IEEE Blockchain Podcast Series: Episode 9
A Conversation with Gün Sirer
Founder, Avalanche Protocol
Listen to Episode 9 (MP3, 46 MB)
Part of the IEEE Blockchain Podcast Series
Brian Walker: Welcome to the IEEE Blockchain podcast series, an IEEE Digital Studio Production. This podcast series, entitled Research Notes in Blockchain, is hosted by Quinn Dupont, former assistant professor at the University College Dublin School of Business and founder of Alumni, a Web3 startup with the mission of putting university diplomas on blockchain. Quinn is also the author of “Cryptocurrencies and Blockchains.” In this episode, Gün Sirer, founder of Avalanche Protocol, explores the risks and rewards of moving from academia to entrepreneurship and shares his views on how blockchain can support credential management. He also discusses the science of distributed and applied systems and how new technologies are impacting governance and change management in blockchains.
Quinn Dupont: Thanks, Gün, for joining us today. The first place I wanted to sort of start talking with you is about your transition away from academia, away from Cornell and moving on to-- well, first you founded Ava Labs and then on to Avalanche itself. Why this actually is kind of interesting for me is-- well, a couple reasons, one, you even tweeted, you said “Academics need to sort of understand that they need to really focus on changing the world rather than just publishing kind of meaningless articles.” I think that’s really important and maybe this even had something to do with my own personal transition. I’m no longer at University College Dublin. I’m now myself kind of working on putting together a crypto startup. So, I wanted to just talk to you a little bit about entrepreneurship in this space and what your experiences were, if you’ve got basically any advice for someone who is maybe thinking about trying to make a real impact in the world. Any ideas there?
Gün Sirer: Sure. I have a lot of things to say to any academic thinking about doing the transition to industry. So, first of all, academia tends to be quite incrementalist and it’s quite often the safe thing to do it’s what we call slicing the salami-- you take a path that industry is following anyway and you look around and then you just decide “Okay, so, here are a couple of things that haven’t been tried,” and then you start publishing these things and a lot of the effort that’s spent like this tends to go nowhere. It’s work that would have been done anyway or it’s work that’s not worth doing and it’s often work that’s so minutely improving on the state of the world, often at some hidden complexity, at the cost of some hidden complexity. So, that’s a very common pattern. There are many other patterns in academia that I could talk about that lead to people not having impact, but on the flip side, when people have a great idea, it’s incredibly rewarding to go out and actually change the world. The mission here in academia, the mission for any professor is to profess, is to go out into the public, to take a stance and to stick one’s neck out and to say “Here’s how the world ought to be,” and that’s part and parcel of the ethos, in my view, of being an academic and it’s absolutely necessarily for us to actually make a dent in the state of the world and of course, the world actually values this. If you do have a genuinely big improvement, the rewards are quite substantial and it’s been a fascinating ride for me at Ava Labs with the Avalanche Systems launch about 14 months ago and when there is a true need, when the work is not, as we call, quote, “academic,” when the work does actually have legs, then it’s incredibly fun and just an amazing journey to go out and change the world. Now, I should also mention the following. There’s nothing wrong with doing theoretical work. There’s nothing wrong with doing work that’s so far out that the world hasn’t caught up to you yet and you can’t really go out. That’s perfectly fine and I have a lot of respect for that, but what-- especially in my area of applied systems and distributed systems, we see a lot of incremental work that’s neither here nor there. It’s not theoretical. It’s not far out and it’s not making a dent because it’s fixing a problem that nobody has. So, as I said, it’s just been a fascinating journey. I encourage everybody else to do the same.
Quinn Dupont: This is also kind of scary, though, just personally as I’m making this transition-- there’s a lot of risk, isn’t there? I mean, I’m-- both whether or not you’re going to be successful commercially, but also just I worry about if I have a research agenda, as you said, that might be a little bit theoretical or really pushing the boundaries, does industry-- is there room to support that in industry?
