Zero-Knowledge Proofs Panel

Transcript

hello okay thank you all for coming to the zero knowledge proofs panel so we've been doing like zero knowledge booths for like five or six years and like today we gather together some of the like key people pushing forward zero knowledge proof technology so I'm gonna let them introduce themselves so vitalik yeah hi so I'm uh I'm vitalik I've written a couple of uh blog posts and like implementations trying to explain various uh zero knowledge proof uh protocols I contributed exactly one idea to the original blank paper so you know I've kind of been on the sidelines watching

this amazing um you know space Grill and it's been really fun the last five years uh hi everyone my name is Crow at scroll we are building a bico level compatible with the kbm so we are using a lot of rare Advanced like they're not approved algorithms to generate proof our dick evm we are using a lot of fancy optimizations to build out of the whole circuit um yeah and personally I'm more doing VK research and some protocol research hello I am Jordan Molina I'm the technical lead at polygon ckvm and besides that I am also the author of silicon and snarl.js among many other things but this is the

main work related to the game hi Ellie Benson I started I fell in love with the cryptographic proofs more than 20 years ago when I was doing my postdoc at MIT originally on things called pcps probabilistically checkable proofs since then wrote a bunch of academic papers which led to things like zcash which I was a co-founder and a star core of which I'm a co-founder so thank you Ali so my first question is for you Ali so over the last couple of years there's been a lot of things that we've wanted

to do with the kitties like the first like main use case was zikash and like the performance limitations there were it was a very small circuit and the performance limitations were were quite important design consideration and it's it's like five or six years later and we have like zke evm and we have like all these other like incredibly complicated things like could you talk about like what's the blocker for us when we try and make SDK applications so I think today the blocker is not the generation of proofs it is the um connecting like having you know

end-to-end Integrations and tooling for developers um because the proofs are not the generation of the proofs even on existing Hardware can be done very efficiently so I think that's the main blocker really just having an end-to-end system with a lot of tooling around it that works uh you know from the developer cycle to testing to deployment to Integrations for end users I think that's the blocking thing okay that makes sense um like my next question is for everybody I'm curious like what is the thing that you want to do that work currently just not able to do do you

have some examples of like applications you want to build that like it's just not possible to build right now for me I I would say I like to compare I'm an electrical engineer so I like to compare the things in between soap water and Hardware okay so we are in an era of writing Hardware so we say circuits circuits is it's like okay we write Hardware not with transistors we write hard words with constraints or polynomials okay but we we are in the era where to build things we built Electronics we want to do a calculator and then we take some components some chips we connect them together and we get a calculator

okay but now we are starting with some How We Invented we are starting bending or we'll start working with a processors that means that we will enter to an Era of uh of the software so where the people will be able to just program normal statement or normal circuit or a normal program in nor in in a regular software language in rust or in C or in in things like that we are at this moment where we are transitioning from one stage to the other so for me the interesting thing and the thing that I want to see and that we are started to

seeing now but this is I think what will Revolution a lot the ZK is that we will not have to write Electronics anymore and the users will be able to use normal uh compilers uh normal compilers that will compile on top of these ZK processors and execute that normally this will open uh this will change a lot the ZK um the designated space and we're in this transitioning space actually a zkbm is very much a processor is a VM processor oh it's interesting to hear that like you both think that like

tooling is a limiting factor um yeah or vitalik do you have any thoughts about like things we want to build but we're not able to build right now with our current proof systems tools things that we want to build that we're not able to build right now um I think I mean there's there's two types of uh DK like synark usages right there's scalability related things and there's privacy related things and I feel like they have some like often pretty Divergent considerations um so like for example like on both sides I mean pervert time is obviously an issue but on the Privacy side like we're talking about prover time on a on a user's computer right like on you know

a phone inside of a browser app and aside from that like the applications tends to be fairly simple and like a lot of it is just reusing very similar code bases to what have already been used but then on the yeah scaling side the uh what people generally look for privacy it seems to often be the case that people just want to keep one particular like the same type of stuff private over and over again like just proving that you're a member of a set for example but for scaling like people want to make totally different things scale right like there's proof of solvency there's eke VMS there is uh like ZK other parallelizable VMS and uh there's just like it just involves writing a lot more

code it involves like the improver time challenges also but in that particular case like you you can parallelize more right and so the kind of stuff that you have the kind of code and like like proof system design that you have to do it's a really optimized for ability to Pro wise and like assume that you have hundreds of gpus and get the answer together very quickly as kind of different from those other cases so I feel like what in both of those situations there's a bunch of different like software things that exist to some extent but that we have to keep leaning on a real and kind of pushing really hard but like what

