Michael Chui of McKinsey & Co.’s global research arm happened to visit the Twin Cities on the last (hopefully) snowy day of the season, and he pointed out a self-driving car would have had a rough day here.
Chui, a leader of the firm’s research in “disruptive technologies” like artificial intelligence, sees autonomous vehicles on the streets of the San Francisco Bay Area, where he lives. But they have a ways still to go.
Recognizing a stop sign is easy for a machine, unless some person messes with the sign by putting stickers on it. Then it might be read not as an instruction to stop, according to a study a couple of years ago, but as a speed limit sign. That’s obviously dangerous.
Chui described self-driving car technology as “amazing” as it is right now, but he knows mass adoption of this kind of technology isn’t happening anytime soon.
But here’s the equally important message he had last week: We need that time, before technologies like autonomous vehicles blossom, to figure out how to get millions of people into different kinds of paid work or how to pay them in new ways for what they do.
Chui is a partner with the McKinsey Global Institute, created by the consulting company McKinsey. He has a Ph.D. and described himself as a private-sector professor, but he also has experience as a problem-solver in other jobs.
He has lately been working mostly on the expected effect on work and business of artificial intelligence, a term that to him means “technologies that allow machines to accomplish cognitive functions, i.e., those that people associate with human minds.”
It was research by Chui and his group a couple of years ago that made news by concluding that maybe half of the work people do could be taken over by machines with what they called “currently demonstrated technologies,” which meant the invention phase was already over. And six out of 10 occupations could have up to 30% of their work done by machines.
One of the things that made this report stand out is how they talked about work, not jobs. A worker could still be on the job but with a very different kind of workday, getting a lot of tasks done by machines. These tasks might be the ones the workers had not previously enjoyed doing anyway.
The research also underscored the point that just because some new technology might work doesn’t mean it’s a practical thing anybody would spend money to implement.
Chui knows of a robotic hamburger joint in San Francisco called Creator, with burgers he described as pretty tasty. But he has no idea how it can be cheaper for the restaurant owner to make precision-engineered hamburgers with a robot than it is to pay skilled workers to cook them.
There are people who buy impractical things just because they are cool, but businesses generally don’t. Off the top of his head Chui ran through some simple math on what it would cost to replace the North American trucking fleet with self-driving vehicles, and got to a few hundred billion dollars even if the additional sensors and computing power on the tractor added no additional cost.
In business it’s commonly assumed that the rate of technology adoption is accelerating, but Chui can’t see much evidence for that. McKinsey has analyzed how long it takes from commercial viability, where it makes financial sense to first use a new machine, through peak adoption. He said it’s still anywhere from eight to 28 years.
For technologies like a fully autonomous truck, of course, the clock has not started yet.
“It’s easy to say, ‘This is possible, someone has developed this technology,’ ” he said. “And then not really realizing all the other steps before it actually affects millions of people.”
That’s time we could use, among other things, to figure out how to have good jobs for people now making a living doing jobs like driving. As Chui put it, “The grand challenge for the next several decades is not mass unemployment, but mass redeployment.”
Chui stressed that he’s not in the prediction business, but the firm’s optimism about technology and its impact on work is informed by American history, as workers left segments of the job market that had gone into decline and found work in jobs that maybe didn’t even exist when they were in high school.
In the United States the share of all workers employed in agriculture declined from six out of every 10 people in 1850 to less than 5% by 1970. A lot of people moved from farm country to take jobs in factories, but then manufacturing jobs declined from more than a quarter of the U.S. workforce in 1960 to less than 10% in recent years, according to McKinsey’s research.
In McKinsey’s view, fields in the United States that should see a lot of growth include caring for an aging population and working with technology. Occupations in what they called unpredictable physical work, like specialized mechanics and emergency responders, could also grow.
On the other hand, the outlook is grim for office support workers and people doing what McKinsey called predictable physical work.
Governments and societies have choices to make on how to respond, Chui said. That includes deciding what to spend on lifelong education and how much value to put on certain kinds of work. Right now there are relatively highly skilled occupations that are not paid well, but Chui said that reflects a choice.
“We know that’s a choice because as we look from state to state or country to country, the same occupation is paid in some cases very different amounts,” he said.
That’s one lesson from the story of the Luddites in England in the 1800s, the name for a group of skilled workers in textiles who saw the threat from new manufacturing machines in Britain’s rapidly industrializing economy. One thing they tried was breaking into the plants and wrecking the machines.
The Luddites turned out to be right, Chui said. Adoption of new technology led to productivity gains, while they saw no real gains in worker pay for decades.
“I’m not arguing that it had to be bad for them,” he added. “Could different choices have been made?”