Hi listeners, this is the 80,000 Hours Podcast, where each week we have an unusually in-depth conversation about one of the world's most pressing problems and how you can use your career to solve it. I'm Rob Wiblin, Director of Research at 80,000 Hours. People who follow me on social media will know that I've been vocally skeptical about the value of bitcoin and the blockchain more generally. But it would be irrational to think I know where blockchain technology might go in coming decades — many inventions take a while to find truly useful applications. And I've been very impressed with a handful of people working to develop what blockchains are capable of, one of whom is Vitalik Buterin. I was first struck by his candidness and insightfulness when I heard him on the show Conversations with Tyler last year. Vitalik is something of a child prodigy, who's been attracting a lot of attention through his research and accomplishments, including creating the cryptocurrency Ethereum back in 2015 which made him incredibly rich. He's been making impressive progress in both engineering blockchain systems that might be turned to useful purposes, and as a theorist developing the so-called cryptoeconomics that explains how these systems can be robust and why they might solve problems that other technologies can't. This conversation is quite related to my episode with economist Glen Weyl from earlier this year – episode 52, Prof Glen Weyl on uprooting capitalism and democracy for a just society. Vitalik has actually co-authored a few papers with Glen, and they're collaborators on the RadicalxChange project, though they have plenty of disagreements as well. So if you've already listened to that episode, or want to go back to it first, that's great — but it shouldn't be necessary to make sense of what we're talking about today. To most people it's a bit confusing why I think this topic of mechanisms for social coordination is of such great importance. This interview also gets pretty technical pretty fast, and doesn't stop to explain what things like blockchains, or quadratic voting, or stablecoins actually are. So I'm going to start here by talking for a while until I think we've set the scene satisfactorily. If you'd rather skip that and go straight to the interview with Vitalik, just jump forward to minute 23. So for those who don't know, the blockchain is a protocol developed about ten years ago, which allows the maintenance of a decentralised ledger, or a set of accounts that define who has how much money on the network. Alternatively you can transfer tokens that don't have to look like money per se — for example a token that indicates that you own something in the real world. The first example of a blockchain was called Bitcoin. In brief, many computers around the world run the bitcoin software. These computers build a long series of accounts, tracking the creation of bitcoins and their transfer between different people on the network, effectively making it a form of money. The technology is called a blockchain because what you end up with is an enormous chain of blocks of transactions, one after another. Each ten minutes or so the network will add a new block of information that encodes all the transactions that are going on to the network from the last ten minutes. For a transaction to be added to a block, the owner of an account that has bitcoin in it, needs to use cryptography to sign a message from that address, certifying that they want to move bitcoin to another address. When a block is created, all the computers on the network certify that there are enough bitcoins at that address for them to be transferred. They all agree on what the block is following predefined rules, then move on to creating the next block. For various reasons it's impractical to tamper with older blocks, so once your transaction has been on the network for a while, you can be confident it will stay there. Everyone can see every transaction on the Bitcoin network, though some later protocols have made it possible to send other kinds of cryptocurrencies in a way that isn't public. The computers running this Bitcoin or other blockchain software are doing so because they're participating in a competition to add blocks to the chain. To get the right to add a block to the chain, you need to solve a very difficult math problem. Each time someone succeeds, the network will create some new bitcoins to give them, paying them for their work. There's a need to have a lot of processing power on the network, dedicated to solving these math problems, because if someone could control more than half of the processing power on the network, they could in principle launch an attack that would allow them to spend any bitcoins they own more than once. Having all this processing power there to make such an attack unviable, famously uses a lot of electricity, which is one of the technical problems that Vitalik has been working to fix. Bitcoin and other similar networks can process 7 transactions between people each second. This in principle allows people to buy and sell goods, without having to trust a bank to hold their money and make the transfer, or anything like that. So even if a bank would refuse to help you buy drugs from a drug dealer, you could use Bitcoin to pay them, and indeed illegal transactions like that have been among the most popular uses of Bitcoin so far. So it functions something like cash on the internet. But it also has the weakness of cash that if someone steals your cash out of your wallet there's nothing you can do. And if you pay someone cash for something, then they never send it to you, you're screwed. Cryptocurrencies have famously had extremely unstable exchange rates with the currencies we're all used to using every day. Sometimes they'll triple in value in a day, or decline to a third of what they were yesterday. This is pretty natural, because their supply is completely invariant to how much people are willing to pay for them. If the demand to hold bitcoins goes up, no more bitcoins get made, so the price can fly up. It's also natural because people have no idea how much people will want to use Bitcoins in future, and a lot of people are buying and selling cryptocurrencies are speculating about their future prospects. But that has made them a terrible way to store a certain amount of value reliably. Some people are trying to create stablecoins — cryptocurrencies that would have a fairly stable value in terms of USD or Euros or whatever. Of course, it's easy to devalue a cryptocurrency by just having the protocol set to make more if the price rises too high. But what about preventing its value from falling? That's a lot harder. One way to do this is by holding lots of USD or other assets, and standing ready to buy the stablecoin at that price. But the ideal would be to make the price stable without having to sit on one USD for every such coin. And pegging currencies like this, as it is called, has a poor historical track record. Many countries have tried to make their currencies have a particular value in another currency, and sooner or later seen their values crash. — OK, why might any of this matter? Vitalik and I talk about whether we can get the blockchain to work to run currencies, but to be honest that isn't where I see much upside. Most of us already have access to acceptably stable currencies, and the costs of sending money between people or countries can be significant, but nothing crazy. The problems that I'm interested in are coordination problems as a whole. The blockchain or other ideas Vitalik is working to develop might help humanity work together to solve the challenges we share. Many many problems can from one point of view be seen as coordination problems. This becomes apparent if you study economics and think about market failures in goods that aren't fully private. But it takes a particular view to see it. Let's just run through a few examples. A first easy one is that most people would prefer to live in a world where others didn't suffer in horrible poverty, simply because they dislike others having bad lives. If their contributing a small amount of money meant that everyone else did so as well, and the problem would mostly be solved, then they'd be happy to do so. Inasmuch as people all around the world value extreme poverty being eliminated, a donation you make to someone who's extremely poor, not only benefits them, but benefits everyone else who cares about them. Within nations we have the structures that allow us to coordinate to solve this issue, which is one reason that the welfare system exists. Many of us agree to contribute to the redistribution of money to the poor, so long as others then have to do so as well, through the tax system. But no similarly powerful voting and coordination mechanism exists at the international level, which greatly reduces how much money gets transferred from rich countries to poor countries. A second related example, is that many people think that eating meat isn't ideal, and would like it for animals not to be suffering terribly in agriculture. A decent fraction would be happy to become vegetarian if meat substitutes were nearly as tasty and cheap as meat is now. If 3 billion people could get together and decide to all become vegetarian simultaneously, you can bet that with the massive incentives to do R&D, we would very quickly invent meat and meal substitutes that were that good. So the problem could be significantly solved this way in a few years, if not months. But there is no way for so many people to get together and agree to do this, so it doesn't happen. And even if they did agree, there'd be the incentive for each individual not to stick to the agreement once it had been made. So the problem of animal cruelty, which initially seems like a very individual issue, can also be seen as a collective action problem. Along similar lines, if we wanted to fund the R&D into green energy that would strike a decisive blow in the fight to prevent catastrophic climate change, it would probably cost less than $1,000 per person in the world to do that. Many many people would be willing to contribute that sort of amount, in proportion to their wealth, if and only if doing so meant that everyone else would pitch in as well, and the problem would therefore be solved. But again, there's no effective way to create or enforce such an agreement. The same is true of funding the fixed costs of addressing many shared problems, like preventing common diseases. Very often those fixed costs take the form of funding scientific research to produce knowledge we all then share, but it can also take the form of a piece of infrastructure that only makes sense at a huge scale, like a sewer system for a big city. I mentioned climate change, but the same applies to global catastrophic risks more generally. It's in everyone's interests to ensure that humanity doesn't suffer some terrible catastrophe that ends up harming most of us, whether that's a nuclear war or an asteroid impact or a terrible pandemic, or whatever else. But for me as a selfish individual, it makes no sense for me to work to solve those problems just to extend my own life expectancy. The benefit of my work would go overwhelmingly to other people. Now of course governments to some extent solve all the problems I've described above. The US government as a whole oversees 5% of the world's population, so it has much more willingness to spend money to prevent a global catastrophe than any one individual. That's why it can spend billions identifying and tracking many of the asteroids and comets in our solar system. But there are serious problems here too, because good collective decision-making is a collective good that suffers the same underlying problem I've been describing. When each of us votes, 99.99999% of the impact that that vote has is on other people in our country or the world. The effect that the wisdom of our own votes has on us is basically nothing. So, unlike when we're buying a laptop for ourselves, the selfish incentive to vote wisely is mostly absent. This is a phenomenon known as rational irrationality. Similarly, if you see the government doing something terrible that affects everyone in the country a little bit, each person has almost no personal incentive to make a fuss about it. Along the same lines, I've discussed on this show so many times how journalism seems like a total dumpster fire. The underlying reason is, again, the coordination problem known as public goods provision. Each reader would be happy to pay a few dollars a month towards outstanding journalism, if doing so meant that the country was well informed and well run. The private benefit they'd get from living in a more sane country would be easily worth it. I have a friend who pays $500 a year to get a subscription to the Financial Times, which both of us think is a very unusually thoughtful newspaper. But he doesn't read a word of it. He just funds it because he thinks it's very valuable for society as a whole for the Financial Times to exist. But he's an unusual person. It's very hard to get people to contribute much to fund the creation of journalism. If I get a subscription to a good newspaper, that doesn't cause anyone else to do so. So I've paid all this money out of my own wallet, but barely made a dent on how well the country is governed. For sensible people who prefer reading fun stuff like celebrity gossip, rather than learning about tragic or annoying things that barely affect them, which should be most of us, the existence of the Financial Times is a public good much more than a private one. So even if people could only get access to the articles by subscribing, that wouldn't solve the problem. So to sum up, here are just some things that have major 'public good' aspects, in the economic sense of the phrase public good, which is to say that people get the benefits of them, even if they don't contribute to creating them: The elimination of poverty. Not hurting non-human animals. Most scientific research. Reducing the risks humanity faces as a whole. Producing information about what's happening in the world and how to fix those problems. Making smart decisions collectively, whether as cities, or countries, or as a species. OK, so that's a lot. Really, it's almost everything. If we could eliminate 1% or 10% of the public goods property of these problems, that would make a meaningful difference, perhaps transforming the world in visible ways. To be clear, I don't want to say that the framework of public and private goods is the only way of conceptualising what's allowing these problems to persist. But I do think it is a powerful one that can give us some insights. OK so that's a big diversion into why when I see something that might help solve coordination problems and provide public goods I get very excited. What does that have to do with today's conversation? Vitalik Buterin and others involved in cryptocurrency economics, or the RadicalxChange movement, have been thinking a lot about what new institutions or incentives we could create that would make a dent on this problem. Blockchain might provide new ways for a group of people to set up those institutions, but they're interested in other options as well. An example of progress on this problem from the past is patents. We took what was mostly a public good — breakthroughs in science and technology — and decided to give people an incentive to invest in it and publish their discoveries, by making it a bit more of a private good, by giving them a monopoly on their new insights for some period of time. I think that mechanism is overrated today and it has major problems, but it's probably better than having nothing at all. One way to get people to contribute to a public good might be to create a program that runs on a blockchain, which says that if 1,000 people each commit $10 worth of a cryptocurrency to a newspaper, it will get all that funding. But if they don't reach 1,000 people, then it won't and all of the people who committed will get their money back. Kickstarter already works like this. But there's a serious free rider problem here — everyone would rather that other people join that group of 1,000, so they didn't have to. An advance on this idea that Vitalik has written about is quadratic funding. Basically, let's say you think it's really useful for society that newspaper X exist and have money for its investigative journalism, but you don't particularly want to read that journalism yourself. With quadratic funding, you could put a small amount of your personal money towards that newspaper, say, $10. For mathematical reasons I'll skip over here, we'd then take the square root of your private contribution, add it to the square root of everyone else's private contributions to the newspaper, and then square the total. The government, or another funding source, would then match those private contributions in some proportion to that figure. The crazy thing is that if the government actually gave exactly that full sum, then each person would have the selfish incentive to personally contribute an amount that would then collectively fund this public good at the right level to maximise the welfare of society as a whole. Which is an amazing theoretical result. There's problems with that model, but if the government just gave 1% of the resulting sum of square roots, it could still represent a big step forward in public good provision from where we are today. Now, if you had the right pool of matching funding, and each individual had a unique identity, then it could be fairly easy to operate a funding scheme like this on a blockchain. Though you could do it using non-blockchain technology too. This is one of many ideas on the drawing board today that might be transformative if we could actually implement them. Another quite wild idea, if we mostly used blockchains to run our money, and could get blockchains to know what was going on in the real world, would be for countries to commit huge sums of money to an escrow program that would only return the money to them if there wasn't a war between them. So, hypothetically the US and China could stake $1 trillion at the start of each year, which the blockchain program would only give back at the end of the year if neither country killed a single soldier of the other country. Now there are probably several fatal problems with this proposal, but maybe there are other incentive structures like this which we can imagine which could nudge states towards cooperation and peace between them, which is probably a key determinant of whether the future of humanity will go well or badly. That's probably enough from me about why this issue matters so much. I'll link to a great and lengthy paper which goes into these issues in the show notes called Recent Developments in Cryptography and Possible Long-Run Consequences if you'd like to learn more. I don't like to talk this much at the start of the show, but some people who listened to this interview thought without it it would be unclear why it matters. So without further ado, here's Vitalik Buterin.
