A DANCE WITH TECHNOLOGY: AUTOMATION AND TOMORROW'S JOBS

Lots of jobs require teamwork. But these days, the team may include robots that work alongside humans. “You’re a team member with the robot, you’re a team member with your colleagues. It allows you to do more things,” is how one Amazon warehouse worker put it in a New York Times video published last fall.

"It’s a team effort between the humans and the robot,” the Amazon employee added. “We all work as one.”

Amazon’s warehouse robots contribute to the efficient order and delivery process that helps fuel that company’s continued growth — and thus the creation of warehouse jobs for humans, too.

But will job creation continue to take place as the technology advances, or is this just an interim step toward a future when jobs are scarce?

In the past few years, rapid advances in artificial intelligence and robotics have fostered lots of breathless conjecture about these technologies’ effects on human work in the future. Some pundits raise concerns about a “jobless future,” while others emphasize that artificial intelligence will create new jobs.

This speculation, which is characterized by wildly varying predictions, does little to clarify the challenges we face. New technologies both create and destroy jobs, sometimes in ways that are not entirely predictable at the outset. But what’s even more important is that many jobs — like those in Amazon warehouses — are already changing due to technology advances and will continue to change, often rapidly, in the years ahead.

"Another fundamental question we should be asking is: How can advancing technologies shape the future of work in productive, equitable, and inclusive ways?"

While some jobs will disappear and new ones will be created, recent research from MIT and elsewhere suggests that the most widespread impact of artificial intelligence and automation in coming years may not be on entire jobs but on tasks within jobs. That raises questions such as: How will we adapt to the new technologies as they change our jobs? What new skills will workers need? What organizational and societal structures will be needed to help individuals adapt and thrive as many jobs evolve and require new skills — and as some jobs disappear, while others emerge?

Another fundamental question we should be asking is: How can advancing technologies shape the future of work in productive, equitable, and inclusive ways? Fortunately, we can influence the major technological transitions that lie ahead. Within our organizations and in society at large, we need to emphasize the potential of technological advancements to enhance human efforts. Instead of thinking in terms of a competition between humans and machines, let’s instead choreograph a dance — a pas de deux — that involves people and machines and combines the wisdom of humans with the capabilities of technology for the greater good.

MIT Sloan Professor Thomas W. Malone advocates for this kind of thinking in his recent book Superminds: The Surprising Power of People and Computers Thinking Together.

Replacing Tasks More Than Entire Jobs

How will jobs change? Two recent studies offer interesting perspectives on that question. In a paper published in the 2018 American Economic Association Papers and Proceedings, MIT Sloan Professor Erik Brynjolfsson, Carnegie Mellon Professor Tom Mitchell, and MIT Sloan doctoral student Daniel Rock used O*Net, a U.S. Department of Labor database that describes hundreds of occupations and the tasks involved in them. Brynjolfsson, Mitchell, and Rock analyzed 18,156 tasks involved in 964 occupations for what they called “suitability for machine learning.” (Machine learning is a rapidly advancing domain within artificial intelligence. As MIT Professor Daniela Rus, director of the Computer Science and Artificial Intelligence Laboratory at MIT, explained in a recent presentation, machine learning “enables machines to improve, make predictions, and create adaptation routines.…Machine learning…starts with a body of data and then aims to learn a rule that explains the data and make predictions.”)

x x

 

MIT Sloan Professor Erik Brynjolfsson (shown here) worked with Carnegie Mellon Professor Tom Mitchell and MIT Sloan doctoral student Daniel Rock to analyze thousands of occupational tasks for their “suitability for machine learning.

From their analysis, Brynjolfsson, Mitchell, and Rock concluded that most occupations have some tasks that are suitable for machine learning — but few if any jobs consist entirely of tasks suitable for it. The authors explain that this suggests that many jobs will need to be redesigned significantly — as machine learning takes over some but not all tasks within the job. They write:

“Our findings suggest that a shift is needed in the debate about the effects of AI on work: away from the common focus on full automation of many jobs and pervasive occupational replacement toward the redesign of jobs and reengineering of business processes.”

Reports published by McKinsey Global Institute in 2017 similarly looked at tasks within jobs and the technical feasibility of automating them, whether through artificial intelligence or robotics. The McKinsey researchers also found that relatively few jobs are fully automatable, but many, many tasks are. They write:

“Very few occupations — less than 5 percent — consist entirely of activities that can be fully automated. However, in about 60 percent of occupations, at least one-third of the constituent activities could be automated, implying substantial workplace transformations and changes for all workers.”

Tasks with a particularly high potential for automation, the McKinsey authors conclude, include “physical ones in predictable environments, such as operating machinery and preparing fast food. Collecting and processing data are two other categories of activity that can increasingly be done better and faster with machines.”

With their focus on changing tasks, these studies are consistent with earlier research conducted by David H. Autor and Frank Levy of MIT with Richard J. Murnane of Harvard. In an important study published in 2003, Autor, Levy, and Murnane found that a prior wave of technology — computerization — changed the task content of jobs. In particular, after computerization, workers in the U.S. spent less time on routine manual and cognitive tasks (such as calculations) and instead did more work that involved “nonroutine cognitive tasks” such as managing people, persuading people, or forming and testing hypotheses.

In other words, that earlier wave of automation resulted in reduced demand for humans to do some tasks but increased demand for them to do others. That’s not an anomaly. The creation of new tasks elsewhere in the economy is one of the forces that counteracts the labor displacement that automation initially creates, according to a recent working paper by economists Daron Acemoglu of MIT and Pascual Restrepo of Boston University. Acemoglu and Restrepo note that while automation displaces some workers, increased productivity through automation has beneficial productivity effects on the economy as a whole, and economic growth increases demand for workers. However, that doesn’t mean that the economic transitions created by a wave of automation are smooth for either individuals or society.