specifically they are ends up differing between the two oh interesting so I suppose I disagree with you about the the assumption that private applications are going to be inherently super simple that is just like proving the same thing over and over again and I think that like I'm the thing I'm a little bit disagree is because I feel like that's all that's been possible to do until now things like semaphore like just proves of group membership or revealing simple things but I think that when we if we lived in a world where like more things were possible people would find like more interesting things to do what's your uh favorite here's case of more complex uh privacy stuff okay so the simplest the simplest would be an ore that I'm an or

gate like I'm either part of this group or I'm part of this group yeah and like and then you can just build on top of that with like more complicated like I guess like my opinions of that sort of stuff is that like all of these more complicated things that involve composition it's like like my suspicion of that category is that it didn't like it it makes the Nerds happy to think about those kinds of things and it makes the Nerds happy to know that like oh you can do Bully enclosures over everything and look it's like logic from second year university but like in reality like 80 of the use cases will be the most simple stuff so I don't know if that will be true in this category but like that would be my fear Maybe

okay I hear you and like I definitely agree with the idea that like simplicity at first is super important that like starting out with building simple applications is important and later we can see if in fact there are more complicated things yeah I mean I think this like this stuff could easily change as time progresses right I feel like I needed like describing current reality then I mean than making predictions and like you know it could totally be that in five or ten years things look very different okay thank you um my next question is for yeah so yeah I want to ask you about zkp security you're currently in the midst of like this massive audit for ZK VM and I'm curious like what are are you afraid

about dkp security are you like what things can we do better yeah yeah yeah they are like you know if you are talking about the security of the owner proof they are like multiple levels right like they are a theoretical level because you are using some proven algorithm so I really really wish people to use more standardized Queen algorithm which is like you know you have a clear soundness proof you have clear in a complete proof you can't do whatever arbitrary things you you want like you need some form of proof on this regular side to prove that this crew system is sound and secure and but but My worry is more on the actually on the implementation side so if you are using the learner proof in practice there will be two things one is on the circuit side so you because you are using different

tools you are you have a totally fragmented you know leaky ecosystem like you know this crowd of people are using this library and we are using a different library and all the toolings are very hard to use so like underwriting circuits it's very easy to meet some constraints or even overcome some constraints so that will be really hard to kind of audit and currently most of people like you know still use some manual auditing like you know line of code and you know country by that and some experiments are you know like we we give guns to some projects which are doing some form of vacation for some simple circuit but it's very hard hard to do a like you know four more like security proof for modification or whatever you know like two links you are using to formally verify that your leaky

VM is is is like you know secure so that will be very hard on the circuit side because I think very few people realize that because more and more people are coming to the stake space right like people are kind of you know like they can spend like two days learning circum and write a ziki circus but a lot of people still don't realize that you know uh like how like the security assumptions behind if you are really missing some constraints and you are using a decade Rob then like you know the program can fundamentally do something really bad to kind of you know break down the whole thing so that's more on the circuit side on the backhand side there are also implementation issue um for example like I think uh previously there like 12-bit find some security bug in many Planck implementations Planck is already very

standard right but showbids still find a lot of like uh like security issues when you are implementing the fear Champion you just didn't hush some of the commitment but it already bring you some security bugs so that's really you know like make people feel worried like because different teams are already using different cooling system and if your proof system has been audited and then it might really you know cause you some security bug and another thing which I think is really important is that people usually ignore the security problem of their owner proof because security is not so explicit because people care about you know how fast ldqvm can be and like you know it's proof generation time in the five seconds versus one minute but they don't care about you know what's a trade-off

like if you take one minute to generate proof or 10 minutes to generate proof maybe you are more secure and so people don't really care about that because they care about user experience that they care how how fast so the The Matrix become very like you know single Dimension which is extremely bad because we really need to educate people like you know if you are even fine then like some of your security actually come from this harsh from some some other you know assumptions and which I think people already previously find some you know security issues for how many security Beats at that you know fry had it's not like you know you have this nice post Quantum security but you also need to read about how many security beats you have in like a normal world and uh and also yeah that's a big big issue like

people need to be really care about the because sometimes you can configure even it's a simple system you can configure different parameters to have like very fast approver but maybe a slightly smaller security you know parameter so that's something which I think like you know different projects need to be more transparent about and educate people like you know what you are really trusting so yeah okay I understand yeah there's a lot to be careful about on the security side my next question is like on a more optimistic note like what are people excited about what what are you excited about in the next few months or in the next years coming or of course and besides scaling ethereum I mean this is the the full point of the ckvm and we are getting to

