ETHParis: Scalable and resilient DAOs with Matan Field

okay hey hey the survivors so I'll tell you about holographic consensus and basically their protocol way to scale up those or generally speaking to scale up decision-making so who is already familiar with those quickly wait who is familiar with our stack or heard about the graphic instance before okay that's good okay so I'll try to run really briefly on that figure stuff maybe even just two altogether skip them but doesn't abduction when we say the unboxing we mean ownership in the assailant cover system too and all that ownership ownership can be external assets such as the Dow holding ether tokens or C internal assets did you know that the tokens off the Dow itself external rights such as reputation or voting power power in another Dow controlling another platform and so on so forth or internal rights now it has its own reputation system contracted it managed and of course governance system is simply this you know a system that makes decision about proposals yes no agents are saying yes no and then the Dow besides the different voices of agents are weighted differently whatever your voting system is if it's reputation based token token base et cetera and then the Dow is just making a decision you know no or yes if the majority of rotation or voting power is yes or no so this is like the simplest form I want sorry I won't get into em many details here on that but this is kind of like slightly sorry this is slightly I want that you talking it so this is like slight slightly few you know and just one or two slides about the technology that were using for the Dow stack basically you have the now you have the Dow stock that I would tell you slightly more about then you have the arc well here the art bean for the time not mentioning there this is just governor's primitives so for some of the whole voting system is just one primitive this is a full framework that takes all those people to in the complete structure of an organization you can ignore that this is a caching layer and eventually you can build easily dubs directly from JavaScript using those frameworks okay this is really really fast this is just one slide how that framework works is not directly related to your graphic consensus but just in case you want to use your graphical sense in your own organization so this is a structure of organization these are logical skins for a yes logic so yes logic means that if such and such will happen such as if I will propose a puzzle and those people will vote yes then the proposal be approved and be executed by the actors so token will be printed reputation will be minted or the avatar that can control any other external contract out there okay and these are negative or no skins which means that no matter what if you are deploying that scheme is that no matter what will happen will not go you know a invalidate that scheme so that if that is no printing more than million dollars no matter what this will approve that that will not be in more than a million dollar ok so basically you have a library of negative positive skins in the library of negative scheme you can deploy which modules you want to have in your organization and eventually you get a full organization that can control anything on the blockchain ok so now who graphic consensus will be implemented via a single scheme over here which itself is made of like a couple of contracts ok so why why scalable dowse begin with what's the problem so firstly before that what do you mean by scalable scalable in what so what what the DAO can be scalable at so of course the simplest is scalable in a number of people participating or effective voters I would call that the centralization so if you have a wide distribution of power to thousands of people I would say it's decentralized organization then you can have a scalability in terms of the governance throughput so people are proposing thousands of decisions every month and it Dow actually need to scan all of those types of decisions and make a decision on each and every of them so this would be the what I call scalable so the stability of governance throughput or decisions made in the unit of time or frequency of decisions so with that so imagine that so imagine you had this machinery so man you had like this magical machinery whether it's AI or its like artificial intelligence that you have a black box that just gets a lot of questions and then just say yes no yes no yes no so this black box is actually a community intelligence rather than collecting integers rather than artificial intelligence but still you need to have that machinery that thousands of proposals can be shoot on that machinery you know those machinery which is made of thousands of agents and somehow magically those thousand of agent need to shoot out decisions quickly right we try to try to imagine that situation like 5,000 collaborators collaborating in a single Dao proposing five thousands decisions every month and running through all of them all the time it's like you know hundreds of decisions a day and shooting mouths okay sounds sounds impossible right I hope who thinks it doesn't sound possible anyway who thinks it sounds is possible impossible okay great we have optimistic crowd thank you okay sounds hard okay great so yeah I also added maybe talk critically right so may totally means that it's not just random you know random is saying yes no that's a but actually actually weighing people according to their you know reputation or whatever you call reputation is like voting power or whatever whatever you define party power with so voting power defied me totally by when you make a contribution to the Dow you gain reputation but it was to be like token based like if you're holding if you're locking your tokens you know the Dow tokens in your contract you're getting voting power so whatever system you chosen okay so that's that imagination over there