#247 – Anton Leicht on how middle powers avoid losing everything in a post-AI world

It feels to me that you could pitch this to the US's. We're going to create fast follower open source model. So such that the best open source model in the world is not Chinese. This is a good thing. We're not trying to eat your lunch. Like why not go for that strategy? If you think the fast follower can actually steadily stay like four months behind, you do two things to the Americans. You put them on a clock for every defensive operation because you tell them, well, whatever crazy dangerous stuff you come up with, in four months, everyone in the world is going to have every terrorist, every non-state actor, every re-organation in the world. Also, you have these like $1 trillion companies that base their entire sort of revenue strategy on like selling these like exclusive frontier models. And you only ever get four months to sell them because after that we'll just undercut them with an available open source model. I'm not quite sure you could sell the Americans that are being a great idea. Today I'm speaking with Anton Light, who is a fellow at the Carnegie Endowment for International Peace, and writer of the excellent policy blog Threading the Needle. Among like many different strengths, I think one of Anton's core appeals is he has this gift to write things that somehow sound both secret and obvious. Thanks for coming on the show, Anton. Thank you so much for having me. Why do you think that by default, countries that are not the US and China look like they have a pretty bleak future? I think when we talk about AI policy, usually we talk about sort of minimizing the risk while capturing all the benefits. I think the default outcome for basically most countries in the world that don't build their own frontier AI models is the exact opposite, which is they capture all the risk that come from these models going through society, and yet they minimize all the benefits they could potentially have from AI happening. Okay, what can middle powers do to avoid this future? Well, I think find some stake and find some participation, not only in this broader AI supply chain, but in what you expect the sort of world order with advanced AI to look like. Because, you know, Anthropic didn't promise us just geniuses in the data center. They said they would cure cancer. So how do you do that? You actually have to sort of, you actually have to test the cures you make. You actually have to like build the production plans to actually make the medicine and so on. The same thing goes for robots. The same thing goes for the same thing goes for all the military assets that you think you might want to produce with that for like US China competition. Lots of middle powers, especially the Europeans and South Koreans and Japanese are pretty good at occupying these bottlenecks. And I think if you have something like that and if you can integrate it into the AI supply chain sort of sufficiently effectively, then that is just sort of long term asset that gives you some economic contribution. And that creates some minimal mutual dependency. There is a very realistic sort of second best slash worst equilibrium where the middle powers are sufficiently useless allies that don't really build up the capabilities that would feed into this broader like USAI model supply chain that would give them this edge, but instead decide to focus on whatever they're like. They're on models or shutting off their markets from American models or just like moving away from this mutually beneficial arrangement. And I think on a very high strategic level, what the goal here should be is to avoid the second best equilibrium and just like move toward the actually sort of best like broad alliance, widespread supply chain integration, the middle powers do what they're good at, the US does what it's good at and then we sort of combine that. But it's not obvious and I think there are a lot of factors pulling away from that and into the worst outcome. What policies do you pursue right now to not waste that window? Like it's kind of easy for out in Japan, South Korea, with the Netherlands, but what if you're not one of those countries, what are you doing right now to make sure you're moving towards that partnership future? I think there are a lot of narrow policy interventions that are very sort of country-specific, right? You want to think about very specifically integrating parts of your industry with American development in a way that is mutually beneficial, right? The American labs want close corporations with legacy industries for a bunch of reasons between like financing and just data and just ability to like very deeply integrate into like broader supply chains that actually deliver on like the things they said they built. I think these partnerships are attractive to the American labs. And then I think there's like the much broader question of, well, what are the actual input assets you need as a middle power right now to enable sort of going down that path in the future as AI models become more important? And I think the most important thing for that is just access to the frontier models. You just need sufficiently good access to sufficiently good AI to then think about how to structure your administration and your business and even your consumer market around that. For securing that access, there are not that many options. I think the best general version we've come up with is compute for access, which is you build out data centers to alleviate the inference crunch and the sort of meet the demand for inference that these leading labs have. And in return, you ask them for guaranteed access to frontier models at least at parity with the US commercial market. Has anyone done this yet? Like have no way in the UAE do you think they've got anything in return for their domestic compute build out? I think those are different cases that were in a position to ask for different things. I think the UAE deals were in many ways an up-and-down for the UAE where they couldn't ask for much. I think it was already quite a stretch for the UAE to even convince the administration that sort of negotiated this deal to even let them export the chips to the UAE as is. And so I think that was already pretty contentious. So I think they weren't in a position to ask for much more. I think Stargate Norway, which is now like a Microsoft build out instead of an actual Stargate because Open AI has partly gone back on some of that, had the beginnings of that. But I also think the negotiator's around that also proceeded this like acute need for frontier access and guaranteeing it. So I think this is not yet part of any of the deal frameworks that are around. I do expect this to be part of the deal frameworks that come up as like bigger data center build apps in middle powers get negotiated. But it is an open question whether that works because you know the draw for the frontier labs in building these data centers anywhere else that isn't the US is well can you get them up faster and can you get them up at least like somewhat efficiently and then the geopolitical considerations of well maybe the labs would like to reduce political concentration in the US maybe they would like to reduce the geopolitical risk that comes was just like being completely sort of having all their infant supply in just one country. Maybe that like to reduce the domestic political risk of like local political backlash against data centers like national political backlash against the AI industry. They probably like to diversify out of that a little bit but they can really only afford to like diversify out of it if it doesn't take like four years to build a data center right. I think the benchmark here is something like 18 to 24 months realistically for a big data center. And so if countries can offer something like the gigawatts built in something like one to two years. I think that's a great base. I think if you pitch that to a lab and say well with like access parity in response you can probably make that deal work. Are they striking these deals with the labs directly or through US government? How do you mission this plan? So I think this is all. Is it George Moore's? They're calling on somewhat somewhat hypothetical right? I think generally what we've seen is that direct deals with the US government are both somewhat volatile and somewhat susceptible to sort of spreading out horizontally in a way that isn't great for the host countries that we've seen the UK US tech prosperity deal. There was in theory a model for some like deep and tech cooperation that also include a data center build up as one pillar of the deal. That didn't work very well all things considered because then next last December the US brought like food safety standards into the negotiations and then open my eyes sort of in the start. It's aspect of it because it's citing like copyright and sort of energy price reasons. And so that all didn't really work out the way they would hope would have hoped. And I think there's the more general risk of well if you strike a deal with the US government then the US can sort of go back on that deal if you do things that the US government doesn't like in any other domain. And that's just that doesn't sound that great if you're a middle power. I think making the deals with labs is easier both because the lab demand for diversifying the inference compute is more acute. And because I think the better way of political influence in some way and of like making sure that the access restrictions aren't as hard and making sure that US government doesn't intervene in your access parity is incentivizing the labs to themselves lobby the US government. So if you have a lab that has guaranteed you access parity and the lab has put in writing well you gave us this data center we're now giving your market the access to the same models that the US market gets. I think that's the realistic case right I think the government contract parity is a lot harder to negotiate for because that's national security relevant. But if we just talk about market access I think this is a perfectly reasonable ask right like if Fable 5 is on the US market it also comes on the UK market also comes to the European market comes on any market that has also come up with the data center for Anthropic right that's the like basic deal structure. And then if the US government decides to come up with some access restriction then the American firms that are involved in these deals and that have a stake in these deals and that have invested a lot of their money into these deals now have an incentive to tell the US government that you'd have Nvidia and Anthropic and ideally sort of other labs as well that have maybe even better standing with US government sort of going to the government saying well we actually have all these contracts and all these like gigawatts of inference compute on another countries can you guys like not restrict access in this way can you at least give a cover to this country that we have a data center in or can you make it make cover for allies only or something because we'd really like to not default on our contracts with them and we'd really like to sort of honor these agreements so we keep access to this inference and so you get a much stronger advocate in DC. But does that count for that much though because I feel like wired I think we're reporting that the Fable export control was imported by them giving methos access to I think SK Telecom the South Korean company which the admin alleged was like had ties to China or something like it just feels like the labs especially in the scenarios where things go crazy and where you most want access to the inference compute the labs will be powerless if the if US government decides that they should just pull the compute like how do you make sure that deal's not reneged on well so first of all I think the Fable thing is I think we're adhering in this conversation just because there is no there is no restricted access for Fable right there is just no access for Fable and so I think Fable doesn't violate access parity in a strict sense right you can't have any less access to Fable than the US market no market is access to Fable so this is perfectly complied with with access parity in some way I think the specific scenario that you want to avoid is you're saying well we're fine with you guys giving access to the US market, but not with giving access to other markets. And I think these are largely not the scenarios where they're like super strong and super restrictive security concerns because then why would you even give access to the US market? This is mostly an insurance against the economic divergence specifically, right? Because what you are worried about on the economic divergence is one market has access to a button in front of TI and another market doesn't have that. And I think for these cases, the security concerns are never all that high because if you are giving access to the entire US market to all US firms that you aren't really doing like super intense KYC and like very high defense contractor style security checks for, then like it can't be that bad, right? So that's not the situation where things are going crazy. That's not the thing that acts commercial access parity and see its against. If we're talking about like glass wing level like access parity or even like government access parity, I think for that, only the inference compute incentive doesn't work out quite as much. I think for that, you also need cooperation on security architectures, which is to say if I am South Korea, for example, and I want to be included in something like project glass wing, then the firms that I propose, my project glass wing is extended to have to pass some security standards that will probably be handed down from the US government. And these security standards will be about cybersecurity architectures, they will also be about like standard, like just like general security of the firms, and they will also be about China ties inevitably. And I think that will have to be part of the corporation if we think about like closed access parity or even like sort of government level access parity, you just have to come up with some shared amount of security standard and only firms and only governments and only agencies that clear the security standard get access parity. And I think for that, the inference leverage still matters a little bit because I think giving and expanding access once this security is guaranteed is not that big of an ask. Like these caravans aren't particularly rare to get. And I think it's not particularly difficult for like to lobby US government to come up with like narrow carvarts for allies and you just have to create some margin incentive for anyone in the room being interested in doing that. So I think the hedge against that is just being like, well, the Australians or the UK or the Japanese