this point and think that we are you know roll ups are something that are real thing there so we finally will have a scaling ethereum and this is I think this is very exciting by itself but besides that there are a lot of applications that for me I'm very excited here is for example the all the identity uh the identity applications self-sovering identity where you can hold your own data and you can prove anything about your data and things around that I think is very important application voting voting and governance systems Anonymous voting I think is an important interesting application I'm very excited on many people working on that direction there is a lot of gamings I mean the gaming is I think it's a full it's a

full industry you know the the decentralized gamings I think it's a full industry that has not take off yet and here Z case is really important uh for that it's a lot I mean a lot of interesting things that can happen with with CK so I I'm excited about three things all around starknet the first is um the performance improvements that are now coming very soon within less than a month we'll see uh TPS for transfers of erc20s on starknet in production of 100 DPS and over in our internal measurements we've clocked even over 250

TPS for transfers and this is all coming next month two months later we'll have volition with dramatically lower gas costs than when the EAP 4844 is ready users will be able to choose from on-chain data dank sharding and volition and so that's the first aspect I'm very enthusiastic about the second is the community that is dramatically increasing and doing terrific work projects like kakawot who spoke earlier Madara Dojo influence the many teams that are already deploying on starknet which is open for business so developers out there please check out Cairo 1.0

there is and the third thing I'm very enthusiastic about is the starknet foundation that is setting up shop it's already been giving grants in the starknet token to early adopter teams and this will expand and increase in scope so I guess all of this yeah please check out Stark net there's a lot to be done there and we need an even greater community yeah I think for me like very what very excited me like you know they as Jordy mentioned like prover keep being optimized maybe eventually we can generate the equivalent proof within one block time which means you can use the KVM on layer one language because you can generate this proof really fast so I think there will still be a huge

Improvement on the process inside they are clearly already to Trend like when they based on like you know the kiligence homomorphism like you know four different instances like people are really crazy about the other like you keep moving to smaller and a smaller field for fry and I think you know it's more and more clear that there will be some path for for people to go into and standardize their proof system so I'm very excited about that and I think a lot of things will happen like you know one depending on another another thing is about Decay Hardware like there are a lot of like VK Hardware acceleration companies or Explorations they are doing Asic GPU and ipj so they can make this proof generation become even faster so I'm expecting that some essay coming out to kind of make you know like some proof

system to become really really fast and so we can get more adoption and on the application side they are more and more like you know applications coming out they are like you know Decay Bridges like you know other Oracle they get called processor like you know so a lot of like those applications are appearing um I'm excited to see that they are being more and more adopted making this world become you know more trustless uh yeah and also like also for gaming like they are lacking some new idea how you can use AK to hide some information for unchain gaming and some strategy gaming like Dark Forest like where are you because every information on chain will be public and how you can do some free downloader game like you know they're on a proof I think there need to be some you know more creative stuff there

um yeah I think a lot of things are you know related and I'm very excited about the future of the AK um but Alec what are you excited about what am I excited by um I mean I'm excited by the infrastructure settling down to the point where like cryptographic verifiability and privacy can like just become defaults in a lot more applications like uh one instead of asking like why um you know add a proof to a thing it becomes why not add a proof to a thing and I mean we're pretty far from that right now because of like basically prover times and um I just uh I remembered but one of the reasons for my kind of short-term like especially

short-term opinion that privacy stuff is simple which is what has to be simple which is just because of like anything more complex is just going to have too many constraints and that's like uh a reality today but Manila could not be a reality five years ago and actually yeah take that like the example of like the Dark Forest style games might be the first uh big category of exceptions eventually but like if we can like make it you know like get a world where regular blockchain applications and users on blockchains have more privacy and that's like a standard for you know Enterprise databases of any kind so like including like balances on exchanges but also lots of other things to just like be some arctin have their updates as an

artist by default then you know that would be a really cool in a whole bunch of ways then it would solve a whole bunch of problems interesting so like is this like Facebook becomes a validium kind of direction right is that what you're thinking about I think that would be cool but like in order to do that like the users of Facebook would have to like sign their their own right well there's uh there's like different degrees to which you can do it right so like one one of the more moderate things that you can do as like you can do some like zkml to uh just for the ranking algorithm and uh like you proved that and you proved that

