How Economies Survive After AI Wins - with Daniel Schreiber

Gosh, guys, with that level of expectations, I can but disappoint. Great to be with you both. Thank you for having me on the show.

So the two are related. My work at Lemonade has been predicated on the power of AI really since the founding of the company 10 years ago. So the very first slide that I showed our management team and our board at the very first board meeting spoke about artificial intelligence rather than artificial delays. Kind of a playful insurance joke, but it really was this notion that we are going to build a company based on artificial intelligence. And in those early days, machine learning, deep learning, other forms of AI were already fairly well advanced. I push us to adopt these technologies. I'm in the thrall of these technologies, thrilled by their power. Everything that drew me to technology and entrepreneurship is manifested in AI. So I'm unapologetic and not in any way hesitant in terms of pushing for the adoption of artificial intelligence. And I say all of that as a CEO of a public company. But there's more to ask than that one dimension. And I'm also a citizen of the country and I'm a father to three boys who are entering the workforce and I have two grandchildren. So the topics that preoccupy me go beyond the success of Lemonade. And the very things that I'm seeing as so transformative and forces for the fulfillment of our vision at Lemonade, those very same forces put some pretty big question marks about the future of labor in general. Just to give you a couple of stats: over the course of the last three years, really since LLMs hit the scene, Lemonade has almost tripled its revenue. Our gross profit has fully 10xed. We've added well over a million, maybe a million and a quarter customers, but our workforce has actually shrunk during that time. So, as I say, in my capacity as a CEO, that just brings a smile to my face. And in other capacities, it says we need to think about how this makes life better for everybody.

So much has been said about this. I doubt I'll say anything new. The blast radius of what we're going through will transform pretty much everything. So I'm focused on labor and the distribution of the abundance, but education and politics and healthcare and families and epistemology and just so many things are going to be impacted by this. But within the context of what I'm focused on, I find myself going back to 2007 when Netflix launched its streaming service and it kind of sucked. There was endless buffering and a very limited catalog and resolution was nothing to write home about. But I think that what Reed Hastings understood then, which Blockbuster did not, is that you dare not look at the snapshot in time. You really have to look at the trend line. In my mind's eye, what Reed Hastings did is he looked at bandwidth and its progression in the three years leading up to 2007 and he just took a ruler and he drew that line out and he said bandwidth is going to be free and infinite. And it's not quite free and it's not quite infinite, but we're speaking on bandwidth without giving any thought to the cost or to its limitations. So at first approximation he was right and he built half a trillion dollars of value based on that insight. And I think if you do the same exercise with AI today, you don't bicker about exactly what GPT-5.2 can or cannot do today. But you just look at the progression over the last three years, whether it's in intelligence evals like GPQA or whether it's in METR's length of tasks that it can do by any measure. You just draw that line forward and you have to conclude that we are just a few years away at most from intelligence being infinite and free. And that is a startling thing because everything about our lives and our economy is organized by the notion that intelligence is something that's rare and has to be invested in and is expensive. And I think all of that's about to disappear.

They're here. They've landed. They're reading and listening to and watching everything that humans have ever produced and getting pretty good at replicating that. And yeah, I said aliens have landed and they're willing to work for free and I stand behind that.

Companies will displace people with AI because there's a buck to be made. And unlike the Elizabethan times when there were a lot of other jobs created which machines couldn't do, machines were dumb. They generated much demand for intelligence. But they did displace our brawn. They were very good at displacing our brawn, right? We don't do that stuff anymore. This time they're displacing what's left, which is our brain. And then with what part of our humanity will we add value in the workplace when machines will do stuff 100 times better, 100 times faster, and for a hundredth of the cost, which I think is rounding to where we're going to see things ending up. But if you take those very same numbers and you think that through, it will create what you're talking about, this AI surplus. If I displace an employee with AI, just think about where the dollars go or what happens there. So let's say there's somebody who's earning X dollars and I displace them with AI and let's round AI to be free. It's not quite, but let's just round it there for the ease of the story. So I've suddenly got X dollars more. That person is X dollars poorer, but I have X dollars more. In return for that, I'm generating more production and I've also pocketed this extra money. There is more money in the system. Even if I actually ended up paying that person that I let go a salary, I'd still be better off replacing them with an AI because for the same salary, I'd get an employee who's much faster and does so much more. So, a surplus is definitely created. The question is who benefits from it. We're accustomed to thinking that we couldn't possibly pay everybody a salary even if they're not working because there's not enough money in the system. And that's true. Historically, our association with the term unemployment is catastrophic because for every 1% of unemployment, we saw about 2% contraction in GDP. But here you flip it on its head because you're not generating less labor. You're actually generating a lot more labor. When you fire somebody, you're not letting them go in return for less production, but in return for more production. So it becomes absolutely, as you say, a question of distribution. How can we make sure that not all of that benefit occurs to the owners of the capital and the labor is ending up impoverished in an era of absolute abundance? And I'll just say if you had to summarize the AI revolution in its most succinct framing, I think you would say that it's displacing labor with capital.