Gün Sirer: That’s such a great question. So, it’s such a-- there are so many risks, so many risks. Let’s try and take some-- so, when you’re an academic, the risks are well-contained. You’re playing a reputation game and there’s some horse trading going around, going on and you’re trying to get your papers in, etc. When you’re out-- but the game, that reputation game, typically is played within well-understood confines. You know what other people are going to do. Yes, I realize that there are all sorts of weird people out there and all sorts of bad situations, but in general, academics have a lot to lose, and the game can be petty, but at least it’s played within certain etiquette rules. The real world is very different. All sorts of funny things can happen. Money is at stake, very, very strange things do happen all the time because if somebody is going to make an extra, whatever, thousand dollars, they will do anything for that. It’s insane what they will do. So, it’s scary and there is all sorts of risk about taking ideas to market. The market might not be ready. You might not understand what the market needs. Maybe you’ve got everything going for you except the personalities don’t click with the people you need help from, etc. So, there are lots and lots of risks and I don’t know how to address any of them other than to say the rewards are great also. So, it’s worth it and so, it’s a calculus that everybody has to go through for themselves with their own idea whether the rewards will exceed the risks that are out there, but in my case and I think in general, whenever you have something that has a good product-market fit, it’s just a no-brainer that-- and there is also some clarity to be had from playing a wealth game, from playing a game where the game is to just sort of change the world and it’s not a reputation game at that point. It’s really a game of taking something to market that people want and the rules of that game are also very, very clear as well. There’s a lot of fun to be had. You’re not subject to other academics’ whims. You’re not trying to appease reviewer B and cantankerous comments and so forth. So, there’s that and that’s also a big win.
Quinn Dupont: Yeah. Absolutely. So, let’s talk a little bit-- I’m still partial to universities. As I said, I just left my university, but you worked at Cornell for a long time and to some extent, that obviously has informed what you’ve been working on at Avalanche and at Ava Labs. What’s the role of universities today, then, big picture?
Gün Sirer: Big picture, I think post-COVID, that role is going to diminish quite a bit. But the role of universities is to equip people with the skills that they need to be able to adapt to a changing world. That’s really the big thing and of course, it depends on the audience of each and every separate university. So, research universities versus teaching universities are slightly different in what they do and what they need to focus on. But going forward, I don’t think that universities as we know them in the form we know them are really going to survive that long. There are so many in the US, as you might be aware and quite a few of these are just unnecessary. It’s just extra overhead and the IT efforts, we’ve seen information technologies simplify a lot of things in life, but it hasn’t really changed the way universities work. Ironically, it hasn’t been applied to universities and MOOCs are one way of trying to sort of do this, but they are really trying to go for scale at the expense of survivability or sustainability of the effort, I think. But I think we’re going to see efforts in the direction of maybe sharing some elements with MOOCs, but really attacking the bloated management that we see at universities. Do you really need to have in every state so many hundreds of universities, each of which with their own associate dean of this and that. So, that, I think, is clearly-- I guess just ask yourself “What would Elon Musk do if he were in charge of a bunch of universities?” It’s very clear what would happen and that’s typically what will ultimately come to pass because the world loves efficiencies and of course, the current administrators will fight it tooth and nail, but that’s the direction that we’re headed.
Quinn Dupont: Right. So, I’m hearing disintermediation and decentralization, which makes me wonder-- does blockchain play a role here?
Gün Sirer: Blockchains do play a role in just about every business workflow going forward, but I think when it comes to universities, the very first way in which they will appear is going to be through credential management. So, universities are credentialing institutions. They give out diplomas. They give out certificates, etc., that attest to the fact that somebody’s been well-trained and has the necessary fiber to take on-- if you’re an engineer, it means not just that you know the necessary math, etc. but you also have some ethics, etc. that make you a suitable person for that field. So, these are important things and they need to be recorded in a way that survives the granting institution, that survives the passage of time and there are many different anecdotes that I can provide about people’s missing diplomas, but it’s very, very important that records of this kind are kept appropriately. So, the very first way in which we’re going to see blockchains play a role is going to be with credential management and after that second step, I’m not sure where it’s going to be, but probably within the internal accounting of universities.
Quinn Dupont: That’s actually the area that I’m moving into commercially. So, that’s wonderful to hear.
Gün Sirer: I was going to ask-- I was going to ask what area you moving to?
Quinn Dupont: It is, in fact, diplomas. It’s a real challenge, though, because I’ve spoken to a venture capitalist who said to me “Well, look, Quinn, diplomas are a Red Sea. It’s a dead zone and it’s impossible to make it work.” But the key for me, I think, is that there needs to be a business model, an actual real business model associated with it, which I think I’ve got. So, I’m hoping that with diplomas comes a big shift in trust. That’s kind of the bigger play, I think, kind of going off what you were just sort of saying about the role of universities.
Gün Sirer: That’s a very exciting area. So, both certificates for people who take continuing education courses and diplomas need to be managed via some external database and I can give you offline-- I don’t want to do it online here. I don’t want to name the country. But there is a country where the President is required to have a university diploma and they ended up-- there’s a bunch of questions about the credentials of the president who got elected and there should never be any such question. It should be easy and straightforward to look these up, but he went to one university, which changed its name, etc. The records were lost and no one knows if he really finished. So, it’s a strange thing. But it changes so much. It seems like a small mundane data record someplace, but it has such great consequences. So, absolutely, I think all record keeping of this kind should be kept by third party institutions and should be easy to consult and access with appropriate privacy measures. So, absolutely, I think that’s a great, great area. So, best of luck to you in your new venture.