against like a hash of the algorithm and then one or two years later you release what the algorithm actually is and that allows some kind of a retroactive auditing to happen and that like gives a lot of security guarantee is already without actually going all the way um so you know like to me it's definitely a like a piece by piece thing and like the other thing you can do is like you can try to be pretty kind of explicit about what the holes are right it's like you know you give your users a key and then you have an operation where if the user voices their gear you still have the exact same recovery process and like Facebook still has a backup key but like there's like a big fat red record on chain of like a yeah a recovery operation was done right and like even

if I you know it's not fully self-sovereign like getting to that level like the getting to that level of transparency is already like a a pretty strong step well probably step two step one is doing the stuff I mentioned before right and the more you can go with them like the more benefits there could be okay that's an exciting potential future like a lot of the applications that we're talking about are use cases we're talking about like depend upon the fact that uh zkps are programmable in that you can make like a a program and a program you can basically put it inside the ckp and like a more complicated program will mean the prover time will be longer which is something that we're

optimizing for during this discussion and I'm curious like what there's other kinds of programmable cryptography in the world that like haven't really gotten the same level of adoption as zero knowledge proofs and I'm curious like if people on the panel are excited about a certain kind of programmable cryptography I mean it feels like the big programmable cryptography that's not zkb is that people sometimes talk about is MPC right like there's others but that seems to be the clear number two and my impression is that the biggest problem with NPC in practice is like you have to agree on this set of people to actually run the thing and like that creates difficulties for new projects to

get in and build things that creates controversy is like it's basically the same kind of controversy we had about around the trusted setup except instead it's a small group of people and instead of trusting them once you have to like be entrust them forever and they can cheat undetectably right like if the NSA compromises two-thirds of them then they can you know reconstruct siphon off whatever information they want to the NSA and like no one else will ever know and so the like the the pro the security property is that are involved with like even just being able to like identify which group of trust to trust and entrusting the group and then like trusting that it actually will stay decentralized in all

of those things like there's hard social problems involved and just like given how much happiness like we've brought developers just by you know going from per application setups to Universal setups and then from Universal setups to no setups like like if those three are like over here on the Spectrum then like NPC is like all the way over here right and so I think that's where the fundamental difficulty is well I give for specific applications you can surpass it then you know it definitely has a lot of benefits too yeah I think that like there's a bunch of applications where it's just like a peer-to-peer NPC where instead of having like everybody agreeing on a certain shared State I have a shared State and

I'm trying to do a computation with you like one example of this is like ZK machine learning where I have a database of health records you have a neural network that you want to train on those health records I don't want to show you the data you don't want to show me the model and we use NPC to do the training together yeah I think you're right like I think like a garbled circuit type two of two stuff I'm very bullish on I think the stuff I'm less bullish on is like trying to create like Global NPC sets that you have to us like that you have to try to convince an entire Community to trust I agree with that programmable cryptography yes so from my view so one thing like which I want to add you know so related to this question

and really to privacy discussion previously that many people still haven't realized that you know the owner proof can Pro they probably say that urinal proof can provide you it's only on the plural side so they prove I need to know all the information to General proof so it's not really the secure computation which you know your secret is known by a lot of people and they are doing some joint computation they are not true can't do that you have to know like all your secrets so that's why it's very hard to make evm compatible with with privacy yeah and therefore the the other crypto Primitives uh like they are fhe their MPC and their maybe sgx I think in terms of the programmability I think uh MPC is actually in some sense

quite similar to the case they are constraints because the fundamental gadgets of a multi-party computation is also audition and modification but you just need you know secret sharing homomorphism to add different shares or you you're a triple or whatever to to multiply different secret shares but the the front-end the application side actually can reuse either the same compiler like there are some academic work talking about how you how you can use the one like Universal compiler uh IR to kind of you know like build one build once and then you compile down to ZK MPC and even smt solver to kind of you know that special constraint so I think in terms of configurity I'm also very bullish on PC but as I had to

mentioned like NPC has totally different assumption another issue uh not related to the data set is that the liveness of MPC is also very different from blockchain like blockchain is you know like people like node can you know go offline or whatever but MPC rely on all they know to to online for a lot of like messaging and they need to compute something together so that will be very you know like why it's hard to adapt and PC directly to this blockchain space and then in terms of homomorphism and I think yeah like homomorphic fhe then it can provide a much stronger privacy but it's not verifiable but if you add fhe to their k then like like your overhead will be significant you still need much

better like proverb to make that happen um yeah I think there are a lot of like more interesting applications showing up in sgx either for you know generating some additional approver like add some additional guarantee like even invicted up like you know there are some idea for using multi-prover like where you generally leaky proof and all sjx proof and you need to verify both proof to know that it's secure so it can help as a training well to help you like you know remove some security assumption and also there are some idea usgx to like flashbot is using that for some other use cases which are also quite interesting but I know like a lot of crypto people really hate fjx because it's like a like a cheating box where it can do everything so yeah that's my