or whatever, like we know that this is a pretty secure country. We're not that worried about the about the security implications of exporting to them or expanding to them. You just meet someone in the room to think, well, maybe we don't screw over our allies in this in this moment. And I think for that, you just see the marginal incentive of the inference thing, for example. But most of it is about the security cooperation instead. There's still films that feels like a really fragile position to be in. Like if all you know, there's all these problems where you need to build loads of compute and obviously, you know, your citizens are going to be unhappy with that anyway because that's going to suck for their life and you need to depend on the US government, you know, being reliable or dependable. Like why not just build your own frontier model? Like, I don't think the best option. I mean, why does building compute suck so much? I think it's perfectly fine for the citizens, especially if we get the hyperscale is to pay for it. And I think there is a realistic chance we do that. Like we just, I mean, a data center isn't all that bad for most of the citizens. It's just like a box with a computer in it that sits somewhere in your country's cape and like Amazon paid for it. Okay, but you're also going to have to relax a lot of regular regulations if you want to be at all competitive with like time to power and like the cost and all these kinds of things. Yeah, but I think you need to be hard to sell. You need to be hard to sell to the like, yeah, maybe you need to like narrow carveouts and yes, you get like some ideological group supposed to, but I think you don't have to give up on your entire regulatory framework. I think there is a way to sell it. That's just what I'm saying. But yes, I think that's right. Building your own frontier model is just really extremely hard and extremely expensive. I think, I mean, if they really wanted to go for it, I think there is a way to do it. I wrote about this recently. I sort of did some educated guesses on how much that might cost. I think realistically, if you wanted for like a four year project split between different middle powers that would work together to throw in the talent at the compute and the funding and the supply chain leverage and everything you would need to come up with an actually competitive way to build a frontier frontier frontier model. It's been like 500 billion over five years or four years. That's a lot of money. I think if you go into a room right now, especially if you go into room in a treasury in a middle power country and not in a digital ministry and tell them you'd like to have 500 billion dollars to build a highly speculative enterprise, invest into American infrastructure, build huge data centers and then try to train a model on them and you don't have a business case for where you'd sell that model. You don't have a clear way to get the investment back and this entire sort of approach is highly vulnerable to both domestic political pressure and to US coercion to stop doing this. I think there is probably no treasury in the world right now that would tell you that they'd be happily by their share and so I just don't quite know how you pay for it. I don't quite know how you do it. There's a long list of political challenges to it. I will say, if there was sufficient motivation to do it and if there was sufficient funding to do it, typhoon-e-billion isn't actually that much money or ex of the g7, sort of Europe plus Canada plus the UK plus Australia plus Japan plus South Korea. If you could get the best version of the sort of like allied liberal democracies sort of alliance set up and you go all of them to pitch in, they could afford 500 billion for that kind of project. It's not an absurd idea sort of on paper and I think if you're sufficiently bought into this specific idea that the frontier matters a lot, I think this is the only realistic play because you're right. If the frontier matters as much as we've sort of speculated it might and if the Americans are as volatile as some people fear they might be especially after the fable incident, then I think any other conclusion really doesn't work that much. They're all hedges and they're all bets that at some point the Americans will be wrong or at some point the commodification will kick in or at some point the fast followers kind of stay around or maybe the sort of the econ just works out in a way where you capture a lot of the revenue downstream. This all might still happen. But any other sovereignty strategy that is not building a frontier model is basically a bet on one of these things. I think you can make enough hedge bets that you probably end up fine. But if you actually want to be certain that you don't get screwed over by a frontier model progress specifically and by sort of super effective like AGI systems taking over the world, then you do need your own AGI. And that's the only way to do it. Why not pitch this? So such that you know the best open source model in the world is not Chinese. Why not make go for that strategy? Why say we're going to deliberately go for the frontier itself? So first of all, I don't think the Americans will be particularly happy. I mean, this is in a very, very real way does eat the lunch of the Americans, right? Because if you think the fast follower can actually steadily stay like four months behind, you do two things to the Americans. You put them on a clock for every defensive operation because you tell them well, whatever crazy dangerous stuff you come up with, in four months everyone in the world is going to happen. Every terrorist, every non-state actor, every sort of, you know, every rogue nation in the world, everyone is going to get access to whatever danger this capability is. Like if you're just like actually committing to just, you know, throwing resources into having a fast follower that open sources things, I think that doesn't sound great to the Americans, not only because of the security part, but also just because of the commercial partners like, well, you have these like one trillion dollar companies that base their entire sort of revenue strategy and like selling these exclusive frontier models. So I'm not quite sure you could sell the Americans that are being a great idea. I'm also just not sure it actually works because what does it actually mean to build a fast follower? I think there's basically two versions of saying like this is a fast follower. The one is you just have to, it's just code for like, we are building like a worse frontier model builder. Like someone who's like just like four months worse at building a frontier model, but there's all of it independently. So they invest almost the same amount of money, almost the same amount of talent. They're just like four months slower. So they never get any sort of exclusive window where they get to capitalize on a lead. No one ever has any incentive to go for their models instead of any other models. So it's just a money drain where you just spend a lot of money to be worse at something than ever. Then someone else and there is no real niche for that. That you can always do. And then there's fast follower in the way that is actually dependent on frontier progress. Right? Where you're like, you're following and you're not only sort of in parallel moving up the same gradient, but what you're doing is you're following because you know what is happening at the frontier. And because you can emulate some of the things happening at the frontier, you can maybe even distill some of the things happening at the frontier. And that requires the frontier to be available to be able to fast-follow it and just not so sure that that's going to remain the case. And then I think the final question about the past follower is like, why do you think that helps in the specific scenario? I think what we've talked about until now, why you need the computer access strategy, why you need these bottlenecks, was the specific scenario where you were very worried about this format gap that the Americans have to use their frontier model very well, where you said, well, we also want the same exclusive access to this frontier model. And we think that if you give any other nation a reliable sort of four months, six months, I think it might realistically rather be 12 months lead, then they will use that to generate a compounding advantage in R&D and software development in AI progress. And all these domains that these models might be good at that eventually disempower us relatively. If you think that that's the case, then that the sort of the fast-follow doesn't actually help you. In the limit, in this scenario, you actually need the frontier model itself. And I think then you actually have to go all in on buying the resources you need for that and actually developing the model itself. If the middle power is the site, they want to go for a building in front of your AI project themselves. What would that project actually look like? I think what you do is you set up the biggest possible thing you can. You don't start small, you don't start with giving money to any existing institution. What you do is you build an actual coalition that has enough money, that has enough leverage, that has enough talent, and that has enough general geopolitical heft to actually back this up. You get all of them to put in their money right away and they commit their money on day one of the project. They don't get it back under any scenario. They can't affect. You put it all in one pot. And then what you do with that is you fund a private company-like vehicle that is structured just like a frontier lab. This is the thing that builds your model and that builds your frontier system. And then this system, this vehicle only corresponds with the funding governments and only takes instructions through a middle layer of people that get what's going on. Every country that is involved in this coalition, designates like a sort of this is like our Zhar for like the project and this is the guy who like talks to the board of the project or talks to the founders of the project and communicates what we say in political language to them in technical language and what the project says in technical language sort of goes back to the coalition in sort of you know in police. critical language translated by the ZARS, I think this is the structure you have, this is what you throw all your money into. And then this vehicle just gets like maximum freedom and maximum sort of buy-in to actually do its thing in the way that we know that works, which is like the structure of a normal frontier app, which is you pay a lot of very, very, very good engineers from other labs, a lot of money. So they do work on foundation model research that is powered by a large, large, large amount of compute. And so I think you don't need to reinvent the wheel. In fact, you mustn't reinvent the wheel on any of this because there is only one way we know that works how to build a frontier model. And we just like exactly replicate that. We give them all the resources they need. We stay out of their hair as much as we possibly can. And we just sort of let them cook, we buy in as much as we can. And I think the more people you get into this, the more like countries you get into cooperating on this, the broader the coalition you get that actually buys into this, the better that is, just because you will need a lot of geopolitical resistance to American pressure, you will need a lot of funding to buy in. So I think the broader you go, as far as you can get the buy-in, the better that is. And do we just expect the countries involved to accept that this will just be a huge call sink for a couple of years? What I expect is that this does not happen. Yeah, well, but again, I think this is one of the reasons. I think there are smaller scale versions of this where basically countries have been okay with making this kind of bed, like Airbus is one obvious example. I think framed and understood as this is just a critical strategic capability we need to have, this would not be understood by countries as like a money sink or not a money sink. Like a war plane is also a money sink. A war plane doesn't make its money back. It's something you buy so you have it. And I think in some way, this sort of AGI project would be the same thing, which is tough swallow because obviously the Frontier Labs in the US also make a lot of money, but they have to justify their cost to like private market investors, right? As opposed to governments, it's just like the American capital market that is invested into the Frontier Labs and soon the sort of broader US and global stock market that's going to be invested into them. Whereas the project needs to be funded by treasuries mostly because the private capital doesn't exist in the rest of the world. So that's a bit of a disanalogy. So you do have to make the strategic asset pitch first and foremost, but that said, like if once you have developed a Frontier model as a strategic asset, that is also something you can probably sell. Like you know, you can do just the government procurement of this Frontier model and just like people buying the Frontier model like governments buying the Frontier model and like domestic firms buying the Frontier model surely gives you some revenue back. And then I think there's some arguments you made why the European project if successful would be somewhat competitive in broader markets, right? I think there's like some concerns about American preparedness that might make people like less excited about buying American models for all this. - You still buy from this world? - Yeah, yeah. Missile still has like a decent American market share. And so I think in international market you could think about competing in some ways. I'm pessimistic about that because I think you can perhaps brute force your way to building a good Frontier model, building a good software product company that actually manages to sell a lot of its product has been historically almost harder for Europe to do than it has been to build big infrastructure projects. So that's gonna be really tricky. So I think to some extent you should expect this to be a money sink for some time, but you know, so are militaries. And I think they still sort of in some way justify themselves. - You give me a way. - Yeah, yeah, yeah. I don't think anyone ever looks at like, well, you know, this aircraft carrier is just sitting there and it's not making any money, why don't we sort of rent it out and why are we so bad at like productizing aircraft carriers. They're not supposed to productize. They're national security assets and that's what we pay for them. And I think you would have to pitch the coalition on a similar logic for why they should go for the project. Who do you think should be in the project? - So, I think should be in the project is the EU, the UK, Canada, Australia, and