ep 8 – Top Tech Topics, Explained

Transcript

[Music] welcome to web3 with Asics and z a show about building the next generation of the internet from the team at Asics and Z crypto this shows for anyone whether researcher developer engineer artist company leader Community manager entrepreneur other Builder seeking to understand and go deeper on all things crypto and web 3. we're back with all new episodes this coming week but I'm sharing yet another hallway style riff that a bunch of our researchers did touring through and defining at a high level top of Mind technical topics such as PDFs or verifiable delay functions ZK Roll-Ups synarks and zero knowledge in general zipping up and down the stack and across themes such as scalability

data availability and reputation in the Creator economy including discussing applications for nfts and more it's a quick tour through explaining key topics that also sets the stage for the next batch of content from us and podcast episodes as well as our newly launched YouTube channel which you can find And subscribe to at YouTube by simply searching A6 and Z crypto research there you'll find videos from researchers with overviews of vdf's zero knowledge snarks and much much more that we discuss in this episode and Beyond but this riff is based on a longer convo that originally took place on Twitter for a live audience on the heels of our research lab announcement a few months ago and it

features all of our current research team except for one person who couldn't make it last minute then but who will be in our next podcast episodes It's moderated initially by Asics and Z crypto General partner Chris Dixon followed by Ellie Yaya the first voice you'll hear is head of research Tim rough Gardens followed by Chris's and then Joe Bono Scott commoners and then Ali again be sure to subscribe to the new A6 and Z crypto Channel on YouTube for more on the topics that follow as well as our newsletter on substack and stay tuned for the next all new episodes [Music] so maybe just to add in a little bit from the researcher's side there are areas and areas of computer science which are mature enough where you can read the latest papers from researchers

and know what the most important problems are to go next but the Science and Technology web 3 is nowhere near that mature yet and so if all you do as a researchers look at the academically published papers out there you're not necessarily going to know what are the most important research problems to work on over the next few years you know and we really firmly believe that this is a new discipline forming before our eyes in real time and so the opportunities to do fundamental research are just huge I expect research done in the next year or two to already be taught in undergrad classes by the end of this decade yeah just to add to that the reason at least for me I always think about deep lines and open Ai and these modern AI Labs as aspirational goals what I find so inspiring about those organizations is I

think of it as at least three things one they have very legitimate and respected academic research they put out papers and my understanding is they do it in almost a completely open source way which is definitely how we want to operate by just advancing the field but two they also then do applied things they build software but then the Third Leg of it they do documentaries they alphago stuff things like this like our viewers this is important because to Tim's point it's striking in Academia frankly how skeptical a lot of computer scientists are of this field and I think one important lever for doing that is use cases that have utilities societal value Etc and that get a lot of attention and I think those AI organizations really did that so I think it's a really great model to Aspire to maybe with that we could kind of get

into some fun technical topics to starting with you if you could tell us a little bit about your background particularly for those who don't know he's one of the pioneers of a field called algorithmic Game Theory which is I think a really cool area that combines the scale of the internet with older ideas and Game Theory maybe Tim if you could talk a little about that and then we'll move on after that sure absolutely so I did my PhD it was like 99 to 2002 and at that time the internet was really just starting to explode is really starting to become commercialized and widely used so it was a really exciting time to be a computer science graduate student because computer science was really changing in a big way it was obvious to everyone across the discipline the computer science needed to be different and one of the things that happened

around that time is your scientists realized that they really needed to understand economics and Game Theory to reason accurately about the new computer science problems that were coming out at that time and there had not been much interaction between computer science and economics prior to then so I was I was lucky enough to get sort of in on the ground floor of that movement and you know my whole career my calling card has sort of been the connections and interplay between computer science and economics so for my PhD I was working on problems closely related to the internet for example understanding the consequences of selfish routing between different internet service providers when I was an assistant professor at Stanford I got really into a field of microeconomics known as mechanism design which kind of broadly is sort of making

good decisions when you don't know in advance exactly what people want what people's preferences are one of the the main application areas of mechanism design is the design of auctions auctions were becoming very prominent in computer science through business models of you know the big web 2 companies for real-time auctions for advertising there are also for example the options that governments run sort of cell Wireless Spectrum licenses and that's another really nice sort of fertile ground for the interplay between computer science and economics so with all of that backdrop naturally when blockchain and cryptocurrencies came along it's kind of like how could I not work on this right I mean I've never seen a technology where the economics are so deeply indebted in the technology itself so Bitcoin I found very striking when it

