When experts talk about the disruptive effects of artificial intelligence, they tend to focus on low-paid laborers — but a newly published study suggests higher-paid, more highly educated workers will be increasingly exposed to job challenges.
The study puts Seattle toward the top of the list for AI-related job disruption.
The analysis, which draws on work by researchers at Stanford University and the Brookings Institution, makes use of a novel technique that connects AI-related patents with the job descriptions for different professions.
Stanford researcher Michael Webb extracted entries from the tens of millions of patents in a Google database, as well as the 964 job descriptions indexed by the U.S. Department of Labor.
The goal was to match up noun-verb descriptions from the patents related to automation and AI with the job descriptions, on the theory that a particular job might face more exposure to future disruption if there are more patents associated with the description of that job.
For example, a patent for a device that monitors the operating conditions of power equipment could affect someone whose job description mentions monitoring operating conditions of equipment at a wastewater treatment plant. If there are a lot of patents for such devices, that would increase the job exposure index in Webb’s analytical model.
Webb looked at job exposure assessments for robotics, software applications and AI, and ranked the highest and lowest job exposure for each category:
-
Robotics: High-exposure occupations included forklift drivers, crane operators and janitors. Low-exposure occupations included payroll clerks, art/entertainment performers and clergy.
-
Software: High-exposure occupations: broadcast equipment operators, water and sewage treatment plant operators, parking lot attendants, and packers and packagers by hand. Low-exposure: barbers, podiatrists, college instructors and postal carriers.
-
AI: High-exposure occupations: clinical lab technicians, chemical engineers, optometrists, power plant operators. Low-exposure: non-farm animal caretakers, food preparation workers, college instructors, art/entertainment performers.
Why did highly skilled workers such as lab technicians end up on top of the AI list? Webb noted that AI is getting better and better at matching human performance in some of these jobs. In contrast, some job categories that may seem closely related — for example, lab researchers — “involve reasoning about situations that have never been seen before.” Others, such as food preparation or massage therapy, require the sorts of interpersonal skills that aren’t as suited for AI.
Lawyers who interact with the public have less to worry about than paralegals who merely review documents. Podiatrists who make judgment calls about patients’ feet have less to worry about optometrists whose diagnostic processes are easier to computerize. “Optometry is the area of medicine that has seen perhaps the most success of AI algorithms to date,” Webb wrote.
The Brookings Institution built upon Webb’s study by looking at how job exposure was distributed geographically, demographically and across occupational groups. For AI, the exposure scores are highest for agriculture, engineering and science. They’re lowest for education, food service and personal care.
Job exposure is relatively higher for men as opposed to women, for white and Asian-American workers as opposed to black and Hispanic/Latino workers, and for workers in the 25-64 age bracket as opposed to workers who are younger or older.
Bigger, higher-tech metro areas and manufacturing centers are more prone to AI job disruption. As a result, the Seattle area and California’s Silicon Valley have high scores on the disruption scale — but not as high as, say, Elkhart, an Indiana city that’s considered the nation’s capital of RV manufacturing.
Webb’s calculations suggest that AI’s effect on high-wage occupations will have an impact on the nation’s income inequality — up to a point. “Under the assumption that the historical pattern of long-run substitution will continue, I estimate that AI will reduce 90-10 wage inequality, but will not affect the top 1%,” he said.
He and his colleagues at Brookings acknowledged that there’s still a lot of uncertainty about how AI will play out. Changes in population trends, investment levels or education programs could reduce job exposure, or heighten it. And the rise of AI could well lead to new products, services and occupations.
“While the present assessment predicts areas of work in which some kind of impact is expected, it doesn’t specifically predict whether AI will substitute for existing work, complement it, or create entirely new work for humans,” the Brookings Institution’s research team said. “That means much more inquiry — qualitative and empirical — is needed to tease out AI’s special genius and coming impacts.”
For the full details, check out Webb’s study, “The Impact of Artificial Intelligence on the Labor Market,” and the Brookings report, “What Jobs Are Affected by AI?”