New Zealand, South Korea, and Japan. And if you can get Taiwan into that, that's even better, but I think that's really stretching the limits of how realistic this is. I think the general logic here is, somewhat aligned liberal democracies, like fairly like even sort of like, you know, value floor for sort of coalition of the willing, good talent, good sort of compute bottleneck, that will help you in terms of leverage just enough capital to build the actual thing. I think that's the ideal version. I think realistically you're not going to get that because I think there are strong incentives to defect from this coalition. And I think countries that still have a very good bilateral relationship with US, or that still think that their bilateral relationship with US is their best play to sort of secure access to models and to secure participation in this economic order and basically the way that we've just described, right? People that have like sort of countries that have the best access to this allied scale play, why would they instead sort of throw in with the European Union that has like time and time, again, like failed at building the things that announced it would build? That's a pretty hard pitch to make, to kind of hit like the UK, to Korea, to Japan, where currently still feeling pretty good about their bilateral all things considered. So I think realistically you have to cut those out and hope that they may be joined later if things deteriorate further, or maybe they join if the project seems more, seems to be going well, but I think for the time being, you probably start with Canada, as much of Europe as you can get, you try to get Australian, as well, and then you start from there. How do we get the US from blowing this up? Like, why wouldn't they just be extremely antagonized by this attempt at a project? They would be, this is part of the reason why it's very difficult to pull this off. I think if you were in a completely free and open-air fair market and you had abundant access to every US asset and no US retaliation to fear, this would not be as difficult. But the thing is, A, you do risk just like sort of broader retaliation from the US, just on whatever the US might choose to do. It's tariff on a related things, seizing intelligence cooperation on all kinds of issues. Like, hitting the security architecture of, especially the Europeans, hard, for example, in the conscious of Ukraine, in the context of intelligence sharing. There are a lot of broader things the US could do to just stop you from doing things that have very little to do with the project or with building AI itself, and they just have a lot more to do with, whatever other issue the US has leveraged. So this horizontal escalation is the first thing to worry about. And I think there is no clever 300 AI policy intervention that deals with that. What you need to deal with that is just, you just need to be generally more sort of strategically powerful, you need generally more geopolitical leverage. This is an open problem for a lot of countries in the world when they're going to solve it in the specific AI policy conversation. But then there are two specific AI-related threats that the US has against anyone trying to build a frontier model outside of the US. The first is you need US chips to build this project, to build this frontier model. And the second is you probably need at least to start with using US coding agents to build this model. I think much of the talent base of a frontier lab today are already coding agents. And the best coding agents by far in the world are the US coding agents. If you actually poach a bunch of the US researchers, if you put them into a lab in Europe, if you give them all the compute they want, and you tell them to start building a frontier model, but you don't give them like codecs or cloud code, they're going to be pretty annoyed. And they're going to be a lot slower at building these models. And as the labs keep using their best version of their internal coding models, and the project still doesn't have access to even basic versions of the coding agents. The gap just widens even more. And the more pronounced this gap is at the start of the project, the harder it is to navigate it off the ground. So you do need access to some version of the coding agents ideally, which is something that both the US and the US labs can restrict. So you have to find a way to maybe do like a carve out for like some coding agent use at the very start of your project. And then your boot strap very quickly to your own model. And you build your own coding agent scaffold around it. And you just like, you just have that be your first step of building. And you just like hope that that works well enough that you can eventually, that you eventually not that far behind the coding agents. And maybe you can get a lab to agree to that. Maybe you can get a lab to sort of not take you very seriously and think, well, the Europeans are never going to get anyway. So we might as well like sell them like, you know, like 3000 codex seats for like a billion euros or something. And then you just like, you spend a lot of money on getting coding agent access as long as you possibly can. But it is going to be hard. You could try to build around like Chinese open source coding agents in that case and try to bootstrap from that. But that's, I think that makes it even more of an Apple Battle. And the second thing that's even worse are the chips. Like, you need, I think that my very like general guess is, you know, something like three millions of black world class chips for building this thing. That's a lot of chips. There aren't even that many chips just to buy at the moment. Like Nvidia supply is pretty limited. So you would actually need to sort of offer sufficient money and buy out contracts and give sufficiently good incentives to Nvidia to actually want to sell you these chips. That's not trivial, but that's the easy part because Nvidia is interested in having lots of different. They love selling chips. Yeah, yeah. They love money. But what they specifically love is selling chips to a bunch of different people. They don't actually want to be reliant on just like one or two buyers. And so I think they generally be excited about selling a lot of chips to this project. I think the US would be somewhat less excited about Nvidia selling chips to this project. There is this argument that, well, there's selling chips to China. Surely they would also be interested in selling chips to the Europeans. I think that's sadly not really true, not to get into the export control discussion itself. But I think this specific style of argument only applies to China because if you're not export the chips to Europe, they're not going to build their own chips in five years. They're just not going to have any chips. So it's it's a lot less geopolitical risky to not export the chips to Europe. And it's a lot more geopolitical risky to export the chips to a project that actually says we're going to reach front your parity. I think the basic motivation is just well, they just said that like, Fable and the cybersecurity capabilities are so scary that they don't want it on the open market because they're scared it's going to get jail broken. So why exactly would they look at a European project that says, well, we're going to build our own Fable and then we're going to use it for whatever the hell we want. And you guys can't control it anymore and then export the chips to them instead. I think that's just very, very unrealistic in the long term. So you do need some leverage to secure access to both the coding and the chips. I think there are ways to do that. I think the main way to do that is to just use the semiconductor supply chain links that these countries currently have. So there are the most obviously the thing everyone always talks about is a sml in the Netherlands. If you can get Japan and South Korea on board as well, you get like some hype up with memory, you get some more semiconductor manufacturing equipment into that. You get all these things that Nvidia and the US are very reliant on to like eventually like build the chips they need and sort of get the chips online they need, there's probably some dealing and threatening to be done around that. Because there is eventually, like, immediately beneficial setup to be had here, right? The European's keep expanding the sort of the coalition keeps expanding its supply of these semiconductor manufacturing, manufacturing equipment and all these upstream inputs, continues giving them exclusively to the US and the US turns them into chips and then in return, the coalition gets to buy some of these chips for its own use. I think that's basically a pretty positive outcome. I think it's less good for the US than the outcome where the coalition gives them all the manufacturing equipment and then the US just uses all the chips it builds with that just for itself. But I think it's a better outcome for the US than if the coalition just decides to turn off the supply, to limit the supply or even the limit starts threatening to export some of the supply to China. But I think there are things that you can do as a coalition to threaten around the supply of the semiconductor manufacturing things that don't only target the US but that also target Nvidia. So they just create incentive on the US and Nvidia side to keep the favorite vocabulary of exclusive access to these semiconductor assets and then in return they give you some of the chips. And I think if you're sufficiently coordinated to actually play hardball on some of the stuff, I think there is an incentive and the path of least resistance for the Americans to keep giving you the chips. And then realistically, if you look at what the American appraisal of like sort of, you know, middle power government led project to build AGI would be I don't think a lot of people in DC or SF would take this very seriously to the extent that they think we have to stop the set or costs. I think they will realistically think this would probably go the way of all the failed like catch our projects before. So they might not just might just not care that much about stopping the chip exports early. And then by the time that the chips are online and the project is going and things are actually going well, it might be too late to claw things back and then the leverage just needs to get you over this like initial phase where you buy and get online the chips. What would convince you that okay, it's go time like let's actually let's actually pursue the project now rather than you know settle for compute access deals or anything else. So I think there are two parts here. The first part is political willingness to do this and the other part is just like government government and policymaker interest in doing this. I think right now I'm not sold that this is the right thing to do just because I still am a big fan of the sort of allied scale integration comparative advantage. It just appeals to the sort of efficiency maximizing part of my brain very much to just like have the yes do what it's good at and then have the allies do what they're good at and then we put all that together and then that's just the best possible outcome that we come up with for sort of an alliance wide like post agi economic and such as structure. I just very much still like to do that and I feel like it's not bad yet and I think what is needed now is for the allies to make a sufficient contribution to incentivize that and then for the US to respond rationally to the incentives that are set by actually sort of effective allies that are doing well at building out these capacities. So I would very much still like to try that and I would very much still like to see what that works and sort of cutting off the bilateral before we've really even tried that would be pretty sad. That said like things are going a lot faster than I thought. I thought we'd have another year to set up this compute for access stuff at least and get some of these deals online get some of these corporations online. The fable thing even if it is a red herring has shaken up a lot of people and has made a lot of people more worried about this and it has at least like now very much made this idea of export controlling models with its weights or with its outputs a much more realistic prospect in the policy discussion in DC. So the conversation on that is moving a lot faster than I would have hoped for setting up this mutually beneficial alliance. I think that's one of the problems and then the other problem is like well how much political awareness do you really have to get started on this? And I think realistically the political awareness is always going to lag what is going to be necessary by at least a few months. That's a problem right? Like right now I've like written this like I've written this piece from the perspective of like this week. I think if we started tomorrow we'd need the five hundred billion. If we started in like six months even and if we like if you know we started a big tour through all the capital's now and we met with all the 300 ministers and like some somehow sold them to take like a hundred billion of sovereign debt each and put it into a project that they don't know will work which is pretty big ask. If we only took six months for actually authorizing this budget and getting the project started I think we already have to increase the price by like at least a hundred billion again because it gets harder and harder. It gets harder and harder to compete for fewer and fewer chips that are going to be more and more shipped by then it's going to be harder and harder to convince the labs to give you access to somewhat leading coding agents to build things as the labs pull further and further and further ahead and it's going to be harder to avoid US retaliation as US political awareness of this risk and of sort of the strategic relevance of these systems keeps and keeps increasing. So five hundred billion is a price tag now. If we convince people to do it we're going to have to tell them in six months that well now it costs 600 billion then we have to do another round of negotiations and by the time we've sold them on getting 600 billion online we're going to need that petroleum and you know I think there is just a very real way in which the political awareness is just lagging the necessary moves too much and I think that is part of why I think it's probably just not going to happen it's probably not going to work and what would convince me on the political side is well people actually want to do this and what would convince me on the strategic/tactical side is well the compute for access of allied scale the type of thing is not actually working the US doesn't want to go for it the US stops these deals from happening the US restrict the outflow of chips even to sort of allied data and this kind of thing so if our play failed add