Martin K

A one-day summit focusing on the theme of d/acc: emphasizing the values of decentralization, democracy, differential accelerated progress, and defensive tech including crypto security, public epistemics, bio defense,  neurotech/longevity, decentralized ai and physical resilience.

Breakout 3

Prediction market panel

 Hello everyone and can you hear me? Awesome. I'm Vaughan Mackenzie-Landell. I'm one of the co-founders of Butter. We've been working in the governance space for a very long time and Obviously very excited about futarki and prediction markets and that's why we're all here We have this incredible panel full of lots of people who my love and admire And I'm gonna just hand over to each of them to give a very short introduction and then we'll kick off Yeah, I'm Martin. So in a previous live venosis, we started with prediction markets and Wanted to implement food hierarchy and other things. So in general, I'm absolutely a super big fan of that, but I think in this panel I will a little bit play the skeptic here because I think as Robin mentioned, you have the idea and then you have all the details that need to be figured out. So I think I can a little bit bring in that perspective. I just talked, so I'm Robin Hanson. So I'm Calvin, also CAS.if. I'm the co-founder of Utarki.fi. We just announced our project a few days ago. Our plan is to help DAOs and other organizations make great, impactful, positive decisions. Hi, everyone. I'm CJ from Limitless. We're building, like, basically we're processing around a million dollars a day in bets right now for, like, zero-day contracts on financial markets. I believe that believe that like with crypto, we can build the world's largest economy on chain outside the jurisdiction of any one particular nation state. And I think that as well as that, we need to build really efficient global marketplaces on top of the infrastructure. And so that's what we're doing at Limitless Labs. Thanks, guys. Really appreciate that. So we're here at DIAC today. And so the first question is, how do prediction markets tied to DIAC, how do prediction markets and epistemic tech make the biggest impact in the vein of DIAC on differential progress, defense, and democracy? Over to any one of the panelists. One ring to rule them all. If you could have reliable prediction markets on whatever policy question you have, then you can just do better on all of your other areas of life. You could better know how to pursue longevity, better know how to pursue decentralization. Whatever it is you're trying to do, having better information on that can make you do that better. So could we do much better? And how does that map directly to, say for example, defense? I can just replicate that. I think the promise is very clear. So the promise is whether we we talk about Even simple things like this community node or so from micro decisions whether to show that community node or not to show that to Extremely large macro macro decisions should we raise the interest rate or should we make this huge? Well defense spending here or there. In general, the promise is that this is a tool for more robust decision-making to bring in, yeah, to have a better quality of bringing information that actually shows the true impact of a decision. Yeah, so I'm going to make a comment here. So over the last few weeks, everybody has seen a lot of people are talking about prediction markets, of course predicting who's going to win the election, and this generated a lot of excitement. But actually, the most interesting thing is not only predicting probabilities, that's what normal prediction markets do, but predicting the consequences of things. So if you could have, for instance, conditional markets on inflation, on GDP growth, on spending, on unemployment, conditional on candidates, right? This would be even much more interesting for democracy. You could actually see like a scorecard for the candidates, right? Like a game, you can pick one. This one's better on this one. This one's better on this other thing, right? So this is the promise of using conditional markets for decision-making. Yeah. Sorry, Monik, go ahead. Maybe let me throw in one kind of concept to challenge this idea. So to say, I mean, I'm just pulling up here CoinGecko and look at prices of crypto. And, well, they change by 6%, 7% a day. So I think prediction markets rely on the assumption that markets are so efficient that out of those prices, we can get reliable information. So I think prediction markets rely on the assumption that markets are so efficient that out of those prices we can get reliable information or better information. But I would argue well there is also a lot of noise in markets and I would say that markets here, prices change by six percent is largely not related to any meaningful signal, but is... Well, I need to be careful in the statement, but there's still a lot of noise. So that's, I guess, my question. How can... Is there enough signal in the noise? So when you compare head to head this mechanism to other ones, consistently this wins. So that means all those other mechanisms have more than 6% mistake. Think about any committee you've ever been on. Think about the gossip network. Do you really think that's within 6% of the truth? Come on. Like being 6% close is great compared to what we usually have. Yeah, I would, I guess