Researchers studied AI layoffs. Here's their warning. transcript

A few weeks ago, I read a paper called The AI Layoff Trap by two professors, Gerry Tsoukalas and Brett Falk. It's a warning about a possible future where AI is good enough to do a lot of our jobs. Companies would start replacing workers to cut costs, but those workers are also customers who spend money at businesses. If every company fires a lot of people, eventually hardly anyone is left with enough money to buy anything and keep the economy going.

What scared me most, though, wasn't the idea that mass layoffs could happen. It was that even if CEOs see this crisis coming and desperately want to avoid it, they might not be able to. Waiting for the firms to figure it out for themselves, I think is the worst possible thing we can do. The CEOs are coming out, you know, from these foundational labs, and they're constantly saying AI is going to replace everyone's job.

And the main question we wanted to understand is who's going to be left to buy products if everyone gets automated and replaced by a robot? So we ended up building our own model and writing our own paper about this very topic. So that's how this whole thing started. What you are talking about is something that you call the AI layoff trap. Explain what that is. Yeah, sure. And here it might be helpful to simplify things down a little.

The paper is 60 pages of mathematics, so we don't need to get into that. Thank you. Imagine there's just one very large company, a monopoly in a market, and they're making widgets. And everyone's working at this company and this brand new shiny technology arrives. AI. And so the CEO is looking at this and saying, all right, well, I could adopt this technology, but I would have to lay off, let's say, 90% of my staff. The problem is these people are also the ones buying widgets from me.

And so when it's just a simple thought experiment like that with just one company, you know, they actually end up doing the right thing, meaning, you know, you take that into account and you don't over automate you, you do just enough, right? You balance these things out. What's interesting and what's different is when you're in a competitive setting, and that's where kind of everything changes. So you're no longer a monopoly.

There are lots of widget companies. Yeah. So now there's a bunch of widget companies, lots of workers everywhere. So your benefit of adopting this technology is great. You save a bunch of costs on producing the products. You have to lay some people off. And now you're thinking on the other side of the equation, which is who's going to be buying my products? Now when there's lots of companies around and lots of other workers, you don't worry so much about the fact that your own workers might no longer be buying your products.

You can say, well, there's plenty of other people around. And you know what? My workers were only spreading their money around anyway. It wasn't that they were only buying from me, they were buying from the 100 other widget companies. And so really what I'm losing now is just a fraction of that because, you know, they weren't spending much with me anyway. And what happens, Gerry, then if the other companies start laying off their workers.

There you go. That was the exact thing I was going to say next. Right. And, and that's the problem when you're looking at this from a market perspective, if everyone thinks in the same way, you end up with this essentially disastrous scenario where at the end of the day, the demand is entirely eroded and every firm is worse off. But but CEOs are rational, right? They they see this cliff coming. It's not that they don't understand the economics that you're laying out this lay off trap that if you get rid of all the workers you've got nobody left to buy your groceries or your widgets or your home furnishings or whatever it happens to be.

So why are they all collectively driving towards this cliff if they know that they could be killing their own customer base? Because it goes back to yeah, it's a great question. It goes back to this idea that when you're looking at the math, when each individual CEO is looking at the cost benefit analysis, the cost side of the equation is almost zero at this point. Because from your perspective, I'm laying off workers, but they weren't really spending much with me anyway.

So from my perspective individually, it makes sense to do that. And if I hold back and I don't, all the other companies are going to automate and now I'm going to go bankrupt. I'm going to lose my workers anyway. So no matter what you do, no matter what the other companies are doing, your best strategy is to adopt as much of this technology as possible. And that's called a dominating strategy in economics. It's been known for a really long time, and it creates some really nasty outcomes, including some classical things like the prisoner's dilemma, which won you know, a Nobel Prize a long time ago.

So that's the trap. You've also gone a little bit further, though, Gerry, which is to lay out what could be done about it. Right. So if you don't mind, let me take you back 3000 years just for a split second. Okay! So I don't know if you've read The Odyssey, but basically one of the high level ideas in that story is, this was after the Trojan War in Greece, where one of the heroes, Odysseus, he has to go through sirens through his boats that are all knowing beings and have, you know, beautiful music, but also divine knowledge.

