Why UBI Won’t Save Us from AI (and What Might): A Linkpost

This is a linkpost to a video (and corresponding transcript) of a panel from Asterisk Magazine featuring Kelsey Piper (staff writer at The Argument; formerly at Vox's Future Perfect), Caitlin Tulloch (Senior Director for Research at GiveDirectly), Nick Allardice (CEO of GiveDirectly), and Saffron Huang (Societal Impacts Team at Anthropic), which I watched a few weeks ago and found compelling. I’d be interested to hear what others think about it. If the discussion surfaces good ideas and/or I come up with more, I may write a follow-up post on AGI and income distribution over the next few weeks.

What follows is a transcript of an exploratory panel discussion hosted by Asterisk Magazine, GiveDirectly, and SwissNex on Oct 20, 2025.

As AI capabilities advance, questions about how to distribute the technology’s economic gains have moved from theoretical to urgent. This event brought together practitioners working on cash transfers at scale and researchers thinking about AI’s societal impacts. Their conversation touched on evidence from existing cash transfer programs, the practical constraints of universal basic income UBI proposals, and alternative approaches like “predistribution” of AI tools and capital.

This wasn’t a debate about specific policy proposals so much an attempt to establish common ground between communities approaching the same problem from different directions. The AI policy world and the international development world use different frameworks, cite different evidence, and often talk past each other. Getting these groups coordinated may be as important as solving any particular technical question about implementation.

What emerged was less a set of answers than a clearer picture of what we still need to figure out — and who needs to be in the room as these conversations continue.

Kelsey Piper: We’re going to talk tonight about how to share AI’s future wealth — things like UBI programs, different ways to design them, which of those work, and we know about what works and what doesn’t. I’m Kelsey Piper, a journalist at The Argument. You guys go ahead and give a little bit about yourselves before we get into this.

Caitlin Tulloch: My name is Caitlin Tulloch. I’m Senior Director for Research at GiveDirectly, but more broadly, an economist focused on the cost-effectiveness of different interventions to address global poverty.

Nick Allardice: I’m Nick, the CEO of GiveDirectly.

Saffron Huang: I’m Saffron. I am on the Societal Impacts team at Anthropic, and where we try to find evidence about how AI is actually impacting people.

Kelsey Piper: AI seemed like future speculative stuff when I was in college 10 years ago, but people were already talking about, “Once we’ve automated all jobs, what are we going to do next?” And the good outcome was always imagined as, “Yeah, we’ll do something like a UBI. We will redistribute the wealth that we have created. Everybody will be able to spend their time however they like.”

Of course, UBI isn’t speculative. In a lot of ways, from the perspective of 1900, we are in the post-scarcity future, and we’ve already tried many experiments with different kinds of cash transfers. Any way I characterize their results is going to be very loaded. What we’ve found is that giving people money when they are very poor causes them to not die. That’s fantastic. It causes them to have better lives in a lot of ways, on every metric that you look at. Giving people who are already not poor money does a lot less, but that might just be because we’re not giving them comparatively nearly as much money.

Caitlin and Nick can both correct me in a lot of detail on that, but it raises a bunch of questions about if we’re trying to redistribute the gains from AI. There is an insane range of proposals that could all fall into the very broad bucket “UBI.” I don’t know — what would you do? Go for it. I didn’t warn them I was going to ask that. What would you do? Caitlin, you’re first.

Caitlin Tulloch: I’m really hoping that my status as a researcher is going to allow me to be a little bit pedantic, and hopefully you all will find that charming. When you look at the evidence on giving people money, you can kind of break it down into short-term and long-term results. In the short term, it’s really, really good at helping them eat and access healthcare, especially if you give people small amounts of money for long periods of time. And for a pretty big fraction of the world, that’s actually what they need. I’m a very bottom-of-the-pyramid focused person, in that my utility function is largely run on babies not dying. Helping people do the very basics is tremendously important to me.