Quinn Dupont: Yes, thanks. So, I’m going to switch a little bit to maybe some slightly more technical questions, if that’s okay, since after all, the audience is the IEEE primarily. As I see it, I see that consensus and security are kind of basically solved today with bitcoin. Scale is basically fixed. Avalanche and other next-generation platforms. I think we’re getting much better usability within blockchain, but of course, there’s a long way to go there. But for me, one of the big outstanding questions and one of the harder ones is governance, change management, and really just leadership, this question of leadership and I’m wondering if you could say a little bit about maybe how you imagine this to work, both internal to Avalanche, but also if I was to adopt Avalanche as a tooling, what kind of opportunities does it provide for answering some of these questions around governance and change management and leadership?
Gün Sirer: Absolutely. There’s so much to unpack there and so just in case there are people in the audience who have only seen blockchains through the lens of bitcoin, let me just say a few words about it though. So, bitcoin did introduce the world’s first decentralized widely adopted currency, cryptocurrency, and so, that’s wonderful and so on in that it serves as an example of what is possible. But it is not a good example of a sustainable green consensus protocol. I don’t want anybody thinking that I think that, and it also has, as you pointed out or alluded to, it also has some severe scaling problems. So, one of the main things that we introduced with Avalanche is a new way in which even consensus is in a distributed system. It’s a sustainable green system. It doesn’t require the consumption of substantial amounts of electricity in the background and it does this by a process of repeated subsampled voting it ends up achieving finality very, very, very quickly and is able to accommodate millions of hosts in the system. Now, going forward-- so, that’s the scale problem of blockchains along with the energy consumption problem have been addressed with the Avalanche system. Now, as you point out, there is a huge problem with governance and change management in blockchains. So, what a lot of systems did, early systems did, was fix everything in stone. So, essentially, what are they trying to do? They’re trying to decentralize and take the human element out of the underlying system. So, that’s a laudable goal. That’s a great thing to want to do and the easiest way of doing this is you just bake in the algorithms. You bake in every decision that could ever happen in the future into the system on day zero. For example, in the case of bitcoin, the number of bitcoins to be minted is fixed for all time. I can tell you exactly how many bitcoins bitcoin will be minting 17 years from now. It’s just all set in stone. Now, the minting rate for a currency is one of the most crucial economic parameters and it ought to depend on external factors. It ought to be able to change. Why? Because macro conditions change. If you look backwards, Satoshi ended up getting the minting rate just right some of the time. He ended up over-minting some of the time and the price fell and then he under-mints some of the time and then the price goes up. You can say “Well, look, the price going up is not a big problem,” but the price going down certainly is and being able to respond to external stimuli is pretty important for any living system if it's going to be sustainable. So, what do we do in Avalanche and what do latest systems do? In Avalanche, what we do is we allow the coin holders to vote on key economic parameters. So, the exact minting rate, for example, in Avalanche is determined by the votes of the people who are holding the coins and it can be sped up or it can be slowed down. The minting process can be sped up or slowed down depending on how the community votes. So, in essence, what we are doing is we are replacing and disintermediating the human element-- the central bankers are gone. The political meddling in the system is gone and instead, the people who are actual participants in the system put their votes down towards however they want the system to behave. There’s a second thing as well that we do that is absolutely crucial to being able to accommodate changes. So, often, what you want to do is you want to introduce something to a system and you may not be able to have the pull or the support of the community for that change and in systems past, this meant that you would have to fork and create your own independent system and we saw many, many, many different bitcoin forks for this exact reason. They want to change the block size and all of a sudden you’ve got seven different versions of bitcoin, each with a slightly different block size or a block time and so, that’s highly, highly undesirable because you’re fracturing your user community. You’re fracturing your liquidity. You’re fracturing your funding. So, what Avalanche does and it’s one of the very few systems to do this is to provide this notion of what we call subnets. So, you can create a subnetwork under this overarching umbrella of the bigger default network and in your subnetwork, then you can enact any rule set you like, you can run any virtual machine you like subject to whichever set of decisions you want to make in that subnet and the subnets are isolated from each other, but they can communicate with each other. So, your subnet can, for example, communicate with mine and I can invoke-- I can send you assets. I can take assets from you, etc. So, this structure is incredibly useful. It’s useful for change management, clearly, and that was your question, but it’s also useful for another use case that was a dire need in this whole blockchain space, which was being able to accommodate jurisdictional differences. So, people with assets, companies or institutions with large, valuable assets, they want to be able to open them up to the public. They want to be able to have them traded on the blockchain, but they can’t get legal clearance because they’re under legal obligations. They’re supposed to follow certain laws and bitcoin doesn’t follow any laws, bitcoin does what bitcoin does. So, with subnets, you can actually create your own blockchain, so to speak, your own universe where the rules are exactly as you like. You can have a US-centric one where certain countries cannot participate. You can have a European-centric one with GDPR rules and so forth in place. So, whatever the legal requirements might be, they can be accommodated in these subnets. So, that gives us two birds with one stone with the same mechanistic scheme. We can support both change management and differences of opinion among people who want to use different parameters and we get to support different legal requirements. So, I’m really excited about what’s to come with this. It’s a very flexible framework and it’s a very fast, scalable, and sustainable framework.