first came out it wasn't obvious to me what would be the most important research question to do at that point but I certainly followed the space from a distance throughout most of last decade in general the way I learned things is by teaching courses on it so it's pretty frightening when you first teach it because you don't feel like you really know what you're talking about but you do learn an enormous amount and so I started teaching courses on black James you know just really as a forcing function to get me to learn it sort of more deeply and in doing that by like 2019 or something I was completely sold on the vision of what I like to call a big computer that lives in the sky and sort of runs as a public good to sort of the vision of General smart contract

platforms what's cool is this super interesting open questions all up and down the quote unquote stack both from sort of very low level implementation stuff to the application layer so I've been trying to work in different parts of the stack you know I've been working some on consensus problems so really developing a theory of permissionless consensus to mirror the classical theory of the 1980s that was developed for permission to consensus and then also thinking some about mechanism design style problems at the application layer right so for example there's a very rich theory of optimal auction theory developed by Roger Meyerson he won a Nobel Prize for that I think one can Aspire toward having an analogous theory around some of the applications that you see for example in the defy space so like an optimal theory of automated

market makers would be one Holy Grail to go for another thing that was significant in my own involvement in the space was in 2020 the ethereum community reached out to me the talent had proposed this transaction fee mechanism EIP 1559 and some members of the community asked me to take a close look at it and I spent several months analyzing it and writing a 58-page report about it and so at that point Not only was I all in on the space but I think it was obvious to everybody that I was all in in the space awesome thanks Tim maybe let's go to Joe next job I know Joe maybe if you could tell us a little bit about your background like your PhD thesis and then how you got crypto pilled one of those many claim to fames as he originally was a crypto

pilled our own Dan Bonnet which is a pretty epic background yeah sure so I started my PhD in 2008 so the Bitcoin white paper came out I think my first or second year the main focus of my PhD was on authentication protocols for the web so all this had been kind of 20 years of speculation about how secure passwords were or weren't and all of a sudden we had data sets with tens of millions of passwords and you can really do fine grain statistics and look at you know not just how strong passwords are but in a whole bunch of different contexts for different types of users for different types of applications so the main thrust during my PhD was analyzing all that data figuring out how to collect it in a privacy preserving way of tying it to

the field of crypto but I had this side interest in blockchains the whole time the story I tell us we had a Friday research group meeting every week and one week somebody canceled and my supervisor on like Thursday afternoon at the end of the day said um you know we need something for tomorrow can you talk about this Bitcoin paper that first came out and I never heard of that so I was like oh okay sure the paper is only 10 pages long and I was like yeah I can read this by tomorrow and present it for the group and then I'm turning into this late night Affair trying to figure out the Bitcoin paper is pretty easy to read but it leaves a lot of detail out so I was up reading the code and reading some of the early Forum posts trying to

figure out how things actually worked in practice so after I finished my PhD I did two postdocs one at Princeton which led to this textbook that it co-wrote with a group of people at Princeton like Tim said teaching is really when you learn trying to write that textbook really required digging way deep into a lot of stuff that was really not documented at the time I realized that a lot of professors were reading the textbook which is a little unusual for a textbook I guess but a lot of people wrote and said oh yeah like I've been meaning to catch up on this blockchain thing for so long but it was confusing and then I you know just decided to try reading the textbook and then I got really interested in this space so hopefully the contribution of that textbook was convincing a lot of

academics in the mid-2010s and then yeah I did a postdoc at Stanford with Dan Bonnet and that was when Dan was first getting interested and he was you know slightly skeptical but I think he was receptive and I was like we can find some real crypto problems that end up leading to the work on PDFs which is probably in the blockchain in space of the technical thing I'm most known for is starting to work on verifiable delay functions and looking at different applications to Randomness protocols sure would you mind explaining what bdfs are yeah sure so VDS verifiable delay function it's a crypto function where the main security property is that you can't compute it in less than a certain amount of time so it's an inherently sequential function we don't know how to speed it up with parallelism so that the