has some problems which is the scalability problem of those by the way which is completely analogous to stability problem of blockchain blockchain is a governance system it just decides one kind of things which is the blocks but the scalability problem of that is the same it's just a lack of attention or the finiteness of the resource of attention it's the usual tree lemma between the centralization scale and resiliency which were known from production but it's also true for governance which means that naive this underground system simply cannot be scalable and resilient at the same time and just one cent that better what do you mean by each our terminology so the centralization we mean exactly why distribution of voting power to a large number of effective influencers which are autonomous so a large number of investors you cannot form decision by two people this will not be decentralized if you actually need attention of hundred people or or 1,000 people this will be decentralized now you don't want really to have that attention because you cannot expect all thousand people to look at each and everything but in principle if you want to get a majority of our reputation to say yes or something this should require like thousand people then when we say scale again we mean the governance throughput the frequency of decisions that are made and finally resilience I want to give really precise definition because many people say resilience all the time but we never have a definition so I want to have a precise definition so when I say resiliency I mean representativity which means by definition if we want to have a scalable down we want to run want to shoot on that machine like thousands of decisions you know or hundreds of decision a day we want to get those answers by the finish it means that not everyone can look at everything right that's for sure so I know maybe that decision will be made by ten people still I want that decision to be representative for the rest of the Dow if it's not it means it's not resilient and can be attackable and manipulable okay so resiliency is exactly is very well-defined representativity such that if we just hold time we say okay time stop please now we just got a decision saying yes now we stop time right now I have all of the diagonals of the world we are asking those five times people what do you think you know there are patient look patient fully look looking at that giving an answer all of them then we are checking if that decision was made is in line with what everyone would say if the answer is yes then we have representativity the answer is no we do not have representative it's a very precise definition okay so you're putting all that together the claim is that you cannot have large number of people that are well why this abuse of power some large number of people deciding very very frequently in a representative manner that's the claim okay naive thing and the proof so this is a big big problem for big house and then a Orin is not a real person I think nobody got a job last time is it so it's just as a result of scarcity of attention and it's very easy to prove lots of fully engage people means lots of proposals of positive means little attention on each one if you have so many proposals you have just you know ten people voting on each and every and then if you have little touch on each proposal it means that each of them one generally speaking will not be representative okay anyone not convinced by that proof most real truth right so in an you know simplest word you say that in small does everyone look at everything well in large does no-one look at anything that's that's very very true by the way I don't know if you have a hunch but how many people with the first like large experiment was I mean we made experiment with real down with hung people will tell you about it or but Aragon actually had one vote with you know twenty thousand voters how many voted on the vote like one not like 5000 volts like a single vote how many voted roughly as far as I know I think I've heard 60 and that's when we have one vote while we haven't we have 5,000 votes how many would vote on each vote just a few if we're lucky okay so that's it's a real problem really nobody looks at anything so let's that's yeah I think your technical enough to skip that what's there yeah hope so the solution by definition the solution we call holographic consensus which means that we must allow must enable decisions to be made locally so when I say locally I mean by relatively small amount of attention maybe five people over here in this room can make a decision for the entire Dow so you decide how to spend money that you're not owning that the entire now is all new you make decision so we have to allow that without allowing that there is no chance for scalability like you're just banging your head to the wall so we have to allow local decisions but by definition for sake of represent activity for sake of resilience we need those Assyrian to be globally representative that's logical definition and I haven't said anything I just rephrased the previous statement in a nicer words and this is what we call our own graphic contents so that's easy again it will jump that but the big question is how you do that right how do you allow for quick quick quick quick decisions by small amount of reputation such as they actually represent the global majority and that's what we call graphic incentives so how graphic consensus is achieved by turning a social arbitrage into an economic event right so why do you mean by social arbitrage social arbitrage is the exact situation we're trying to avoid we're a small group of people are making a decision that is not representative of the majority I would call that a social arbitrage now how do we solve for not having those situation well we are translating we're mapping that into a an economic arbitrage and we know that the markets already know