00:44:17 · Eric Seufert
[00:44:17] S2: Well, so that's where you actually do need the frontier. Right? Like coding tasks, like actually benefit from having the frontier. But if you're having customer support stuff, that doesn't need a frontier model. That could use a three or four release back model. And it'll be producing reliable results that you know work. You've measured, you tested, and you don't need to upgrade, like, the customer support or, like, the chatbot, you know, for customer support integration to the to to the the bleeding edge model every single time. But, like, coding, yeah, if you're actually using these models to build models, you probably want the best of the best. Right? And, like, what Meta is doing is really actually at the frontier with, like, integrating agents into the coding workflow. If you look at, like, their system they built called Confucius, it's like a self learning agent, like, that actually helps them deploy better. They've got a whole pipeline for data science and machine learning tasks that helps them decide, like, okay, which which which where should we even apply this? Where should we even do testing? Right? Because that's actually takes a lot of time. Like, you're just figuring out what kind of experiments, what kind of tests you wanna run. They built a whole pipeline around that that's all driven by agents. Right? So my sense is, like, there's where you want the the the top of the line, and maybe their own models don't perform best there. But, like, also, don't know how how excited investors are gonna get when you say that we cut expenses. I think they really need to see the revenue growth at the top line. And so my sense is, like, you get that you get more of that by just making the ads platform better, and they've done that. All they need to do is say, look. We can if they could forecast out that growth and say, look. We're really dedicated to this. This is what we're pointing everything at. My sense is you could get investors excited over time. You keep printing 33% or whatever every quarter, like, you're gonna get investors excited after some time. If you start saying we're gonna compete with Kalsi, we're gonna build an AI pendant hardware, like, they're not gonna get excited. They're gonna think you don't know what to do, and they're gonna think you've got all this compute capacity. You don't know how to use it.
00:45:57 · Jordi Hays
[00:45:57] S0: Help help us understand the the prediction markets play. Is that I think the the answer that we landed on was it is just in Meta's nature to copy the new hot thing. So regardless of what it is, we're just gonna do it. Right? And you can see the history of, like, these shots on goal with, like, every new hot thing in consumer. They just build a version of it or they try to buy it. So I think that's the most simple explanation. John was trying to explain it as like maybe there's some way to do it with like, you're there there there is no dollars and it's just for like social status, you know, who who can be, you know, an oracle. I think that that was we got more news that maybe showed that that wasn't But the then when you look at again, look at the market, like, you look at the total market for, like, gambling and again, even if they got a meaningful amount of that market, they would not really move the needle in the way they need to on the core business. And they would invite all these new regulators that are now saying like, you're not only trying to harvest, you know, my teenager's attention and making them, you know, sad about their life, you're also getting them Like it just feels like it opens up this huge can of worms for no reason.
00:47:07 · Eric Seufert, John Coogan
[00:47:07] S2: It's just a totally misdirected move. Is that really the space you wanna go into right now? That is so politically fraught. That is such a political hot potato. Why do you even wanna touch that? I would steer clear that I would say, look. Facebook abs is time well spent. You connect with friends. All this gambling, you're gonna get addicted to that. Don't spend time there. Spend time in Instagram. It's more wholesome. Why are you gonna touch that at all? Are you just gonna open the door to more scrutiny? That's insane. I really have no idea why they're even talking about that. It doesn't make any sense. But they've got, like know, look. They said we're gonna we're gonna publish a lot more apps. They've got the new Pocket app. I mean, it seems kinda interesting. Maybe some of this stuff sticks. I think they could just try a bunch of stuff. But why would you touch the most politically toxic area right now, like, when you could just be touching anything else? [00:47:49] S1: Yeah. Take us through the Prosperous Society. I wanted that's why originally why I wanted to bring you on. I want the thesis. I wanna dig into it, because I I I found the it's a three hour, four part podcast series. You've written a lot about it, but, introduce it for those who haven't been following along.
