The impact of AI on the labor market is not so clear-cut, say the authors of a new scientific paper titled “Weak Package, Strong Package: How AI is Rethinking the Boundaries of Occupations.” They challenge the idea that the active implementation of AI almost automatically leads to job losses.
The impact of AI on the labor market is not so clear-cut, say the authors of a new scientific paper titled “Weak Package, Strong Package: How AI is Rethinking the Boundaries of Occupations.” They challenge the idea that the active implementation of AI almost automatically leads to job losses.
According to Luis Garicano, a professor at the London School of Economics, and Jin Li and Yanhui Wu of the University of Hong Kong, it would be inaccurate to simply call work a “task” or a “to-do list.”
Robots are packages, they say, and it is illogical to simply calculate how many tasks AI can perform in a particular profession and then conclude that greater automation will inevitably lead to worker displacement.
“Most discussions about AI and labor markets start with ‘job vulnerability’: if AI can perform more functions within a particular occupation, then employment levels or earnings in that field should fall,” the researchers note.
“This approach is incomplete because labor markets value jobs, not individual tasks. A job is a bundle of interconnected tasks, and the impact of AI depends on how costly it is to break that bundle.”
A good example is radiology. A radiologist doesn’t just “sell” medical image classification—he performs a multitude of other functions: triages cases, consults with other doctors, trains interns, makes complex decisions, and officially confirms a diagnosis,” Garicano explains.
At one time, Jeffrey Hinton, one of the so-called "godfathers" of AI, made a loud statement in 2016: medical schools should stop training radiologists because AI would soon surpass them in reading images.
Of course, that didn't happen: According to Garicano, there are more radiologists now than ever before, and they earn more than they did a decade ago.
This is because the job of a radiologist is a “strong package,” and it would be too expensive to separate all of these tasks from the position. In contrast, the “weak package” seems to be easily separated, leaving the person with a much narrower range of responsibilities—and a much more modest compensation.
“In professions with a weak task package, AI automates individual functions and narrows the boundaries of the position, leading to a standard job replacement scenario,” the study says.
“In occupations with a strong package, where tasks cannot be redistributed independently of each other, AI improves productivity within the job, but does not displace humans from that package. Thus, task integrity becomes a force protecting jobs and labor’s share of the economy.”
