You"ve probably heard this one before: "AI is going to take your job." But how true is it? Researchers from Stanford University claim they have found what they call "early, large-scale evidence consistent with the hypothesis" that the AI revolution is hitting the American labor market, especially entry-level workers.
Researchers Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen arrived at their conclusion after analyzing administrative data from ADP, the largest payroll provider in the US. They looked at employment records from late 2022 to July this year. The team"s method involved tracking millions of workers and ranking their jobs by how likely they were to be replaced by AI (high AI-exposed) and how unlikely they were (low AI-exposed).
The researchers provided a metric based on earlier studies and found that those in what one might call the "email job" caste, such as software developers and accountants, were most at risk. Maintenance and repair workers and people in nursing, however, were fine, possibly because their work mostly involves interacting with the physical world, not just a computer.
The study found that workers aged 22-25 in those highly exposed jobs saw a significant drop in employment. The paper presented six facts that paint a detailed picture of this shift in the job market.
The first fact centers on specific roles like software developers and customer service reps, where young workers experienced a pronounced employment decline while older colleagues did not. For software developers between 22 and 25, employment plummeted by nearly 20% from its peak in late 2022.
In the second fact presented, the researchers argued that the job losses for young people in AI-exposed fields are so significant that they are dragging down the overall employment growth for the entire 22-25 age group, making it look stagnant.
If you look at young workers in low-exposure fields, like healthcare aides, their employment actually grew between 6 and 13%. The problem is that the bloodbath in entry-level white-collar work is canceling out those gains.
The third finding distinguishes between AI that automates (replaces) tasks and AI that augments (helps) them; the job losses are concentrated in the automation category.
The fourth and fifth facts address the argument that this is just a tech recession in disguise, as the researchers controlled for firm-specific shocks, like a company having a bad quarter, and found that it"s the job role that"s disappearing, not that the company is struggling.
The final fact claims these patterns hold up even when the data is sliced in different ways. For example, the researchers removed the entire tech sector from their data, checked for effects from remote work and the pandemic, and the conclusion was the same.