I would come with the, like the whole question from, from a little bit of like a different angle. I think that definitely prediction markets have a lot of power in terms of forecasting and decision making. But I think in a crypto context, what's really interesting and powerful is the fact that essentially we can build these global efficient marketplaces. And so I really believe that markets in general accelerate human progress. If it wasn't for markets, we wouldn't be where we were today in almost every field. And even they have positive trickle-down effects to science funding, for example. Brian Armstrong sells his Coinbase stock to fund moonshots. And science funding, that's a trivial example in a global context, but still kind of proves the point. And I think that building prediction markets onchain can be defensive in the sense that we actually have a capital formation and wealth generation outside of the nation-state in this kind of global on-chain economy and I think why that's defensive is because we can build this kind of resilient independent decentralized systems that manage to kind of form their own capital, like without, for example, being taxed on it, also without the very heavy regulation, which is a big issue. There are only 16 designated contract makers in the United States. There are 16 license derivatives exchanges, but money cannot move globally at the speed of info in the traditional system. Here it can, and we can build these really powerful marketplaces that are global and accelerate human progress. I completely agree. I think markets are a very important piece of technology. I actually do believe that one of the reasons that we are able to accelerate defense, particularly with prediction markets, is because we're basically getting rid of any authority's ability to lie to us, right? Because we can delegate information and truth to markets. I think this is great. And actually, speaking about this, if you think about Futaki so far, as you said, we've had lots of messy details. I'd love to know what you think the key challenges in implementing Futaki will be for a world organizational decision making. Right. So some challenges, of course. Robin mentioned the matter of the autist in the boardroom. So there's always this risk that prediction markets and also decision markets are, in some sense, threatening to the power of insiders and some organizations that maybe are not making the most efficient decisions. So this is for sure a big challenge. So another challenge that is marching up, I'm going to try to connect the previous question from Martin to the combinatorial market idea from Robin, which is part of our basically our grand vision on fitarki, which is so the matter of noise When we're evaluating potentially a small or medium-sized proposal or decision in a company Many of those if there's a smaller medium, you're not gonna have a clear impact, visible impact on the share price, so that makes it really hard to use this directly to estimate, but if you've had a full network or a full tree where you can see, like you could have markets for share price or token price, but connected to these you also see the KPIs, markets for different KPIs, number of users, revenue, all the kinds of metrics important for the company. So you could actually see, okay, maybe I don't, there's a lot of noise, I don't see the impact of this decision on the token price, but I see the impact of this decision on the token price, but I see the impact of this decision on this minor KPI that has, but it's minor KPI associated with this major KPI, which isn't then related to the share price. So I think that's a, this is a big trend region for many decisions is actually have the markets, all the information connected to each other. Maybe to kind of reorient the question, what problems are you running into building and operating these things? Prediction markets, futakis, what are the hard problems, what are the solutions? I can say one hard problem, relatively hard problem is to just find traders that are comfortable going into those positions. And just to give you an example of, let's say we would do this for Ethereum. We would say, okay, conditional, we might go this roadmap or that roadmap and then we want to use this food mechanism, so essentially people can say, conditionally to this roadmap, I buy Ether and conditional to that I sell. But then to practically do that you need to believe that going back to this correctness of prices, I mean yes 6% is in a day but I mean we really, what is really the fair price of ether right now? And I think there could be a range of some people believe it should be 10,000 and some people would believe it could be less than 1,000. So if you are in the camp of it should be 10,000, then kind of probably in both decisions, you would say, well, I want to be long on Ether, I want to hold Ether and if you're in the camp of well this is really not worth it, it's just worth 10,000 and again in both sides, so you need that person that says well the price is pretty much right now and really that decision makes a marginal difference for me to hold the asset. And then let's say there's a really good proposal and you really like that, so you go long on that. Now you hold this conditional yes. Now the next day, some external completely unrelated event happens where the whole crypto market crashes