Quinn Dupont: That’s really impressive how much you’ve been able to sort of fix with the couple of changes there.
Gün Sirer: I know. It’s cool, isn’t it? It’s just a few technical tricks, really. It’s more than one.
Quinn Dupont: Yeah. I’m sure it’s harder than it seems.
Gün Sirer: There are lots of one-trick ponies in this game, where it’s just one idea and then they have a new chain. I’ve been in this space for a very long time, and I ended up packing everything I knew from just distributed system science into this system as well as economics. I consulted-- we consulted many different experts, and the result is there were lots of pretty obvious sounding ideas but that nobody else had thought of before and certainly nobody else had combined.
Quinn Dupont: I want to get to questions around sort of the science of all this. But before, I just want to push you a little more on this change management just so I can get a better sense here. Subnets seemed to be really an alternative to forking, in some sense, and offering you these really excellent differences, which I think is really useful. I feel like on the issue of these endogenous changes, we seem to have a paradox, though, right? Vitalik Buterin, a little while back, he wrote about the challenges of coin-based voting. Do you see anything on the horizon that we can get away from this, this idea that simply-- since we don’t have true identities, we don’t have one person, one vote, we’re just going by whales potentially making big changes?
Gün Sirer: Right. That’s a very good point. There are many instances where you want to weight votes differently. So, there are many different things one can do. So, there’s obviously one coin, one vote, which ends up weighting the whales heavily and there’s obviously also one person, one vote, which seems like an ideal except then how do you determine what a person is? You’d suddenly need an oracle or some kind of authority on who’s a person and who’s not. So, that brings us to this whole notion of digital identity. It’s a topic that’s been discussed quite a lot and it’s-- if you look around, there are absolutely no solutions that have been deployed yet, but there are technical solutions whereby people can create self-sovereign identities, that is identities that they themselves, that they are in control of. They can get these identities vetted and they can use these identities for purposes of voting and other things.
Quinn Dupont: I think that’s a pretty good answer, to be honest. I want to just, in the last couple minutes I’ve got you here, I want to ask you a little bit about the science of decentralized systems. Are there big, wicked problems that are still out there that we’re unaware of or are we finally now-- I mean, arguably one to three decades on in distributed systems, are we at a fairly mature science at this point? How do you feel where the world is right now?