the political awareness majesty fixes itself we should go for this until that happens I think I'm still fine or at least hopeful to do the alternative plane instead. Okay I mean one question is why is this so catastrophic so the US has you know Frontier AI before the rest of the world does but the US already outmatches you know most countries on most domains like and they're doing kind of fine like why is this so bleak for France or England or anywhere else well I think the question is what can you do with like two or three or six monthly time like all the secrets of the world right I think if you just have the cyber capabilities now the effect is somewhat limited but you can already you could already see the Europeans freaking out about this and I think right be so and then I think we expect models to be able to do more things well than just cyber right and I think a lot of them could matter a lot economically I think there's an open question here which domains specifically in which domain specific differentiators that much but I think R&D is one of the other obvious examples right if you have a few months lead time on innovation and you find out what material you're supposed to use a few months before the euros do and then the euros find out as well and they build their thing and like sort of it takes 24 months for anything to get off the production line anyway for anything to get shipped anyway so what really what what do the few months matter I think that's a realistic near-term future where the lag is just not all that relevant for the R&D stuff for example but the faster the world moves and the more integrated these systems get the more automated manufacturing get maybe there is ways to sort of lock in these innovations in a way so if you for example you can like register patents for all the innovation that your that your AI finds before the euros get around to them or you just get a quicker and quicker and quicker iteration cycles on building things and testing based on them and the faster the world starts moving as we integrate these systems and as we automate more and more things the more you will expect these initial advantages sort of compound because then you build the better thing then you check how well it works if you build the better thing then you try to you feed the data back into the system and you try what the system makes of that and then you feed in the innovation again and again and so you just get to start the flywheel and you get to start the exponential of this innovation just a few months earlier and you're always a few months earlier so your product is always better by a few months and I think if the world just moves sufficiently fast for that that does compound into pretty substantial advantage but but I mean the world you paint where okay if most of the you know marginal returns on intelligence are in areas like cyber or R&D then I mean Europeans just get access to better and better products from American firms or purchasing power probably goes up if these products are also you know being able to be produced at like cheaper levels again it doesn't sound catastrophic to me like what's bad about that situation I mean I think this is not catastrophic in the near term I think just general increasing economic growth and general increasing innovation and productivity is usually usually starts out being pretty good for the world in an open market I think there are two conversations the first conversation is well how does that lead to adverse economic effects and the other question is how does that lead to sort of bit by bit relative disempowerment and I think the the economic part of it is not as catastrophic as you say I think the relative disempowerment stuff is a problem so these frontier AI systems themselves become a really important input into your economy you would think that they pull away further and further and further from the AI systems you yourself built in your country for the same reason that the R&D effects compound the software R&D and the AI R&D effects also compound right the Americans not only build better models they also get much better at building their own AI models because they can use their own AI to both sort of broadly improve their broader AI supply chain from everything from my chips to data to everything and because if there's a little more narrow recursive self-improvement effects that might start within the labs so between those you just expect the American software to pull further and further away why is the open source software like open source models not compounding at the same time at the same rate I mean a you already have a growing divergence in compute that just means that the American labs can move faster can explore more new innovation can sort of you know can just spend a lot more compute on building better models and then be because there are these recursive effects inside the labs they can reinvest all the revenue they make into building more infrastructure and then their models get even better they can reinvest all the usage data they get into making even better models and of course they can just use their superior models to apply them eternally use them to make their models better but then prevent other competitors from using the models in the same way we already see this with Fable which is limited for a frontier model research specifically which is I think something that we're going to see more broadly because why would you build a tool that helps everyone else make models that this ruptum. your own business model, I think at some point they will stop doing that. And then I think the question is and the challenge is exactly the other way around. It's like, why on earth would you expect this format gap and open source to remain in place forever? I think the specific mechanism that has led to this in the past was just there was a lag because you could see where the frontier was and the frontier was sort of moving, four months ahead, and you could see, well, what's the shape of the frontier? What kind of model are they building at the frontier? How quickly is it moving? What are the capabilities they're focusing on? What is this specific that what are the specific capabilities they're throwing all there are resources on what benchmark shape comes out when they build a new frontier model. And from that you consider guests like what are we exactly chasing here? What are we going for? What are the next architectural breakthroughs that we have to find as a fast follower and so on to say nothing of the fact that I think a lot of fast following is also about like outright distillation, which is even harder to do. And so if you get to this point where the frontier a pulls away even further and also b becomes a lot less visible because some of it gets restricted, then it's just a lot harder for a fast follower to even know what they're supposed to be doing. It's just like find the road to the frontier. And if that's the case and they also have less and less compute relatively to the frontier, I would just expect that gap to up further and further. And so I do think we should expect the rest of the world not to have their own fast followers that can really keep up. So the frontier models stay relevant input to these economies. So you have a lot of reliance from that because you need at least the frontier or the sub frontier models from America as just like your basic economic input. But you also just need all the AI produced goods that they're giving you right to the extent that your economic mechanism is well, we'll be fine because we have something to offer to the Americans like, you know, like the Amalfi coast or like sort of Melanie suits or like whatever you might come up with, right? Or like maybe even like sort of very sophisticated machinery that integrates the AI very well. There is still asymmetric leverage, right? And the US could cut your off at much lower cost to the US than you could sort of be able to threaten the US. And who's to say I don't know, you can go into like pretty crazy futures from there on. And I think at the limit, there is just very few things that you as a relatively disempowered country can't really do to stop this stop this power concentrated country from doing basically whatever it wants. Maybe land on earth becomes very scarce and they decide they really want more land on earth. And then what do you really do about that? You know, maybe they would like your country as a military base on the way to somewhere else. And then you can't really say anything about that. Maybe they want a natural resource you have. And then you can't really say all that much about that either. I think if the relative leverage becomes like pronounced enough, then at some point it's really difficult to look at these European countries and still think that there are sovereign nations in a relevant sense. And I think that is a realistic threat if you just have a relative growth divergence. Even if the Europeans are fine on an absolute level as you describe, I think at some point of divergence, the relevant, the sort of the relative geopolitical leverage is just bad enough that that is still a pretty existential threat. Let's go back to the world where you don't go for the project, but you have insured continued access to the frontier models. In that world, how is your economy capturing value when most of the labor is done by AI's created by American AI firms? I mean, maybe the labor is just sort of commodified in a way that makes the labor not matter all that much, right? I think the Asians are just like one input into your economy in that world. Maybe it's a high revenue input, maybe it's a low revenue input, maybe it's a high margin input, maybe it's a low margin input, but they surely are going to be other things that are also important to contribute to this, right? The Asians do the labor, but what do they need? They probably need at least for the start, like a lot of sort of proprietary data that you feed into them. And as they continue to do the work, they probably want to interface with sort of, you know, with production that generates more data as they try things out. Like for like the sort of very classic loop that you might imagine here is you have like sort of a supermodular super automated sort of factory that builds whatever like, you know, let's just say it builds robots and it's sort of connected to US frontier model and the sort of idea like this, like US frontier data center here, you have this like robotics manufacturing plant here and sort of sort of corresponding coordinate very closely and like you so the model comes up with, well, this is how we should ideally sort of buy this component, build this component and then some ideas of the of the AI sort of implemented in the factory and then like a lot of sensors and a lot of monitoring in the factories of captures how that exactly happens. So it feeds back into the model and tells the model where this has worked, this hasn't worked. The model comes up with another way to optimize the supply chain. It sort of feeds back into factory. The factory sort of corresponds to that and sort of makes the modifications and then it's to feed back in and at the end you have to like very good robots coming out of this process and that's just obviously a valuable economic asset. And the question is like, how big is the role that the AI company has played in that and how big is the role that the manufacturing has played in that? And I think it turns out it's almost as hard to build a factory that can do this as it is to build an AI model that can do that. And I think a lot of the world is going to bottleneck on factories that can do this because you need a lot of like inputs for that, right? You need the factory to be staffed by people that are good enough at like making the modifications at heart that the system is telling you to you need factory that is like good enough on like sensors and all these like reporting loops that it can actually give information on like how well the AI ideas are working back to the AI. This is just like a very rare asset and very important asset. And I think if you have that, it probably generates a lot of revenue for you. This is like the idealized version of like industry that is like highly AGI relevant that is integrated into a lab. But I think you can come up with a bunch of other small ways to do just the same thing. Like people that adjust like, you know, that adjust like providing the way that they provide services, the way that they advertise services based on their integration with AI apps in the same way. You can come up with things that are maybe not quite as AGI pulled as robotics, but they are still just important to build. Maybe it's drones. Maybe it's just like random hardware devices that are used to, you know, whatever you want to do, right? You know, basically any industrial output in the world, you can integrate into AI in the same way. And it turns out these industrial assets could just turn out to be an important bottleneck. And I think the same story in the study broader sense can be told for a lot of other downstream bottlenecks that can be integrated with AI. And that maybe just captures a lot of the value. And then there are a lot of things that just don't interface with AI that much that also capture a lot of the value. This goes back to the balance stuff that we've talked about, right? I think like once you have your economy isn't super disrupted, once your country is somewhat stable, once your government is sort of doing fine, you can just come up with things that are scarce as that becomes important. You can create products that people still want. You can sort of, you know, you can go back to the tourism and the music and the sort of and the artisanal goods and all these things. Like all these things still capture substantial value. I will say it's sometimes overrated, the extent to which that is important. Even in countries that we think of as very tourism heavy, tourism really makes up more than like 10% of their GDP. So it turns out you just like, you just need to do all the legacy things that your economy is still doing in some ways. And I think that's going to be somewhat hard in the limit. I think there is just a lot of like a services industry that is susceptible to being displaced by AI eventually. And I think there are the question of like saving some of your production and saving some of your economic activity is more about comparative advantage. Like why exactly would it be worth it for AI agents to completely replace like some random aspect of like the services industry in like Germany or in Greece or whatever instead of investing like this margin in France budget you have, this margin are in do you have like why are you spending this on like displacing like random German services industry instead of doing things that the countries of geniuses in data center like crazy innovation that could come up