It was after the Elizabethan times that the current social safety nets were put in place. The early days of the industrial revolution were terrible for the masses of people. They were very good for the owners of the factories, but you just read Charles Dickens and see what it was like for child labor and there was no trickle-down economics back then. Policies eventually were put into place to prevent child labor and enforce health codes, etc. We need policies now to make sure that it ends well. It can end superbly well, but I think left to market forces alone, no.

I'm always a little bit leery of giving precise predictions. It's much easier to talk about broad strokes than it is about when things will happen. The progression of technology is relatively predictable. Things like Moore's law and the scaling laws more broadly in AI do allow us to think about what the next few years might bring. That's that line that I suggest that we should draw out. I think the most conservative thing to do is just to assume that the next few years look like the last few years. But socially there'll be some countervailing forces, right? There'll be regulation that will slow us down. There may well be labor disputes that will slow things down, politics that could slow things down. We should come back to those in a minute because Israel should do everything it can to prevent those things. And then there may be the limitations that you're speaking of of energy or just not enough tokens in the world to do what needs to be done. So there definitely may be some forces, but whether it takes 2 years, 3 years, 5 years, or 10 years to get to the point of an AGI that can displace labor at scale, I think for our purposes doesn't matter. It's not a hundred years. It's any of those years is very short, very quick for this kind of transformation. And it means that our children who are in school today, whether that's primary school, secondary school, or university, they are going to be impacted by this tremendously within, you know, our field, just engineering and software. It's being felt already. People who are studying computer science degrees today are rightly worried that writing code is a perishable skill or one that's actually no longer relevant in 2026. I'm not talking about 5 years hence.

I think a negative income tax may be more palatable. The idea of giving a check to everybody just seems very far-fetched. So people kind of struggle with that idea. But a negative income tax ostensibly says let's put a floor beneath which nobody will have to go in terms of their earnings. In other words, today you have to earn a certain amount before you can pay income tax. Now let's reverse that and say below a certain level, income tax pays you. So really what you're putting is a floor. And in our modeling work, if we put the right mechanisms in place and maybe we can talk about them a little bit because you have to capture all that abundance otherwise it will just go to shareholders. But if you put the right mechanisms in place to capture it without scaring away capital and enterprise, that's the kind of balancing act today. Before you've got that abundance, the state of Israel could afford to pay a rather measly 1,500 to 2,000 shekels, that's below the poverty line, so that's not going to be Elon Musk's universal high income. But when you model it forward, if you're doing a good job of generating and then capturing some of that abundance, I could easily see how within just a few years, we could see that 10x. And if you could 10x it, that would mean that a household where both parents are unemployed would be having the purchasing power that today is reserved for the top fifth of the population. So you'd eliminate poverty. This would be something extraordinarily powerful for the state of Israel if you could really get to the point where everybody is upper middle class by today's standards just 5 or 10 years from now. But I think that that is the kind of thing we should be striving for. That is the kind of vision that is perhaps realistic but will not happen if we don't put policies in place in order to capture and distribute all of that wealth.

Well, I guess these are moral questions. But some things that we are accustomed to in the meritocracy, we think in a few terms, many of which are going to break down. One of them is that you get your wealth because you earn it, because you worked hard and because you're super smart. Parenthetically, I don't believe that. I believe that there's a lot more luck and determinism in making a fortune than people credit and that therefore redistributing it is actually a good policy for everybody. But however true that may or may not be in an era of AI, that's definitely going to be true. When you're using AI models to do the hard work, it's going to be hard for entrepreneurs or others to claim all of that to themselves. But more to the point, what kind of country would we become if we generate the kind of wealth that AI generates and we don't ensure that everybody benefits from it? We don't share it. It would be such a dark and divisive place with trillionaires, billionaires on the one hand, but an impoverished masses. And that's not a country or vision for the country that I would like to see. And finally, I think that if we don't, just kind of the broader self-interest of the capital owners, if we don't put these mechanisms in place, I'm not sure that the AI era ever comes to fruition. I mean, we remember the Luddites and Yonatan was talking earlier about what happened and there was some kind of vision of alternative labor for them and for their children. Take away that hope and see what happens to society. I don't think anybody would benefit from that. More broadly, when policies are adopted and the compensation principle isn't part of them, in other words, you do great, enjoy, but it shouldn't be at the expense of others. At a minimum, they have to be no worse off than they were before. That I think is a basic policy attitude that serves everybody's interest and is entirely appropriate. Otherwise, you're making money on other people's despair.