But there’s also emerging research on how much bang for your buck you get if instead of giving people a little bit of money every month, you give them a much larger amount of money that lets them develop their own income stream. And that kind of liquidity seems really good at doing things beyond just eating and accessing healthcare. That’s when you see people making productive investments or moving to areas where there’s more opportunity. I think the thing that is in favor of redistribution in this way is that people are just infinitely creative when we let them make those decisions.

Kelsey Piper: Totally reasonable starting point. Nick?

Nick Allardice: We have a wide range of policy options for responding to labor market disruptions caused by AI, but there’s a lot that some form of cash transfer has going for it. It’s incredibly simple to administer. It’s incredibly flexible.

And to build on what Caitlin was saying, a lot of people find the idea of retraining or skill-building programs very intuitively attractive. But when you actually think about it for a second, what institutions would you each trust to identify exactly what you needed to learn and then teach you effectively what to do with it? Most of the research that we see is that people don’t need to be trained. They already know what they need to do. Often we’re actually quite bad at training people to do things. But when you give them some sort of access to capital, then the incredibly diverse things that they can do with it really unlocks quite a lot.

Like Caitlin, a bugbear of mine is that the discussion around how AI is going to affect society is heavily concentrated on rich economies. The amount of conversation about jobs in the U.S. drives me up the wall, because the risk is not labor market disruption here. The risk is that you have 800 million people still trapped in extreme poverty who are just going to be left behind because they’re not even able to get on the ladder to start with. And at a time when we are unlocking unimaginable wealth — the greatest concentration and growth of wealth in human history — the opportunity to make sure that people aren’t left behind is enormous. And it’s actually fairly affordable.

There are economists who put the estimate of how much it would cost to literally lift every single person in the world out of extreme poverty in the sub-$300 billion range. Now that sounds like a lot of money, but when you think about just the market caps of companies in the AI space that have grown in the last 5 to 10 years, it’s a pretty small proportion. When you think about the amount of economic value that might be unlocked by this kind of revolutionary technology, I am hopeful. There are some cynical parts of me that say this is going to be really hard, but this is a moment in which it’s possible to not leave people behind.

Kelsey Piper: Saffron.

Saffron Huang: I’m going to take the anti-cash stance just for interestingness on this panel. I work in AI, figuring out what its societal impacts will be and what we should do about them. And I’ve seen all of these tech CEOs throw around the word UBI a lot, but also put a lot of money into UBI programs in Oakland or in the U.S. To, as someone who is pretty uninformed in this space relative to you guys, I was like, oh, this seems like it’s not the right solution here. It might just be we’re taking a policy solution that works very well in a different context and trying to apply it to the global north.

Here, the evidence in support is weaker, but also it feels like a pretty unambitious idea. Like, the idea is that we are going to create this unimaginable wealth, and the result is that we’re going to cut everyone out of the economy and put them on a sort of—the cynical part of me says, it will be a floor— just enough to live on or something. And people are going to have to hope that the payments don’t decrease, since they won’t have much political leverage anymore because the government doesn’t rely on their revenue in order to keep doing things.

This is a very extreme scenario of labor automation that I’m talking about, and it might not even be possible in the next few decades. But in the case that it is, I think it’s worth thinking about, okay, do we have any policy tools that could do anything about this? To me, UBI doesn’t really address the root cause of the issue, which is that the wealth-generating assets are owned by a small number of people, and that divide is just going to continue to go up if you don’t spread ownership around. This is politically precarious, I think, and probably quite bad for democracy.

Then I came across the idea of pre-distribution versus redistribution. Pre-distribution is the idea that we should try and spread the wealth-generating assets around, or at least give people more opportunity to earn a more equitable distribution of income ex ante, rather than doing a lot more redistribution ex post. This is appealing for a bunch of reasons. It’s more politically feasible. People just don’t really like being taxed a lot. It requires moving less money around often. Putting more money into things like education and healthcare upfront so that people are more able to earn more income actually costs less than redistribution, and it seems to, for me, get at the root of the thing a lot better.

