In the early stages, artificial intelligence AI can increase worker productivity—helping them do more and faster, which often leads to higher wages before automation begins to replace some of these jobs.
But Ioana Marinescu, an associate professor at the University of Pennsylvania’s School of Social Policy and Practice, says that artificial intelligence may have already pushed wages to their peak.
She co-authored a new paper with the Brookings Institution on “smart saturation.” Citing her research, she stated, “ Our automation level is over 14%, which is closer to a potential wage decline than when there was no automation.”
Brookings' model suggests that wage growth could begin to reverse once approximately 37% of cognitive or "intellectual" tasks are automated. She notes, "But when the decline will occur depends on several parameters."
warning sign
Marinescu stated that the first signs we are approaching our peak will be a slowdown in wage growth and a shift in employment from knowledge-intensive to physical-intensive jobs.
She said that more than 14% of “intellectual” tasks in the economy have been automated—a proportion that has been steadily rising since the 1980s. Her estimate is based on a study that showed the proportion of routine cognitive work fell from 49% in the late 1970s and 1980s to 35% in 2018.
Marinescu interprets this 14 percentage point drop as an indicator of increased automation in smart work. She also derives the so-called "AI pay curve": wages rise as workers use smarter tools, then plateau, and finally decline as machines take over more work.
"For automation to lead to lower wages, we must see a decline in the proportion of workers performing intelligent jobs. This would happen, for example, if job losses in the tech sector are not compensated for by increases in other intelligent jobs, and the number of manual labor jobs remains stable," she wrote.
For example, if software development If human resources, translation, or marketing personnel are replaced by artificial intelligence without the emergence of new high-value jobs, the overall share of employment in these industries may decline, thus laying the foundation for wage declines.
Which jobs will disappear first?
Marinescu says the model assumes automation will begin with the most easily replaceable “intellectual” tasks, where AI can quickly surpass human abilities in areas like pattern recognition, translation, or basic writing.
“We expect these jobs to decrease first, such as translation or basic marketing copywriting. If these positions are not replaced by other related jobs, the overall proportion of these positions will decrease, which may lead to a decrease in wages.”
The speed at which this transformation occurs depends on how tightly the material and intellectual aspects of the economy are linked. Wage losses accelerate when they become more easily interchangeable.
Marinescu stated that wages tend to decline faster when artificial intelligence can easily replace more types of human jobs: "As intellectual and physical labor become more substitutable, workers concentrated in manual labor sectors benefit less from further automation."
What should we focus on next?
Marinescu says the most important metric to track is the proportion of workers engaged in manual and intellectual labor—in other words, whether humans are moving toward jobs that are easily replaced by artificial intelligence or moving away from those jobs.
If this balance begins to shift, with more people leaving desk-based cognitive jobs for manual, slower-growing work, the pay curve for artificial intelligence may have already peaked.
(Article source: CLS)