by 50%. You hold this yes proposal and hold it under the price from yesterday, you still think the proposal is good, but now your incentives are for the proposal absolutely to not pass because then you would hold all the dollars and not the asset at a price that's now 50% or where the real asset is now 50% cheaper from what you bought it. So I'm not sure if anyone could follow, but those are kind of the challenges I see. So 20 years ago we had a burst of applications of prediction markets, including in corporations. So I think we saw typical failure modes there that are instructive about today. Typically if you went to work with a company about setting a prediction market, you would say, well, what are your most important issues and numbers you'd like to track, and let's set up markets on those, and they would usually say, well, that's a little sensitive. Let's do something a little safer, and they would pick safer topics, and then they'd get accurate estimates on them and satisfied users, but they go, yeah, but we didn't really care about those. So, you know, that would be one outcome. Or, you know, they would say, talk about something important, and then management couldn't resist having opinions that disagreed with the market, and the market gets proven right, and the management wrong, and they're just really mad and want to kill it. So, for example, the US government missile test agency had prediction markets on which tests would actually go forward. So they put up, they try to make a lot of tests, and a lot of them don't actually happen, because they have to coordinate a lot of different parts of the military, make a test happen, and if one part isn't there, the whole thing has to be scrapped, and a lot of money is lost. And so they wanted to know which tests would actually go through it and maybe save money by cutting back earlier but then it was more legible that they knew ahead of time that it wasn't going to work and that was a management problem so they didn't want that anymore they'd rather pretend they don't know how these things are failing yeah uh like i guess i'm gonna like talk about this from more like consumer product building perspective, less like the future side. And I think definitely like historically, it's absolutely true that marketplaces are like very hard like products to bootstrap and to activate like the atomic network effect. I think that like it's definitely like to incentivize like market makers or liquidity providers, especially for things like pop culture markets, right? I mean, you can always build the model for, for like elections or sports markets, or especially in limitless cases like the financial markets, right? You can use the volatility to understand how to price them. And so it can be attractive to institutional market makers because it's essentially like a short term retail retail option flow but at the same time you know you may look at Polymarket how much they spend per month to incentivize market makers to come like into their order book actually I spent a lot of time previously and like I spoke to Robin about it in Berkeley like kind of obsessing about incremental improvements in the AMM which we see like for example like paradigm just launched the PMAM, right? We're actually going to spin up a contract for it to see how much of an improvement is it. But I definitely think that challenge is like, how do you incentivize the market makers? Either you use cash or you use your token or what, but it's still a huge expense for bootstrapping like the initial network, yeah. But maybe over time it will be worth it. All right, brilliant.

https://streameth-develop.ams3.digitaloceanspaces.com/transcriptions/673595919dbb7a90e13b23f7.vtt

Proof of personhood panel

 Hello, hello. Great. So we just had a nice overview of the three protocols. And what I want to do in this small amount of time we have is just challenge each one. I think there's something to learn from each approach here and see if we can come to some greater synthesis. So I'll start with Remco. So Remco, I actually wrote a critique of proof of personhood. It's called Compressed to Zero, The Silent Strings of Proof of Personhood. And what I did is I looked at a parallel protocol to WorldCoin, a different authentication mechanism. But they had succeeded in verifying unique individuals through cognitive tests. And they had avoided the problem of account trading, which WorldCoin has had to grapple with because each account has money. And then, of course, there's an incentive to sell accounts early on and so that's something you're grappling with. They had solved that with this novel mechanism called sublinear identity staking, which I'm happy to talk to you about later. But they were still left with this information game theoretic problem around puppeteering, which is that when you have an account and it has monetary value, it has a UBI stream coming towards it, which is one of the goals of WorldCoin, you create incentives for other people to try to control those accounts and extract from them. And so the kind of pithy way to put it is when you try to differentiate humans from bots and use UBI incentives, you actually create an unintended consequence of humans trying to control other humans as if they were bots, right? And so the control problem becomes an information problem, and it's an information game theoretic one. So I want to know how you think about this challenge, because there are two goals that are very explicit with WorldCoin. One is both UBI and a kind of rails for democratic governance and even maybe even AI alignment. But if ultimately, you know, pursuing both of those goals ends up in coalitions and oligopolies and basically politics, right? And one person, one vote breaks down very quickly in a coalition. It's like, how do you deal with that? How are you thinking about that? So we're at 16 million users now. 16? 