And so it’s like, okay, what if we applied this idea of pre-distribution to AI? What would we need to pre-distribute now to ensure more equitable ownership, more ability to participate in this flourishing, abundant AI future for as many people as possible? Part of it is skills and education—people being able to actually use these things to start businesses or to achieve their own ends. Part of it is productive assets, financial capital, things like that.

But what I find really interesting about this group of people is that you’re trying to solve a different problem that I’ve spent less time on. I’ve spent more time thinking about the global north, partly because AI is more threatening to knowledge work than anything physical, and there’s a lot more knowledge work in the global north. And these are the industries that people aspire to. What happens, not just economically, but culturally, when that is much less available? So I think I’m definitely thinking a lot more about relatively very privileged people who are knowledge workers. But I think those are also potentially the people that AI will threaten the most. In terms of benefits, I really agree that we should try and use AI to benefit as many people as possible.

Kelsey Piper: So, Nick, do you have something there?

Nick Allardice: Yeah — one thing that I think is important is that UBI can mean many things. When many people hear the word UBI, they picture a monthly payment forever to a universal population of some kind. Saffron just said something really interesting, which is actually consistent with our experience. We ran—we’re running—the longest-running trial on UBI in the world right now. There are three wings to this trial. It’s in western Kenya. There is one set of communities that received a large lump sum amount of money, equivalent to one to two years’ worth of income all at once, something in the range of $50,000 to $100,000 for an American.

Another set of communities took that same amount of money and then dripped it out into monthly payments over the course of three years. And then the final set of communities received a small monthly payment that’s lasting for 12 years. We’re eight years into this trial, and we have results on the first two wings. And what it showed pretty unequivocally is that the people who received the money all at once versus dripped out over time outperformed on basically every indicator — every single one of them, whether it’s income generation, health outcomes, et cetera.

Why is that true? The reason is, and it speaks to what you were just saying, is that they invested the money productively. It didn’t get eaten up in their daily budget. Maybe it went towards their education, maybe it was in starting a business — the types of things that allowed them to productively participate in an ongoing way. So this idea of UBI can encompass many ideas. And for us as an organization, we’ve actually shifted more towards this model of giving a large, one-time amount of money that catalyzes people, gives them the startup capital they need to then support themselves long-term.

Kelsey Piper: Yeah. Saffron, what you just said reinforced for me that I feel terrified of the world where the broad plan is an ongoing transfer which most people are reliant on because—well, because of watching the destruction of USAID this spring. If the system is that we are giving out aid on an ongoing basis that people rely on for survival, and then somebody’s going to say, “Well, we don’t want to do that anymore,” then they can stop.

I do also feel nervous about a one-time transfer, though, especially if it’s—so, okay, maybe we need to start discussing a specific proposal. Maybe Anthropic should give out shares — I’m picking on Anthropic because you’re here — everybody should give out shares of potential future revenue in the event that they meet some insane benchmark, which a lot of people think they might. These wouldn’t be voting shares. These wouldn’t have to do much with the traditional investment structure, but you can own it. It’s a property right. It is enforceable.

I have a couple of hesitations. One is that most people have no idea what this is, and probably it is very easy for me to go around and say, “This piece of paper that is about a speculative sci-fi thing, can I buy it from you for $5?” Wait, do we want to ban that? Do we want to say it’s non-transferable only until the future? Do we want to say it’s non-transferable forever? I’d be curious if you thought about that, and I’d be curious if you guys have thought about what structure works best for people.

