Studies disagree on whether AI is causing job losses

In short: The best evidence so far shows AI is not causing mass job losses across the whole economy, but it is linked to smaller job hits in certain entry-level roles.

What's going on

Researchers are trying to answer a simple question, is AI taking jobs. The problem is that the data are still early and incomplete, and different studies measure different things. Some track “exposure” to AI, meaning a job has tasks that a tool could help with, while others track actual changes in hiring, pay, and layoffs.

Several broad studies find little change in overall employment so far. A Yale Budget Lab look at US labor data about 33 months after ChatGPT’s release reported no clear economy-wide disruption. Economists at Goldman Sachs also reported no strong link between AI exposure and things like job growth, unemployment, hours, or wage growth, which they partly attribute to slow and uneven adoption.

When researchers zoom in, the picture changes. A Stanford analysis cited by the Carnegie Endowment found a relative drop in employment for early-career workers, ages 22 to 25, in jobs most exposed to AI since ChatGPT arrived. Other summaries say entry-level hiring in highly exposed office roles may be slowing, even if it does not show up yet as a big rise in unemployment.

Part of the confusion is that job change can look like reshuffling, not a clean loss. AI can remove some tasks while adding others, like a new tool in a kitchen that speeds up prep but still needs a cook to plan the meal.

What to watch

The next few years should bring clearer signals as more companies move from testing AI to using it daily. Watch entry-level hiring, especially in routine office work, and watch for better “net” numbers that count both jobs lost and jobs created.

Source: NYTimes