8 million in the highest tier or verified unique human. And in the process of growing to this scale and being active in dozens, if not nearly 100 countries right now, you kind of see everything that people can throw at you. And while the vast majority of people is honest, there are some very dishonest actors out there and you have the Byzantine generals' problem of you need to make your system resistant against malicious actors. I think this is not, as you mentioned, this is for every proof of personhood primitive out there. There is this issue. As I said, proof of personhood is not a CAPTCHA. So there is this idea that you can delegate an account. This is not something I think we should prevent, but it is something that you need to give the user the tools to properly manage, which means that it needs to be explicit, it needs to be consensual, it needs to be revocable, and so on. So I think the emphasis is on developing those tools, empowering the user to control their world ID in the way that they want to control it. Blockchains as a whole are not particularly good at the moment at these kinds of things because a lot of decisions you make on chain are one-way streets. But there are a number of things we are already doing here, like allowing you to recover your account by going back to an orb. One thing we've rolled out that is very powerful is using the phone's front camera to do additional interactive face authentications to make sure that the wallet hasn't been stolen. It's still in the possession of the original human that signed up and so on Right Maybe I can Comment there directly so I made in this presentation my claim that I think this absolute identity They're actually not that many use cases for it and kind of just to give the extreme example Why I'm making this claim is let's say somehow North Korea does get hold of one of those orbs and and can can we are properly authenticate they're all there 25 million I'm not exactly sure how many people they have probably no one knows but 25 million people so let's just say in WorldCoin now you go from 8 million fully verified to now 32 but 25 of them are just people from North Korea where it's not even clear whether they have control of their key or probably it's somehow centrally organized that someone ultimately has full control of those keys. Now you have 32 million properly verified actual humans. The question is, for what application would that set be useful? So my claim is, in absolutely most applications, and probably actually this gas subsidy that you're doing is one of the few where I think it is useful, but in absolutely most applications, I would say you need more context than just being human. If you want to go even more extreme, we can think of this Matrix movie where they're literally, I mean, in the Matrix movie where they're literally farming humans, or it kind of has those farms of humans that basically just sit there and yes, they might have an eyeball, you might be able to scan that, but that's really not what you want. Right, and even just to add on, as long as there's money at stake, I mean, we have synthetic biology, it could also be just sort of synthetic eyeball forms to hack the system. Okay, hold on, there's like a couple different things going through each other here. Static eyeballs don't work because with biometrics it's critical that you verify that it's actually part of a real live human body. If you get possession of an orb and you abuse it, while the signup process between the user is absolutely anonymous and privacy preserving, the orbs themselves are not. Their behavior is public and there are methods you can use to actually revoke the fake accounts that might be... But there are no fake accounts. Okay actually revoke the fake accounts. But there are no fake accounts. Okay, that's the second part. If they are not fake accounts, then the system works as intended, in a way. And you make a very good point that proof of personhood is just a primitive, and you need a richer ecosystem of things around it. The example I like, a very simple one, is let's say you have an on-chain DAO. You want to implement one person, one vote. If you just do this and allow all of the world ID verified users to vote on this and your social community is only 5,000 people, you can easily get swamped. You don't have a very strong governance system that way. So you do need to incorporate this in more complex systems. Lascha? I 100% agree. Lascha? Can I add something? I kind of like... I get all the points, but I think like forgetting the orb or the eyes and stuff. So I agree that overall identities, and I don't believe in universal identities, because identities have to be built for the governance purposes. So we should come up with frameworks that increase the coordination quality among the humans and then decide the depths of the