Saffron Huang: The devil is in the details. How exactly do you actually give shares, and to who? If we’re saying that we want to benefit everyone, how do you actually give a share to every single person in the world? We don’t really have infrastructure for that. I don’t really want to be scanning people’s eyeballs, although actually we do have infrastructure for that now, thank you, Sam Altman. But I don’t know. There are still some real logistical difficulties. It’s just easier to do things on the national level. One idea that I think is cool is in the One Big Beautiful Bill—Trump has this thing called Trump Accounts, where every child born between 2025 and 2028 gets $1,000 seed money in that account from the government. The parents can invest like $5,000 every year after that, and it grows. It’s invested in a diversified fund tracking the US stock index, and it grows tax-deferred until they’re 18.

And I think that this is a pretty cool idea. This could end up tracking AI stocks if the S&P 500 basically just becomes AI companies. So maybe this kind of thing is the way to go.

Caitlin Tulloch: I think this conversation about how it’s structured is actually really important. I lived through the collapse of USAID earlier this year, so it’s very fresh to me. But one of the things we’ve observed is that it’s actually relatively hard, particularly if you’re observing the rules, to completely end foreign assistance flows. And that hasn’t happened. It will be dramatically reduced, but it will not be gone because there is bipartisan consensus supporting it. And I think that’s going to be an important difference when the wealth is explicitly in private hands or in corporations.

So, who gets to decide what happens with this wealth? The more people are involved in that, the slower and the more painful it gets. But even as an economist, I’m very willing to say that deliberative bodies are a good thing because it makes it harder to suddenly have this future where you’re like, “Well, we’re just cutting off the income stream.”

The other thing I would want to add to what Nick was saying is that when you take the long view of economic development, what you see is not people who spend more in the same households and the same occupations and the same places they’ve lived their whole lives. Long-term economic development is the structural transformation of the economy. What does that actually take? I mean, this is the million-dollar question for all of economics since the Harrod-Domar model was published in, I think, 1945, so I’m not going to be like, “Oh yeah, once we figure this out, we’ve got it sorted.”

But the framing of this doesn’t just have to be, “How do we help people in the places that they are?” but, “What is it that helps this process of structural transformation that has actually gotten a large fraction of the world to what would look, to someone in 1900, like a post-scarcity society?” As Nick says, we’ve come up with a lot of really clever ideas about how to teach people to fish in the last 50 years, and when we study them, unfortunately a lot of them really don’t work as well as we thought. Some of them do. But this doesn’t need to be an either-or thing — you can think about both the role that redistribution plays and the structure that it’s happening within simultaneously. That’s actually my read of a lot of the evidence and where we see the biggest impact.

It’s not as though cash transfers are acting completely independently of the context in which they’re happening. You see them having the greatest effects when there is an education system to which people can send their children to get more labor market skills, when there are roads that connect them to markets they can make something that someone a reasonable distance away is going to buy. These things are interacting.

Kelsey Piper: So that all works as long as that big structural transformation has at the other side of it economically productive work that they can do for themselves from there. Right? I’m a little worried that that might not happen.

Caitlin Tulloch: How would you define economically productive? Ireland has a UBI program for artists, and they’re a center of an incredible art scene. One of the highest aspirations I think of humans is to have enough leisure time that they can do something that is absolutely productive in a human and meaningful sense and that other people would pay money for, even if it’s not producing, I don’t know, meat and potatoes and AI models.

Kelsey Piper: I’m worried about the world we’ve ended up in if all of the things that people are doing and buying and selling are totally disconnected from the parts of the economy that produce all of the goods and all of the ability to use violence and all of the ability to enforce and make laws. Once we’re no longer an input to the parts of the economy that have material power, that’s where I feel scared.

Saffron Huang: Can you also clarify what you mean? I don’t think I caught that.

Kelsey Piper: So let’s say that we have a system where we’re giving people money, maybe in the form of shares in the S&P 500, maybe in the form of monthly payments of some kind, maybe in the form of a one-time transfer. They can live off the interest or they can use that money and the flexibility it gives them to do something—to retrain, to do whatever it is that they’re doing that allows them to participate in the new economy.