only way to speed it up is by having a really some processor with a really fast clock speed which is expensive and hard to do you can't just throw a lot of processors at the problem and it turns out to be really useful in all sorts of different protocol designs the class of example is there's a lot of commit reveal protocols where everybody's supposed to commit to something which could be random input to a Randomness protocol it could be a bid in an auction and then you know you have to have some mechanism to have people actually open their commitments and reveal with vdfs you just have a slow function that opens the commitment so you don't have to rely on people I effectively opening them like anybody can open them at a certain period of time and the time delay it covers Network latency and made sure

that people don't have to post it exactly the same time but that's kind of okay because they don't have enough time to open everybody else's commitments or you know test out different random values they can post or whatever the security property is for the specific protocol for crypto it's a bit weird you know that you design something to be slow usually you know we want everything in crypto to be fast but uh yeah so we identified there are a couple applications and then kind of the way crypto research usually starts is you you spec out an idealized functionality like if we had a function that had these three properties then you know we could build some really really cool stuff and then you know we propose the first construction and said yeah you can actually build this thing connect up the space often since then a lot of other

researchers have proposed different vdf constructions three or four years now since the first paper came out we've had like a something like 100 papers on vdfs and whole workshops on it there's about 10 different sub types of edfs with extra additional properties so it's been a really fun crypto space to see developed and hopefully they'll be deployed more in Practical protocols soon great thank you Scott we'll go to you next if you could tell us a little bit about your background and early academic career and then we could jump into your interest in the crypto sure I studied Market design actually one step back I did a lot of work in number theory in college so I am also excited to be getting back into the world that borders on number Theory but my economics research work has been on Market design

which is the field that thinks about how you structure and design economic systems to make them run better you know in the same way that economics is a way of looking at the world and thinking about what the forces and incentives that drive different outcomes are Market design is sort of like the engineering lens on top of that it asks okay how can we use economic theory and Analysis in conjunction with a lot of technical tools computer science and market clearing algorithms applied math and sometimes like a little bit of ethnography and sociology to better understand the markets we're interacting with how can we use all of these tools together with a fundamental understanding of the incentives to make the system work better and my PhD where work was very much about fundamental

research on the structure and market clearing mechanisms for what are called matching markets for settings where you have potentially very complex transactions where different individuals want to care about whom they're transacting with and how I spend some time thinking about these fundamental theories and market clearing mechanisms and auctions like Tim was talking about but then I also spent a lot of time helping companies figure out how to use those ideas to better make their platforms or marketplaces work so when your two-sided Marketplace platform for online education or something of the store and you have like these challenges around like to what extent do you keep pushing the content from the creators that are already well established versus like bringing new creators and sort of

give them a way to to gain traction you know and they're often tensions right you're short-run incentive is typically to push the Creator that's going to get the most views that you think are most successful but the long run health of your platform depends on your ability to optimize ice for the long run to bring in new creators to constantly enable someone to be successful on your platform and so Market design thinking is sort of about trying to understand how to balance these trade-offs and how micro details like how those incentives produce the outcomes you want as a platform and this type of thinking of course translates very naturally to crypto most crypto projects have some system of incentives that a protocol or a crypto based platform is going to create and

you can set them down transparently you know you can commit to them by putting them in code that everyone can audit and interact with sort of as guaranteed to function at a fixed state but you still have to get the incentives right and so what's drawn me into crypto is that so many of the businesses and public projects have at their core a fundamental intellectual challenge that is exactly the type of challenge that we face in Market design but I should say also early on I was aware of this stuff in the late 2000s early 2010s but I was actually on record as a big skeptic because so many applications that people were pitching at that time right like entrepreneurs would come to my office hours and say hey I've got this idea for a project but like the projects weren't

using the technology in ways that were fundamental right there was a lot of we're trying to take this thing and just now make it decentralized but decentralization is costly and they weren't gaining a benefit to offset that cost and and what's made me so excited about the space now is that we're causing huge waves of applications and again the design and balance between degree of decentralization and like what things live on chain what things live off shape economic ecosystems that matter are there like particular projects that turned you well sure so the first project that I got involved in Christian catalini invited me and also one of my former grad students Ravi jagadisan to work with him and many others on the

design of the DM protocol the idea that you're going to have a fundamental transaction layer that sort of is backbone infrastructure but the same way that the networks that manage the internet are backbone infrastructure for like all of information these days like that's a case where the underlying technology is just like the only way to do it in some sense right where everyone can transact with anybody else in a way that's completely trustworthy independent of like how you're interacting but who you are it sort of has to be in a system that is heavily robust and many times replicated and also baked into software in a way that people can understand and interpret and that's fixed and committed and the other really cool thing about the idea of an infrastructure layer like that that lets