identity, how it should be built. When we talk about the proof of personhood, it has like two dimensions, right? One is the proof of uniqueness. And I agree for the real world and especially in multiple cases, just uniqueness proof and badge does not guarantee the quality of decision making or coordination. And then we have to rely on passports or like citizenship or the age or like all the other stuff. And the thing is that like if we, like for example for voting, refer to the passports, the dependency on these document is so big that we can generate the uniqueness based on data and we might not need any type of biometrics additional to it. And I think, in a way, that's the purpose of identity. So I don't believe in universal identities. I believe of defining whether we like fighting AI or we like fighting the misinformation stuff. The only systems that are dynamically evolving and improving are the ones that have this kind of fluid identity aspects and not just the universal ones. Like on the example of Twitter and the community nodes. So just labeling is just creating this very low barriers in terms of defining the permissions. What you can do or or you can't do as a human or not. And like you said, yeah, there's a kind of North Korea scenario. There's kind of a dystopian scenario where we might be like creating market, right, for the eyeballs and stuff. And like you got two kidneys, you got two eyes, and then what happens next, right? Well, what you raised is an interesting point about governance, right? Because in some sense, the point Martin was making is highlighting an enforcement problem, right? And this gets back to the information problem. Yeah. Right? And so even if you do succeed in authenticating unique individuals in North Korea, how do you actually enforce that? And what you're saying is, well, you need richer governance mechanisms. We need to express identity in different ways, right, as a complement. And if you just have that one identity system, ultimately enforcement would require surveillance. And, of course, surveillance results in everyone becoming informationally the same. So I want to shift gears, Martin, and poke a little holes at circles. I think it's a great experiment. But one of the things I noticed on your slides is, you know, you're talking about currency and money. I think of them almost as community currencies and like nested networks of cooperation building in a nation states and like, you know, even the globe. But, you know, not everybody in a community should receive the same amount, blood, sweat, and tears, and contribution. It shouldn't just sort of be like an hourly rate. That's another kind of UBI that has its own pitfalls because it's not tied to incentives. And you mentioned a 7% interest rate. These things seem arbitrary, and they don't seem to leverage incentives or people's voice, choice, and stake in a community. So how do you think about that? Yeah, I mean, should they receive the same amount? I would say if a bunch of people together comes together and decides to use something as money, then just that coordination effort to some extent creates value to whatever they decide is money. And I would argue that the fairest form of distributing this value that has been created by just this coordination effort would be to distribute it equally among those that join this coordination effort. Super important to understand that that doesn't mean that everyone needs to work for the same hourly rate or doesn't need, I mean, that's absolutely not the case. So it's just that the coordination effort is in a way or the benefit of that is equally distributed. Sorry, there was a second question that I... Just, you know, non-equal distributions within, you know, acknowledging that there's different levels of contribution. I believe it's the 7%. Yeah, 7%, right. So that is probably the only parameter in circles that has been set and that's therefore debatable. And yes, you could say every community should be able to set that themselves. That would make coordination much, much, much harder. Why? Well, because then you would need to, or in circles, it is meant to be so simple that I issue circles, and then I know I can kind of trust that other account. I can exchange my money to their money one-to-one because we are issuing under the exact same rules. If everyone could define their own rules it would be much harder to negotiate of how much should my money worth compared to yours. So I would say for a stable equilibrium, the goal of circles is to have the stable equilibrium where the exchange rates between circles are just one-to-one. The goal would be to say one circle is one circle, and it's kind of abstracted away the fact that under the hood, we have all those different... Just like in the dollar system right now, technically speaking, if you have a dollar at one bank and a dollar at another bank, those are not the same dollars. But the system is built in such a way that usually it feels like one-to-one because if everything goes well, you can exchange them one-to-one. But if it breaks, you will realize, oh, the dollar at the bank that just went bankrupt is different from the other one. But the US and Euro have floating exchange rates. Why can't you... Say again? The US and Euro have floating exchange rates. Oh, no, no, for sure, for sure. So, in circles, yes. I mean, the expectation