Thanks for coming on the podcast, Vitalik.
Yeah. Thank you. Good to be here.
So, I hope to talk about your views on long-termism, how to improve the world, and how to better coordinate humanity as a whole. But first, what exactly are you spending your time on at the moment and why do you think it's really important work?
Sure. So most people know me as the Ethereum person, and I do spend a lot of time working on Ethereum. The biggest thing in Ethereum land right now is Ethereum 2.0, which is this big multi-year long upgrade that we've been working on to try to massively increase blockchain's scalability with sharding, and a massive increase in security with proof-of-stake, and to improve its programmability, change a whole bunch of technical things that we probably got wrong the first time. So, that's been the big thing that's excited a lot of us for years, but now it's finally coming to fruition. There's these big teams around the world that are working on – not just thinking about what the protocols in Ethereum 2.0 should look like, but actually implementing it. So, the first phase of Ethereum 2.0, which has the proof-of-stake, is going to probably come out very early next year. There's already some test networks. They're starting to link different clients together and make different implementations written by different teams to talk to each other. A lot of really exciting stuff there. And yeah, and I think I've said it publicly a lot of times that I believe in the potential of blockchain technology, but on the other hand blockchains as they currently exist are in many ways a joke, right? Like 15 transactions per second… you're not going to run the world economy on top of that. And then the insanely gas-guzzling proof of work that I've criticized many times as well. So I'm definitely just really excited about turning Ethereum into a system that we can really fully be proud of and, and expect it to push ahead and actually become useful for a lot more things.
Well, we'll talk about some of those criticisms of blockchain technology as it stands today in just a second. I guess it sounds like you've managed to carve out time to actually continue working on a lot of the technical sides of things and not get kind of bogged down in the politics or perhaps the funding or bureaucratic aspect of … I guess there's organizations built up around Ethereum and blockchain now, which I could imagine just absorbing all of your time and management.
Yeah. And in theory, having more organizations is definitely interesting and weird because … and first of all it's definitely not like a conventional company, like we don't really have bosses that hand orders from the top-down to the entire structure. We have teams, and teams have leaders, but the way that the teams and the people inside of them, their way of communicating with each other is this fairly unique thing that we've just had to learn and figure out over the course of the last couple of years. And a big part of that does mean that the teams that are working on the different parts of the protocols have a lot of autonomy to them, which is also good for the people running the Ethereum Foundation because it means less work there. But I mean there is still a lot of technical coordinating work that needs to be done. And the other good news is that I'm not the only one doing it. So we have a lot of wonderful people, like Danny Ryan from the research team deserves a huge amount of credit for the work that he has been doing. Talking to the different development teams and getting them to coordinate together more.
So I guess to give some structure to the conversation today. I think first, first we're going to talk about kind of the blockchain and cryptotechnology as it stands today and how that's working out. And then we'll move on to talking about potential future applications for blockchain technology and how it could potentially end up having a big positive impact or perhaps not. And then we'll talk later on about long-termism and existential risks and effective altruism and things like that. So I guess for this interview, I'm in the slightly awkward position of being a huge fan of you and your writing and all of your research. But I'm pretty skeptical of blockchain technology, maybe relative to a lot of people who might interview you, at least in terms of what it's managed to accomplish so far. Though it sounds like you're perhaps even more skeptical of its accomplishments so far than what I thought. But all of that said, I have no idea how useful it could turn out to be in the future, which is one of the reasons I'm very keen to have this conversation, to try to see whether it could potentially accomplish a lot more good going forward. If I can start with a skeptical quote from the cryptographic expert Bruce Schneier, who I imagine you're aware of, who wrote this in Wired in February. "Do you need a public blockchain? The answer is almost certainly no. A blockchain probably doesn't solve the security problems you think it solves. The security problems it solves are probably not the ones you have. Manipulating audit data is probably not your major security risk. A false trust in blockchain can itself be a security risk and the inefficiencies, especially in scaling, are probably not worth it." "I've looked at many blockchain applications and all of them could achieve the same security properties without using blockchain. Of course, then they wouldn't have the cool blockchain name though. Honestly, cryptocurrencies are useless. They're only used by speculators looking for quick riches, people who don't like government-backed currencies, and criminals who want a black market way to exchange money." So your collaborator, Glen Weyl, is also pretty skeptical of blockchains as he mentioned when I interviewed him a couple of months ago. And you've been admirably candid about what they can and can't do so far. So to me, kind of like Bruce, it seems like so far that the main application for blockchains has been to facilitate illegal activities like tax evasion or avoidance of capital controls or selling drugs. Which may or may not be good, but it's not going to transform the economy. In terms of what it's useful for now, what kind of functions can it deliver for me that I can't get some other way through like saving my investments in a bank account, or in equities, or transacting using my credit card. Is there anything?
Yeah. I probably want to break my answer up into a few parts. So the first part is, what blockchains have done up to today, which I would say is realistically pretty much entirely cryptocurrency. Blockchains beyond cryptocurrency are themselves a big sector that I'm very excited about. But being realistic, there has been very little actual deployment of blockchain applications outside of cryptocurrency so far. And I do think that's going to change and it's starting to change. For example, in Singapore there is this thing called OpenCerts, which is basically using blockchains to verify that certificates like university degrees haven't been revoked yet. And they've managed to get partnerships with a lot of universities and different institutional entities. So there's things like that. But now and in the past, cryptocurrency has definitely had the biggest impact. And it's definitely had both good and bad consequences. On the bad side, there's obviously hacks and thefts and ransomware and all the things that people already talk about. And on the good side, I think people do underestimate the extent to which it's actually helped people by making it easier to move money around. For example, I visited Africa two months ago and there was this hackathon in Cape Town and I talked to one of the local African community members. This was a guy from Nigeria. And I asked him, "What do people use blockchains for today?" And his answer was actually pretty simple. It's just "There is a bunch of people here in Africa that are working remotely as software developers or something else, for companies in first world countries and they need to have a way to bring that money back to their homes so they can feed their families and live where they are." And the nice thing with cryptocurrency is that it's actually a lot cheaper than many traditional channels, including things like remittance companies. And sometimes people try to measure this and they say, "Oh well, these little quaint remittance companies are going nowhere, the sector is tiny". But the nice thing about cryptocurrency is that it's about decentralization and you don't need a frigging remittance company to do remittance. You just take the money, turn the money into Bitcoin, then sell the bitcoin or ether or whatever, sell it on a local exchange and you're done. You have it. So it acts as this global layer that basically says, instead of needing to have institutions that operate both in, say, North America and Africa, you have one class of institutions that operates in North America, lets you convert money over to cryptos. Then you have another class of institutions that operates over in Africa, that acts as the local exchanges and apparently the local exchange situation is actually pretty good there. Then you sell the cryptocurrency and you get the money. So there is definitely genuine value in just pretty simple and dumb things like that. And otherwise, you definitely correctly identified that kind of censorship resistance flavored things are another kind of category of usage that just naturally is attracted to this sort of thing. Because that's a big property that the technology has. And there's downsides to that, but then there's also upsides to that as well. And especially as we enter into this less favorable geopolitical environment where you're going to have a lot of actors trying to restrict people's ability to do things with each other for reasons that have more to do with zero-sum competition than making humanity better, trade wars and all that. So I definitely agree it's a complicated issue. But the second part of this question is 'blockchains in the future'. The big thing that I think will make blockchains in the future more interesting is basically if we managed to actually solve a lot of the big problems that make them hard to use today. Where scalability is probably the big one, and because 15 transactions a second is like basically nothing. And if you want to run global economies, they need to go into the tens of thousands of transactions a second. And Ethereum 2.0 with sharding is hopefully going to let you do that, which is nice and hopefully will provide enough capacity that people can actually run a lot of interesting things on top of it.