you extrapolate from places that have really strong institutions to places that have weaker institutions right there's potential to enable a stable medium of Exchange in places that have weaker monetary policies they can sort of lean on the institutions in the places that are doing the processing so that was like the first thing I got really excited on and then more recently I've gotten very very engaged in nft communities and consumer crypto and what's fascinating here and again this is an Insight that Christian and Ravi and I wrote about briefly in a short paper what's fascinating about nfts to me is that you know a currency is only really useful as a medium of exchange once lots of people that really do exchange goods and services for it you have to build a really powerful infrastructure layer that everyone

agrees is going to work and they coordinate on before like everyone can gain the value from a given cryptocurrency by contrast nfts are formed around small communities right all that really matters initially is whether the people who might want to own that piece of art agree that the token means ownership and so this means that nfts can establish their value among a small community of holders and then grow outwards which is a really powerful consumer design Paradigm actually I think Scott what you were just saying was super fascinating maybe for the rest of the time it'd be fun to go down the stack and talk about some of the open problems at each layer we should start at the top which is started the application layer and at the risk of keeping you talking for too long I would love to start with news you've done a

lot of work thinking about incentive structures around things like nfts and around reputation systems one of the things that I find so important about that is that one of the key problems in the space is that because we don't have a good mechanism for reputation it's as if every interaction is forced to be a one-off prisoner dilemma Style game as opposed to a repeated or iterated prisoner dilemma Style game and so that essentially forces you to assume that everyone you interact with in this page is an adversary which very much constrains the kinds of mechanisms that you can design because it's a very severe assumption that you have make awesome yeah so certainly one of the things I've been working with teams on the most I'm thinking about the most in

my research is how you optimize your token design for the use case you imagine in your ecosystem and this could be as simple as if you have a utility token there are challenges around money supply right if you print too many units of the utility token then suddenly the value crashes and maybe people don't want to collect it but on the other hand they are willing to spend it and meanwhile if your utility token is too scarce or people to see too much of the utility Downstream then maybe they all hoard it and no one spends it and you have no circulating Supply and so you know the overall ecosystem can collapse so there are simple macro balancing questions of that form and then there are much more subtle ones you alluded to this I wrote this piece on reputation tokens with one of my former students

Chad esper and we talked to lots of different companies that are thinking about how do you use tokens to generate a durable system of reputation right and they're a Creator platform and they want to reward people who've contributed content or they are job search platform and they want to like reward people who are helping other people find jobs or something of the sort and a lot of platform Architects pick a token system where they are hoping their token is going to do two things one is it's going to identify some people as having high reputation like they're contributing a lot of value to the platform but at the same time they're hoping to like allow people to liquefy some of that value it's a way of rewarding your contributors to the platform what Chad and I wrote about is there's a funny

Paradox which is if you want to give people reputation that's durable but also want to make it liquid and you only have one token then suddenly now it's not a good signal of reputation so we propose like a two token model and we talk some about how you balance these sort of like a durable reputation token that then spins off like a liquid token that you can use sort of as a medium of exchange but so optimal tokenomics is one let me just point to one other application layer Challenge and big opportunity that comes with it and then let's maybe move down a lot of the promise of web3 comes from the idea that we're going to enable people to use the interoperable layer that is the blockchain to sort of create established

digital assets and Forbes of digital ownership that we can then point lots of different applications to and people can take with them from application to application right and that has a lot of problems in terms of competition policy for example like consumers gain a lot more leverage if they can take the content they've created it with them and move to another platform but for all of this to work well we really need the transaction costs of the layers to be low and consumer technology around the blockchain to be much much more accessible and so there's a ton of design at the consumer access level that we don't normally think of as part of the technical challenge but it's very much part of the economic design challenge Scott I'm really glad you brought up the user-facing design

problem that's clearly one of the biggest challenges facing the space among the more sort of technical types it doesn't get spoken about enough but one of the things that needs to happen to reach one or two orders of magnitude more crypto users is exactly some advances in the user interface design part yeah we can all chime in with other ideas other interesting problems at each layer and we can just sort of take it from there one natural place to go next would be the scalability problem so Layer Two Solutions so the way I think of this back to this analogy of the big computer that lives in the sky we want two things right one we want really cool applications deployed on that computer but then we also want that infrastructure this virtual computer in the sky we want it to be as powerful and