is they are within a connected cluster, the exchange rate is one-to-one. But, yeah, there can be different clusters where within the clusters the exchange rate is one to one but between the clusters there are floating exchanges. Why not just rely on supply and demand? Well it will be I mean the price between those clusters will be defined by supply and demand and what we are actually supposing in our upcoming relaunch I'd say, or restart, or reactivation, it's a continuation of the system, is that people or groups can back their very own circles with reserve assets they want, Bitcoin, dollar, whatever they want, and then there will be a free floating exchange rate for those circles, so defined by supply and demand. But as long as people trust each other, then they are essentially creating additional one-to-one demand between those currencies. Maybe we need to go back to the identity topic. No, but you're expressing identity in a different way, and so I just want to dig on that. But I want to turn to Lasha now. So how do you see yourself and Rarimo in this constellation of identity protocols? What like differentiation do you see that you're bringing? I think we try to use the both sides of the world, such as passports as an instance first, but then extended with graph and capture more depths of the interactions. What I would say is how we try to solve the identity problem or enable the buildup is very use case centric and like I said, the identity problem or enable the buildup is very use case centric. And like I said, the governance centric. The fact that we understand this voting stuff has to be done that way, it gave rise to solving the uniqueness problem in a certain way. But this doesn't mean that the same uniqueness should be reused universally across other use cases. And so going at a primitive level of these kind of registries and the building blocks and having more like this kind of flexible framework, I think that's the way forward. And as well as we've seen mostly on the regulatory pressure or this kind of collusion risk, I think the permissionlessness and empowering of the users going forward should be this kind of collusion risk. I think the permissionlessness and empowering of the users going forward should be like this kind of core principle that no one should kind of violate or cross the line for. So we only have three minutes left. I want to quickly do one minute each on where, you know, AI is a part of identity is helping us leverage AI to communicate with each other, also to govern AI. So starting with you, Remco, where do you see the most immediate use case with WorldCoin in the near future? And world or world ID interfacing with AI will, you know, open AI, for example, grant free access if you have a world ID. What is the sort of immediate near-term goals on that interaction? I mean, I can't speak for open AI, but we do share some money in common, so we'll see what happens there. Regarding AI, I think we need to embrace the fact that it's going to be some sort of human-AI hybrid operator. I'm a big believer of AI is just making humans better versions of themselves. So this is why I think it's so important that you distinguish between a CAPTCHA and a proof of personhood, which are different primitives with different use cases. And we might not even be that interested in the CAPTCHA use case. What is going to be a very important problem is distinguishing real world sensory data from faked data and authenticity. I don't think proof of presence alone is a solution there, but I do think it can be an important part of a functioning solution that allows us to authenticate content and information. Lascha, any immediate applications? Yeah. In my world, I don't have this like us humans and AIs and kind of like the separate camps of the camp. I more see and question like, okay, what is the first thing? Like I own multiple AIs. So the first dimension is how do I control it so how do we build this my me my AI and identifying these relations and control and then zooming out like all the AIs of individuals or like the systems how do they coordinate and what are the mechanism for us so I don't know like lower down the computational capacity or shut certain things down and abstract them. This is the next level of governance problems that we should be figuring out. That's where the identity aspects will be born uniquely. Martin, I know you're thinking a lot about AI? Yeah, I think it's certainly good to have a robust identity framework in place that also is relational because, yeah, for the reason in the very beginning. So I like to see a version of something like Circles that's completely permissionless as a base layer built on this trust graph. But then, yeah, projects like DK Passport permissionless as a base layer built on this trust graph, but then projects like DK Passport or WorldID that can essentially provide additional attestations, what I would call it, to those open base layer identities can bring us a very robust system overall where humans can coordinate if necessary maybe against AI. Will each circle have its own AI agent? That's a good question whether AI agents will also use circuits or not. I don't know yet. They could. Alright, let's give them a round of applause everybody. Thank you.

https://streameth-develop.ams3.digitaloceanspaces.com/transcriptions/673592a79dbb7a90e1088fac.vtt