general opinion it's very interesting but it's very early and like you need some other efficiency Improvement for other like primitive maybe maybe hate is a strong word but like definitely very oh yeah yeah yeah thank you very very yeah what do you think Jordan for me the program programmability is what I said before so the interesting part is when it's not something it's not cryptography itself is when you can write a program with uh with Russ uh and and you have the proof that this program executes correctly that you have a proof of execution of a normal program this is for me what uh

it's a game changer but like like the problem with that is that like a lot of the time the approver performance the grades a lot when you have a compilation step but this is I think that this will get lower uh every time I mean it's a matter of time but it's there are new gpus Asics new hardware so it's something that I think that we'll see it like very soon with zkp is for sure but with like NPC and like are we going to see the same trends it's a matter I think it's a matter of time I mean this all a step by step it will it will come all these new or more

encryption I mean it was something that was unrealistic to think about uh five years ago no it's getting a little bit better but not much better but why not in 10 years we'll have computers that run homomorphic encryption we will have fhevm in 10 years maybe not Ellie what do you think so if we talk about programmability with respect to um scalability or you the use of scalability then I think that Cairo and systems like it give a very interesting example of virtual machines that basically take us their main

constraint to generation of proofs and we'll probably see more Evolution along that if we're talking about privacy programmability and privacy then I think in the near term we'll see two things one is zcash style limited functionality I mean in tornado cash and things like that where there's a very simple limited functionality for which you have very good end user privacy and they're the you biggest challenge which is a really non-trivial challenge is making the ux uh you know digestible by users that's the biggest challenge the other kind of privacy I think we'll see in the near future is this good enough mid ground how does it

look like you have as an end user you have full privacy from the rest of the rest of the world but you choose a centralized operator of your choice and you reveal information to that operator and then that operator will generate zero knowledge proofs for those computations I think it's a very good sweet spot for the next things because in most cases users will be fine with this sort of uh privacy and it's like good enough for them and you'll have very very good ux because from the end user you'll just have a simple you know interfaces just like you have let's say on ethereum and some of the starkec systems that we use with our customers

already offer this to their customers customers do have privacy from all but their operators there's a small caveat that we don't have ZK Starks we have plane Starks but you could make that ZK and even give the formal Assurance of zero knowledge proofs further down the line I think that NPCs are not coming anytime soon and the reason is because of the interactive complexity the round complexity and that increases with the complexity of the circuits you can't have you know Turing complete architectures like Cairo or evm they they break very quickly under um MPC so you're basically uh restricted

to circuits and you pay very dearly in the number of rounds with the complexity of the circuits so because of all this I don't think we'll see them anytime soon and if we will see them and this is a really cool research problem is how to bring in economic incentives into the interaction that is necessary for secure NPC so this is an area that has not been researched cryptographers have not looked at it but if you throw into the model the economic incentives then you start to get very interesting things so maybe a way to make progress on this would be would be to come up with a protocol that also puts you know you stake some things you have to walk through the steps or else you get

slashed and things like that that might be a way forward but still I don't think in the next five years we'll see secure MPC okay thank you um so since where I had gone my last two questions are about education so the first question is for like new people wanting to get up to speed on like ckps like how would you recommend that they get started what I recommend them is just go to a snark.js and do the tutorial that's in there you will create a full circuit you will create a proof you will verify the proof you will verify the proof and chain and I think this is a very good starting point in one hour

you can do this tutorial and you can have a good idea of what means programming uh silver of course from there there is a lot of literature a lot of things to read but I think this is a very good starting point I can go so it depends what you want to what you want to understand if you just want to get a high level Spirit of what zkps can offer then I often recommend I think vitalik explains everything very very well including Starks and snarks and zqps in general so you can read some of his writings also on the star course site we have our own writings about

zkps if you just want to experience their power I think the simplest route is to go to starknet you know to hello world on Cairo and then abstract it away you already have proofs generated for you so it's really a question of what you want and then if you really want to understand then there's no other way but to start reading some very heavy material in cryptography and theoretical computer science it's not something simple but still yes so yeah it also depends on whether you are just want to develop some thick application if you want to develop some of the application I would just recommend you like you know following either a circum Halo 2 those we have hosts a bunch of workshops and