robust as possible and there's going to be a co-evolution of these two things so the better the infrastructure like the more transactions per second it can process the bigger the design space for applications similarly you know once certain applications sort of Catch Fire there'll be extra motivation and capital for improving the infrastructure so it's been one of the biggest areas of intense work over the last couple years Layer Two Solutions including side chains including Roll-Ups really trying to sort of boost the transaction throughput of a platform like ethereum by ideally at least two orders of magnitude and hopefully eventually forwarders of magnitude or more and so here again you know there's a lot of I think really nice competing designs for example in the roll-up space you've got the game

theoretic approach taken by optimistic Roll-Ups you've got the sort of more proof theoretic or cryptographic approach taken by GK Roll-Ups right just understanding better the pros and cons the trade-offs between those two approaches I think there's opportunities for fundamental research and certainly there is ongoing a lot of really amazing deep research about how to get VK Roll-Ups like ZK snarks as practical as possible you know for me someone who's trained as a theory of computer scientist this is just the most amazing thing to see because these proof systems on which ZK snarks are based they were introduced in the 1980s just by mathematicians really just to study the mysteries of the computational universe really just to understand what does it actually mean to prove something and now

fast forward one thing that's been amazing about crypto space is all of these 20th century areas of computer science have just gotten you know it's not like they ever went away but they all just have these completely new lives at this point through the new applications and so snarks which can be thought of as a specific form of proof system which now billions of dollars is going into heavily optimizing you know that's another great example of the interplay between fundamental research on the one hand and very real problems in fact technological challenges that are literally essential for web 3 to realize it's full Vision damn I think in the context of Layer Two it may be interesting to just briefly touch on the data availability problem and also on the applications of proposals like

eip4844 and how that might significantly improve the performance of layer 2 Solutions today sure it's a really nice sort of research question here so let me focus in specifically to ethereum as I'm sure many listeners know in ethereum transactions cost gas so basically there is a single metric single type of unit which is called gas and it's mints as a proxy for all of the burden imposed by a transaction on the ethereum network and that burden is really multi-dimensional okay so it includes things like computation it includes storage it includes bandwidth that the full nodes running the protocol are using to communicate with each other so you have this really multi-dimensional cost of a

transaction and yet that projected down into this one-dimensional space you have to use this one number gas as a proxy for all of those burdens so one of the key ideas in Roll-Ups is to solve the data availability problem so what is the availability problem so basically you have this separate blockchain this layer two and it claims it processed the Thousand transactions it tried to help out layer one it's like hey you know layer one let me take some of the burden off your shoulders let me sort of process these thousand transactions and let you know what are the end results right and that's fine I mean that's that's very nice but then later on how does anyone know actually what transactions were executed or whether they were executed correctly and so data availability is like where do you find

out the information of what actually happened what were the transactions that were executed off chain and so in the roll-up solution that stored is what's known as call data so it's basically just published to L1 using a particular type of relatively cheap data and anyway so some of the proposals going on it's like well if having room for lots of call data is essential for you know solving the scalability solution maybe we actually want to think of things in a more multi-dimensional way and so maybe call data coming from Roll-Ups at layer one we can sort of treat as cheaper we can increase the amount of capacity at layer one to handle the publication of call data while leaving the other parts unchanged rollups and call data is one particular instantiation of this but

much more generally thinking multi-dimensionally about the cost of transactions and pricing accordingly I think it's a super exciting direction for research I've thought a lot about measuring decentralization and more of like a maybe not even quantitative but at least like a systematic way doing transactions Layer Two with like ZK Roll-Ups there's like a little bit of a trust placed in the roll-up server if you do optimistic roll ups the trust model is slightly different there's a whole lot of other Layer Two protocols you know like payment channels is another one that has a different set there's clearly a trade-off between you can get better performance at the cost of introducing some new parties into the mix that at

the very least you rely on for reliability you might also rely on them for some other properties too that you hope that they'll exhibit for the rollout to work the way you want but we don't have a really systematic understanding of like how much performance can you gain for what trade-offs in the trust model and aren't these like actually fundamental limitations worries it's just like the tech we know how to build now so that's kind of like the high level that might actually be a good segue to thinking about how Layer Two Solutions on top of ethereum compare to just alternative layer ones here's what you guys think and I'd say quickly there's a really similar high level question just between layer one but what the real trade-offs are well we're still at the point where people don't even really

