding Chinese speech, answering questions about Chinese culture and recommending a good Chinese restaurant.
“Very different AI systems would be needed for each of these tasks,” the report said. “Machine performance may degrade dramatically if the original task is modified even slightly.”
Alan Mackworth, a professor at the University of British Columbia who holds the Canada research chair in artificial intelligence, said he thinks it will be at least a decade or two before researchers make any real progress in artificial general intelligence.
Some experts don’t think general AI is possible at all, but Mackworth cautioned against dismissing anything as impossible with a long enough time horizon.
“It’s very risky to say AI won’t be able to do this, that and the other thing,” he said. “Almost every time that prediction has been made, it’s eventually been proven false.”
It’s fun to speculate about what general AI might mean for humanity in the future, but the task-specific AI already present today has the potential to do a lot of good.
For example, Toronto-based Blue J Legal sells AI-powered software that analyzes case law to help lawyers predict the likely outcome of lawsuits, which founder Benjamin Alarie hopes will make the court system more efficient and improve access to justice.
“It gives both sides more confidence in the potential range of settlements,” he said. “And that’s great for clients. It’s also good for lawyers. Those cases can be resolved more quickly and more confidently by counsel.”
Analyzing case law is a job usually reserved for junior lawyers, but Alarie said clients that use his software don’t find they need fewer attorneys.
“It’s not a case of this sort of technology taking work away. It actually just channels the effort in the right direction,” he said.
In addition to making work easier and more productive, AI has the potential to lower the barriers to entry for creative pursuits.
Say you always wanted to be a DJ, but never had time to master the required technical skills. Waterloo, Ont.-based Rave Inc. is working on software that can take a playlist of songs, analyze places where the tracks sound good mixed together and turn them into a professional-sounding set.
Rave chief executive Mike Pazaratz said the software won’t replace big-name professional DJs, but it will make it easier for amateurs to go from idea to execution without having to buy expensive equipment and programs.
“We don’t want to take a human being out of the loop. That’s not at all what we want to do,” he said. “AI is your helpful friend. You say, ‘I want to do this thing,’ and the AI goes, ‘We can totally do this thing.’ ”
Artificial intelligence certainly won’t be a helpful force for everyone. Most jobs involve multiple tasks and human interactions, but there are some positions that could be completely replaced by the type of AI systems that already exist.
Erik Brynjolfsson, director of MIT’s Initiative on the Digital Economy and author of The Second Machine Age: Work, Progress and Prosperity in a Time of Brilliant Technologies, said he is confident artificial intelligence is good for the economy overall, but some workers will suffer, particularly those who spend most of their time doing repetitive information processing, such as bookkeepers, middle managers and travel agents.
For people who are worried their jobs are at risk, Brynjolfsson recommends focusing on the type of tasks AI won’t be able to perform for a long time, such as interacting with people. If there are parts of your job an AI system might be able to perform better, people should embrace the machines and use the extra time to add value in other areas.
“The more we think about ways of structuring work so we’re racing with the machines, using the machines, the better it’s going to be for human labour and for total value creation,” Brynjolfsson said. “A combination of humans and machines is often going to be better than just humans or just machines.”
It’s very risky to say AI won’t be able to do this, that and the other thing. Almost every time that prediction has been made, it’s eventually been proven false.