00:48:05 · Eric Seufert
[00:48:05] S2: Yeah. Prosper Society has started out I wanted to write, an economic bull case for AI. We've heard all of the, you know, bear cases. We've heard all of the, you know, doom narratives around the around the economy. It's gonna just basically displace all white collar work. You know, you're gonna get DoorDash created with the vibe coding session and so all these companies are gonna go out of business. And I wanna make the case that it's probably not gonna that's probably not gonna happen. And, actually, there's lot of reasons to be optimistic. Right? And I think, you know, in writing that, it ended up becoming you know, we live in a in a sort of, like, very pivotal moment, I think. You know, if you look at the elections, you know, in the sort of the the house primaries in New York, you look at what's happening with the the, you know, the New York City mayoral election, you look at what's happening in the LA mayoral election, like, there's there's this this this sort of moment where these sort of, like, impulses against capitalism have become a lot more popular. Right? And, like, there's reasons for that, and, you know, I you know, I'm I'm not an expert on those reasons, so I won't delve into them. But I think, you know, ultimate's a mistake to go down that path. And the thing is, like, my sense is a lot of the AI or the anti AI narratives are actually have nothing to do with AI. Right? That's just seen as an avatar or like a bogeyman for capitalism. And so what I wanted to do is sort of, like, anchor this economic defense of AI, this economic bull case of AI in, you know, the sort of like liberal tradition of of of the Western world. And and and in in doing so, like, you could say, like, you know, you could anchor it to these these sort of like these these great thinkers, you know, this sort of like enlightenment thinkers and and sort of the economic giants that have built, you know, built this sort of intellectual framework that our that our western civilization is is based upon. And people could say, look. Well, look. You've you've misinterpreted them and so go, okay. Well, maybe. But but if that's not the case, then they you're gonna make them drop the mask. Mhmm. You're gonna make them drop the mask and say, no. That's not my problem with AI. I just don't wanna live in a liberal economic society based on these western thinkers.
00:49:53 · Eric Seufert
And I think if you actually kind of force that to be articulated out loud, you make a lot of progress against the anti AI narratives. But the whole point of the Prosperous Society is my sense is, you know, a lot of these AI investments, they're going to push, the economic constraints, away from production and towards just distribution. Right? They're gonna make distribution the binding constraint. And so because you just have this this this flourishing of of content creation. And and and and so when that becomes the actual problem with distribution and these AI investments go into things like ads platforms, digital advertising, you know, REXIST, recommend recommendation systems, then actually commerce, the economy, becomes much more efficient. Mhmm. And it's a really good thing, and you generate a lot of value by pushing that binding constraint to the distribution layer. Mhmm. And you get then as a result, you get a lot more heterogeneous product development because you can actually reach those people economically. Right? So, like, what is the constraint now? It's like, well, can I reach a big audience? Right. Right? Can I reach a big enough audience to support a business because, well, I've just got this kinda, like, blunt tool? But if these AI investments go to making recommendation systems better, digital ad systems better, reaching these these these these pockets of people that have these very specific interests that were totally un unserved before, then you enable a lot more commerce. Right? Yeah. And and then and then, you know, you support this, like, flourishing of people making this wide diverse variety of goods. You get rid of this idea of, like, the Pareto principle. We have
00:51:12 · Josh Lindgren, Eric Seufert
[00:51:12] S7: to serve, [00:51:12] S2: you know, the the 20% that supply 80% of the commerce. Well, no. Now you can serve everybody, and they can pay what they're willing to pay for these things. You reduce consumer surplus sorry. You reduce consumer surplus, you just you create this flourishing of of of everyone getting exactly what they want. Yep. Right? And that's the prosperous society. And so the way I frame it is kind of a reaction or, like, call call it a conversation with John Kenneth Galbraith. He wrote, The Affluent Society. Right? But this was written in the postwar economy. It serves kind of as the degrowth or handbook. And my sense is like, you know, John Kenneth Galbraith is a brilliant man. I'm not saying he's wrong or he was wrong when he wrote [00:51:46] S7: the book, but I'm saying