This works well if the difference between the current economy and the post-AGI economy is something like the difference between the 1900 economy and the 2025 economy. Those are ludicrously different from each other. Almost all of the jobs that people were working in 1900, we don’t have anymore. If we suddenly got the population of 1900 America dropped on us, it would be a refugee crisis because they would all be trained for roles that we don’t need them in, and they would not know how to do and not have the skills to do the stuff we’re doing. But fundamentally, the stuff we’re doing is still stuff that is important to the functioning of our society. If we all stopped doing it, everything would grind to a halt overnight.

That is, I think, the source of a lot of our power. Obviously our democratic institutions are also a big source of our power, but I do think a lot of power is economic in the sense that it comes from the fact that if we did not do our work, society would stop. It’s very easy to imagine that that is not the shape of the transformation from 2025 to post-AGI. It’s not that there wouldn’t be lots for us to do, like I would happily spend all my time writing stories that only 100 of my friends read and running the homeschool co-op much better and raising kids and making art and all of that.

But if none of the stuff that any of those people do is actually essential to the functioning of the economy, that feels very precarious.

Saffron Huang: Yeah, I agree. I think that’s also another thing I found kind of weird about UBI as a solution to job loss. It’s like, oh, well, we’ll all just do other stuff, but something is humming under our feet. That’s the economy that we have no say in, and it’s shaping our lives. We’re detached from it, and nobody’s actually participating in it, but it’s our entire material reality. That’s also quite weird and bleak.

I also think it would be odd to build an economy for humans with no human involvement. And I would hope that we still have a robust supervision infrastructure. Another thing is that we might have AI infrastructure — the training of the foundation models or something — that’s owned and overseen by relatively few, but potentially a lot of applications on top that are much more bespoke and smaller. And so different people can have different small businesses, and they can get their AI agents to do stuff, or leverage less labor into more returns or something like that. (I also think we should try decreasing the length of the work week, but that’s another thing.)

Nick Allardice: So I just don’t care that much about this problem, which is not to say it’s not a problem—

Kelsey Piper: But it’s not the problem that you’re motivated to be working on.

Nick Allardice: When we look at the order of magnitude of the problems that we should be focused on solving, there are 800 million people in the world who have to make choices between eating more than once a day and buying medication for a sick child. Maybe in 10, 20 years’ time there’s going to be mass white-collar disruption. Cool. Welcome to the 20th century for blue-collar workers. Maybe desk workers are about to experience what call center operators in the Philippines are experiencing right now. And I just think that it’s a very hard, abstract, intractable problem to forecast the future.

I don’t say that we shouldn’t do it at all. I just think there’s a very tractable problem that I see getting far too little attention, and which is orders of magnitude more important on the scale of human flourishing. And so it’s hard for me not to be distracted by that when we start forecasting into the future of, you know, “Will we find productive things to do with our time?” And we have some evidence that when people get access to resources and freedom, they don’t turn into slobs who gamble and drink it all away. They find creative endeavors. They create entrepreneurial activities. I don’t know what that’s going to look like, but I have a lot of faith in humankind’s capacity for ingenuity.

Kelsey Piper: It’s fair to say that it feels to you like the most important mistake we might make here is like coming up with some kind of national-level scheme for distribution and completely ignoring the fact that a lot of the people who are—

Nick Allardice: The default scenario, right? Yes. The default scenario here is that there is unimaginable wealth captured in the US, or maybe a few other global north economies. And a couple of things happen from there. Either that just drives productivity growth and the economy adjusts in the same way that every other technical transformation has happened over the last half-thousand years. That’s Plan A.

Plan B this time is different. But even then, there’s still a post-scarcity society, and, oh no, we have to figure out what to do with our time. And, yeah, there’s issues with democratic power and accountability and various things. All of that is plausible. But still, the default scenario right now is that all that value is captured at the top of the chain and that there’s an extraordinary number of people, and not just those in extreme poverty, who will be left behind, like the call center workers in the Philippines. The pathways to middle income, manufacturing jobs and things like this, which have over the last 20 years lifted billions of people out of poverty—those are the things that I’m stressed about.