But the problem, of course, is that in that story, it's a trap. It's a literal trap where humans end up dying, who go on this island with the sirens. And so he asks his sailors to tie him to the mast of the ship. And he says, no matter what I say or what I scream or what I do. Keep me tied there. And that's what we're kind of advocating as a solution here. It's the fact that companies need to slow down with the firing or the replacement of AI, you know, human workers with AI.

So we're calling for self restraint, just like in that Odyssey story. But Gerry, I mean, I love the idea that you're calling for self restraint. And I wish I saw that in the tech world emerging, but I don't. I don't see - I see companies rushing like sailors to those sirens trying to get to this as quickly as possible. So what can be done if companies decide not to exercise self restraint? And let's be honest, companies are not best known for their self restraint if they don't think that it's in their own shareholder interest.

Absolutely. And that's exactly the point of the paper and the point we want to make to our other economist colleagues, but also the policymakers, is that they will not self-restraint. That's literally the prediction of the mathematical model. And so what you need in that case, just like in the Odyssey example, you need an external force. And the instrument that we propose, you know, we went through about six different popular proposals that economists are discussing right now, things like universal basic income and giving workers equity.

But the only one that really works is the one that directly addresses the incentive to replace a worker with a machine. If you're going to want to replace a massive amount of your workforce with AI, we're saying to either impose a tax on that so that it's costly to replace to fully replace the worker, right? It wouldn't be costly to augment them with AI, but it would be costly to fully replace them with AI. So it's basically a tax on firing.

And the companies will therefore have to bear some of the responsibility for replacing human workers with automated workers effectively. Has it ever been tried, Gerry, before? Has has there been a moment in history where this kind of a tax has been used and been successful? The closest thing I can think of is the carbon tax, which doesn't directly penalise the firm for polluting a river, but, you know, allows it to offset some of the pollution, you know, through a through a government mandated target.

And so that's what it would have to look like if we wanted to implement it. But because the word tax is pretty toxic, it might not be directly implementable, implementable this way. And then you can start thinking of other things that might be a little bit more realistic. You could also provide subsidies to firms that keep their workers instead of firing them. So there's all sorts of other instruments around this that might not be perfect, but that we could use.

If companies don't exercise self restraint, if policymakers fail to follow your recommendations, what does the United States look like in ten years, let's say. What does it look like if we don't avoid this trap somehow? You're putting me on the spot here, Katty. So I'll answer it. But let me preface by saying again, no one agrees about this very question. Half people are on one side, half the people on the other side.

My view is there's going to be a tremendous amount of wealth inequality, something that we've never seen before. And we already have lots of problems with that. And that's my prediction. If we do nothing, this technology is going to keep on improving. And I'm very confident about that. And then we're going to have massive wealth inequality problems, which could lead to political instability and all sorts of other things that we've seen.

For instance, in the past financial crisis in 2008. Are you confident, Gerry? That's the dystopian view. Are you confident that actually CEOs are smart enough and rational enough that they will understand the trap you're laying out and try to do something to mitigate against it? No, I mean, if I believe my own work, it's not that they it's very hard to create agreements across companies because as soon as you agree, let's say you agree, let's slow things down, right?

Let's say Anthropic and OpenAI do slow things down. And then Microsoft agrees. You know what? Let's scale our efforts down. And Google agrees to do that as well. As soon as you walk out of that room, your incentive, unless you're you're penalised somehow, right? You would have to put in place a system of sticks and carrots and the sticks have to be pretty significant because once you shake the hand and you walk out of the room, you immediately look at the gains you can get from adopting AI.

Now that you know that no one else will, right? That's your opportunity and that's your duty to your shareholders waiting for the firms to figure it out for themselves, I think is the worst possible thing we can do. Gerry Tsoukalas, thank you very much. Thank you for joining us. Thanks for having me.