how to deal with economic Mart price arbitrage okay so basically we're making a situation that anyone that is successfully identifying and staking against this mismatch this misalignment there are five thousand things happening like this week and there is now a crowd of predictors there's not crowd of stickers ready just to go hungry to find those mismatches and there are snuffing oh this decision here is not reflective of the majority I'm betting against it and they will make profit from identifying that okay so translating a social charge in the inter canonic arbitrage and we've designed that protocol about a year ago this was genesis zero point why it was not sophisticated was pretty simple we knew that there are some well not major major but some holes or some failure modes and recently we've upgraded into Genesis 0.2 it's still not you know perfect in all means but it's it's it's quite closed and we would say that there is no like severe failure modes or tech sectors so it's called the Genesis protocol Genesis your point which is the low graphic consensus I would say minimal viable protocol basically okay we have three kind of actors in this game we've come of course having proposers anyone can make any proposal I can repose the doubt to spend money to that project I can propose it down to sign that transaction I can propose out to vote in another dow in that way i can post it down to widely spoken on that chicks but firm if the dow is the owner happen to be the owner of the Dodger platform and so on so forth I can I can actually propose to the Dow to open a sub company of Argonne platform I can do anything okay so the it's complete model or interoperable and the Dow can do any transaction in the blockchain and it's not cause it doesn't cost anything so we'd have a lot of ideas because it doesn't cost anything you're not risking anything just make a proposal of course if it's not interesting it will be down the line somewhere and nobody will see it so the beauty of it is that you're not really spamming anything is just becoming a queue of like thousand proposal and you'll be somewhere down the queue you just paid for the guy that nobody care right it's your you'll see you'll get there will get there you'll see you need to you need to see how that could float up that's the point but if you know if you just want to propagate your proposal in your social media you have a link to the proposal great what you know welcome to do that so voters voters are basically this DC permissionless this is permission has this permission in a sense that voters are weighted by their reputation and again weather doesn't matter if the relation was derived from meritocracy contributions or from token locking holding or something else each now has its own repetition system or voting power system and then voters vote swayed with their voting power and if you don't have voting power well you cannot really weight kind of relief or make you can vote but it will not count and then the new actor is the predictors so anyone can make a prediction so there is a so I'm proposing a proposal it's now live and now you can make a prediction deciding if you think that is representative or not representative if you think that is not representative you can make money from predicting that it's not actually sorry oppositely if you think that is representative you can make profit from identifying and betting that it's going to pass that's the idea now if you have made that bet now you can also make profit from saying no no this is actually not representative and I'm betting against that okay so in order to bootstrap well I mentioned that later but you want to bootstrap the profits for this prediction game that now itself is down staking each and every proposal so the DAO says each proposal that comes up says I know it's not representative it's with one-pass I'm just putting that you can see you can follow the logic as we'll go and you see that eventually that that will pay that money only if eventually the proposal will pass which means that though is really paying for the governance cost for each successful proposal that will go through the mechanism that I will tell you about the boosting then that will pay and it's really a pay for a real resource that was consumed it was paying for outsourcing the navigation of the collect intention into an external Network of predictors and they actually do real job so how it works predictors there are placing predictions on what voters will decide this is really important prediction are not making decisions they are not they don't have the incentive to say what they want to happen the only incentive they have is to say what they think voters will say ok so there are completely touched this is different from if you Turkey for example or from token qat registries where the vote that the predictors as takers are also eventually making decisions here they're not they're just predicting inside the only have the incentive to predict for the future for the future of voters the way that I do think is that by staking Jen tokens they have two ways to stage a talk Jen tokens either as up staking or down staking which means staking claiming that you know that it is representative or staking claiming that it's not representative and of course at the end of the game once the voters cast their vote the right predictors win the Jen of the losers that's great and in order to bootstrap that crypto economic game the Dow down stakes every each and every proposal and of course the higher the down stake of the Dow in a way the more incentive for predictors to predict and such better it's kind of like security a level that the devil requires and that's a process so the process is that every well every proposal that is issued by anyone is by definition a regular proposal which means just sitting in this infinite queue that you know nobody cares about if it's just stays there you can still