understanding how to use that I think the best way just to try to use and trying to hack it and then like you know you will figure out how to write circuits and you look at some you know like what other projects are doing look at their circuit code and understand better how to write a circuit um and if you want to go deeper uh like you can check out some kind of you know uh read some very new recent paper and talk I think talking with the people and attending some conferences and figure out what you know what what's the problem that industry is looking into like for example people are looking into folding then you need to reason read some recent folding paper and figure out you know what people are looking into what what problems have been solved and trying to you know figure out those

problems and talk with with the people and if you just want to generally get a very comprehensive understanding uh there is also some book from from Justin tailor like which is a very comprehensive book and yeah I think another thing is also what we are doing is that we are hosting some bi-weekly uh JK application discussion which a lot of our applications will prevent their application into very detail like how how things work and you can check out those videos and uh you know like understanding better like how the application work um yeah and also there are a lot of like on online resources like I think a very good one like besides you know ZK podcast by the way is a really good resources for you to learn deeper about papers there and also they can hack where you know because hiking something can really

think of like if you think from the other side like how like I can prove something generally proof or something which is not approval you definitely understand why like something is provable so I think hiking that is a really good way for you to to understand yeah and then if as a developer you want to really go down the deep end at some point you can try implementing your own prover and verifier for one of the Z case networks games it's actually somewhat less hard than it seems like it's still very hard but it's almost hard than it seems my advice is to try not and be intimidated that all of these ideas boil down to like simple intuitions and the struggle is just to get there but at the end of the day it's a lot of simplicity

so we have a little bit of time left my last question is for all the people on the panel you everyone on the panel is at like different stages in their like zkp Adventure they're like learning and exploration of what's possible and my question is basically if you had to say it again what would you do differently sorry can you repeat a few my question is that like if you had to start again on your like learning exploration what would you do differently I would do the same thing exactly again I don't

sorry [Applause] I would come up with the concept of fry in 2012. I would I would do the math career at some point so as an engineer okay Engineers we have I think a good degree of uh understanding the math but when you go deep to the cryptography and let me hear like the cryptographers it's I mean it's getting hard and and it's getting tough and the the foundation of Engineers

I would say it's enough for implementing and for using it but having a better foundation in math is something that helps and it's something that if I mean if you ask me I would study more math when I was young yeah for us I think uh because you know being open and Community Driven are two important values we already have I think we even want to make make that more open uh which is not like which means a more Community Driven which means I think we even need to spend more time like you know our developer our league circuits besides there are no more like you know leaky engineering they keep them with the KVM they also need to spend much much more time thinking of Dev tooling

how to make like more people like you know writing circuits like way easier because the country is writing Halo 2 circuit really suffer it you know you can have a lot of configurations but it still suffers like whether how you can improve this efficiency to make even even more people to join your ecosystem join this open source community and how can think of keep constantly thinking of how can we make everything more transparent and more open and more Community Driven that will be our goal we already started doing a lot of like you know workshops and a lot of like you know Dev tooling there are a lot of being a lot of people even building more you know convenient toolings like you know Xmas building hello to lead for people to write case it's a lot of things that happen in this ecosystem we

want to make that bigger and bigger than like you know the RB Network effect like you have a lot of academics implemented uh even more and then like you know Hardware might want to support this framework so that's something which I want to we are already doing but I think we want to do like a lot more a lot more I think my serious answer is that it'd be a just like I'd be less scared and more willing to just like go and do things very good I modify my answer so if I had the time machine to go back so I still stand that I probably would have done everything the same but not related to proofs so I think it was in 2014 or maybe early 2015 actually vitalik reached out and said do you want to read

the yellow paper comment on it and be more involved and I said nah no I'm not interested I don't have enough time so definitely you know that I would have done differently [Applause] um So my answer to this question is that like kind of what we spoke about a little bit earlier about like simple things being super important I wouldn't I wouldn't worry I find a lot of the times I get sucked into this Rabbit Hole of just like optimizing things and like when I go there it takes me on like this journey that I don't need to go on and it's actually kind of harmful that like I think that I'm I'm I

sometimes over optimize things I think that's true though I think um like there is an aspect of this which is sort of the fault of the researchers more than the faults of the newbies which is that like often I find you know the researchers end up only writing out the optimized version of the thing and like not bothering to write out the noob version of the thing which is like maybe two times a lot sufficient but ten times easier to understand and like I know in my intellectual Journey like first understanding the new version and then understanding Improvement by improvements would have gone through would have like helped me get along and you get toward understanding things much more quickly so like that's something that like I would love to see people on other on the research side also doing

