Massage therapists, breathe easy. Robots won't steal your job. They don't have the empathy, manual dexterity or...
ability to ask the right questions. Radiologists, take notice. A tool wired with machine learning algorithms can easily interpret medical images using computer-aided diagnosis systems.
But there are 26 other tasks radiologists do that machines can't do well, such as administering conscious sedation during procedures or developing treatment plans, said Erik Brynjolfsson, director of the MIT Initiative on the Digital Economy. He was sharing results of a study on tasks in various jobs that can be done well by machines, at the MIT Sloan CIO Symposium, in Cambridge, Mass.
"Machine learning was able to do some tasks but not others within a given occupation," Brynjolfsson said to the audience of senior IT executives in May; he moderated a panel discussion on future work in a world where AI and machine learning are the norm. "And that means that most of the jobs in your organizations will be partly affected by machine learning, but there will also be things that the humans need to continue to do. We'll have to have partnerships of humans and machines."
As AI and machine learning advance and begin to match and even outdo humans in fundamental skills such as recognizing images, Brynjolfsson and other academics on the panel agreed, business processes need to be re-engineered and tasks reallocated among people and machines so that technology amplifies, not substitutes, human potential.
I, robot; you, human: We're in this together
Panelist Elisabeth Reynolds, executive director of the MIT Industrial Performance Center, called the idea complementarity. A member of the MIT Task Force on the Work of the Future, convened to study the impact of digital innovations on the nature of work, Reynolds spoke about a narrative that has been shaping in the public sphere of robots becoming so advanced they steal people's jobs and spur widespread hardship and despair.
Erik Brynjolfssondirector, MIT Initiative on the Digital Economy
In the media, "technology is this thing that is happening to us, and we're all trying to figure out how we're going to survive," she said. "We're really interested in trying to change that narrative to one in which we understand technology as a tool."
Reynolds cited a recent McKinsey & Co. report that estimated that robotics and AI would fully replace people in less than 5% of occupations by 2030. The technology would instead enhance or change 60% of jobs.
One example is from the manufacturing industry. Cobots, or collaborative robots, are being put to use alongside humans at companies such as FedEx, as reported in March by The New York Times, replacing routine work like lugging heavy items around a factory floor, and allowing workers to do other things, she said.
While automation will put some people out of work -- "and we do have to deal with the displacement," Reynolds said -- the need for human workers will not go away.
Future work on the shop floor
Amazon's huge distribution centers are another testing ground for human-machine collaboration. In the past, when someone bought an item from Amazon's online marketplace, a worker called a picker would have to walk to its whereabouts on a shelf somewhere and back again to fill the order. Today, the facilities are buzzing with breadbox-shaped robots on wheels that identify the right shelves stocked with goods, lift them up and cart them to workers.
"They didn't just do a one-for-one substitution of 'Here's a human; we're going to replace them with a machine,'" Brynjolfsson said. "Instead they reinvented the whole process."
That innovation has resulted in good and bad, Reynolds said. "People are not moving from A to B, long distances, etc.; they can stay where they are, and they can sort of pick and pack."
The downside is the reformulated job of the picker isn't ergonomic, she said -- workers are standing in the same spot for eight-hour shifts, doing the same thing over and over. That's difficult for people to do and potentially harmful. The challenge now is to design technology with humans in mind.
"We've solved one problem, but maybe we've created a few others," Reynolds said. "We need to think about designing in the technology a way in which humans are actually advantaged and using all of the skills that they can bring to a job."
A precarious future work?
Of course, Amazon's robots are not the only AI-machine learning innovation with negative consequences for humans.
Take Uber. The ridesharing app's platform uses machine learning algorithms to match riders and drivers. But Uber's drivers, like all workers in the so-called gig economy, are contract workers, not full-time employees, and that means no fixed hours, no fixed pay, no employer-provided benefits. It's a notion MIT's Jason Jackson, assistant professor of political economy and urban planning, called precarity. It's a form of the word precarious, and it refers to having insecure income and employment.
"Their worker status is now much more precarious both in terms of being able to be moved completely out of the firm, or the number of hours that they work becomes much more flexible," Jackson said. From a managerial point of view, that flexibility can be very efficient, but for workers it means their cash flow becomes unstable.
"Most of us have fixed expenses -- rent, mortgage -- every month. So if your income becomes variable, then that creates a huge problem," Jackson said.
Perpetual learning for survival
Conference attendee Theo Kornyoh is optimistic about AI and machine learning and future work. Kornyoh is CTO at Kaleida Health, a hospital network in western New York State. The technologies are transformative, making work and life much easier, he said. And they have different effects in different industries. In finance or retail, companies can look to AI tools to determine their customers' needs, quickly address business concerns and expand.
In Kornyoh's industry, healthcare, "it gives us a chance to be able to provide better services to our patients, also provide better services to our providers."
As for whether technology will decimate entire job categories, like the panelists, Kornyoh doesn't buy it.
"I think people just have to be able to be ready to learn and expand their knowledge base," he said. "Because when you keep learning, I see automation and technology and ML and AI complement what you do rather than take your job away."