with. It's just very unclear that that's the thing you primarily do. And so I think a lot of this is also just a bit on the division of labor shaking out that way. And some of these activities still staying around. And I think in the limit if they all get displaced and if all the service industries just get done by like cheap and abandoned AI, I think as there is no good answer to that. But then there is also no good answer to what the US economy does all day. Yes, they probably get like they get to tax the agents or whatever. So it's not as like economically catastrophic for them. But as sort of as a matter of like economic structure disruption, I think if the eyes just do all the jobs, there is no good answer to what does your economy do because current you kind of was still do hinge on doing all the jobs. Okay, what policies do you think this nets out in right now for middle powers? So like if what you want is to own the robot factory, does that mean Germany should be banning baselesses new company from like buying out new germ like all German action firms like definitely, definitely screen for the FDI on because I think like this is a very risky situation right now where I think a lot of outside companies are starting and like you're a lot of US based actress specifically are starting to become aware of how important the bottlenecks are. But the countries in the rest of the world haven't internalized how valuable and how important these things are just yet. So there is a lot of incentive for outside is to come in and buy up all these bottlenecks and then sort of you know, productively redeploy them once the assets and have become more important. The baseless fund that pervious thing that you've mentioned is I think one obvious example like they just buy all the industry capacity and then they sort of redeployed when AI systems are stronger and you know, obviously you shouldn't sell these things right now. Obviously they're just like you know, if you think about them now and if you think about selling them now like a prize that you know, something like the Bzos fund would be able to offer or like something that like basically any sort of investor that would be interested in running this play would be able to offer sounds great if you're not sufficiently aware of what is coming in AI you get like a few billion for an asset that you thought was only worth a few hundred million the current economy that sounds crazy good to yourself as data. Well, this is just a windfall right you know, you're a family set for life that sounds amazing. But it turns out like sort of alternatively this would have been like your slice of the light. This is like a super important piece of like manufacturing. This would have been like what if they sort of secured your stake in the AI economy for like decades to come and it would have been so worth so much more than a few billion and this sort of awareness gap and how much these assets are worth is a very very threatening phase of the sort of AI economy transition right now. So yes, I think you just need to get ahead of that. You need to screen for like foreign investments into this. You need to make sure that your important assets aren't taken over. I think that's like the very baseline of what you should be doing as a country. I think you can also think about how you can make these assets AI compatible because like as I said before, I think the risk here is it's just become so hard to work for example with the manufacturing that is like AI compatible that it's just more efficient to just rebuild this thing in the US to begin with. And I think to make it efficient and almost sort of like bindingly interesting to the AI labs to cooperate with your robotics faculty. some where else in the world rather than just rebuild in the US is to just make it AI compatible in all the ways we just went through. So make it sufficiently modulated and quickly react to what the AI Amput is. Like, have workers around that is sufficiently skilled to adjust the production process to whatever the new idea of how we're doing this now is. Like, actually manage to not sort of regulate yourself out of providing the data to the US developers in a way that creates these fast feedback groups. Make sure that the plans are sufficiently set up with like, you know, censoring equipment and all this kind of stuff that actually reports the data back like there are AGI compatible production plans for a bunch of things you can build right now that are extremely enticing and attractive for cooperation both on a geopolitical level for the US government and on a commercial level for the AI labs. If you make these changes now, you have a very, very good position to then ask for these corporations and integrate. If you don't do that right now and you just have a basically defunct like plant somewhere that is like still staffed by people that don't really know what they're doing that isn't really sort of modernized in ways. Then the pitch becomes a lot harder and then it might the only thing you might be able to do is sell some of your IP you still have sell the like physical plants and then have the US take them over reintegrate them eventually indigenize them. And so I think the way you stay alive is just anticipating where this future is going and making sure that whatever your production process is actually becomes compatible with that. I think that on the specific industrial policy stuff is I think the most obvious intervention. This is the sort of micro example for the broader economic structure because that's the sort of cleanest and direct most direct sort of integration example. I think one slightly more abstract version of this is what you do about your labor market. I think currently labor markets in a lot of middle power aren't very well set up to deal with like disruption and quick iteration based on what AI systems do to the labor market. To the extent that there is a future labor market that still includes humans in important roles along like bottlenecks and things that AI can't do very well and you have this idea of this like very jagged frontier and the human sort of fill out the spots where like the AI isn't particularly good and the AI does the thing that the eyes very good at like moving into the sort of dips in the jagged frontier of AI capabilities requires you labor market to be extremely flexible right. If your labor market isn't very flexible then people just stay in their jobs even if the eyes better at them at their jobs and then you have huge efficiency losses people don't move into the things that they would otherwise be good at quickly enough your firms become uncompetitive because you're not doing a big particularly sort of efficient distribution of the workforce across the task profiles that your human workers would actually be good at and eventually you just risk being disrupted in a few ways like either. So fast competitors that are just like integrating AI agent sort of entirely sort of vertically just start disrupting you because you are just not using AI to its full effects and then at some point just like these like quick and more agile competitors just start out competing you and then your entire business model crumbles and then you actually just like you have to do mass layoffs your company just goes under things good really bad or you just get a scenario where well other countries in the world that are better at sort of having a more flexible labor market that are allocating their workers along the jagged frontier more efficiently. These firms just get more efficient and then they out compete internationally and then you have the same effect your firm gets under pressure maybe gets out competed maybe you get forced to just put a higher and higher trade barriers to compete to protect that firm but either those are pretty bad outcomes so what you need to do is you need to make sure you're a market is flexible enough that the workers can move to the sections of the jagged frontier where humans will still be required so you get the general economic efficiency gains and you don't expose your workforce to the broader international disruptions from AI agent. Just being deployed everywhere and that's a pretty counter to finding right because in a world where agents get really good and you get really worried about the workforce displacement the first thing you might want to do is you know lock up the labor market and make sure people can't get fired for now and make sure that sort of people stay in their place and stay in their positions and make sure that the disruption doesn't get too bad too quickly because that's politically really risky that socially really sort of disruptive and sort of you know that that is sort of having like mass layoffs and having people go to like 6% unemployment. For some time like 7, 8, 9, 10% unemployment like some of the labs are calling that is extremely politically scary but what's even more scary is well we lock this labor market in place we stay at 5% unemployment for another year or two until our firms go under then they do mass layoffs then we end up in 15% unemployment but it's also the case that like none of these workers can really reallocate into more useful jobs after that. And so I think you have to buy the bullet on the count and two different conclusion of make your labor market more flexible as the address comes in otherwise you risk sort of going to the second best like very inefficient structure in the short term and then get really disrupted in the long term I think maybe one framing to think about this is that this is sort of race between automation and augmentation where like you have this you have this threat of this impending automation like labs are building these agents they might be sort of you might use these agents to build like entirely vertically integrated firms that need very few. And what you want to do is you want your firms that still employ a lot of humans to stay competitive with these vertically integrated sort of very agent heavy firms and you need to do that by moving towards the most efficient like distribution of labor between human workers and agents and if you don't do that you lose the race between the augmented firm and the automated firm the automated firm wins and you have no use for the workers anymore and so rather try to win the augmentation automation race try to incentivize a flexible labor market that makes humans move into the market. This humans move into the parts of the giant frontier that are actually like human favorite and then you can deal with some of the disruption. And how do we how do we do this without causing too much social unrest like is there anything else we can offer well I think you just have to buy off the losers of this of this trend basically I think this is going to lead to localized disruptions there are like inevitably going to be people who are currently very good at the things that agents are going to be very good in the future and that are only good at these parts of the narrow task profile they're not very good at like selling software then that maybe they don't have like lots of great research. But they're very good at executing on building software and these people you need to compensate in some way you just need to buy them off basically like these are the people that generate your social unrest it's not trivial like how you would buy people off it's not like it's not entirely clear like you can do something like wage guarantees and wage insurance is like has been done in like past sort of globalization sharks digitalization sharks to economies that has worked. OK, but it still has created these pockets of political resentment you can't actually get around with it just by giving people money I think some people have floated the idea of giving them job guarantees instead and just ensuring that these jobs stay around but basically just subsidizing them into existence. I think in the short term that is not the worst idea but I think there are real limitations to how well that works in this specific case which is mostly because like you can do wage guarantees and you can do job guarantees fairly well if the only thing that people care about. Is there actual job as they have it right now but a lot of the disrupt that workers were talking about here are like in their early to mid 20s they they're not interested in their job because they want to do their specific job they're interested in their career trajectory and they want to get somewhere with this. And if you just give them like their wage and their job for ever now they're not going to like no one who's like 23 years old and like doing software engineers are going to do this forever like I just want to sit here and like execute doing this code no no I don't ever want to sort of time up the ladder I don't want to ever make more money just like giving my wage guarantee it's very different than like a 55 year old like sort of the guy who works in an industrial pan that you can just like sort of pay off and be like well you get the wage you got you can stay home for like most of the day you get like you can work for 20 hours like this is the sort of the deal that a lot of sort of western economies try to make with sort of displaced industrial workers it also didn't work great but it kind of worked because that those were like upwardly socially mobile workers that really wanted more out of their career. So I think the tricky thing here is you don't only need to buy off people's current state you need to give them something that is like sufficiently attractive to slot in for the career they used to have. I think one of the things we can do about this is hope that if you just keep them around by force with something like a junior job like subsidy which like incentivizing firms to keep them around on top of the AI agents they use. You just they just pick up enough along the way they just learn enough along the way that they eventually just become like you know they become good at other things and they become good at things that will be valuable in the sort of broader context of this like augmented economy and for that you just need to get them through these first few years where they probably don't have a lot of useful skills to offer and they probably primarily have things to offer that are good for agents so maybe you can brute force your way through that with some sort of subsidy I think that is a start and I think that gets rid of some other disruption. And then the other question is like how much of the disruption just happens as a result of generally bad vibes and I think that you just can't get around like if you have these AI systems entering the economy at the same time where you just have like bad macro economic indicators and bad macro economic sentiment and you just have like high is unemployment rate even if it's for unrelated reasons people are just going to hate this trend no matter how good or how bad the