more of in the future hear you like another thing that I think is becoming more important is like product thinking in the ZK space that up until now we've like built a whole bunch of applications largely as like research projects but to actually get to the point where we have Facebook as a validium we need to transition to more like user-focused thinking okay so we we have like a few minutes left if there's any of the panelists want to ask some questions yeah maybe questions from the audience it's good too does anybody have a question for our

panel there's someone way up there you probably need to come down I think too or you can throw the mic oh thank you hi um I'm from China my name is James and uh so I'm just wondering so score is one of the top three players right of top five and so wondering when is school gonna

launch the main net and what's the reason for scroll um either what's your okay panelists what is your vision for scroll Jordy please go first for scroll s [Applause] look um well first of all scroll is an amazing project I mean they are building uh they are building as the KVM right now I think they are just thinking a lot in the in the being a type one uh the KVM uh very much

is uh it's a project that uh um so it has been very engaged in the community has been built from the has been built very much from the from the community perspective and it's fully written in Rust of course they I mean if you ask me you probably uh okay uh I would actually we did the things different I mean we had different architecture and different approach but uh I think that we learn a lot each other uh I can tell you that we learn a lot uh of uh of a scroll uh in the way they're doing their meditation and the fact that it's an open project an open

source project and this helps a lot the the collaboration means the contribution the learning one each other you need to think that doing sdkvm is not an easy in the City cast no nozik avms exist uh before I mean this is something that's really new and um it's not a single way to do the things I mean when you're doing engineering you need to take a lot of decisions and some decisions sometimes are good sometimes are bad and even there are even some decisions that depending on the time that you take it you can be in because you are taking this decision very early you can go to the wrong path and then you have to go back and move forward and the fact that there are not only scroll

but many teams uh many people in the space uh working uh working trying to build the same but with different with different Optics with different perspectives with different ways the freedom to do it because this has been very very very helpful sometimes I mean I would like to increase and my question is how we can do to increase this collaboration and to increase this uh knowledge transfer not between the scroll on us I mean between the scroll between the kissing between uh takeover between nurses people between or I mean

again of course yeah I mean but Star Wars ckvm well yeah but uh yeah but I want to say but this is is that it's look we are using we are using all the technology we are using in polygonies mainly Starks I mean if if this guy would not create stars at this point the polygons in KVM would not be possible at all it would be a different thing so and this is this is this is the cool thing this uh stealing if you want to call it that way this borrowing ideas I'm borrowing uh uh um Concepts and learning together this

is what allows us as a community to go faster and we need to understand that and here for me is a call special call especially for marketing peoples and product peoples on that people we are in the same we have the same goal we want to scale ethereum or if you want we want to scale blockchains we believe in this technology and we want this technology to happen and if we don't believe if we don't understand that we need to collaborate together it seems so we are going to go at least slower and this is hard because I mean when you see some people in the same in your company that's just fighting in

there is is is is hard and and uh we need to do the effort to keep contributing to keep collaborating to be meeting to keep asking questions and to keep uh and this is the sense and and I I'm working hard I'm trying to work hard on this direction I believe in this it sometimes it's hard but this is what I believe and and scroll I mean you can you can say for your in your own voice but this is the the the the the the open source spirit and if you want the famous street that we want to grow together thank you thank you Jordy [Music] [Applause] okay so we're out of time but I just

want to let yeah Zhang have the last word for the question I'm sorry to joke around but I think the question was like uh what's the vision for scroll yeah I think scroll want to you know truly accessible be truly decentralized Builder by collab of the KVM and we and also not only building you know technology wise your Skilling easier but you also need to inherit all the good values that ECM really has like being open being Community Driven be incredibly neutral so that you know influence a lot like how we develop our Wiki evm like how like when you open source whether it Open results from day one or day like 100 like whether you prioritize like you know Community contributions or you want to just ship paths so a lot of things related to our

decision making and but I think we will always uphold those values we have to build the you know the KVM and also like we are very collaborative with I think every conferences with majority and we discuss a lot about you know how we can improve this with the security problem we even hold some kind of security audit like you know sessions together to teach people what's the difference and uh like you know make that more secure so oh okay do we have time I guess yeah so um in my last weeks of exploring end user application when trying to build something

a lot more powerful I I saw myself going towards more using blockchain just as a something that holds the state in a decentralized way but doing the most complex computation outside and just proving State transitions for example with zkml and and then I I question myself like oh a lot of people are talking that change where you need to do computation on chain but with ZK you you have um you might have like same security as a blockchain in terms of computation itself and I'm just curious about like how do you see the future of like like a lot um more complex computation