The Work Ahead for All of Us

These studies about tasks and automation have important ramifications for both business executives and policy makers. For businesses, a key implication of the findings is that companies will need to think less about plugging technology in to completely eliminate specific jobs and more in terms of widespread job and process redesign. Executives also need to think about how to prepare their employees to work with new technologies and acquire needed new skills.

In a 2018 report, Antonis Christidis of Mercer, Axel Miller of Oliver Wyman, and Professor Thomas Kochan of the MIT Sloan School offer some important insights about how managers should think about collaborating with their workforces in the ongoing adoption of digital technologies. Instead of senior executives making decisions with vendors and then essentially imposing new technology solutions in a top-down manner on the workforce — an all-too-common approach — Christidis, Miller, and Kochan argue that companies are better served with a model that brings employees into the process from the outset. In this inclusive approach, there is a continuous cycle of not only training employees in new technologies but also seeking their input in applying technology to business problems.

Such an approach is particularly well-suited to the type of job and process redesign that is likely to characterize the coming wave of automation. Kochan advises managers to take the following steps:

1. Be proactive and determine how you want to deploy technology to achieve your business goals rather than waiting for vendors to present you with options.

2. Integrate the design of technology with the design of work processes.

3. Train and educate your workforce about new technologies on a continuous basis — and help your workers develop what are called “hybrid skills”: a combination of knowledge of the new technologies and skills in communication and problem-solving.

4. Engage and involve the workforce early in the process of choosing and integrating technologies, so that they can, to use a Japanese phrase, “give wisdom to the machines.”

5. Engage with policy makers and institutional leaders to ensure that you will continue to be able to draw on a skilled workforce in a society where workers feel secure enough that they don’t have to fear change.

Policy makers at all levels of government will in turn need to display leadership and develop innovative solutions. We are in the midst of major change — and major change can be destabilizing and detrimental to individuals, families, and even society as a whole. During such times, the policy makers’ role is to create the conditions that maximize the benefits of change and minimize the harm. They have a responsibility to enable all citizens to thrive in an era of dramatic technological transition. That means building the stabilizing structures that will make it possible for people to embrace this new era with energy and creativity. For example, policy makers need to consider how our crazy quilt of education and training programs can be made relevant and accessible to far more people at different points throughout their working lives.

What are the policies and programs that will support our society during a period of major technological advancement? Simply tweaking existing government programs that date back to the 20^th century won’t suffice. Workers will need new skills to adapt to a 21^st-century world of work, and policies that support people in a lifelong journey of skill development are imperative. Training must be high-quality, relevant, accessible in real time and available over a lifetime. A strong and flexible social safety net must also be in place so that workers can successfully navigate through career transitions and economic uncertainty.  Such programs enable people to feel more comfortable about new technologies. In Sweden, for example, which has far more generous training and unemployment programs than the U.S., people have a much more positive view of automation than in the U.S. — presumably because they have less to fear from technological change.

The level of policy change required may sound daunting. But as MIT President L. Rafael Reif wrote in a 2017 op-ed column published in The Boston Globe,  “reinventing the future of work needs to be a whole-society effort — and finding long-term solutions will require ideas and initiative from every quarter.”

Reif further reminded us that we have, as a society, tackled big challenges before. He wrote:

“Let’s remember that ideas like universal public education, the GI Bill, and the post-Sputnik focus on science education met…resistance. However, it was such broad, far-sighted investments in human development — by the nation, for the nation — that allowed the country to mitigate the immense pain caused by previous technological and societal earthquakes.”

Jobs throughout the economy are already changing due to technology advances. For example, at Stitch Fix, a San Francisco-based online personal styling service, human stylists work in tandem with machine learning algorithms. At 99Degrees Custom, an entrepreneurial company based in Lawrence, Massachusetts, incorporating advanced technologies into apparel manufacturing — and developing a specialty in clothing embedded with wearable technologies — enables the creation of new apparel manufacturing jobs in the U. S. Some hospital nurses today interact with robots (rather than people) that deliver medicines from the hospital pharmacy. And businesspeople involved in international trade take advantage of the artificial intelligence that now powers online translation services like Google Translate to communicate more easily with customers and suppliers all over the globe. In fact, one intriguing new study suggests that advancements in AI could increase international trade. In a National Bureau of Economic Research working paper published in August 2018, Erik Brynjolfsson, Xiang Hui, and Meng Lui analyzed what happened when eBay rolled out an improved machine-learning powered translation service for Spanish-speaking Latin American users of its platform. The improved translation service, the researchers found, resulted in export sales via eBay from the U.S. to Latin America increasing 17.5% in volume and 13.1% in value.

As these examples suggest, we can’t stop the dance with technology, but we can choreograph it in productive ways. In fact, many leaders are determined to use these new technologies in ways that benefit society. For example, the MIT Initiative on the Digital Economy’s Inclusive Innovation Challenge is recognizing entrepreneurs across the globe who are using technology to shape a future of work that brings shared prosperity.

Our collective task is to create the structures and context that will enable society, companies, and individual workers to flexibly adapt to rapid technological change. MIT is committed to helping shape this future and has established an Institute-wide MIT Task Force on the Work of the Future to explore such questions.

As MIT President Reif put it:

“Automation will transform our work, our lives, our society. Whether the outcome is inclusive or exclusive, fair or laissez-faire, is up to us. Getting this right is among the most important and inspiring challenges of our time — and it should be a priority for everyone who hopes to enjoy the benefits of a society that’s healthy and stable, because it offers opportunity for all.”