specific safety nets for the I disruptions are the vibes are going to be bad the stories are going to be bad there's going to be politically incentive to put it in the right way. And if that happens I think there's just not all that much you can do you have to try to counterbalance with some like social policy interventions you have to try to kind of balance with some economic you know yeah I think at the limit this is just going to take courageous like leaders that say well yes this is hard this is economic hard trip we're still going to have to go through this because if we don't do this the the equilibrium we get to if we don't do this is just even worse and so just you know gratitude this is pretty bad economic situation but this is the only path to a better economic situation and then you just hope that these leaders don't get sort of politically outcompeted by someone who has a better message of well no this guy's selling you guys out of the air companies will just go and shut down our market we're going to keep the air companies outside through these guys we're just going to do the firms I will always have and we're just going to like put frictions in we're going to like pass human in the loop laws we're going to pass laws against agent adoption we're going to like a particularly restrictive token tax that disincentivizes the agent update and you can come up with all these restrictive policies that ultimately just make you lose this automation augmentation race and mean you get like really disrupted and like two or three years, but these are going to be attractive political solutions and getting around them is just a matter of like beating them in the political arena. That's going to be hard, but it thinks it's going to be necessary. What's your preferred way to pay for these junior jobs such as subsidies or whatever else it is that you're offering? Like I think you said, you prefer corporation tax or something like a token tax. Yeah, so I think the general worry here is anything that narrowly targets AI use makes us just more sort of makes us do this automation augmentation race even faster, right? Because what you want is you want maximal uptake of the AI agents at the things that they're good at in the firms that in the legacy firms that exist right now. Because otherwise these firms get less and less efficient and ultimately do get displaced by either international national competitors. So anything like a token tax that basically tells a corporation, well, if you use more AI, we're going to text you more. I think that makes it very hard. And I think that just disincentivizes the kind of adoption we want. The other thing you might think about is like, can you tax the AI companies that the AI developers specifically? And I just think maybe that works. You can take the maximum version of that is taking the equity stake in the AI developers that your government has recently started to think about. Bernie Sanders has written about a lot. And then the president has also started looking into this, I suppose, and calling meetings on this and they have conversations on this. So we'll see what that goes. So I think it's a very bad idea. Because I mean, A, I think it's a bad idea for our narrow economic grounds, mostly because we're just not sure where that works. Like even if these AI companies become extremely valuable, like 10X or like 100X their value in the next few years, that's still not even remotely enough to pay for like large scale social disruptions. Like the economy, even the economy is a lot larger than the AI labs right now. The economy is still going to be a lot larger than even AI labs that become 10 times as big as they are right now. So whatever equity stake you have in these labs, it's not actually going to be enough to pay for any of this. So if it isn't enough to pay for any of this, like you still need alternative modes of taxation to pay for part of it. But then by taking the equity stake, you just create a massive misincent for the government in basically doing anything to the AI market and regulating AI in any way. I know we've mostly talked about the economic part here before, but we also do have like, we also do want AI models to be regulated in a reasonable way for like other sort of risk-related factors, obviously. And if the US Treasury and also just like general US economic incentive and maybe like a sort of Trump account filled with AI equity stakes or something is like directly tied to the stock of AI providers, like what kind of incentive does that place on the US government in terms of regulating models in terms of, you know, making sure these models are safe. You're going to give a lot of power to these labs to say, well, you can do this, but then like the Trump accounts are going to crash and how does that look politically? And I think you don't want to create this incentive basically ever. So I'm really happy. What about other countries? Well, I think the question is like, do you even get the equity stake? You can buy it. That's I think one way to do it. I think in the US, they're still thinking about just like seeding some of the stakes to the US government. I don't think any other country where any AI would be interested in like giving them stakes for free. So you just have to buy them. At this point, this is just a sort of, this is just a sovereign wealth fund with extra steps, but in the sovereign wealth fund that is like much less diversified than any other sovereign wealth fund. So yeah, this is like sort of the stupid eye only by a topic stock version of the sovereign wealth fund. And yeah, I think in the limit that works. That's hard to sell politically, right? Like it's already hard to sell the idea of we'll take on a lot of debt and build a sovereign wealth fund, even though it is a good idea because turns out stock markets are pretty good that they're generating a lot of revenue. But then maybe you just invest into the broader stock market and you just do an actual sovereign wealth fund instead and you probably put in the alives as a hedge. But then you need a pretty big sovereign wealth fund to pay for widespread like social disruptions like that. And so I think in the limit, you need something that captures where the AI growth accrues. I think it's pretty good guess that corporations, it's just like corporate income taxes, just like a pretty broad net that captures a lot of value accruesion and like all kinds of different steps of supply chain. Maybe it's the chips and then we then the tax hits Nvidia. Maybe it's the model development and then tax hits the model developers. Maybe it's the downstream users and then the tax hits the downstream users. But we get some of it anyway. We just increase the corporate income tax. That is what we have right now. That is how we capture I think a lot of the value. I think there are other ideas around that are even more elegant in some ways. I think a consumption tax, just like a very broad based consumption tax for whatever consumption is one of the other things that people have talked about. I just think I have, I just have a political issue with that which well, it works elegantly, but then people open their door dash app and they look at their sort of order and then there's like a one dollar consumption tax on like the breach of the order. And that's kind of aggressive in any way as well. Yeah, and I just like, I think that's just not a great vibe for a new tax and it makes it feel like the burden is on the consumers even though most of the consumption tax probably does fall on the companies instead. I think it's just politically more elegant to just campaign on this as well. It was raised to corporate income tax and that's just the thing. Isn't there a big risk of capital flight? Also, if you have like vertically integrated firms, they're not tied to a workforce in a particular country. Does that just make it capital like a lot more mobile than ever? How do you deal with that situation? Yeah, I think, I mean, as with many of these things, a lot of like the basic low-hanging fruit solutions that we already have for general issues in corporate tax and capital flight just become a little bit more important, you probably need like global or at least between alliances, you need sort of minimum floors that just work. This, we got pretty close to that. Like that seems like a tractable problem comparatively to me. Just like, you know, like a 10-15% like minimum corporate tax rate floor across like major allies that could potentially host this. I think these firms are mobile but they're not that mobile compared to just like a sort of post box somewhere because it's just, well, they do need a physical anchor. Like a firm that uses a bunch of AI agents needs like a data center that runs these agents somewhere, needs access to a data center that runs these agents somewhere. So the fact that, you know, the most disruptive firms, we would expect them to have like huge compute spends and the compute spends need to be anchored in a data center somewhere. You do get a revenue anchor for these firms in more general terms because they have to be linked to a data center somewhere and then they can't really escape the sort of gravity of this data center in terms of where they're located. But you also want, yeah. But you also want to incentivize them to build a data center in your country too. Like doesn't that cut against that somewhere? Well, that's great, right? I mean, you just, you know, I mean, you will probably be incentivized to give them some carve out to build a data center in your country. But it also means that they can never entirely run away from the tax. You get like minimal raise to the bottom dynamics as you always do with company headquarters. But that's still like we currently have that sitting and like the floor is yes, corporations don't pay as much taxes in the ideal scenario where they all sort of sit in like the high tax countries or whatever. But it's also not a pure raise to the bottom. It turns out not every company is like listed on the Cayman Islands because there is like there are a lot of a lot of factors that pull against it. And not that many countries can build a data center and not that many countries are willing to master the political energy and the capital backstops and everything to build huge data centers. So you will have the data centers, hopefully diversified across a spread of countries that isn't entirely incentivized to just put the corporate tax rate to zero, especially if you have global minimum floors. And then you can't actually escape the pull of the data centers having the data center give you some minimal tax revenue anchor. And the fact that this is also infrastructure bound means the capital can't actually fly away from the minimal tax rate anywhere. And so you would think that the economy just sort of like smoothly distributes across the countries that have data centers that can serve as revenue anchors. And I think things probably shake out somewhat fine. It's going to be hard to capture a lot of this value, especially if you compare this to labor taxes, right? It's just a very easy tax labor because laborers aren't particularly mobile. And because we just have pretty high taxes on labor, all things considered, it's like both direct income taxes, but then also all the costs that come with payroll, especially in some of your paying countries. So if we move a lot of the revenue away from labor and towards basically capital, which is to say AI agents and the infrastructure they run on, we do get much fewer taxes to talk about. It's just, I mean, you have rates of 25% maybe on corporate income and everything to do with the capital. We also give a lot of tax breaks to capital investments compared to labor investments. So we do lose, like if we just like naively move like one unit of revenue from labor to capital and AI agents, I think we lose maybe like half or more of the tax revenue we get off of that. So yes, there is going to be a massive revenue crisis for most treasures in the world. Yes, there is no good tool to think about this right now. And yes, like at minimum, corporate income tax changes are probably going to be necessary. At maximum, we have to basically rethink how we do tax revenue on like private sector activity, all things considered. And that is right now it all hinges on fairly high taxes and human labor that we are not going to be able to sustain within the US. To what extent or what kinds of nationalization do you think are in some way inevitable? Like how do you expect anthropic to be governed by the time Claude is running robot factories? Well, I think I think my main line scenario, especially on the current demonstration is there is no like there is no strong framework. Like there isn't like there isn't at any point like any or a bill that says, well, this is how we're soft nationalizing or like hard nationalizing anything. It's just you have slightly more embedded observers that are linked to the US government within the labs that get what's going on. Maybe this is like maybe these are like third party valuators that report to the US government. Maybe these are just like directly sort of placed assets from, for example, the intelligence community that just sit in the labs and monitor what's going on. And then you just have these feedback loops where the labs communicate more and more of what they're doing to the US government specifically. And then there are people in the US government that think about whether they like what's happening or not. And if they don't like what's happening, then like the executive calls up the lab and like says, well, have you guys considered not doing that? And what you'd like guys not like to do this differently? Because if you don't, then we have like all this range of like potential threats that we could come up with what we never quite do, whether that's export controls on the models or on the model way. So whether that's sort of defense production act related interventions into the data center or like compelling the labs to turn over their models or compelling the labs to like give information as to what they're what they're up to and what they're doing. Like you have all these theoretical policy backstops that are already available to the executive today. And I think most of the nationalization that happens is just going to be the admins that are vaguely gesturing at. Well, you guys don't want us to pull on any of the things that you're doing. these levers, right? So just maybe just like do what we want in some unspecified way. And then there's going to be a push and pull. Like, there are going to be sort of commercial forces pulling against them. The labs are going to say, well, we'd like to, but this is pretty bad for our market position. And then the admin is partly going to say, well, then maybe don't do the worst version of this. And we're going to see a lot of messy negotiations like around the fable thing