STEVE INSKEEP, HOST:
Our next guest questions how many jobs are really lost to artificial intelligence. It is true that many people have been laid off. It is also true that many companies say they're cutting jobs because of AI. Sayash Kapoor asserts that what they say is not always the whole truth. Kapoor is a computer scientist who has worked in tech and currently studies at Princeton University. The title of his blog gives his point of view. That title is "AI As Normal Technology." Kapoor and his co-author looked more closely at corporate layoffs in software engineering.
SAYASH KAPOOR: I think a lot of companies when they laid off workers have reached out to this convenient excuse that generative AI and advances in AI systems have been the leading cause for these layoffs. But when we looked at the data, we found that in most cases, companies that have resorted to these excuses have also had other pressures. For example, many companies overhired right after the COVID-19 pandemic. Many companies have had a lot of pressure from shareholders in order to increase their revenues or decrease their spending. And I think at the same time, generative AI comes along as a convenient excuse for these CEOs. And so what we found is that in a lot of cases, companies that have made bold claims, these claims were actually also coincidentally sort of timed with the shares not doing well or there being a lot of pressure from investors in reducing costs.
INSKEEP: So these are not layoffs that are early signs of changing the world. These are just the same kind of layoffs that companies have always made.
KAPOOR: We believe so. That's right.
INSKEEP: And how many examples of this did you check out?
KAPOOR: I mean, I think we've seen over half a dozen such examples. We've seen analyses from economic analysis firms, saying that when companies are preparing for AI-driven layoffs, in most cases, in 9 out of 10 cases, they don't even have an AI application that's ready to fill in those jobs.
INSKEEP: Wow.
KAPOOR: And when we've seen surveys of thousands of global executives, we've seen that in many cases, when companies do make these headcount reductions as a result of AI, they are mostly about the potential for what AI could do in the future, rather than what AI systems can already do today.
INSKEEP: I was interested that you decided to focus on a field that is presumed to be extremely vulnerable to losing jobs to AI - software engineering. That's supposed to be something that AI should be super good at. Is it turning out there is something intrinsically human even about coding?
KAPOOR: You know, AI systems are getting a lot better at coding. What today's AI systems can do for writing code is nothing short of revolutionary. This is perhaps one of the most important changes to the profession in over a few decades' worth of time. And yet, the impact on the software engineering profession as a whole, we think, will be far more muted than what many people have said. And that's because the job of a software engineer is not just to write code. It is also to understand and decide what needs to be built, what is it that'll provide value on the market and also to take accountability for what comes out the other end.
INSKEEP: Are you then seeing evidence from the point of view that AI is not going to eliminate work; it's just going to create different kinds of work?
KAPOOR: That's exactly what we think is going to happen. I think as AI automates certain tasks within jobs, we think - like previous waves of automation, we think the definitions of those jobs will change. One other example that comes to mind is that of radiology. About a decade ago, Geoffrey Hinton, who's a Nobel laureate, said that we should stop training radiologists now because it's clear that within five years, AI systems are going to do better than radiologists. Ten years later, there's a global shortage of radiologists. That's not because AI systems can't read X-rays or can't read radiology scans. They can. They are very good at the job, often better than radiologists. But radiologists do so much more than reading X-ray scans. In fact, even if AI systems automate that specific task, radiologists can do a lot more. In fact, in many cases, we can increase the number of radiologists because it's now cheaper to hire radiologists because they don't need specific kinds of skill sets.
INSKEEP: If companies are not looking at the opportunity to eliminate their workforces, why are they investing trillions of dollars in artificial intelligence? Isn't that the point, ultimately - to save money?
KAPOOR: I think we can look at it two ways. Of course, one of the ways in which companies can realize or recoup their investment is by saving money. The other option, though, is to grow the pie. It is to create so many new products and services that enter the market that this is how you can pay for all of the investment that has been put into AI. Now, it remains to be seen whether these investments would be recouped using substitution for existing labor or creating entirely new markets for creating value, but we would place our bets on the latter.
INSKEEP: I want to ask about one more aspect of this, and it's not firing existing workers but companies creating new junior-level employees. Is there evidence that there is, in fact, a soft market for young people just getting out of college, looking for their first job because companies are less and less interested in people at that level?
KAPOOR: I think this is certainly a possibility. And I think as opposed to the story of AI causing mass layoffs, this is one where we should be putting more emphasis on. A lot of what junior employees did at a firm is to carry out these specific tasks while they learned on the job, while they learned the ropes or what made them more competent across bigger array of what constitutes their job. But now that AI systems can carry out these small, well-scoped tasks more easily, I think this does reduce the payoff for a lot of firms for hiring junior workers. And so in some sense, it creates a collective action problem.
Of course, companies might hesitate to hire junior employees, and we're seeing some very early initial evidence of that. But at the same time, this is the very workforce that will go on to be the future senior employees and the decision-makers at the firm. So I think there are lots of ways that policymakers can step in to resolve this kind of collective action problem. But I do think this is one that's worth taking more seriously than the story of mass layoffs from AI.
INSKEEP: Bottom line, you're saying if you've already got a job, don't worry too much about AI, but if you do not yet have your first job, you should worry about AI.
KAPOOR: One other thing that I would add as a wrinkle to this is, if you are in a field which is exposed to AI in some way, it is worth figuring out how that field will evolve in the next few years because even if you do keep your job, I think software engineering and many other fields will face dramatic changes over the next decade or so. This is something that is coming your way. Regardless of whether it leads to job losses, it might lead to the nature of many different professions changing significantly.
INSKEEP: Sayash Kapoor of Princeton, thanks so much.
KAPOOR: Thanks a lot for having me.
(SOUNDBITE OF JON KENNEDY'S "ALL A DREAM") Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