Caitlin Tulloch: And I’d love to jump in on that. I really liked Kelsey’s comment at the beginning that from the perspective of the 1900s, we live in a post-scarcity society, and a huge fraction of that is thanks to the Green Revolution. The Green Revolution is this wonderful thing that happens in the 60s and the early 70s. Before that, people were worried that the world population was growing too quickly to feed everyone. Then we figured out how to double maize yields in Mexico. We learned to export that technology to Bangladesh, to Pakistan, to India. And this problem just went away…except for the continent of Africa, where for some reason the benefits of this immense technical leap forward were not made available to the systems and the people who live there.

That’s part of where we’re at today—there are still people who are left behind in their very basic food security for what is essentially a problem of distribution. There’s a very famous scholar of famine who has made the point that since roughly 1955, all famines are man-made. But we still have famines. Honestly, I get into arguments with my husband about this. He reads a lot of speculative sci-fi, where there are post-scarcity societies, and I have a hard time believing in a future in which everyone is post-scarcity because if we’re in that world now and there are still 800 million people left behind, we’ve clearly missed something.So the problems of distribution are, to me, significantly more interesting than the amount of aggregate wealth.

Kelsey Piper: So on a policy level, let’s say you travel back in time, and you know everything except which stocks go up in the 1940s. Is there something we could have done differently going into the Green Revolution that would have made it such that we didn’t currently have 800 million people living in poverty?

Caitlin Tulloch: I have spent a lot of time thinking about this, because I worked on agricultural policy and what’s cost-effective for improving yields. And you’re just stuck with this fact that average yields are two to three times higher in some countries than others. It’s been stuck that way since the mid-sixties.

There are a few reasons why. One—and I’m going to tell you this detail not because I think you need to know about local ecosystems in Africa, but because I think it may come back to your point about small businesses — is that the amount of heterogeneity in microclimate, soil productivity, all of these things that determine agricultural yields is massively greater on the African continent than many other places in the world. If you design a high-yield, drought-resistant type of maize that works well in one place, it is simply not as effective in certain communities, certain microclimates. So that’s a technical failure, and I think speaks to this idea of, “Well, how are you going to adapt AI to the actual problems that people have on the ground, not the problems we would think they have in the US or even working in Mexico or in Pakistan?”

The other thing, as I understand it, is at a certain point, either MacArthur or Rockefeller stopped funding Norman Borlaug’s agricultural research. They did Latin American varieties, they did South Asian varieties, and at a certain point the funding was cut. And that also comes back to the point also about how quickly the tap can be turned on and off. When you look at the amount of the grants that went into CIMMYTin those early years, the ROI in terms of human welfare is insane. And then somebody just decided to stop.

Kelsey Piper: I completely buy the analysis that we are much more likely to solve the problems that are experienced by Americans in the course of the AGI transition than the problems that are experienced by desperately poor people in other countries. Even if the problems that are experienced by Americans are way less bad, those are the people who vote in America, the people who you see when you walk on the way to work, the friends and family of the people working at AI companies.

But this makes me more in favor of a system that doesn’t rely on ongoing transfers. But I still don’t feel excited about a big one-time transfer as a solution to a next major structural transformation of the economy, because I’m just not convinced that it can be sufficient to get people through and into the new economy.

Nick Allardice: I do think it’s worth differentiating the types of problems that we want to solve here. One problem is, can we raise the floor? Can we get to a point where the most awful, abject, unnecessary suffering doesn’t exist? And I think we could do that with a one-time large transfer. There’s quite good evidence to suggest that. But I don’t think that gets us to the point where all of these people are now participating in the next structural transformation of the economy.