Also, solutions to privacy. Another thing is solutions to security. So, one of the negative consequences of this is just all these cryptocurrency thefts that keep on happening. We're like, "Another day, another $34 million gone from an exchange." And the nice thing about that problem is that it's one of those problems that I also do think that with better technology we can mitigate over time. So with a lot of those things being improved, I expect that, first of all, a lot of these non-financial applications. Like as I mentioned, there's a certificate verification project which is quite interesting and promising. Just the stuff involving replacing things like online identities and HTTPS and all of these things with more secure and more decentralized alternatives. And blockchains are a big part of that. So, I guess the way that I view blockchains in non-financial contexts technologically, is that with traditional cryptography you can prove to someone that you did something. But with a blockchain, and not with cryptography, you can prove to someone that you did not do something or, so, you can prove to someone that you have some certificate which did not get revoked. Or you can prove that, you know, you did some computation that included the outputs of everyone who submitted an input and you didn't censor people. So one interesting application I wanted to highlight here is there's … I published this post on ethresear.ch, our research forum, where I had talked about the possibility of using blockchains to do collusion-resistant mechanism design infrastructure. And we'll talk about what that means a bit later. But for now think of it as just secure online voting. And online voting right now is like notoriously insecure for a bunch of reasons. But, one of the reasons why is that it's hard to secure because there's all these different requirements to it. And so I came up with this design where you combined together the properties of a blockchain, which basically gives you a kind of non-censorship guarantee and says what the result is, you can prove that the result actually is the result taking into account everyone's votes. And so the authority didn't just decide not to include some people's votes, together with encryption for privacy, to get… and coercion-resistance, which has this strong level of privacy that says you can't prove to other people how you participated, even if you want to. Which is really important for voting to prevent things like vote selling. And together with zero-knowledge proofs which allow you to prove that the tallying authority actually computed the results correctly despite all of the votes being private. So you actually need to combine these three components together and a blockchain is one of them. And if you treat a blockchain as this cryptographic tool, then, you can create these different kinds of systems and they potentially would allow us to design these different gadgets that incorporate different kinds of mechanism design into our lives in a way that's hopefully actually secure and not going to break.
Okay. Let's just back up a second. So I guess me and maybe other listeners who know a bit about blockchain, but not that much, might be familiar with a couple of issues that there are. You mentioned the security problem, so that's like Bitcoin and other cryptocurrencies getting stolen a lot. Then there's this issue of lots of electricity being used up in order to secure the network, which I guess that's because you have like a proof-of-work method of securing it and the hope is to move to a proof-of-stake which doesn't require all the computation. And then there's this issue of the kind of bottleneck on the, on the number of transactions which you mentioned that you're hoping to solve with sharding, which I guess is something like breaking up the blockchain into many different streams that each can process that number of transactions and then they interact between one another sometimes. How confident are you that each of those technical problems can actually be solved? Are you sure that ultimately we'll come up with a technical solution or is it still an open question?
Yeah, so proof-of-stake, very confident. And so we actually had a spec freeze of a phase zero of Ethereum 2.0 which is basically the proof-of-stake part a couple of weeks ago. The spec is theoretically frozen and the only thing that can change it is if people discover security bugs. But the fundamental concept of proof-of-stake and the basic details of how the design works is something that's been thought about for years and we've figured out everything that we could figure out in theory and things have been fine. Now, then there's also proof-of-stake in practice and with proof-of-stake in practice, the main things that worry me are basically just how the incentives in the protocol interact with the real world. So things like, for example, are people just going to be lazy and run all their staking nodes on AWS? Are people going to be lazier and just do all their staking by sending their money to Binance, and Binance are going to stake for everyone. Are people going to stake in ways that are insecure and lead to a bunch of people getting hacked at the same time? And how decentralized is it actually going to be. Are things like 'bribe attacks against validators' realistic? All of these different issues that have to do with the finer points of how incentives defined in the protocol interact with human motivations. That's probably the most uncertain thing. The good news is that some of the uncertainty is starting to be resolved already. So we're first in a proof-of-stake chain of the class of proof-of-stake that we're doing to launch. So Cosmos got launched a couple of months ago. And Cosmos has these penalization slashing mechanisms, they have rewards, they have all of these things and they've already been having some issues. And that's been interesting because it both validates the concept and the couple of times that they've had problems, they even validated some of our design decisions. So, and we project wish them well and, it's been like we've learned quite a bit from that already. And of course when, Ethereum 2.0 launches, we're taking this kind of multi-pronged strategy where we first launch proof-of-stake, then we let it run for a bit, prove itself, then do sharding. I feel very confident about these things in theory. But in practice, I'm not expecting zero-to-one flavored kind of problems. I'm expecting one-to-n flavored problems, if you know what I mean. Like just 15 different edge cases that we just have to think about, that all leads to kind of small bugs that are fixable every time, but then just do end up becoming problems. So in the case of starting just a peer-to-peer network that can handle many megabytes a second passing through it and that ensures all the data reaches everyone who needs it within a few seconds. There's just things that we're doing that haven't really been done before. And also how that would interact with potentially powerful actors trying to attack it, all of these different uncertainties. So I definitely think that they exist, but they're more the sort of thing that we just have to cross when we get to it rather than showstoppers and at this point I'm definitely pretty confident that there aren't showstoppers.
Do you want to comment on the computer security or theft issue? Whether you think we'll be able to fix that one well enough?
Sure. Yeah. So there is probably three kind of sub-problems in computer security and theft in the crypto space. One of them is people giving their money to central authorities that screw it up. So like known bugs and getting hacked in all these other exchanges. The second is people keeping money in their own wallets and that getting hacked. And the third is putting their money into smart contracts and those contracts getting hacked. So, for the first and… It definitely keeps being a concern. And the idealist in me wishes centralized exchanges would go away and we'd be able to entirely decentralize the ones for everything except just interfacing with the banking system. And there's still the question of how close to that dream we can get reality to be. The technology behind decentralized exchanges has improved a huge amount over the last 12 months and there's even things like Uniswap running that are fully decentralized but that have even better UX than the centralized stuff. So there's reasons to be optimistic. But then I think cross-chain like between Ethereum and other blockchains is the one area that decentralized exchanges haven't really gotten a good handle on. So technologically speaking, there's been a lot of work on trying to come up with these designs that don't require central exchanges to keep custody of a huge amount of money, which is good. But, a second problem, which is if you have custody over coins, then what if your phone gets hacked, your laptop gets hacked, you lose your folios, your laptop and all of those things. So pretty much ever since I entered the space back in 2011 I've been a huge enthusiast for multisig, which basically is, instead of having one key that controls the coins, the coins are basically controlled by a program and the program says "If two out of three keys or three out of five keys sign a message, then those coins can be moved somewhere." And sometimes you can even have more complicated hybrid policies. For example, the Ethereum Foundation wallet has a policy that says that we have seven keys and if you want to send less than 1000 ether a day, then one of them is enough. But if you want to send more than 1000 ether, then you need to get together four of the seven. And that's served us pretty well. Basically I want multisig to get adopted by just everyone, individuals, organizations. But then the problem is usability. For like who are your counterparties, what's the format? So I've been an evangelist for this thing called social key recovery, which basically says that you have an account and you have one master key that can do things, but then you also specify some set of other keys. And these other keys would generally just be like your friends. And the idea is that any majority of those keys could come together and switch your master key to something else. So if you lose your master key, they could recover it. In case your master key gets stolen, then the idea is that you would store most of your money in a savings wallet that would enforce like a one day delay before you can take money out, and so you'd get your friends together within the day so that they can switch the key and cancel it. And it's a security model that exists in the real world already. For example, WeChat uses it for account recovery. So I definitely want to see more experimentation with social key recovery. And then first for the third problem, which is smart contracts, like formal verification, I'm obviously excited about it, but then the other thing is that we just have to be kind of humble and realistic about the kinds of things that we can expect to be smart contract defined. And we want fewer 5,000 line contracts and more 50 line contracts. And one of the reasons I'm a big fan of Uniswap is that the design is so simple it basically is a 50 line contract. In practice it's 200 but most of those are just convenience options, like you can get away with 50.
So it sounds like if we managed to solve all of those problems that we're still pursuing the dream of cryptocurrency being a medium of exchange where people might engage in lots of everyday transactions using cryptocurrency. Which so far has been something that has been … People are using it for remittances and cases where it's transacting between people where the banking systems aren't very well integrated or people are kind of unbanked. But I guess if you can solve all these problems then it might be possible for someone like me to actually want to use cryptocurrency to buy things on a day-to-day basis.
Yeah. And I definitely think the quality of blockchain technology can get there. And not just cryptocurrency, also non-financial things. So you'd own things, like self-sovereign identities. So basically having something that you can use to sign in to just regular web services that doesn't depend on Google or Facebook.
So why do you think people like Schneier are so vitriolically negative? Do you think they just aren't quite seeing like the technical advances that you think are likely to come eventually?
I think first of all it's important for us to address that they have valid points like proof-of-work sucks, blockchains as they currently exist are unscalable, a bunch of the security issues haven't been solved. So it's important to be honest about those things. But it's definitely, I think, also the case that a lot of the people who are kind of famously bearish on blockchains aren't following the space as it's going to be in five years and all the newer developments that have been happening there. And I think there really are a lot of things coming down the pipeline that really can solve a lot of the problems, definitely not literally all the problems. Like having users protect their own money in their own accounts isn't going to solve ransomware for example, but if we can at least start with the problems we can solve easily. If we can at least start with using crypto blockchain technology itself to deal with things like user account and exchange theft, then that's a pretty big start already. And for the other things, I'm sure they can figure it out over time.
Are there any problems or applications for cryptocurrencies or blockchain technology that people talk about which you think are bad applications of it, where it's not the right solution?
Yeah, and there's definitely a lot of things. First of all, there are these applications that just say, let me put these entire really big files in like 2.6MB PDFs on the blockchain and I think even with sharding, that's just going to be crazy expensive and not very pointful, especially when you can just like stick the hash of the file instead. Also things like using blockchains to just store messages that don't have any kind of value components to them. They just don't really need this property of being verified by a large number of computers at the same time. There's a lot of people that just use blockchains either just inefficiently or way more than they need to. So there's definitely quite a lot of that. I mean there's blockchain applications that people do that I think are stupid regardless of whether or not they're on the blockchain. And even just the ideas behind the large portion of ICOs right now just don't make any sense at all. So there's definitely a lot of things that I can argue against and often pretty publicly.
Yeah. Some applications that I've heard that seems strange to me … Well, voting where it seems like pen and paper, at least for like national elections is a lot more trustworthy than anything on computers, at least for me.
Yeah. As I mentioned, like you can use blockchains to make voting kind of better, but like you have to really think hard about how you're doing it and if you're just rolling votes on chain, that's like counterproductive. Like if you do that then someone can make a smart contract that says if you vote for the Purple Party, I'll give you a 0.01 ether and you're just going to get a bunch of votes for the Purple Party, and guess who's running America in 2020, so-
Yeah, I guess people have talked about it as using it as like auditing supply chains to make sure that products came from particular places. Which just seemed like it was solving the wrong problem with that one where it's like the issue is like fraud along the line where people could just put lies in the blockchain, right?
Yeah. So this is one example of a problem that people sometimes think blockchains solve, but really they don't solve, which is an Oracle problem. People often have this background impression that blockchains are this technology that provides trust. And the reality is that they provide a specific kind of trust for a specific kind of claims. And a blockchain has no idea whether the temperature in Toronto is 10 degrees or 20 degrees or 80,000 degrees. It has no idea whether the thing you just shipped me in the mail is a real phone or a fake phone or a real phone that's secondhand and is going to break in two months. A blockchain has no idea what the price of real world currencies are, and so on and so forth, and there's this big long list. And so a lot of blockchain applications basically do require either combining together blockchains with other data sources and you using a blockchain more as this kind of place where the different things get put together and organized or sometimes the data is just so subjective that like it doesn't really make sense at all. Like for fire insurance for example, like what are you going to do there? The problem is evaluating what the level of damages require. That's like incredibly subjective and there's no way a smart contract is going to help.