agree between the existing Solutions just between proof of stake and proof of work we don't have a really good common understanding of what you're really sacrificing in the trust model by going from who work to proof of state that is not even clear I guess that the sacrifice in One Direction or the other like they're each have a different trust model in which there are some advantages to both under what conditions the consensus protocol is going to behave correctly and then you know and all sorts of different different proven states of limitations for example it's like we don't really have a good way of saying like this one's slightly more robust to the cost of worse performance or that one you know is getting performance that is making likes things longer trust system functions so yeah I mean pretty much the same problem like

even we don't know how to compare the existing Solutions in a way that everybody agrees on and pardon at us because there's you know a lot of conflicts of interest and people want to say their Solutions great and kind of Overlook maybe the sacrifices they're making and Trust model and we certainly don't know what the theoretical limits are on like do you have to sacrifice a little bit of like do you have to weaken the security Model A little bit to get good performance we don't know the the fundamentals of it it might be trying to have a brief internude about zero knowledge groups specifically because they are very fundamental building block that is also very general I feel like uh the applications of zero knowledge proves being a thing may be as dramatic as the implications of smart contracts when

they've emerged back when ethereum was created I feel like you're only now starting to appreciate just how powerful their knowledge proofs could be and they also feel a little bit like magic and so maybe you can unpack what does zero knowledge proof is at a high level what it allows you to do and what you think are some of the interesting applications or implications of them as a technology Joe's more the cryptographer so I'll defer to him but Joe would be great to comment on but there's your knowledge proof systems have been around for several decades and I wonder your thoughts about I mentioned earlier about these 20th century parts of computer science finding new life with blockchains it's been in crypto and and I wonder if similarly somehow black actions in their public nature may be

really what makes zero knowledge proof systems go completely mainstream I'd love to hear your thoughts on that yeah sure blockchains have been the thing that brought zero knowledge proofs mainstream the core idea is just that you can prove that some mathematical statement is true and that you actually know what's called a witness so basically like a solution to some mathematical equation and that it doesn't have to be a math equation because you can express a lot of other things as a math equation so the example that I deal in class A lot is that you can prove that you know the solution to a Sudoku puzzle without actually revealing the solution so the zero knowledge part is always that you're proving something's true but not proving anything else you're not revealing the solution or revealing what you know but

you're convincing the other party with overwhelming probability that you actually know it I mean yeah the idea has been around since the 80s and it was a turing award-winning idea it's a pretty clear fundamental breakthrough that you can do with zero knowledge proof at all you'll be known for a long time that you can do it and you can basically do a zero knowledge proof of anything I mean you can do a zero knowledge proof of any NP statement which is almost anything interesting it doesn't guarantee that there's actually an efficient zero knowledge approved protocol you know especially like a lot of the early constructions were sort of more proof of concept it was like you could do a zero knowledge proof but it might take like 10 000 compute years to compute the thing so the application started with really simple things

relatively like proving that you know a discrete log of some group element or something like that but people have come up with generic proof systems where you really can compile any and a statement down and there's all sorts of natural applications in the blockchain space you know you can prove that a transaction is valid without actually revealing what the transaction is like who's paying who or how much so you can prove that the result of executing some really complicated smart contract is like these changes to the state this money moves from that party to the other and maybe don't even reveal the amounts because all the amounts are hidden and that means you can really potentially solve the big privacy problem of blockchains which is that you have to make all the

data publics that everybody can verify it if you're just verifying zero knowledge proofs the transactions rather than publishing the transactions and re-executing them yes obviously much better for the Privacy so you can get the benefits of public accountability and also have everything be private kind of a long road to get there technically though the current tools that we have just aren't aren't ready performance wise they have anything like the transaction throughput that we want for a practical system yeah that's exactly right Tim I thought you actually get a pretty good breakdown of the properties of is there a knowledge proof which may be helpful maybe even actually just starting with like what the acronym even means in the case of a snark sure yeah no happy to talk about that so popping up a level

and this may tie some of the different threads together a little bit is I mentioned these proof systems and there's really two different aspects that are relevant for blockchains one is the sort of succinctness aspect and the other is the actual zero knowledge property so when I was speaking about challenges of Layer Two and optimistic versus ZK Roll-Ups there I was focused on proofs that are succinct by which I mean someone does a whole bunch of work and convinces you they did a whole bunch of work without forcing you to recheck their work you know you don't have to redo their work from scratch so that's what I mean by succinct you can kind of save someone a lot of time you do computation and then you just sort of convince them that you did it all that's not about privacy just that technology and then zero knowledge is what Joe was