The reality is that nation states have power and that nation states are accountable to their local constituents. That is just true. But one point of hope is I think that we are living at a time where what can be done outside of the nation-state architecture is greater than it has ever been. We are seeing the greatest concentration of wealth in human history. Now, I would really prefer not to rely on billionaire philanthropy. That wouldn’t be in my top 10 list of things that I would wish for as the solution. But I don’t have a lot of faith in the generosity of nation states either.

Kelsey Piper: I mean, the more individuals there are who are in a position to individually fund ending extreme poverty, the higher the odds that one of them will go, “Oh, yeah, I guess I could do that.”

Saffron Huang: I have a question. There’s this buzzword that people are throwing around, which is universal basic capital, as opposed to income – that is, productive assets. This might be something like the Trump Accounts thing, or like the GI Bill, which gave a lot of Americans access to assets through education and home loans and things like that. Are there studies that try things like that? I know that cash preserves a lot of optionality, but I’m just curious if there are any ways that people have tried to structurally incentivize putting cash into productive assets or giving productive assets directly.

Caitlin Tulloch: Oh boy. So the latter one, yes. Everybody’s got to understand capital is a horrendously abused term by economists. At some point we figured out that if you frame everything as capital, you can get finance ministries to be like, “Oh, oh, it’s an investment in capital. Gotcha.” The classical model we all learn is that you have labor and you have capital, and capital is like machines that make, I don’t know, screws and grain mills and things that are big and metal and very 1940s feeling.

Then this expanded to include financial capital—the most liquid form of capital that you can hopefully turn into all the others— and then human capital. We had social capital for a while in the nineties, and we’ve kind of walked away from that. So when I say this term is abused, that’s why it’s worth being specific. I really strongly agree with the idea that the GI Bill and the availability of quality state universities in every state in the United States were UBC programs, but to a large extent, many of those are not available anymore.

There were these tremendously generous social programs that were largely financed by taxation, and massive government investments in R&D that ultimately became private companies that make people money. Those are all forms of capital the government did provide. This is not a new concept. That’s how we got here. From the US perspective, I would be very interested in how we can rebuild the social contract that we’re going to tax people enough that anyone can afford the University of California, State University of New York—the evidence around the amount of economic mobility from people just moving through SUNY campuses is bananas. So that’s in the US context.

And there are a lot of studies in other contexts as well. I mean, Nick was citing our study in Kenya earlier. I think this explains some of the discrepancies you see in the research on cash transfers in the US versus other countries. By and large, what we’ve tested in the US is UBI. They are small amounts that will support consumption but don’t have as big of long-term structural effects because people are meeting their basic needs, and that’s getting harder in a world where there’s a cost-of-living crisis, and so the amount that we thought was going to maybe let you also make more investments is just going to paying rent and food and childcare.

Overseas though, we’ve got a lot more studies that give people amounts of financial capital that they are otherwise very unlikely to have. These are contexts in which they don’t have savings accounts, the credit markets are really, really messed with very high rates, all of these things. And there, what we see—we’re getting into the 10, 15, 20-year studies—is that they’ll turn this capital into all kinds of things. It may well be goats, it may be a roof on your house, it may be the choice to move to another city, it may be greater education for yourself or your child. It’s interesting. There’s a very nice meta-analysis from Northwestern recently that breaks out the impact by stream transfers versus lump-sum transfers, which would look like capital, and finds that there’s asset accumulation for lump-sum cash transfer programs but not really so much for these stream programs.

There are also 1,000,001 experiments with giving people the assets we think they should have. I mean, I’ve raised money for the Heifer Project in high school. Those don’t do as well because we’re not as good at knowing. I’ve been at multiple conferences where we had to talk about how many chickens died in Guatemala because in Graduation — Sorry, so this is a mode, the Graduating from Ultra Poverty model, which involves giving people a big set of complementary inputs. It tends to include a stream cash transfer, a large financial asset, usually cash, but it didn’t used to be, training, everything you could possibly want. It’s both pretty effective and really expensive. And I always think it’s funny because the one from the original study published in Science, the one country where it didn’t work well is Guatemala. And it’s specifically because all the chickens died.