Yeah, and another application that seemed odd to me was people talking about using the blockchain to kind of get more autonomy over the kind of data that online services like Facebook and other websites hold about you. I don't understand how it would function as kind of a data storage system because it seems like anything you put on there is getting duplicated so many times that the cost becomes enormous. Is there something that I'm totally missing there?
There is a version of that which actually makes sense, and the version of that which makes sense is that you have services where the data is kept client-side by default, but then like people have tested the data and there's zero-knowledge proofs over the data. And then if you want to sell the data, you can use cryptographic protocols that basically say "I'll give you the data," or "I'll let you compute one particular function over the data if you pay me some number of points." And then for that to work like you can piece together like cryptocurrencies and blockchain hash commitments in zero-knowledge multiparty computation. But for something like that to work, that's a pretty serious challenge. You're not just going to shop patches of stuff on the chain and sell it.
Don't those zero-knowledge proofs require a lot of computation power or have I misunderstood that?
They do and that's definitely one of the biggest kind of downsides of them right now. But one of the really nice side effects that blockchains have is that they are leading to a huge amounts of money being pumped into making these zero-knowledge proof protocols work better. And so back in 2013 there were a couple of academics working on these things in the shadows, but now you have like 20 different protocols and there is dozens of engineers trying to figure out how to cut the proving times down. In some cases, working on using ASICs to cut proving time down, trying to cut down the size of the proofs and verification times. So there's definitely a lot of work being done there that can make things possible in a few years that weren't possible a few years ago.
What do you think are the most widely held but incorrect beliefs in the crypto community as a whole?
Oh, that's interesting. The crypto community isn't really a unified thing and there's definitely kind of different sub-tribes inside of the crypto community. And I think different sub-tribes have different misconceptions. Like on the Bitcoin side for example, like this idea that 2% inflation is this thing that's wrecking the economy and the next… And the next big step for progress would be turning humanity into a type one civilization that roams the stars is to get rid of existing fiat currencies and replace them with this 21 million fixed supply kind of thing. I think that's crazy. On the side of things like EOS and Tron and these more kind of semi-centralized projects. I think they have this belief that basically says like, "Oh these blockchains are not super decentralized anyway and so we're just going to be even more centralized and so we're going to get more efficiency without getting any of the downsides." But I think they don't really realize the kind of, some of the more subtle benefits that de-centralization gets you and a lot of those things are going to come out like if political environments become less favorable. And on the side of things like people trying to use blockchains or … Well it depends on the use case, right? But like if it's payments then, there is just the fact that like number one existing payment systems work in a lot of contexts but existing payments systems also don't work well in a lot of contexts and kind of seeing both sides of that is… I see a lot of people kind of making a mistake one way or the other where they think that either cryptocurrencies are going to completely overtake Venmo or they think that there is no room for them at all. Oh, here's an interesting one actually. I think a lot of people in the censorship resistance space in general, so like both people in the blockchain space that care about censorship resistance as a topic and people outside of it, they tend to kind of not have very good models of politics. Basically, a lot of people have this mindset that says, "We're going to create this thing, and it's going to just on its own completely be able to overpower governments and they're not going to be able to do anything about it and it's just going to turn the world into a crypto-anarchy and that'll be great." And I think they underestimate the power that governments do have. So you look at things like uh, the great firewall, like different countries attempts at enforcing copyright infringements on the Internet. Crackdowns on the sex industry — a lot of these different kinds of censorship. They, there are many cases in which governments try much less hard than they could if they really wanted to. And so, you know, you have these restrictions, then you have people using technology to get around these restrictions. But then you have counter measures against those technologies. And then you have counter measures against the counter measures, but the parties on both sides and particularly on the government side are limited by much more than technological constraints, right? They're also limited by their populations opinions on basically whether or these restrictions are just. They're limited by enforcement costs. They're limited by social costs. Schelling fences. Like if we allow this, would that lead to allowing this other thing, can we ban this without risking banning other things? All of these different complexities. And if you're building censorship resistance software, then you're not just playing this kind of great battle of technology versus a state here. You're also really interacting with these complicated political equilibrium and what does that mean? Well the answers to those questions are really, really complicated and specific. It's not any sort of single answer. It's a category of questions that I think you could get some good answers out of if you think more about it.
Do you think it's possible in principle to create a pegged stable coin that's not actually backed one-for-one by the asset that it's meant to be tethered to? I guess from both from a technical logical standpoint but maybe also from an economics standpoint.
Yeah, and makerdao is definitely trying and I think it's not possible to make one with a risk of zero, but I think it's possible to make one with a risk that's low enough, but then the problem is that if you need one with a risk that's low enough, then you have these capital inefficiencies, right? Basically it needs to have a double or triple collateralized with Ether or with other cryptocurrencies. And then, there is some of the problem of how secure is it actually? And the problem is that because the whole risk model is about tail risks. You don't really know. Yeah. And it gets really complicated and a lot of people I think can measure this sort of incorrectly. They say things like, "Oh, the price of Dai went down to 0.92 for like 20 minutes today, that means the thing is not stable." But no, that's exactly the wrong thing to measure. Right? Things that stay at exactly $1 almost always until one day they don't are like exactly the sort of fake stability that Nassim Taleb likes to criticize. So the thing that you're trying to optimize for isn't a lack of fine day-to-day movements, it's minimizing the probability that the thing just suddenly one day does something catastrophic and yeah, I mean the problem is that it gets hard to estimate the probability of catastrophic things. And I definitely think that cryptocurrency is interesting because it does create this environment where like you can profit from attacking one of these things then you can and someone will. And so far it's definitely looking kind of very soft-promising but we'll see.
Yeah. So the basic problem there is that kind of the less collateral you have, then the more you're at risk of someone prompting, trying to do a run on the currency and breaking the peg. Whereas, but then the more capital you have to have, the more expensive it is to establish a thing and then like, what's the business model here? Is that basically it? Yeah. Do you think there's ever going to be a way to kind of reduce the bad illegal uses of cryptocurrency without preventing the good uses other than like traditional law enforcement, or is that just not gonna be possible?
As I mentioned, I think there are a few specific categories of bad uses that you can prevent because people have the incentive to prevent it, which is basically like stolen funds. And I mean stolen funds definitely have a lot of kind bad knock-on effects. So for example, you have the North Korean government, and there's people claiming that the North Korean government is figuring out how to steal crypto and those funds are potentially being used to fund their nuclear program or other unsavory things and, if they didn't have those funds, then some of those unsavory programs could end up proceeding more slowly, or not proceeding. And so if you can solve the problem with theft in general, then you can also reduce the problems that come with criminal organizations having more money. So there's some things that you can do. There's probably other things that you just can't do basically, except by kind of increasing surveillance on something other than the blockchain system itself. So one example of this, right is that people sometimes say that … Both the proponents and detractors sometimes say, "Oh, cryptocurrency is going to lead to making it impossible to collect taxes." And I think that's crazy because people get caught for cheating on their taxes in a bunch of different ways all the time, right? Like if you're a millionaire, then you sometimes get caught because you registered an income of $23,000, but you have two big swimming pools in your backyard. So things like that. Another example is that you could even potentially use crypto technology to help make tax-enforcement systems that are better. So for example, use your knowledge groups to collect sales taxes without the government having to have knowledge of every single transaction and things like that. There's things that that you can do there. Yeah. And so like basically I think in some cases the answer's definitely yes, there is a bunch of different ways to mitigate it and in some cases, there is definitely kind of things that just genuinely can't be.
Why do you think we don't see kind of more of a crypto-driven crime wave? I remember back in 2013-2014, I was kind of worried that like Bitcoin could potentially be used to create kind of markets for assassinations or like other illegal behavior. But it seems like that basically hasn't happened at all. Do you have a theory for that or they could otherwise kind of help organize criminal organizations?
Sure. And I have a couple of theories. So one of them is that I think people underestimate the difficulty of successfully running a kind of, anonymous, kind of Tim May cypherpunk style illegal organization and like, because these days, you can get de-anonymized in so many ways from the time zones that you log on and make posts, from like text analysis based off of what you write, from just the ways that you make orders for servers, the ways that you log into the internet. So like people just make a lot of these mistakes and so it's really difficult to avoid all of them. Even if you're a technically skilled person, you can try. Like if that sort of thing became something I was interested in, then I would probably resort to crazy tricks like writing all of my forum posts in French and then shoving them through Google Translate to just throw off the text analyzers. But it's-
It's really hard to stay anonymous. So it's hard to patch every hole.
Yeah, I mean it's really hard. And the interesting thing is that there's also kind of diseconomies of scale. If you want to add up a scale of five to 10 people, you generally can and you could get away with it because these law enforcement bodies, as scary as they seem, are often quite under-resourced. But then if you go up to 100 thousand people, then eventually you get figured out. So I think there's that. And also I think people underestimate the extent to which the online software industry in general kind of relies on just goodwill and people being willing to kind of volunteer their time out of passion. And so there just aren't enough technically skilled people in the world that are like interested in writing the software to make assassination markets work really well. Right? So that's one thing. And then another thing is that if you want to make it work well, then you need collaboration between different people, and collaboration between people rests on this assumption of a kind of honesty and goodwill and if you're talking about people that are already willing to make assassination markets, are they already going to have that? Criminal coordination is definitely kind of harder than it seems and a lot of the successful criminal coordination kind of rests upon existing structures of coordination things like families or these Mafia organizations that have existed in "meatspace" for a long time and like … And coming up with a new one of those on the internet is not exactly trivial. So yeah, and basically I think anonymous internet technology, it doesn't solve a lot of the key problems that someone trying to make that kind of anonymous internet enterprise that wants to evade everyone that are trying to go after it would have.
All right, let's move on from kind of crypto and the blockchain today and kind of how it stands onto thinking about how much better it might be in the future and the problems that it might be able to solve. So you and Glen Weyl, and others involved in this radical exchange project have been … Yeah, thinking about how blockchain technologies could be kind of used to give people the right incentives to solve problems like collective action problems that humanity faces and in a way that existing institutions are failing to do. And I guess that includes a public goods provision or avoiding the tragedy of the commons. For example there's that liberal radicalism idea that Glen and I discussed, which is kind of an attempt to give people reasons to contribute their resources to providing public goods. I hope our listeners will forgive this one quote, but I think it kind of helps to
Motivate the whole discussion we're having here, and it's from an unpublished paper on possible uses of cryptographic advances that I hope we'll be able to link to when the episode comes out. And under the heading 'disorganized shared interest groups,' it says: 'It's generally the case within any given political system that only a very small portion of groups with common interests successfully organize to pursue those interests. For example, private industries are often organized to lobby the government, but the consumers of these industries' products do not. Influential grassroots movements have arisen around some issues like abortion and police violence, but not others of comparable concern like economic policy. The collection of people who would be happy to become vegetarian if it meant everyone else did, or to pay for news if it also meant everyone else did, and so on, is large.'
