Economists use the term hysteresis to describe the phenomenon that, when conditions in an economy change, the effects of that change often remain even after the conditions return to normal.
COVID and its impact on the workforce may provide a good example of hysteresis, said HAI Distinguished Fellow and MIT professor Erik Brynjolfsson, who will join Stanford faculty in July 2020 as the director of the new Digital Economy Lab.
To keep workers safe and continue functioning, companies have ramped up remote work and are aggressively automating some operations and exploring machine learning.
“Some of these changes are going to be permanent,” he said during Stanford HAI’s recent online conference COVID+AI: The Road Ahead. “The question is, what parts of the economy are going to be most affected by the adoption of these technologies, and which parts will be less affected?”
Machine Learning’s Impact
Brynjolfsson worked with Carnegie Mellon professor Tom Mitchell, MIT postdoc Daniel Rock, and others on a series of papers identifying the tasks most suitable for machine learning (ML). They applied this rubric to 950 occupations and 18,000 specific occupation tasks.
More tasks in lower-wage jobs could be replaced by machine learning applications, they found. For example, ML today can better recognize a cucumber or a banana and handle some cashier skills. But some high-paid jobs can also be impacted, such as airline pilots. “No occupation is completely immune,” Brynjolfsson said.
Certain industries are also more impacted than others, he noted. Manufacturing, retailing, transportation, and accommodation and food services have many tasks suitable for machine learning.
Additionally, different areas of the country will be affected unevenly. “The kinds of work that people do in Wyoming are very different from what they do in Manhattan or Miami,” he said.
The researchers’ data also allowed them to examine ML impact on individual occupations. Roles like tellers, executive assistants, and personal bankers have a large percentage of tasks that are suitable for ML.
“Our tool gives them a way to have a path for what to do next,” Brynjolfsson said. Personal bankers could develop more skills not subject to machine learning, like leadership, product development, or customer relations, and move away from the skills more suitable to ML like credit authorization. Another option: Find new roles with similar skill sets. A personal banker might transition to business analyst or mortgage loan officer, roles ML is less likely to disrupt.
“With a little bit of training, they’re in a position to be much less vulnerable to the machine learning revolution,” he noted.
Credit: Google News