happening time and time again. I don't think we're going to see a big framework for any of that. It's just going to be this the continuing push and pull of the admin threatening to use some of these tools. And then the labs just sort of, you know, are creating in some way in the lab, sort of doing what the government kind of doing, what the government wants in some way. And we just like sort of try to figure out what the balance of power there is. And then at some point, we're probably going to come to some sort of law, there is going to be executive orders that sort of like come up with some minimal version of this that are going to sort of somewhat like standardize some of these processes. But I think a lot of this is just going to be the soft back and forth between the admin and the labs. Under what conditions would you advocate for something like a pull or a slowdown? Are you excited by this idea? I'm not. I think there are two questions here. The first is what is the politically realistic version of a pause that we would see happening. And then there's the question of, well, what kind of policy intervention do we like in the abstract? And I think on the on the sort of politically realistic version, I think, especially if this comes from the sort of general, all well, this stuff is really dangerous. We got to like get a handle on this and we got to get all this coalition of like people who all vaguely don't like AI for somewhat unrelated and out correlated reasons, which is get all of them together and we'll build this constituency. They're going to ask and push for the pause and then we're going to do the pause. I think this is very likely to not come up with a very good pause idea. And I think you need a very good pause idea for a pause to work, right? If you just pause in the US in a sort of naive way, then of course, the first thing you do is you crash the stock market because that is just like, you know, super dependent on the infrastructure investments that would sort of run directly counter to a pause. It's contingent on the IPO expectations of the labs and everything. So this is like domestically, like the naive version is pretty harmful policy. And then you need all this international sophistication for the pause to not actually mean that all the compute reconsentrate somewhere else that like China doesn't catch up in terms of semiconductor manufacturing in the meantime. And then you know, just like builds the thing later on, like there are all these complications that require you to have like a very sophisticated treaty in place and not just the domestic pause. I think to their credit, a lot of the sort of most intelligent pause advocates know this and they have developed like treaty architectures that they'd like to see. But I think politically speaking, it's going to be very, very difficult to get this kind of treaty done, right? You have this broad collision of everyone who like doesn't like AI, you get into the boat, they're worried about the job impacts, they're worried about the environmental stuff. And then you get them to the point where they like pass a bill or where they have a majority for a bill that says, well, like shut it down, let's stop this like some vague sense of no more new data centers, like maybe that's like tear down some of the old data centers, no more new AI models. And then like the question is, does this coalition stick together to actually also negotiate the treaty to actually advocate like doing the most sophisticated version of this to keep compelling policy makers? No, no domestic pause is not enough, you now have to get under the treaty, you have to get the Chinese on board, you have to fly it there again, you have to do another round of negotiations. And if they if they're not going for it now, you have to give up more to get them into this treaty. Like, is this coalition really going to be interested in the nuances of like the safety relevant part of the pause treaty or is the broad coalition only there for like the vague like, you know, symbolic like policy of we're doing the pause and does the rest of the coalition jump off the ship at the moment that they've got sort of gotten the pause and then like, you know, then you just left like negotiating for the treaty with like, you have to get like 60 senators on board for like ratifying like some crazy treaty or yeah, and that's just isn't there a version of a pause which is a bit more like a essentially a slowdown where what you're really doing is just reallocating compute. So you're making commitments about reducing the amount of compute that you're using for R&D, increasing the amount of the compute you're using for inference, for state capacity or societal hardening or something like this. Like, is this the system you support more evenly? I think that's why we're getting to like, that's less of a pause, that's more a slowdown, and I think that's also like that is like sort of what what sophisticated vehicle would be like in the abstract, which is just like, let's find some elegant way to reduce the pace of furniture capability specifically and increase sort of, you know, increase adoption instead and increase rollout and societal hardening instead. I think that probably works pretty well if you can coordinate around it in some reliable way. I think there is still a geopolitical issue to that in that a symmetric pause in frontier capability development is not a symmetric pause in US China competition for AI because right now China doesn't actually do a lot of frontier development obviously. What they do do is a lot of diffusion and what they do is a lot of catching up on semiconductor manufacturing. So if you pause specifically frontier development, symmetrically throughout the world right now, and if you slow it down in some way, in some coordinated way, I think most of the Chinese ecosystem is extremely thankful for that because they can, the slower we play this, the quicker their semiconductor manufacturing catches up, the quicker their diffusion plays sort of work out through all of the economy. And then if you catch up pretty fast. Yeah, I mean, the question is like, how long is the slowdown? Like, what does the slowdown give you? Like, what do you think that sort of the question is like, how much would the like pause like three years ago when they asked for it for six months would have really helped you? We would have still been like, well, we really have built that much resilience. We'll really have in that time anticipated the kinds of things the systems can do now. We have passed the policies that would have worked now. I think that you have to do a pretty long slowdown to actually get on top of all the things and then also make the predictions of the things you would like to in the future get on top of. So like, the slowdown has to last for a while. And I think right now as it is, it would be an asymmetric sort of geopolitical asymmetric slowdown, which is an extremely difficult to ask to make of sort of to make of the US government in the context where the super, you know, they do see themselves in like a geopolitical competition. You might think that like AI is not only something that the US needs to sort of narrowly win on, but like AI is the advantage that the US has to like win the broader great power competition. And the frontier lead is like the one thing they really have ahead of China plus like the compute lead. And so if you're in institute policy that specifically hits the one thing that America is like very much in the lead on, I think that's a very hard geopolitical cell. So that's probably something that the US government is going to be very skeptical about. If that's something the US government is very skeptical about, then you'll need broader Chinese concessions to make up for that. Like for this to be a fair deal geopolitical speaking, you instead have to ask China to also slow down on similar doctrine and organization and on AI diffusion because that's how you get like a geopolitically even version of this case. And good luck getting China to do that, I think, and good like verifying sort of not only the fact that they're not using their frontier development very well, but verifying all these unrelated industrial policy efforts. I think like a lot of good work has been done on verification of like a narrow pause treaty or like a narrow slowdown treaty where it's like, well, can we have do we have some way to figure out whether the Chinese and the US are both like sticking to this like narrow idea of what you described, which is like we put more compute in the deployment, we put less compute in the training runs, we don't exceed certain sizes of training runs, we don't put that much into the RL and everything. I think maybe we can come up with a way of verifying that, but can you really verify like what kind of progress the Chinese semiconductor industry is making while the slowdown is happening and make sure that the geopolitical parity of the deal works out that way. I think that's a very hard case to make. I think sometimes the discussion about this underrates, like it basically just makes a case, oh well, the US would still be ahead and the US is ahead, then there's no way that China would catch up in that world. But I think that misunderstands the geopolitical dynamic here. The geopolitical dynamic is China is better at most other things that matter for great power competition right now. And the one asset that the US has right now is that basically much of its economy and also some of its strategic environment and an increasing amount of its strategic environment is leveraged that leveraged bet on AGI development. And if you slowdown specifically on the AGI development, that is just a huge asymmetric hit to the US. And if you ask that of the US, I think it's just not going to work. Okay, why not rally around a slightly more stupid, but maybe still appealing idea of just ban on superintelligence. There's not a pause, it's just like a permanent thing. And there's enough, this coalition is broad enough, the US is interested, China is interested. Yeah, I just honestly have no idea what that means. Yeah, I think that is just an incredibly under specified thing. Systems, that doesn't mean that systems aren't allowed to be like super intelligent at everything. No, alpha zero. So aren't anything because like systems are definitely super intelligent at some things right now. Like obviously they exceed human capabilities at some things. So like that, like narrow super intelligence bandings, like, you know, like silly. I mean, I don't want to like this is like the lazy version of this is heard, you want to bend like calculating or anything. But like even if you go into like slightly more like broader human domains, there are already like important task profiles, even like sort of strategic relevant task profiles, like sort of large scale data and image analysis and this kind of thing where AI systems are already super human. So we have narrow superintelligence even so strategically relevant tasks already. And we don't what we don't have yet is like a broad super intelligence better at humans and everything. So like, but how do you operationalize that? Well, you just keep one thing behind the giant frontier. Like it never gets quite as good as like sort of, I don't know, like coloring coloring books or something. And then it's not a super intelligence because it doesn't have the broad capability set. I think the operationalization of this is just very, very difficult. And I think you can just even under general super intelligence band, but like extremely dangerous sort of very jagged systems. And I'm not entirely sure that you're in a much better place with that. I think maybe the most charitable way, charitable way to just to sort of to operationalize something like, can we, is there something we can ban is maybe banning like closed RSI loops within now specifically, like basically ban anything that would realistically lead to like an uncontrollable and unsupervisedable software-only intelligence explosion or even that would just even ban anything that even if it's not a software-only that just like bands like closer and closer RSI loops that just make systems much better and much quicker than we used to do. How I think that's, I think there's something to like about this kind of policy in that it doesn't only ban a thing that seems very scary. There's like a more principled reason for why you're going to do this, which is like retain visibility into what these systems are doing. Like ever quicker RSI loops are just extremely difficult to monitor. It's very difficult to come up with a way that a government would actually have like reasonable insight into the risks that emerge from this as the RSI is getting faster and faster and faster. So I think, I think there's a good argument to be made that this should not be something with our current evaluation. infrastructure with a current oversight infrastructure that labs should really be doing at the moment. Again, this is not going to be as obvious and as easy of a cell because this is one of the like theories of victory for the USAI developers and therefore will also become one of the theories of victory for like broader US China competition for the US like closing this loop and actually making these systems go like sort of go exponential like capabilities much faster than they would they otherwise would is one of the few ways you actually translate like vague advances like vague few month advances in AI capabilities into like very broad geopolitical leverage by actually getting like super super super super intelligent systems that can like you know both carry out all these important tasks but then also maybe by themselves prove to be some version of a decisive advantage and convincing the US to not click the decisive strategic advantage tomorrow button is I think still going to be extremely difficult but yeah I think especially like sort of like the strong closed loop versions of our SI are maybe the most operation sort of the most sort of easily operationalized ways of thinking about what we should like ban or slow down or prevent and I think I'm sympathetic to doing that. How do you personally feel about the prospects for the American decisive strategic advantage like whether or not it's possible like is that something is that something that would actually be net positive for the world in your view? Yeah I think I'm still unsure about like what exactly a decisive strategic advantage is and I think you can come up with like softer and harder versions of it that work or do not work in the real world but they've taken the strong framing of it which is basically just like you basically get like unilateral power over like doing to any sort of strategic rival or any other country basically sort of enforcing your will on them in any little way. I think I'd feel a lot better about a Western alliance with this with a strategic