But people do not form robust organizations into such packs. And similarly, even the voters as a whole would have clear common interest in coordinating to become more informed. The possibility of such a widespread agreement arising seems wholly implausible, and these kinds of problems seem very widespread. So groups interested in funding lobbying efforts could use smart contracts that result in their making donations only if enough other people do as well, and perhaps scaling down the donation size as the number of people who commit grows. So I guess I just want to start by opening... by asking you a very open question, which is perhaps kind of the main topic that I want to discuss today, which is: what incentive design mechanisms, or kind of smart contracts or distributed autonomous organizations, are you most excited about, whether they're blockchain related or not, and kind of how could they make a difference to the world?
Yeah, sure. I think like first of all, the whole kind of decentralized autonomous organization thing is interesting, but also there is a limit to it that we didn't, I think, quite realize and appreciate at the beginning. So the kind of challenge there, right, is that at the end of the day all it is is people putting money into a pool and then using some pre-agreed rules to distribute the money in the pool around.
The problem we can solve, like for example, is giving people incentives to put money into the pool, right? And so like you have something like MolochDAO for example, which has a few dozen people putting money into a pool, so sending money to Ethereum public goods. But the amount of money that's in the thing is fairly small, and most of the money that is there comes from rich people, which an economic theory would predict are kind of the ones that are, at least from a selfish point of view, are more naturally inclined to be willing to send their money into these kinds of things because they benefit from a larger share of the benefit.
It's not solving a lot of the fundamental public good problems that we would ideally want to solve. So that's one challenge. Another challenge is this whole kind of collusion-resistance issue, right? Which is basically that one of the main benefits of a blockchain is that everything that happens on a blockchain is public. It's publicly visible. And even if you have zero-knowledge encryption for privacy, these kind of traceable records and receipts that you can audit and use to prove what happened. But the problem is that if you analyze a lot of the mechanisms that we use in the real world, many of them like really crucially rely on an assumption that people are not colluding with each other, and that if a person takes an action, the only consequence of that action is the consequence that's kind of defined inside of the mechanism, right?
So if you go into an election and you go make a vote for Andrew Yang, then like the only consequence of that is that it slightly increases the probability that Andrew Yang is going to win. And that's something which affects everyone in the country, and like ultimately in the world, right? Like there's no kind of specific benefit that you get from it. But if you start having votes on a blockchain, then other people can see, 'Oh hey, you voted for this guy.' Then people can make smart contracts that say, 'Oh you voted for this guy and I'm gonna give you a set amount of money,' or that, 'Oh, you voted for this guy and I'm going to penalize you,' or, 'I'm not going to sell things to you.'
And so the problem with a lot of these DAO constructions is that they run into these kind of twin problems, where one of them is plutocracy, which is that only the rich people really have an incentive to participate. And the second is collusion, which is basically that really if these things do become sources of substantial economic power, then people are going to try to use these techniques to manipulate each other into voting in particular ways. And we've seen both of those examples. So in a lot of the coin votes that we've had in Ethereum, it's been the case that just one guy had over 20% of the entire share of the votes. Meaning in EOS, there was a lot of interesting evidence of collusion happening. And in the case of EOS, well what happened was that there was this kind of collection of pools, like seemingly based in China, some in other places, that had convinced each other that just that, 'I vote for you, you vote for me,' or 'I pay you money and you vote for me so I can become a validator and I can give you some of that money.'
And all these different things that just basically fundamentally disagree with the usual economic model, right? Like the extent to which economists implicitly assume non-collusion is pretty severe. If you just open up any, most game theory textbooks, they just... people often think of it as the worst thing that these things assume is perfect information and perfect rationality, but like really no. You also make this weird assumption that like everyone is making decisions independently. There aren't groups of people that are thinking about each other's incentives. There aren't people pooling each other to make decisions. And so even something a second-price auction, for example, there's all these famous mathematical proofs about how optimal it is, but in reality if the top bid or the second highest bid are colluding, they can work together to push down the price and it's not efficient anymore.
So trying to figure these... basically, once the possibility of collusion enters the picture, then just a lot of things break and the space of mechanisms that you have to work with ends up like really decreasing by a huge amount. So this is where the work that I have been, that I mentioned earlier on collusion-resistant mechanism design really comes down, right? You can think of it in its simplest form as being just an online voting system. But really it doesn't have to be a vote, right? It can just be, select any function like F of a bunch of votes, and let everyone provide their inputs and then compute the output and provide the output. And you do this computation in such a way that nobody learns what the actual inputs were and nobody has the ability to prove what their inputs were. And if you do things that way, then you actually can kind of instantiate in reality this game theoretic ideal.
That says that people are making decisions independently that a lot of these systems rely on. And so we can start actually doing things like quadratic voting, quadratic funding, and all of these really wonderful things at larger scales.
Explanation of quadratic voting. It's not going to solve all problems. But I definitely think something like it is necessary to kind of prevent those schemes from completely breaking, right? Another example of collusion actually, this is a fun one.
So quadratic funding, which is also called Liberal Radicalism, this mechanism proposed by Glen Weyl for funding public goods. And the basic principles here is that you have projects and each of these projects represent some public good and people have the ability to donate money to projects.
And if multiple people donate money to a project, then there is this central pool of money which basically adds a subsidy to the donation to kind of take into account the facts that each of those people are contributing money to something that benefits all of them together. And the more people donate, the more the subsidy grows, kind of quadratically.
And so we have this thing called the Gitcoin CLR which is basically trying to do that at a small scale for public goods within the Ethereum ecosystem. And right now the matching pool is about $50,000 every couple of months. And the problem that it had was that there were a couple of cases where there were just accounts that were kind of clearly either not legitimate accounts and like big accounts, or in some cases were kind of related accounts that were voting for the same thing. The biggest example of this was that there was a bunch of accounts that were all employees of the same company that basically donated to the company and lots of the CLR matching. I think like these kinds of exploits, and they're definitely going to happen to some extent, but there's things that you can do to mitigate them.
One example of a thing you can do to mitigate it is, you just encourage more kind of honest participation. So the problem with the Gitcoin CLR is that there are only a few dozen people participating. But then if you increase that to a few thousand people, then a company with 10 employees is not going to be able to extract subsidies to the same extent. Another thing you can do is again have the collusion resistance. A third thing you can do is you can try to change the formulas to try to be more, kind of robust against medium levels of collusion that actually exist. So I wrote a research post on this. I forget the name, but if you just search like quadratic funding under my name, you can probably find it. And that's definitely been like interesting to kind of just see in real life like how this market for public goods works and what it does well and how it sometimes doesn't work and from there you keep going in and improving it.
Yeah. So we kind of jumped to the problems right away. But, I guess for listeners who aren't so familiar, is it possible to kind of paint a picture of how, in a good future, kind of the work that you're doing on incentive design, mechanism design, incentive design and these automated smart contracts, or automated processes on the blockchain, how they could help and what problems that could solve?
Yeah. So I guess the big problem that quadratic funding is solving is this public goods problem, which is that like there is just so many different projects and like things that all of us rely on, where they're not a thing that one person does specifically for another person. There are things that a person does and what the thing that the person does benefits a huge number of people. And that person really has no fine-grained control over which specific people it benefits. And so, one example of this would be open source software. Another example might be even just all of the blogs and public things that I've provided in this podcast. In the kind of less info and more physical world context, like public parks and roads. And sometimes you can charge for roads but sometimes you can't.
So yeah, I mean there's like just so many things in the world that just don't come in this kind of convenient format where it's one person doing a thing for another specific person and so you can have a nice efficient market where they're selling it to each other. We just don't have good institutions for kind of encouraging the production of these kinds of things. And so, I mean basically what we have right now is either governments or large corporations or whoever bids. And that's basically all and it's good that we have each of those and they actually provide those things. Because if they didn't, then there's no way we'd have a civilization. But at the same time, there's a lot of things that they leave undesired, right? These mechanisms just like don't do a good job of kind of incorporating all of the information, the kind of highly distributed information that we have in the world about what it is that is useful and that people find valuable.
So quadratic funding is basically trying to make a sort of more market-like alternative that encourages the production of public goods that basically says like, we're going to be neutral, we're not going to have our own kind of specific opinion about what is a real public good and what isn't. We're just going to say, people can donate money and stuff. And if a lot of people are donating money to the same thing, then that's clearly a project that benefits a lot of people. So that's a public good. And so we're going to detect that and we're going to automatically subsidize it based on a formula. So, the idea here is to really think about the final goal of these as being like basically your social institutions, in the same way that like markets are basically a social institution and democracy is basically a social institution.
It is the same kind of thing and could potentially work at the same scale except, at least hopefully it doesn't have laws that some of these other existing mechanisms do. So that's the kind of high level idealistic pitch for it. And of course, quadratic funding is one of the ideas. There's also quadratic voting. There's also Harberger taxes, there's also all these other things. And the idea there is basically that we have all of this really interesting work on mechanism design that happened over the last century. Let's actually use it and let's use it to build things that can encourage people to cooperate better.
So, you mentioned that there's these issues with quadratic voting and quadratic funding, that they're vulnerable to kind of one person pretending to be lots of people and potentially benefiting from that. I guess there's also this issue of, at least with the quadratic funding, it needs matching funding in order to give people sufficient incentive to like vote or provide their own funding to the right extent. Is there any alternative to just using kind of taxes to provide that matching funding?
Yeah, and ultimately the funding has to come from somewhere and basically it is like in a blockchain context you could... If this becomes popular in Ethereum or in like other blockchains like Zcash. I talked to Zooko and he's definitely interested in the idea of just like printing a quantity of Zcash that gets automatically put into a matching pool. Potentially governments and tax revenue are definitely the thing that you would need to use of some kind if you want to do it on a large kind of really society-wide scale. I mean, it doesn't have to be the same taxes that we use today. Like Glen also has been doing a lot of research into Harberger taxes as one example. But in general, the idea that instead of efficiency-decreasing taxes, we can try to have efficiency-increasing taxes.
And there's a lot of really interesting work being done there. So the kind of big slogan that Glen likes to use is 'Tax the congestible and subsidize the increasing returns.' So basically we take all of the things that people do that are zero-sum activity, put taxes on that, then take that money and then basically shard it over and use it to subsidize public road projects that people do for each other. And that's a really wonderful kind of one-sentence formula for like how to do political economy well, so that's interesting.
Yeah. What are Harberger taxes?
Harberger taxes are this idea that, for a thing that gets taxed, basically the owner has to set a price for it and they pay a property tax that's like some percentage of that amount every year, but then the price that they set is a price at which anyone else can go come in and basically buy it from them.
Are there any mechanisms that you're excited about other than kind of quadratic voting or quadratic funding?
I mean, prediction markets are probably one big example. And I think in addition to kind of public good providing institutions, we need better info institutions, and things that are prediction market-flavored, definitely seem really interesting there. And they don't have to be exactly kind of prediction markets, in the way that Robin Hanson designed them. They could be even just the idea that you would have to put down a bond, and that bond gets taken away if some negative event happens, in exchange for putting down a bond, you would be able to have the ability to increase the extent to which something gets done that does have some risk.