Literally, this is a paper published in Science. It’s a five-year study with at least one person who now has a Nobel Prize. And the success or failure comes down to whether they chose the right breed of chickens. There’s a lot of risk when we make these decisions for people.

Kelsey Piper: Cash programs in the US do not have nearly as much of an effect as we see in poor countries, and you just observed that in poor countries, maybe we do the best with these lump sums, and no one has tried giving Americans lump sums. This would be the equivalent of $100,000 or something. Do you think that’s the only reason why we’re not seeing results in US cash programs, that lump sums are better than increasing consumption somewhat?

Caitlin Tulloch: I don’t think you’d see quite as strong results because at a certain point you’re dealing with a selection effect where if you’re in an economy in which there are or have been more opportunities, people who are going to have the easiest time getting out of poverty will have already started to get out of poverty. That may be changing as time goes on as some of the social safety net erodes. But in general, the people you’re finding in the US who are still in pretty extreme poverty are there for overlapping reasons that cash alone may not solve.

Kelsey Piper: All right, so then since we’ve just got a little bit of time, I’d like to hear last takeaways from everybody. Mine is that the thing that’s going to happen by default is absolutely not going to be just for billions of people. I am also not sure it’ll be great for people even in rich countries, but there’s certainly more forces pushing towards justice in that case than in the broader global case. It seems difficult, but not impossible, to do anything in advance that changes that. But I’d be super excited if somebody had concrete plans there.

Caitlin Tulloch: We’ve talked a huge amount about the impacts of the concentration of wealth that AI will create. I think some of that is perpetuated when a huge amount of the focus in the community is on applications that are relevant for the people in our community, the people in the country we live in. I like this idea of riffing on the Green Revolution and saying, “What would it have looked like to actually make that more broad-based even than it was? To make that more holistic?” And I think that should really be a challenge. I’m from Oakland, and I’ve lived in San Francisco long enough to say that the amount of techno-optimism that assumes AI is going to help everybody is slowly going to give me an ulcer. But there’s so much interesting stuff to dig into that I think actually these kinds of tools can address. Being so much more curious and creative there could go a long way.

Nick Allardice: I think I’ve been a little bit of a pessimist on this stage, but I actually have quite a lot of optimism right now because, and again, I will be just honest, I’m very focused on the floor of extreme poverty. I do care about the other problems, but they’re second- and third-order problems for me. And I actually think that it’s more possible than it’s ever been to raise that floor. It’s not going to happen by default, but, you know, we are running a grand experiment at the moment of trying to do a large lump-sum, capital-like transfer to every single person in the country of Malawi, and it’s going to cost us $5 billion to do. We’ve raised $250 million to do it. We’re marching our way down that pathway. This is a time in human history where that is possible, because of technology and because of the amount of capital that is available. This just wasn’t true 10, 15, 20 years ago. And so the default pathway here does have a lot of downsides, but I think there are possibilities that give me a lot of hope for raising the floor of human existence in a way that I think we should all be very excited by.

Saffron Huang: One of my takeaways is similar to Caitlin’s, which is about AI literacy. It’s not that everyone needs to use AI all the time, but it’s really important that people are informed and know where it’s useful and know where it’s not useful for them. This is for democratic reasons. I think people being well-informed means that more people can have a say in what’s going on. But it’s also so that more people can take advantage of it financially, and also knowing where it’s not useful, so that we can kind of all collectively sense-make about the whole thing. So yeah, I think literacy is a huge part of it. I don’t think we’ve necessarily talked about that as much today, but I wanted to bring that up.

Kelsey Piper: All right. Well, thank you all so much. Thank you all for coming out here. I think that’s the end of the panel part, and now we send into the crowds and chat with everybody.