um because I see like if even if you do it on a private server and then you just prove and you have the state on chain and you just do the state transition on chain that would be sufficient um and I just curious about like how do you see like do you still think like blockchain computation is necessary or is it enough to have States and do the computations on even private servers and just prove it on chain okay so the question basically is like if dkps get powerful than us enough do we need to have like a blockchain or is the world where we just have proofs that everything is correct enough so let me try and get my answer and I'll open up the panel so like my answer to this is that like zkps give

you validity but they don't give you like data availability that like there's no way to know that like the data that you need to make the proofs has been published unless you publish the data and that's kind of like the idea like for me that's a central problem that blockchains tried to solve the hardest problem is the data availability problem and for that reason I think that there's a really nice um interaction here where like blockchain solid state availability and and dkps or fraud proofs handle validity but I'll open it up I I think so so I refer to this a little bit in my talk this morning so if you think about um what a blockchain does it gives you this very high level of Integrity

through uh very good social consensus and some distribution of values to operators okay and then what validity proofs allow you is to invite arbitrarily powerful operators to bring on more computation that is done not within the scope of this thing without toxin all of those operators but having the same level of Integrity now it's true that if you want to save data you may want to save some of it or all of it on chain but for the processing you can sort of dramatically broaden the scope of Integrity that a blockchain gives you with the use of validity proofs and you can do so even recursively and that's exactly what I think all of us l2s are going to do to ethereum

by the way there's a I just want to say later on if we have time there's a question there that I don't know if we have time but okay one last question from the two ladies with their hands up thank you um so my name is Gwen and I'm quite nervous at the moment yeah so I'm from orbit of Finance so I would like to hear

um your guys the point like your view on the crossroad Bridge because basically we support like um scrolls.net and polygons in UVM and the ZK era as well so yeah what's your point on like mainly the security of the crossroad Bridge thank you so so this is an important question that comes up because people see across blockchain Bridges like layer one to layer one being hacked all the time and it's a very big concern and the reason why we did all the work to make like ZK and optimistic Roll-Ups is because the bridges have like fundamentally better security because

instead of being based upon trusting like a set of people in a multi-sig it's based upon verifying like the cryptographic guarantees that that this day transition has been happened correctly and that no one is able to see align this money so like my opinion is that like uh roll up to roll up Bridges will be fundamentally more secure than like layer one to layer one Bridges but I'll open it up again yeah there is one thing that I think here we have a kind of a naming problem we are referring to a bridge to three different things that have nothing to do in in my opinion but they are bridge maybe from the user perspective but one thing is I would say a trusted Bridge that's probably what we understand we

have two layer ones and then there is some kind of multi-seek or whatever that's just approving from going from one side to the other okay then we have this trustless reach which is there was a word that no I don't see it very used but time ago people use a double pegging between two chains I I like much better this this this this idea but there is that there is no I mean this automatic there is no intervention there is no multi-seek it's just uh comic swaps between changes yeah okay and and then the the other is would be between roll ups but these prolapses goes more in the in this double pegging category which is I mean this is you don't need it's a smart contract it's

just uh it's a bridge because it goes from layer 1 to layer two but it's it's a different thing of what people understand from a bridge people hear breach and they think about this Committee of approving things so here I like more the double pegging and roll ups I would not even talk about bridge I'm just just transferring from layer one to Layer Two so there's no bridging I mean it's just it's it's it's just it's very ethereum you are sharing the security of ethereum I I agree that we should have another name for the bridge from layer one to to roll up or to Layer Two any other thoughts about that yeah I I

think I think another thing depending on how the bridge is implemented recently there are some you know efforts implementing some ZK Bridge which actually you know ZK snarky consensus or whatever like the other channels have but it's still like you know fundamentally different from the key VM because they can improved the whole you know computation and then verify through smart contract which all rely on these ecm's valid data set but even if you use you you know you think it's not the other chain Two Chains still fundamentally have different valid data set so you can make sure that all those validators are agreed on some kind of you know consensus about this change did but they depends it then depends on like like ecms value set so the the the the

problem that player two or they can wrap solve is that we can move trust to Smart contract on e2m which we can reuse all the ecms you know while the data set and another strong reason why you know even like between is called champ bridging there are some idea for you know if you are too thicker show up maybe you can read state route through the same you know chain on Smart contract and then like doing some more trustless operations but that's another different maybe you need a different name than like you know those cross chain Bridges which have totally different value data set so I think the the fundamental of Layer Two and the choice of the group come from the base chair the Same by the data set or smart contract is not from the other you know set or whatever and digital can provide you this nice

publicly verifiability which you know like the public is all the ECM validator thing yeah okay thank you everybody thank you that's let's thank our panelists again