talking about that's really a cryptographic aspect which says that not only do you convince someone that you did something that you know something but you don't even actually leak what it is you know you convince them that you know it but you don't actually teach them what it is it's so that's just the first thing so ZK Roll-Ups is a bit of a misnomer often ZK Roll-Ups are used to describe proof systems that don't actually have zero knowledge and so sometimes people prefer to call them validity proofs instead of GK proofs for that reason and so snarks you can add ZK on top of a snark but already just the basic sort of definition of a snark is pretty amazing so let me pop up a level so the print system like I said originally it was meant to study the computational mysteries of the universe

really like what does it mean to prove something to somebody and so the formalism you use is that of two parties sort of a prover who's relatively powerful and a verifier who is relatively computationally limited this is actually an unbelievably good mapping on the layer one layer two interactions in a blockchain where the slow low throughput layer 1 plays the role of the verifier plays the role of the actor that has very limited computational power and then the layer 2 would play the role as the prover so that would be like a chain that maybe has higher trust assumptions but also has higher throughput and so the idea of approver trying to get into verify or something true again it maps on perfectly Layer

Two wants to convince layer one that it did a bunch of transactions accurately and so that's why these snarks which came out of the proof systems literature sort of the perfect fit for layer twos that proactively to layer one prove that they did what they said they would now this context usually that just means like there's a thousand transactions we executed all of them and they published a new state route to layer one that would be the typical computation okay so as far as the acronym snark so the S stands for succinct what that means is that it better not be the case that for layer one to check layer two's work it has to literally redo all of layer 2's work from scratch that would totally defeat the purpose you're trying to take some of the load off of layer

one's shoulders and so the snark should be small it should not take up much space at layer one and it should be cheap to verify for example in ethereum you shouldn't need much layer one gas in order to verify that in fact it is a correct proof so that's the S succinct so basically just you can check work much faster than redoing it from scratch the n stands for non-interactive so in the original definition of proof systems in the 80s the Fable that was often told back at that time was like imagine a tutor and a student having a conversation so the tutor here would be the more powerful all-knowing prover the person being tutored would be the verifier and there's a definition of like what kind of conversation is a

convincing conversation so you have these rounds of communication back and forth so that was the original notion of a proof system but if you think about it you don't want Layer Two and layer one to be talking back and forth like for a long time you can't afford to have this interaction you want it so that anybody can download these software to run the ethereum protocol spin up a node and like immediately sort of verify everything that has happened up to this point even if it was a snark proof published by some L2 that was defunct as of five years ago so non-interactive just means one and done right so really the layer two just says here once and for all publish to L1 this is something anybody can verify now or in the future and then the arc the AR stands for argument so that just means it really

should be a proof meaning that the layer 2 can only generate a snark proof if in fact it's correct your detected really did do those computations and conversely it should not be able to spoof a correct proof so if it didn't do the computations it claimed it did then the verifier should not be tripped the verifiers should say aha I see the bug in your proof I don't believe you did all of those transactions so that's the AR and the K which is the knowledge part that's what Joe was talking about that was like the witness and so that again comes from zero knowledge considerations originally but actually even in validity Roll-Ups that are not zero knowledge that's actually a really useful optimization for compressing the layer one Footprints for example normally to execute transactions you need to make

sure you have the appropriate signatures from the people initiating the transactions and you can actually kind of sort of not publish those signatures to layer one in the validity roll up because they can serve the role of that witness so in effect Layer Two promises that it new signatures generating that sort of authorize the Thousand transactions that it executed along with the final state after it did all of that execution so that's kind of the guts of what one of these stock proofs actually looks like and that's the meaning of all the letters in the acronym you might say the term snarring sometimes instead of star so there's a difference between a proof and a proof of knowledge like in the Sudoku case I can prove to you just that this Sudoku puzzle has a solution that it's not a broken puzzle versus