There is a lot of different ways to permute and apply the concept. Whether we're talking about immigration policy, or whether we're talking about trade-offs involving different kinds of risks. Whether we're talking about just, what policy to apply to achieve some particular known measurable objective in general. They're definitely not a tool that you can use for everything. I think the reason being that prediction markets work well when the objective is known, but the objective is not always known.
And figuring out the objective is often more than half the challenge. But at the same time, I do think that they are quite underused today. And I think one of the reasons I'm excited about blockchains is just the possibility that they could open up more room for using like prediction markets, and kind of security deposit-like things.
You mentioned earlier that there's this really important issue of establishing online identity, to prevent collusion, or to limit collusion. Can you explain, or give an example of how collusion can damage these mechanisms? And maybe what kind of work is being done, maybe by you or others, to create a trustworthy identity system, such that we can get around these problems?
Sure. First of all, these are problems that are wrecking the internet, even today. You have these click farms where people just buy up a few thousand phones, they buy up thousands of accounts, and then they use these accounts to just vote for things, to push messages that they want to promote, to manipulate a lot of different kinds of mechanisms.
And once we start using these kinds of mechanisms to allocate resources in more substantial ways, with things like quadratic funding, then there's a huge incentive for people to buy up a bunch of accounts, and pretend to be a big crowd of people that all cares about some specific problem and all throw a bit of money into it and then just extract subsidies. For any kind of voting mechanism, you can use these sorts of things to swing votes, and you can use these to manipulate online conversations, and all of these different problems.
Basically, we need to have mechanisms that identify who is a unique human that are more robust and that are more resistant to people just either creating a bunch of entirely fake identities, or taking existing identities and end up buying them, or renting them for different things. And that's something. It's a really hard problem. It's a problem that even governments have a hard time solving.
It's a problem that corporations have a hard time solving. Centralized services have to keep on fighting, and even they don't really have especially good solutions. It is the sort of problem, though, where if we don't solve it in some kind of reasonably secure way, then a lot of these quadratic funding-based and other kinds of mechanisms just can't end up working well.
So I guess with normal voting, people show up. Yeah, people show up to the voting booth, and they say their name, and I guess for some reason, there isn't as much of an issue with voter fraud, at least in the countries that I've lived, than what you might expect. I guess a bank uses passports, which are maybe too hard to copy. I mean, are we going to have to go for something as a system as costly as that, do you think, in order to solve this problem? Or is it possible that the problem is insurmountable, and it will just make it non-practical to use quadratic funding or voting in the real world, for anything other than, something where you can require that everyone actually rock up somewhere personally.
I think there's definitely always going to be some dollar value at which you can obtain fake identities of different kinds on the black market. So I think it's also definitely important for us to continue to research on making things like quadratic funding be more robust to identity theft. And as I mentioned, one of the recent posts I made on ethresear.ch is about changing the quadratic funding formula, so that it does put a bound on the amount of subsidies that you can minimally extract from the system, if you have 'n' identities. So yeah, things like that are important.
So that might be something like a minimum contribution for each person?
It's not a minimum contribution. It's more a maximum subsidy for a pair of people. Yeah. So there's that, and there is also of course the possibility of trying to make identities that are more robust, and that work better. I mean, there's definitely things that you can try to do, to solve it. One other problem that you'd end up going up against, if you try to make systems for this, that are universal, is hostile governments. Like, North Korea prints fake hundred dollar bills, so why can't they print fake passports? And so, you've got Kim Jong-Un with a click-farm of a million fake North Koreans that are claiming there's this obvious really important public goods that we have to give money to.
So basically things like that. And it's definitely a challenge, and it's definitely the sort of challenge where once again, it's not fully insurmountable, but also not fully surmountable. And so we just have to get out into the real world, and like keep trying, keep seeing what the issues are, and keep improving it over time. We'll see. I mean there's definitely a chance that the identity issues are going to be too big, but I hope not.
Yeah. Who's doing the most impressive work on this? And is it mostly, do you think, a social or political problem, or is it more a technical cryptographic problem?
Some of both. Like the anti-collusion stuff is definitely cryptographic. And to figure out who's a real person problem, is partially mathematical, because often you can use things like graph theory and minimal cuts, and all of these things to try to figure out how to actually aggregate the information to figure out who's probably a real person, and who isn't, but also partially a social problem of, how do you present these mechanisms, and how do people participate in which mechanisms you use. So it's some of both.
I've read that you're more excited about quadratic funding than quadratic voting. Why is that?
So quadratic voting is interesting too, but there's unsolved questions with it. So, one big example is that with quadratic voting you still need to have someone who figures out what's going to be up for a vote. And the ability to put something up for a vote, is a really huge amount of power, right? Because for example, one of the use cases for quadratic voting is that it protects minorities, right? And so, if you have quadratic voting on all the laws, then if someone comes up with laws that benefit a majority slightly, but really hurts a minority, then regular voting is going to pass those laws, but quadratic voting is not. But then here's the challenge. Let's imagine we have a world where 10% of the people are blue-eyed, and 90% of the people are anti-blue-eye-ists. And they want, kind of, but not to the point where they die for it, to have a genocide of blue-eyed people and kill them all.
So, this is sort of set up, where if you set up a regular majority vote to just kill all the blue-eyed people, maybe two thirds would vote in favor, a bunch would not care, and all the blue-eyed people would vote against. And so the blue-eyed people would lose. In a single quadratic vote, the blue-eyed people care about not dying more than the anti-blue-eye-ists care about killing them. And so they would win. But if one person had the ability to just keep on putting things up for a vote, and that one person was an anti-blue-eye-ist, what he would do, is he would just keep on raising the question. And so every time there's a vote, everyone would have to make these votes, and everyone would have to pay money to make these votes.
But the problem is that the anti-blue-eye-ists would lose, but because there's more of them, they're more spread out, they would have to pay less, and the blue-eyed people would have to pay quadratically more. Like the blue-eyed people, their votes are maybe 10 times stronger, but they'd have to pay a hundred times more money for that. And so if you keep raising the issue, then eventually you just bankrupt the blue-eyed people, and then you raise the issue for a vote again, and so blue-eyed get killed. So...
But this seems like a pretty devastating problem.
Yeah. I mean it's... this is the problem with voting mechanisms in general, like there's this very large design space of what can be put up to a vote and when and under what conditions, and even democracies have very different rules for when elections can happen with very different consequences. And you know, it wreaks very different outcomes. Whereas quadratic funding is really nice because the mechanism really is unopinionated in that way. Right? With quadratic funding, you just have projects. Anyone can spin up a project, anyone can donate money to a project and then you have subsidies. And that's it. It is really nice and clean and neutral and that's that. So that's a property that I really like about the mechanism.
Yeah. There's this website, schoolvoting.net which has a couple of pages which have pretty vociferous critiques of quadratic voting, I think including the one that we were just talking about. To be honest, I found the articles a little bit hard to follow and a bit ranty. But I saw you responding to them on Twitter. Do you want to comment on whether there's any other problems and I guess how convincing you found those articles? And maybe also do any of the critiques from that site carry over to quadratic funding?
Yeah that critique on the site was definitely really ranty but the critiques of quadratic voting in general that I think I hear the most about, I mean one of them, and this doesn't apply to quadratic funding, is this issue of who makes the vote and who decides what's up for a vote. And I think that's definitely a big issue. It's not a deal breaker, it doesn't mean that you can't have quadratic votes, but it does mean that they're not a universal replacement for things. Whereas quadratic funding it doesn't have that issue, which is nice. The other class of critiques of all these public good mechanisms that I feel the radical exchange movement doesn't pay enough attention to is, kind of rational ignorance and rational irrationality issues.
So basically the size of incentive that an individual person has to form a correct opinion of any of these topics is tiny. I mean Bryan Caplan wrote this big book on 'The Myth of the Rational Voter', where he basically... A big core part of why he likes capitalist libertarianism which is basically that people are good at making decisions when the consequences affect just them personally, but they're terrible at thinking through issues carefully when their individual contribution to an issue, it's so tiny. They don't really have any incentive or social pressure to be correct. And so all of these other irrational human biases, instead of just having an influence, end up dominating completely. And so if you think of some farmer in Texas, if you think about what incentive he has to form a correct opinion on anthropogenic global warming, the answer is basically none.
His incentive is to have opinions that make him look good in front of his friends, essentially. And there's definitely plenty of cases where people do have very correct opinions on things despite not having an incentive to. So I'm definitely not claiming incentives are always necessary but at the same time, there is definitely this issue and quadratic mechanisms do mitigate it, but do they mitigate it completely? Definitely not. And the score voting guy even did the math and pointed out how if you use quadratic voting to run an election, the average person would end up paying two cents. And are two cents worth of incentive really going to end up changing anyone's opinion? Probably not. Right? So that's one issue. And then also in general, if you go out into the world, there is plenty of unsavory organizations that ended up exploiting some combination of biases, lack of education, all of these things to pretty successfully extract money from a lot of people.
You have Ponzi schemes, MMM and pyramids and all of these things. And then if you started talking about, 'Oh, well we're going to create a scheme where if enough people donate money to someone that person gets even more money' then that could end up supercharging cults.
Yeah, it's definitely something that I worry about. But this is one of these sorts of empirical questions that you can't really figure out easily ahead of time. You pretty much have to run the mechanism at small scales whether or not the results of the mechanism are of high quality and if they are, you keep on going from there. If this sort of thing ends up being a problem, then one category of a solution that, once again, I think radical people generally don't think about enough is sortition. So if the problem is that you're asking people to express opinions about all of these issues, and they don't have enough incentive to really come up with the correct opinions, and they don't even have enough time to come up with correct opinions, then why not just force them to specialize?
So for any project that has more than a million people that participate in it, use a random number generator select a random 0.1% of people, only those people are allowed to contribute. Then you multiply their contributions by, I forget, it's either a thousand or a square root of a thousand, whatever the right factor is and compensate. And then suddenly each one of the 0.1% of people that do contribute is making a significant change to the output. And so they have more of an interest in actually figuring out what a sort of correct answer would be. So that's one category of a solution that could be worth exploring I think.
How serious a problem with smart contracts, I guess any process or mechanism that gets locked into a blockchain, is it that we have to then establish clearly ahead of time exactly what we want to contract about? And usually figuring out exactly what the goal should be, there's usually a lot of ambiguity in contracts. A lot of like negotiation that goes on between parties after a contract is signed on exactly what they intended in order to avoid them having to like bargain forever up front. How serious a problem is that with smart contracts in terms of limiting their potential application in future?
Yeah, so I think this is the same problem as the Oracle problem, which is the contracts don't understand the real world very well. And it's definitely a limitation to what contracts can do, right? You're not going to modify an employment contract and then get rid of employer and employees, that's not how it works. Smart contracts are a much more specific tool that is about, 'Hey, here are the mathematical rules of an economic interaction that we want to have and let's formalize the things that we can formalize.' And there are going to be inputs that have to come from the outside. And if there are, then we can formalize where the inputs come from. But that's a specific encapsulated thing that you can worry about separately.
