Given the heightened anxiety over AI’s impact on jobs—viral essays warning about impending “massive disruption,” software coders fearing they’ll become a “permanent underclass,” questions about the market for entry-level workers—we turned to Martha Gimbel, executive director and co-founder of the Budget Lab at Yale University, to ask what the data are really saying.
A recent analysis from her group finds that AI’s impact on the labor market has been minimal so far. “While anxiety over the effects of AI on today’s labor market is widespread, our data suggests it remains largely speculative,” Gimbel and her co-authors write. Still, they caution that they cannot “explicitly determine what the future holds.”
Here are highlights from our conversation, edited for length and clarity:
You recently told me that reading CEO statements about AI and jobs might be the worst way to understand what’s actually going on. Why?
Maybe I should say one of the worst ways. I’m sure there are plenty of other very bad ways to do it. First of all, it’s just not the case that [people] always have the best sense of exactly what is happening in labor markets, even labor markets that are very close to us. The other thing with CEO statements, specifically, is that there’s [a] huge [bias] in who even makes a statement about hiring. Your local hardware store doesn’t make a statement whenever they decide to hire or fire people. If a CEO just says, ‘Well, we’re seeing business as usual on the horizon,’ that doesn’t tend to get covered.
Finally, there are incentive[s] for CEOs to characterize things in a certain way. If you have to lay off a bunch of people, are you going to say, ‘Listen, a lot of you are going to lose your jobs because I’m not a very good CEO and I screwed up over the last couple of years, so now you’re getting laid off.’ No, of course you’re not going to say that. You’re going to say, ‘[The environment is] changing and we have to adjust.’ That’s not to say that all layoffs have happened because CEOs have screwed up, but there’s always an incentive to characterize layoffs in a [positive] way.
What do you think the impact of AI has been on jobs so far?
Where we are right now is that AI is not having a large impact on hiring. That’s not surprising. This has not been going on for particularly long.
When these generative AI tools first came out, people could see the potential, but it wasn’t like people were relying on ChatGPT for tasks at work in November of 2022. It’s really important to keep in mind that on the [timeline] of technological adoption, this is just not that long, [especially] when you think about how fast the tech has evolved.
Some studies have identified a drop in employment for roles most exposed to AI since the release of ChatGPT. Your analysis at the Budget Lab suggests things have been stable. How should we make sense of those different findings?
Some of the findings with the initial results [showing that AI has already affected hiring] are still a subject of debate. Some people think that what they’re finding is happening too quickly to really be reflective of the tech and that something else might be going on. ChatGPT was released in the middle of a Federal Reserve [rate]-hiking cycle, so we would’ve expected the labor market to slow down anyway.
This is going to be very hard to perfectly track. We use these measures of [AI] exposure [for different roles], including here at the Budget Lab. Do I think [the exposure scores] have perfectly identified the jobs that are going to be affected by AI? Probably not. None of these analyses take into account new jobs that may emerge from AI or new industries because they simply can’t.
We should also say it would be [atypical] for us to have technological change and no labor market disruption. I’ve seen some people using the Budget Lab’s work…to say, ‘Therefore, workers should not worry about AI.’ That is not my view. Just because something is likely not happening right now does not mean that it can’t happen next year, three years, five years [from now].
What type of work do you think will be particularly disrupted by AI?
I think it’s clear that routine computer tasks are probably going to be pretty severely affected by AI. What’s unclear is what that means for specific workers. Does that mean that you then just become more productive? Or does that mean that’s actually a large chunk of your job and things go away?
One of the things that I’ve been thinking about is the combination of economic pressures in an industry with technological capability. For instance, I worry a lot about arts, entertainment, and media, because that is an industry that AI does [well]. Is it perfect? No, but it does have a lot of capabilities there. It’s also an area that’s been under a lot of economic strain.