proving that I actually know the solution and turns out mathematically those are two different concepts so there are like what are called snarves which are arguments that don't have this knowledge you're just sort of proving something's true without proving that you know something which can be a little bit easier to construct and a lot of the time that's what people actually use so I think we've got to the point where people say snark for everything even if the a is not really there but that's kind of like an upselling property one thing I I wanted to emphasize which is something that you both said but I think it's just so important it's just like the concrete implication of this at least one concrete implication of zero knowledge proofs for blockchains is that if you look at the way blockchains work today every single node in the network

essentially has to re-execute every transaction of every computation in order for the network as a whole to build confidence that the transactions will perform correctly and one of the implications of a zero knowledge proof is that with this technology you can have a single node run all of those transactions run all of those computations and then generate a proof a zero knowledge proof that those computations were done correctly and then everyone else in the network just has to validate that proof and that can be done much much more efficiently than actually running the computation itself that's a succinct property that Tim was was referring to and so that has massive implications to the way that watches get architected to the potential performance

and scalability improvements that we can see on blockchains in the future it is also a key aspect of the implementation of Roll-Ups like zero knowledge Roll-Ups that are based on this insight so just wanted to emphasize that Scott I know you've done a lot of work on the Creator economy and there's like a lot of overlap between that whole body of work and crypto1 with three awesome it's very much ties in with these questions because it a lot of it comes down to access and infrastructure so first of all just go back to nfts as a conceptual example like what are they doing they're enabling types of transactions that have been either non-existent or just very difficult to do in the past so if you create a digital good sort of pre-modern

crypto infrastructure there was no way to like Define the owner of that good absent uh centralized intermediary that was actually just defining ownership right so we had ledgers of digital Goods in computer games other sort of like closed economic ecosystems but if you were a Creator whose creative form was digital whether it was digital art or digital music or certain forms of writing and things you didn't have a way to sell the digital asset because there's no way to define the owner and so one fundamental use case for nfts is a sort of digital deed and because they exist in the blockchain in a way that is transparent and verifiable and builds on all of this interoperable transaction architecture one can create an nft or can create a series of them around like

whatever your creative activity is now suddenly you can have a digital Goods economy and you don't need an intermediary like a label in the case of music or a gallery necessarily in the case of art certify you and your assets and people can buy and sell it and if you can build a community of enthusiasts around your work just like I was talking about before you could spin up a broader ecosystem as an enthusiasts and so we're seeing this emergent Creator economy built on digital Goods and so that's one really powerful opportunity but of course we've pushed trust to a different level right so talking about consumer protection how do people curate and keep safe their digital assets while engaging in this Marketplace all of these are contacts where there are incentive design

challenges wrapped in technical challenges right like what can we do to give people the ability to interact their digital assets with a variety of different online applications without putting them at risk right now you don't know what you're signing and so you don't know what you can interact with necessarily can we create technology that wraps your assets in a way that keeps them safer and you can like do more of these asset interactions with less concern and then on the Creator side this enables creators to enter more easily and also it enables them to be platform independent right if you decided suddenly you wanted to like transition from YouTube to tick tock maybe you could export your content you could reformat it and so forth but you can't bring that reputation with you web3 at least in principle lets you do that all of these assets belong to you

and the reputation that builds up around them like you control it and you just point different applications to it this semester I've been teaching a lot of our first ever HBS course materials of this stuff I wrote the first case that nfts or the board a yacht club case I wrote one on the luxray or vampire attacks anyhow what I show to people when I'm trying to explain to them what makes crypto powerful to me is all the ways in which crypto is either enabling new types of transactions that we always wanted to do but didn't have a technology to do or enabling us to do existing types of transactions in more efficient ways and all of this at some fundamental level comes down to if you want to have markets in something you have to have property rights and we didn't have property rights over these digital

assets but now we do or at least we have a way to create them and so similarly this idea of smart contracts as trusted code it's funny Tim talked about going back in time into classic computer science we're also going back in time to Classic economics right like the earliest models of contracts assumed trust and assumed the contract was going to behave as planned and so now we're living that again or finally rather thank you for listening to web3 with A6 and Z you can find show notes with links to resources books or papers discussed transcripts and more at asicsandcrypto.com this episode was produced and edited by Donald that's me the episode was technically edited by

our audio editor Justin golden with seven Morris credit also to moonshot design for the art and all thanks to support from Asics and see crypto to follow more of our work and get updates resources from us and resources from others be sure to subscribe to our web 3 Weekly Newsletter you can find it on our website at asicsnzcrypto.com thank you for listening and for subscribing let's go [Music] thank you [Music] foreign [Music]