So there's definitely cases where smart contracts make a lot of sense. For example, there've been people recently trying to compare their freak insurance which is basically contracts that automatically pay out if there's a drought or if there's hurricane or if the temperature goes above some number and there is a project doing this for Puerto Rico, there's a project doing this for Sri Lanka. So that's the sort of area where you do need to get some data from the outside world but the amount of data you need to get from the outside is going to be limited. And it's a thing that you can design an entire system around and make sure the system works well and then once you have the system then it becomes easier to build these things on top of it and at least conditional on the Oracle working, the rest of the system is also guaranteed to work. So you know, and that's the sort of things smart contracts can do.
This made me think. So this issue with quadratic voting where you have the problem of exhausting people's votes by putting up the same idea again and again. The UK parliament for example, has a rule that you can't put up the same substantive thing for a vote more than once in the same parliamentary session. I wonder if you have enough of a human element in it where people, trusted elected representatives are acting like the speaker does in the UK parliament to control the flow of who speaks and what gets voted on. Did you think that something like that could help to make it work a bit better in practice?
Yeah, I mean the human element is definitely the reason why the formal systems that exist in the real world today don't break completely despite there being the possibility of edge cases that could break them completely. Right? The reason why I don't need to go around getting signatures from 200 million people to turn on a light bulb is because we have courts that have common sense and that have decided that me shooting photons at other people's houses at that scale is not a property rights infringement. And the reason why shareholder governance continues to work okay is in part because we have these minority shareholder protections. The reason why votes continued to work okay is in part because we have these constitutional rules that say here is examples of things that you can't set vote for. And these kinds of exceptions are definitely important in improving outcomes but if you have too much room for them then there is a really huge cost of that as well.
So one example of this is the famous fact that there's so many federal crimes out there that pretty much everyone is guilty of doing at least some felony everyday. And so you have no way to not be vulnerable to a legal attack if the legal system really wanted to go after you. And so the more you have exceptions and these kinds of case by case situations that sometimes they help you but also the more they're something that's exploitable and that becomes really hard to work around. And in this case, sure, you could say that you can't vote on the same substantive thing multiple times, but then who defines what is substantive? Is that something going to turn into a partisan weapon in the hands of the wrong activist judges?
And I don't know. Right. So the idea that I think is the spirit behind the radical exchange movement, the radical exchange movement has many spirits, but the one that I want to narrow down on here is basically that if we can design formal systems where the formal system comes somewhat closer to being optimal, then you still will need the rules to kind of figure out what the exceptions are. But the number of rules are going to have to be fewer in number. And so because of that, you still get a lot of benefits. You know, in the case of quadratic voting, you'd have to have rules about what it is that you can vote on, but alternatively in the case quadratic funding, you need fewer of that. And so that's one of the reasons quadratic funding in particular is this flavor that I'm really excited about.
So yeah, I guess I'm feeling a little bit more pessimistic about all of this, well this is radical exchange project, than I did half an hour ago.
There's definitely a lot of challenges. But at the same time, it's important to remember that, first of all, if we can make it work, then this is really huge. We're talking about a general purpose infrastructure for funding public goods in the same way that money is general purpose infrastructure for funding private goods. Like this is something that really could supercharge the information society. Second, the fact that we're starting to think through problems is not a sign to be pessimistic because really every existing social institution that we have has a huge number of problems too.
And we just don't think about them because they're sort of normal, right? So for example, I talk about like 'is this supercharged donation gadget going to empower cults?' And then you realize that honestly the largest country in the world kind of is run by one already, right?
So yeah, and then with all of these things like identity issues, it's a challenge but then just regular voting depends on identity issues too. And we've eventually managed to stabilize it.
Even by my own standards, regular corporate governance like governance of companies with shareholders. I think it's completely economically broken up, right? For the same reason that coin voting is broken. And over the course of a few decades we've managed to figure out patches and shareholder regulations and all these things that sort of make it okay. So I think the goal here is to try to experiment with things in these smaller contexts, try to see if they work in these contexts, try to identify very quickly what the problems are and try to move towards solving them. And then I think it can definitely turn into something really interesting in five to ten years time, right? There's utopian versions of it that are going to be really hard. But then there's the less utopian versions of it, sticking Harberger taxes on copyright, that could easily end up being really valuable and are one very specific and well encapsulated thing. So I think it's a push for the easier things at the same time and analyze the moon while shooting for it. I think it'll get somewhere very interesting.
Yeah. So I guess, if we could solve all of these problems then the reason to be excited is that we could potentially solve a large fraction of all public goods provision, which is one of the fundamental reasons that the world sucks. But I guess you're saying even if we only at 10 or 20% of the way there, just by muddling through and fixing these problems to a sufficient degree, then that could still be a very big reward.
Yeah. I mean 10 to 20% of the way in the same way that zero to one is 10% of the way there from zero to ten. It's both 10% and also not 10%.
Yeah. Speaking of corporate governance, have there been any experiments to try to use these innovative governance and incentive mechanisms in the governance of Ethereum and other crypto projects?
So I mentioned Gitcoin grants, right? It's this quadratic funding gadget that's funded with about $50,000 every couple of months. So that's been running and it's been, despite the occasional problems, it's been giving outputs that I think are pretty good and match projects that I think are good and valuable. So yeah, and the main challenge at this point is we need a whole lot of work on figuring out identity. On collusion resistance, the projects that I mentioned, it's already being worked on by some cryptographers and they should have something else pretty soon.
Are there any mechanisms that you might be able to design where people don't benefit from collusion or you know, pretending to have multiple identities or pretending to have fewer identities than they really have?
Yeah, I mean money is probably the simplest example of that kind of mechanism. Right? Because people paying $10 to something is not treated differently from one person paying $20 to the same thing. And the way that I think about the importance of collusion resistance is that in a lot of cases you can show that a large class of mechanisms, if you allow for the possibility of collusion, basically reduces to being money, right? If you have a mechanism that says people can give each other points, but then people give each other points publicly and so I can give you points in exchange for you giving me points. Then the entire thing collapses into equilibrium and what's left is a money system.
And I guess the reason we don't want that is that we want to limit the power of the people who have the most money?
We want to have money systems, but money systems are good for private goods. They're not good for public goods. So we want to have these other than money systems as well. And the other class of things that doesn't require an identity, which is nice, is info institutions. So prediction markets and things that look like prediction markets are probably one big example there. And I'm really interested in those as well.
Okay. Yeah. Hopefully we'll have time to talk about futarchy a little bit more in a minute. But to get even more ambitious let's talk a little bit about whether any of these mechanisms could replace governments or even have an influence on international relations. So from that paper I quoted earlier, it moves on to talking about what influence a cryptography technology might possibly in the future be able to have on international relations or in promoting international cooperation and peace.
And it says, 'In addition, the need to establish the credibility of one's commitments in the absence of other mechanisms has often been cited as at least a reason for participating in prolonged wars, the Vietnam war, as a notable case. The number of lives lost, risked and harmed because of an absence of good tools to make credible international commitments would difficult to overstate. Therefore, given the stakes, the possibility that smart contracts could have applications in the international sphere is very much worth investigating. For example, smart contracts might enable countries to more credibly commit to international agreements that include penalties for non-compliance, for instance, those that seek reductions in greenhouse gas emissions.' So yeah. What's your reaction to that? Is this something that is probable enough that it's worth pursuing?
It's definitely interesting, but then there's also challenges. As I mentioned the Oracle problem and the challenge that on geopolitical scales, the Oracle problem is this serious Oracle because regardless of who's the Oracle, North Korea can probably find some way to hack that. So, you know this is not as easy as saying, 'Here's a smart contract that pays us a pile of money if we denuclearize', because how the hell does a smart contract know whether or not you've denuclearized? Yeah, I think that's the biggest challenge now that could get alleviated somewhat when we have more things that are running on the blockchain. So when we have, for example, currencies get issued there and a lot of identity systems running on it, like carbon credits being traded on blockchains and things like that, then you can start making smart contracts off of it, which I think could be really interesting.
And there definitely are projects that are trying to do things like putting carbon credits on a blockchain to make them internationally tradable more easily. So I definitely put a lot of value on things like that. I mean also in international contexts. One way to look at it is to try to improve coordination ability between state actors. The other possibility is improving coordination ability between people who are not state actors. So like making carbon credits tradable internationally, is one example of that sort of thing. Otherwise there's also these issues with certain certification and trying to make those work securely and blockchains could definitely be part of that. There's definitely people really actively thinking about it.
Yeah. Are there any other possible ways that the crypto technology could help, like with international relations or peace or international coordination or government that you'd be interested in highlighting?
See, credible commitments are definitely one you... Being a backbone for quadratic mechanisms and these other kinds of things at scale is possibly another, you know, and those are probably the big ones. A third one is also just making regular kinds of markets more and more international. But the challenge there is just, it's inherently a more long term thing because when the stakes are high, that would imply that people have a large incentive to attack a blockchain or attack an Oracle. And we're definitely not at the stage of blockchains being resistant to attacks and definitely not at a stage of Oracles being resistant to attacks on that kind of scale.
Yeah. So, so the point of Oracles is that sometimes, well for many functions, the blockchain has to know what actually happened in the real world. Like, 'Was there a hurricane so should we pay out the insurance to people who had hurricane insurance?' How close do you think we are to having trustworthy oracles and is that a problem that you think can be solved or is that something where again, we just have to muddle through and do a tolerable job?
It depends what level of trust you're looking for. For kind of small scale things I think we are there already. For large scale things, I mean we'll figure out things that kind of work over the next couple of years, MakerDAO has a huge incentive to come up with a model that's reasonable and right now they have some kind of multi-stake out of 13 Oracles and maybe they'll find ways to iterate and refine on that. And Augur has their decentralized Oracle for a prediction market, I mean Kleros the blockchain based decentralized port is doing oracle stuff.
Basically the decentralized oracles work by letting a random sample of token holders vote on any issue. And then if some random sample votes incorrectly, then a bigger sample can vote. And if it comes down to everyone voting, then basically the token kind of splits in half and the theory is that the token on which the token holders voted correctly is going to be the more valuable one.
And that's been working tolerably?
I think so. Yeah. And Augur is definitely been working okay for now.
I guess it will be vulnerable to someone who has a huge incentive to manipulate it and is willing to spend a lot of money, potentially.
Augur is definitely not secure, or markets that are larger than the value of all the Augur coins into existence. In crypto economics it's going to be a common pattern that the systems are going to be as secure as the market cap or their token but not higher and maybe just have to wait for the market cap to hopefully go up.
Are there any kind of other initiatives or projects out there that you think could benefit from crypto economic research, which will be particularly effective from kind of an effective altruist point of view that it would be worth directing people to?
I mean, so as far as specific applications go, I mean one idea that I have talked to GiveDirectly people about is that you can have GiveDirectly as a universal basic income, but you can also have GiveDirectly as a quadratic funding gadget.
Basically instead of giving every single person X dollars and letting them spend it how they want, just create a quadratic funding gadget and seed it with the identities in one of these communities and pump money into that. And because we are talking about communities where people are very poor, so small amounts of money could help them quite a lot. But these people don't just have a need for private goods. They also have need for public goods and so if you can just seed a quadratic funding thing with a few million dollars, then potentially it could end up being a way to really transform them.