advantage with the DSA then then just an unconstrained US. I think I still have I think a little bit more faith in the ultimate balance that will shake out between the US and the labs but I think there are a lot of ways that US decision-making around how to deploy this could go very badly. I think we've seen the administration's very volatile and how it chooses to deploy its assets. I wouldn't be entirely sure that it deploys this asset very well. I think we've also seen like unfortunate tendencies in the securitization of the labs that would lead to this like I think there are a lot of like unfortunate amalgamations between the labs and the US security apparatus specifically that you might come up with that you know are maybe favorable to some narrow version of American strategic interest as determined by whatever the political mood of the day is in the administration but that aren't like necessarily furthering the world very much. I think that being said if you think this advantage is something that is like slightly constrained within the context of both like a separation of power within America between both like you know the sort of classic like constitutional separation of power also the separation of power between like the private businesses that build these models and the government that sort of oversees how they're used and then also a broader sort of separation of power within a broader Western alliance that controls this advantage and there are some more like checks and balances on how exactly it's deployed and it's mostly like sort of you know it's mostly like an ultimate enforcement stick and to sort of like last piece of leverage that just like wins geopolitical competition. I think that is a much more tenable future that I'm much less worried about because then you get rid of all sort of you know you you still get all the positive effects which are like sort of you know you get like sort of you get a GI deployed or like ASI deployed in a way that is sort of ultimately backstopped and backed up by functional somewhat functional like liberal democracies that sort of control how it's used that seems pretty good then that seems like a better way to control how these systems are deployed than basically any other way we can come up with it also seems like broadly geopolitically good and also geopolitically stable for this advantage to to exist but to rest in the hand of like a broad coalition that has like a decent high bar of adversarily using it so the best version the alliance and separation of power constraint version I feel good about the narrow version where you have like a closer and closer amalgamation of labs and US government and then this like sort of you know this you know there's like big public private partnership or whatever of like labs and usg just like can use this on like a base notice I think that at rather not happen and from the perspective of like a risk from loss of control is there a reason to believe that middle power is being involved having access to compute is that going to be helpful from a safety perspective I always think so I think so my general view of this is what it takes for us to like lose control over these very powerful systems is fundamentally some people making like stupid decisions well just on the policy front I know that the technical problems are hard and I think the technical problems don't only come down to people being stupid I think that we will have to like actually figure out a lot of the technical stuff that we haven't figured out yet but on the policy side like we're choosing people from reducing mechahitler choosing yeah choosing to deploy a system that's like not save like giving it like sort of abandoned access to compute and the internet and all these things like all this sort of like classic sort of preconditions to loss of control they do also come down to just like bad policy decisions being made around both the governance of the developers and also the sort of governance of the specific systems so the question is like in what environment do you get the least density of like stupid decisions that can really square the wrongness and I think the chances are like I think whenever you concentrate power within like a single volatile actor the variance of the decision quality just obviously gets higher right it gets less constrained like you get like more daring sort of moonshots and like cool things you might deploy these systems with so the upside of just like letting like one guy run well with his eyes maybe even higher because it's like less constrained by process and whatever and you can just do like risky bets but the risk of just doing something monumentally stupid with the very intelligent system is also much lower so like by is much higher in way if you have the crazy guy and it's much lower if you have the constraint version so I think just like if we can narrow the option corridor a little bit by involving more people in the decisions of how these systems are deployed and more people can sort of unilaterally shut them down in some way or like unilaterally threaten to like blow up the structure that makes these systems possible by withdrawing a bunch of the compute withdrawing a bunch of the supply chain inputs or whatever like if there is more mutual control over what is happening with these systems I think the band of outcomes just gets a little bit more narrow and we avoid like the like the stupid policy decisions that seem to me to make up a lot of the policy risk for loss of control so I think just you know broadening the inputs just seems robustly good on that case and then on a much more general note I feel like just having middle powers in the room and just like having sort of geopolitical leverage more evenly distributed just is a less chaotic and volatile world in a lot of ways I think if the US pulls ahead but the US pulls ahead alone and the rest of the world sort of gets disrupted in all the ways that we've talked about the labor markets get disrupted the strategic position gets like seriously threatened the like people lose their jobs people get hit by AI misuse every day they start protesting on the street they start there's mass migration flows happening because they want to move to countries that actually have access to the defensive AI that keeps them out of that countries they get into conflicts because one side of the conflict has access to good AI systems and the other side doesn't have access to these systems then these like temporary technological advantages sort of re-open old conflicts and make for an opportune window for one country to strike against the other all these scenarios were like a world that doesn't have equitable and good and sort of mutually beneficial access to AI systems just becomes a volatile and chaotic and messy world and I think just from a very sort of general perspective that just doesn't sound like a world where we'll be more likely to make good technical and policy decisions than the world where we have you know pretty sort of complex and well functioning supply chains people are generally not entirely the sort of like discontent and disrupted by the situation there is less AI misuse happening the world just generally seems less like it's going to hell outside the borders of the US I think that is just a robustly it's just like if you just like think about which world would you rather make like the ASI policy in and will make the important decisions in in what world do you think the people in the room are going to respond to better incentives and what world do you think the decisions that are going to be made are going to be more sane and less reactive to a world in crisis then I think you just want the more stable world and I know that that's not the sort of like finished causal model that people usually like when they talk about these things and when they talk about like the impact that work specifically has you you really want this like very like step by step by step and like we build the compute in Japan and then therefore the ASI rather does this I can't give you that but I can't give you is just like well I can give you an offer to like build a much more stable world that does much better on like any basic categories of like what we would want the world to look like and what world we would think good decisions might be made in and then I put to you like I think that sounds like a much better world to make these important to Simpson and I think that's good enough reason to work on this. What are you doing in this world are you just going to be writing all the way up to the singularity like do you have a plan? No people keep reading. I think there is a real open question that anyone who's doing writing and work sort of in the outside game right now is facing which is like how is the outside game changing relative to the inside game where I think a lot more information that is basically only available within the labs and within the governments is becoming more and more relevant. The labs have exclusive access to the best frontier models and they're going to like the gap between internal and external models is maybe going to increase the information siloing is going to get worse and worse you would expect as more of these things enter like national security relevant domains. So I think fewer of the discussions are going to happen on the outside. Dean Ball had this piece I think it must have been two years ago now but the sort of Republic of Letters that we had when SB 1047 was being discussed and everyone was sort of writing their little essays and writing their little tweets and people were going back and forth and it was all an open discussion and like there was like a battle of ideas and you know you could make a real contribution by writing about this and saying things about this and one by one the people writing the letters are choosing the inside game instead and the Republic of Letters is sort of getting starved of both its core input but also of its core output in that the decisions are just moving further away from like the idealized discussion on the outside. So I think that does ask a a real question about the value of the outside game all of this. But I think there is also something pulling in the other direction, which is that this still needs doing. All of this needs explaining, all of these needs analysis by people that have sufficient access to the information by having the conversations with the people that are on the inside, but still not looking at this from the biased institutional view of anyone specific place. And it also just needs doing what is effective of what you need to make these points and these findings credible to the rest of the world. And I think, especially in the rest of the world where people are still skeptical of the narrative that might come from inside security has environments or inside the labs, there is real value in making this case and making this argument in the context of the outside game and not just in like a sort of privileged conversation between intelligence officials between different countries or in a pitch that reads like a marketing pitch made by a lab to a foreign government. And so I think, especially for the work I do, there is still real value in doing it from the outside. But yes, I think there is an ongoing trend that pulls against the outside game. And I think that is something to be not for myself, but for the ecosystem to be somewhat worried about and I think something to at least look at. Are you long term bullish or bearish on Germany? Well, I think bullish on the macroeconomic basics and on just like sort of, I think there is also the sort of regression to the mean argument to be made here in that this has long been a very successful economy and a very like very sort of at least in the context of the West been like obviously a particularly important country. And maybe you just think that like well, incidentally this is not going so well right now it used to go better 10 years ago, maybe we'll go better like 10 years in the future and maybe there's like this like the very high level historical view. And then the more you zoom into the details of what is actually happening the more worried you get about some of the specific policy features not to expand this to the general like domestic politics conversation, but just keeping it to the eye policy conversation. It's just Germany has a lot of great assets for doing well in AI in very general terms like very skilled workforce less service heavy economy than most of the other like you know most other compare point comies still a pretty good industrial base. I think a pretty good network of both bilateral and sort of organization based relationships that are I think the right set of relationships to have if you think that US is building a GI but you still need some counter leverage against that. I think all of these are a great position to be in and the only things that Germany needs to do is like make all the politically hard decisions to get back to like 3% economic 4% economic growth, flexible labor market all these things. So like in a way like if Germany only solved all of its other problems, it would be fine on AI, which is like a fairly contentious thing. There are a lot of countries that have just like bad assets for AI specifically and if they do well right now it doesn't mean they do well with AI like if you have like a very service like say like maybe like Singapore or Hong Kong or maybe even the UK if you have like a very service heavy economy that is very susceptible to disruption by agent specifically. It just seems much scarier than if you have economic assets that matter in the sort of broad set of AI futures. So a lot of the German like a lot of Germany I policy and I think a lot of European and policy in general is just much less about AI and much more about general policy. So in a way it's very easy to solve because you just have to do all the things that like hundreds of thousands of people are thinking and writing about anyway but in a way it's also much harder to solve right because you can't go there and you can't give them like a two page memo of like here are all like clever ways to like pull the ropes sideways come up with like all these like very specific interventions. And then you fix your AI policy and then things are good no you actually need to go through like the messy politics of doing the labor market reforms and sort of you know getting the sort of industrial based back on track and all these things and that's really hard in a way but that's you know that's at least a known and well scope challenge and I think if Germany can get a handle on that then it's in great position for doing the AI stuff well as well. And but TBD and honestly way out of my depth on whether that is going to happen and how that should happen but yeah. Well it's been a pleasure having you on Anton and we'll see you next time. Thank you so much for having me.