Gün Sirer: I think we’re just getting started for real. I think the opening salvos have been fired. I think bitcoin paved the path for us. It showed us what is possible, but it just lacks the scale. I think Ethereum, similarly, is a fantastic system. It shows us what is possible with smart contracts, but it’s got a lot of downsides. It’s just not how you would design it if you had to if you could design it from scratch. So, the outstanding problems, long outstanding problems were scalability. I think that’s been addressed. The second problem was flexibility, being able to change, being able to add new features in a way that’s compatible with other people using the system and I think that’s been addressed with subnets. Governance is an ongoing issue. I don’t think the last word has been spoken, mostly because decision theory is-- there are all sorts of tradeoffs in it. There isn’t a single outcome that I can just go out and take and adopt. It’s hard. It’s a fundamentally hard field. But I suspect that we’ll be able to make headway in that space, especially as we figure out a way to solve the digital identity problem and I believe that solutions to that problem are coming. What other issues are outstanding? I think there is this entire discussion around Web3 and how that’s going to take place is going to be fascinating and Web3 is this big push to take control of information and data away from centralized service providers, like Facebook, Google, and so on and give it back to the users. These centralized service providers, they essentially compile dossiers on all of us. They sell our data and it’s just a terrible, terrible way of doing business and it’s a terrible way of living one’s life as we constantly leak digital information about ourselves, and these other people are compiling it and trying to monetize it. So, the alternative, where the user is in charge of their data and they reveal selectively exactly however much they are comfortable with is a far more exciting universe and I’m really excited about the possibilities of Web3, but there remain many different problems ranging from cryptography, zero knowledge proofs, and so on, all the way to game theoretic issues that are outstanding. So, that problem is long outstanding and it’s an exciting one. Another problem that’s outstanding has to do with what we’ve seen in D-fi. So, in decentralized finance, what we have are essentially financial instruments that are implemented as programs. So, whereas normally, you would go to a bank and interact with bankers and go through a very slow process for doing whatever you want to do, what you can do with blockchains is take advantage of their ability to run code and you interact with a program and these programs can do all sorts of exciting things these days. They can do-- they can lend money to you even without having to see your ID and there’s a whole discussion to be had about how that might work, but you and I can, even at this very moment, take out about $1 billion and use it and return it. As long as we return it, it’s all good within five milliseconds or so, but if you want to take out $1 billion and run an arbitrage trade, you could do that with D-fi. It’s an amazing capability. You do it without showing ID. You do it without any history and high school kids from Eastern Europe are doing this to make money and they provide liquidity to markets. So, these are all great. But what’s happening in D-fi constantly, and this is where the challenge lies, is that the smart contracts end up having bugs in them and they get exploited. So, program verification and reasoning about correctness of programs is of paramount importance. This has always been true for humankind. We’ve noticed this starting in the 60s and on when the field of program verification started. But it’s changed a little bit and I want to pull out for other academics what exactly is different this time. In the past, people focused on safety properties and they focused on liveness properties and both of these are important. You want your programs to be safe. Bad conditions should be encountered. You want them to be live. You don’t want them getting stuck. They should be able to make progress. But there are other properties, especially in this domain and what are those? They tend to be game theoretic. You don’t want your contracts or your systems that you create to have dynamics that are undesirable and that process of being able to reason about the dynamics of a system set up by a program has not gotten the attention that it deserves. So, I’m really excited about that area as well. So, those two areas I see as long outstanding, actually deep problems that will require some smart minds.
Quinn Dupont: Are these because money and value is just everywhere? I think this is actually part of the Web3 promise. It includes this idea of ownership, taking ownership, self-sovereign identities, so on and so forth, but it seems to me also the future is just going to be money, forms of money or just more generally value everywhere, which seems to be the undercurrent behind these challenges, especially these game theoretic challenges.
Gün Sirer: Absolutely. The world was always about value. It was always about value extraction. So, I think as academics, we kind of shielded ourselves from it, but deep down, that’s the game that everybody’s playing out there in the real world and so, when you see it so starkly play out in the blockchain space, it’s quite easy to spot and the nice thing about blockchain is it’s such a contained environment that you can actually take some tools from engineering and computer science and apply it to them.
Quinn Dupont: Are you worried at all about-- some people talk about financialization as this thing that’s going to erode humanity. Are you just a realist about this and you say it’s going to happen anyways, or you look forward to this? How do you feel about that?
Gün Sirer: No, no, no, no. I’m a huge humanist and I worry about that exact same thing. I hate transactional people and I hate financialization, where every single interaction is a financial decision or people think about expected value from every human interaction. I think these are terrible. They erode the society in which we live, and I certainly don’t run my life that way. In fact, I run my life in a way where I actively avoid such people and certainly, in the blockchain space, we see a preponderance of this kind of thinking. It attracts a certain type that’s kind of like this. But moving forward, I believe in the sort of inherent goodness of people, and I think yes, sure, there will be people who try to financialize everything. There will be people who will build systems that require you to make a financial decision, “Do I interact with this web service? Do I give them a tenth of a cent?” etc. But I suspect that those kinds of systems will not find traction. I suspect that the winning systems will be things that make sense to reasonable people who have a reasonable operating system. That is, they don’t view other humans as people to extract value from and they don’t view every interaction as something that is financial. So, I think we just aren’t there yet. I think the early tries, the naïve tries, etc., they will be in this direction. They have been in this direction, and I certainly share the concern. But deep down, I’m a humanist and I believe that people will prevail and they will build systems that are nice to use and will not require you to have to make financial decisions every time you click on something and there will be new models that emerge that make it all convenient and seamless.
Brian Walker: Thank you for listening to our interview with Gün Sirer. To learn more about the IEEE Blockchain Initiative, please visit our web portal at blockchain.ieee.org.