Yeah. So that's one example and public good funding in general is obviously a kind of feed-in for effective altruist problems because pretty much all effective altruist problems are public good ones to a really large extent, and so me, I guess public goods funding as a kind of a general problem is definitely the thing that I'd focus on at the moment.
What kinds of people do you think are kind of necessary to try to get as many benefits from this strain of research? Is it kind of people who need more technical crypto skills or people who know economics and political economy and can develop the crypto economic side of things, or was it perhaps something like people who can build projects and make them work? Or something else?
I actually think at this point the thing that we're lacking is product people so to speak, like those who specialize in taking these systems and making them work for users at scale.
Yeah. We have the economic reasoning. We have a lot of cryptographic and mathematical reasoning, what we need's, actually, figure out how to deploy an identity system. We need to figure out how to deploy quadratic funding of things with the right user interface so that it actually works well kind of fairly quickly without people not knowing what they're supposed to do, and all of these other things. It's definitely a problem and at that stage at this point.
Crypto these days kind of has a, blockchain in general, has a bit of a reputation as being a bit of a scammy space and I wonder, how much is that a limiting factor or a challenge for organizations that are gonna try to do these really novel projects that people might find it hard to evaluate?
It's definitely a challenge and I definitely think that in general, blockchains and crypto as a whole, are still viewed quite positively to the point where you have large institutions getting excited about it. Like you have these UNICEF groups making trials on watches and all these other things. So I definitely don't think that the existence of Ponzi schemes and scams and all of these things is preventing the space from growing at all. It's definitely a cost and if we can find ways to solve the issue as I definitely think it can attract more people. But for me, the kind of larger value of fighting the scams is basically that we'll have less scams.
Let's talk about futarchy again, which you mentioned earlier.
Explanation of futarchy. Yeah. You wrote about futarchy back in 2014 and discuss some of the pros and cons of it. How do you feel about it today and do you think kind of yeah, blockchain technology could be used to put it into practice, in any case.
Yeah, I definitely think that futarchy and using prediction markets as an input to decision making is really interesting and as I mentioned you know, security deposits are sort of a sudden loss of liquid but like related customs or prediction markets. And those are really interesting too. I think the main challenge with say, using futarchy to run a country, that I maybe didn't realize a few years ago that I realize more now is basically, how do you set the objective, and if you talk to AI researchers then you know that setting the objective is a really, really big problem and if you set it incorrectly in the face of a really powerful optimizing force, then you created very bad consequences and that exists with AI. But that exists with prediction markets as well, right? Because if you're going to create a prediction market of say, the average life expectancy and you're going to use that to run health policy.
Well, what some guy is going to do is he's going to come up with a policy that basically says kill all the people that are statistically expected to have a low average life expectancy and that's going to push the number up, and so if you just pre-coordinate an automated mechanism, then that thing is going to basically voted for them. So that's the sort of thing that definitely is a bigger challenge. But there are definitely cases where the objective like is known. Right? So one way in which you can kind of solve this problem sometimes is you combine a prediction market together with some other form of governance, it could be a quadratic vote or a regular vote or something else.
So I adjusted this idea that if you want to create a unique identity system, then one of the ways in which you can do that is you can have basically a prediction market where you can just predict, which of these addresses correspond to real people and you need to have some threshold amount of money to vouch for you before you can be inducted into the system. And then the prediction market just is on this mechanism where if enough money challenges a person, then it comes to some vote and then the vote decides is this person actually a human or not? And if the vote decides that they're not a human, then you penalize everyone who supported them on the prediction market.
Right. So the idea there is that the objective, which is basically, 'Is this person a human according to this other mechanism that we assume we trust, and mathematical rule defines?', Then the problem doesn't exist to the same extent. But if your problem is, 'I have money, how do we use this to do the most good?' Then that's harder to define. And so prediction markets are not just magically going to solve that for you.
So yeah, very often kind of collective action problems persist because of transaction costs. Like search costs, bargaining costs, the cost of making and enforcing commitments. Is it right to imagine that one of the main ways that the blockchain could help solve these problems is by reducing those costs kind of one way or another? Is that, is that a good framing?
In some cases, yes. I think the word 'transaction costs' is one of these really misleading things because it refers to really unrelated concepts. The reason why we're not working collectively to fund public goods more is not the same category of thing at all as the 2.5% being charged by PayPal. Right? They're not the same. Even though both are technically transaction costs, they're totally not the same class of thing. And the reason why that's important to emphasize is because as markets become more efficient, one of those kinds of things goes down. But the other kind of thing is definitely not at all guaranteed to go down. And sometimes it even goes up if you end up accidentally disrupting traditional existing structures or coordination.
The kind of inefficiencies that are preventing more public good provision are in a specific kind of thing that you do need to approach it a specific way. Just like the efficiencies or the inefficiencies in private property leading to monopolies are a specific kind of thing that you have to approach in a specific way and commitment problems are a specific thing that you have to approach in a specific way. And then the same thing with inefficiencies in information systems and all of these other issues. So they definitely kind of solve problems, but I do think that you want to have a framing that recognizes that we are dealing with many different kinds of pieces here.
What's your view of Facebook's Libra project?
Ooh, that's a fun one. So I guess in terms of what their motivations are and I think a big one is just that Facebook is seeing that WeChat is getting into payments and that WeChat is really successful. So Facebook needs to make their next move, and so Facebook should be getting into payments as well and try to become this kind of omnibus app, and then if they're going to get into payments, the question is how do they do it? And the problem is if you do it through traditional banking and you want to be available in all these different countries, then there's just so many of these regulatory complexities that you have to go through. And so the theory is that if they do it as a cryptocurrency, then it's simpler for them in a bunch of ways.
So that's kind of my read of part of what their motivations are. And another thing is, they kind of have to get into payments because their existing business model, which is basically monetizing people's data is something that's kind of coming under political attack and we can expect it to come under political attack even more. And so they need something else. So from that side, the incentives definitely line up for them to do that. And on top of that, I think that people inside of Facebook, they are Silicon Valley people and I'm sure many of them take the critiques about how Facebook is wrecking the world and everyone's privacy to heart. And so they want to do better. And Zuckerberg did that release that memo a few months ago about how he wants to push for Facebook in a more pro-privacy direction.
Libra's base layer privacy at this point definitely sucks, but pushing Facebook in a not hegemonic direction. That might've been their intent though. And at the same time, obviously they failed in a bunch of ways on delivering on that because everyone just still associates it with being Zuckerberg coin, which has the exact thing that they really tried not to be by creating this association with 28 different members. So yeah. And I guess given that the thing is, to a large extent, this kind of political and regulatory play, it's the political and regulatory responses that are really going to decide it's fate.
So you think from a technical standpoint it's okay? So it's more a question of the politics?
Yeah. I mean from a technical standpoint, a lot of it's just the same as Ethereum.
Are there any blockchain projects that you think are just stupid and that you wish didn't exist?
It starts with a T and rhymes with cron.
Well, no, realistically more than half of them.
Interesting. Okay. Is there no kind of summing up what their main problem is?
And there's lot of projects that just basically see blockchains as a way of doing a money grab.
Yeah. Speaking of politics with these projects that, how have you found the challenge of making happy all of the different kinds of constituencies that Ethereum has? I guess like people who hold it, the developers, the people who were involved in the foundation. I imagine this like a lot of challenges and potential tensions there.
Yeah, it's definitely a big challenge. Especially more recently now the community has been growing a huge amount since around 2017. Before it was much tighter and everyone was talking to each other really frequently. But now it's gotten to the point where you can clearly identify sub-tribes forming and that's both inevitable and there's risk associated with it. So even just figuring out the next steps in the Ethereum roadmap are something just keeps on requiring more and more in a consensus building. And first of all it's good because it means the platforms getting more decentralized, but on the other hand it's important to make sure that it continues to be able to actually move somewhere.
I guess to wrap up this section, do you think that the kinds of approaches to making the world better that we've been talking about, crypto economics, blockchain, these incentive and design mechanisms, are they kind of underrated by people and that we should be seeing more smart people moving into them? Or do you think they're maybe appropriately rated or maybe even overrated and maybe too many people are trying to make them work relative to the difficulty of the challenges?
By the world at large, it's definitely underrated. The world at large I think at this point still sees crypto as just being a coin thing and maybe a trust thing and the subtleties around, 'Let's come up with ways to make mechanism design work and let's get quadratic funding running on this thing, all of this other stuff.' I don't think that's kind of gotten into the public consciousness yet. There's definitely an opportunity for it to, it's definitely the sort of thing where it does feel like there is a sort of political moment right now that's favorable to it at least among some communities. Inside of the existing groups, there's definitely large portions of the crypto space that are overrated. And as I mentioned, the whole 'Let's rescue the world from the evils of 2% annual inflation thing' is possibly one of the parts that I'm the least excited about. But there's other aspects that are really interesting too. So I guess it depends what aspect and who you're talking to.
Maybe this is an aside, I wasn't sure I wanted to go into, but the people who are so concerned about the federal reserve and inflation, what is driving them? Because just as someone who's studied economics, I feel like monetary policy is worked out reasonably well. Like 2% inflation is maybe if anything too low, what's going on here that this is such an interest for a significant number of people.
I think first of all there's this entire heritage of Austrian economics and goldbugism and these kind of sister movements to libertarianism that have existed for a really long time. And then a lot of people in the crypto space like libertarianism and just historically speaking, it has coevolved with these positions, and I can see why they'd be to these kinds of positions because they basically say that we should get the government out of the money supply and it definitely is a sort of ideologically aligned viewpoint in that sense.
Also, the monetary stuff particularly appeals to people given that the main thing that kind of distinguishes this kind of decentralized projects from the ones that came before it is the fact that there are coins inside of it. Yeah. I guess if you're going to ask me why Austrian goldbugism attracts people in general, it's a political ideology. It attract people for the same reasons as other political ideologies that kind of claim to know big secrets about how certain kinds of existing institutions are totally screwing up and wrecking everything. It attracts certain kinds of people.
Do you worry that if cryptocurrencies did become the main store of value or medium of exchange, that the lack of a central bank that can engage in monetary expansion could lead to more financial crises or more macro economic destabilization?
So it's a good question. And I think first of all, governments do have a lot of policy tools other than monetary expansion and contraction to use. So if, for example, just one random thing in Singapore obviously that I like is that apparently at one point, the government adjusted the mandatory pension contribution rates for employers downward during...
A recession. So they basically use that as a lever to reduce employer's side costs during a recession, which is when you theoretically want employers' high costs naturally to go down, so you don't have to lay off lots of people. Otherwise there's fiscal policies and all those tools that you can use. So I definitely don't think that it's going to, cryptocurrency policies succeeding would take monetary policies' effectiveness down to zero.
And another thing also is that if people use cryptocurrency as a store of value but don't use it as a unit of account, which seems extremely likely because cryptocurrency is just way too volatile to be a unit of account, then just adjusting the unit of account is still a highly effective policy tool that governments are going to have. So there are risks, but I think there are opportunities and options as well.
Of all of the potential new applications that we've been talking about, which ones do you think might we start to see in the shorter term perhaps in, you know, some more modest form, like an actual application that might get people excited and start really helping with a real life project?