execute that proposal so if somehow voters come and say oh that's so amazing and you get absolute majority so 50% of your updation holders say yes yeah absolutely do so just execute the proposal you you have ensured representativity right because if you have 50% of reputation holders saying yes you have ensured representativity x' and it leaves for a long time is its parameter it could be a month two months three months now you have a new status which is called boosted proposals and the new thing about that status is that it requires only relative majority of the voters in order to be approved so there's a shorter time frame for the voting period okay so maybe a week maybe three days depending on the use case and what are the things that you're approving you have like a week for this voting you know whoever votes votes you can you in the attention you know the shooting of the proposals on the topic and so on so forth and then at the end of that time if majority of those who voters say yes thou execute if they said no they'll reject so you have guaranteed resolution and now the question how proposes how do proposal gets boosted to this situation how do they move from absolute majority long time into a relative majority short time and the answer is well we have boosting conditions so let's see what the boosting conditions are so we said before that out down stakes every proposal and now people can play the game can up stake down stick and so on so forth and we're looking at this C which we call confidence which is the ratio between up stakes and downs like the total up stakes over the total down stakes and why would we call it confidence because it is actually the confidence that predicted need to have in order to not change something so if this is for it means that a rational player should dict if he thinks that but as well should should think that the confidence should think that the statistics of that to actually be approved by voters should be 41 and if we think that it's five to one he has incentive statistical incentive to actually increase up stick if you think is three to one he has incentive to increase down stick so the eventually there is a saturation of the Berkshire market and there is a cumulative confidence level in the market that says see now if that sees large enough and after in a moment what large enough means it does see large enough only then that the proposal gets boosted so there is a large economic signal that predictors things and not just things but we're going to stake that that this is going to pass now notice that is really important that you have the initial down stake such that you know large enough is defined with respect to that initial down stake that's why I said that it's a security level and of course after you have that level of up stakes maybe you get more down stakes by others who think that so it can it can spread above but this is with the minimal required up stakes the minimal stake it will be being a system before it can boost it and now high enough this is just parameter this is the difficulty parameter basically and we added here a very aggressive protection formula you can lower it to polynomials but this is exponential so it's the parameter this is part is slightly larger than one so let's say 1.2 1.3 and it goes exponential with an and B where NB is the number of proposals are currently boosted because you the more proposal you get that are already boosted in a way the more the collective attention of voters is diluted the more it needs to protect from you know having a mismatch the more you need the staking to fix for that miss much possible mismatch so that's why it makes sense and we took exponential protective aggressive format to begin with and now well the resolution is very very simple well as I said before a regular proposal can be resolved in three ways either there are boosted in timeframe or they're resolved with absolute majority or that are dismissed if timing out let's say after two months then booster propose are resolved by definition at the end of the voting period and they resolve with the majority with the relative majority to one of the other ways and I'm adding here given a quite ending so you have to in the last period of the voting you need to have quiet so not not shifting and if you have shifting the votes keeps extending every time there is a shift it keeps extending in another quite ending period let that ensure saturation and as I said before the right predictors gain the stakes of the wrong ones pro rata and let's just keep that last point it's not relevant now the most interesting world kind of like that that's the promise I just want to deliver now on my promise I'm not always doing that but this case I will try to deliver my promise so I promise that this this process will be a creating representativity right so why do we have now representativity so let's let's select like one you know step by step so a proposal can be decided by relative majority only if it gets boosted right so where the non represent if the risk comes from if things has been decided by absolute majority by definition they are representative so the only case where a proposal can be achieved be executed in a non representative manner if it got boosted right but a proposal can be boosted only if it had sufficient up stake remember that the ups take over the downs take confidence was large enough so it has sufficient ups take at least in relative of the doubt down stack which is the minimal s - is the minimal down stick but once you had this sufficient up stake you just created the crypt economic incentive for people to identify that this is a hack that this is a mistake right because if it's for example let's say that the down stake is $20 and now you need the confidence level of five so people are up staking on the dollar now I can earn $100 from simply putting 2000 I can actually earn $50 because I will compete with the Dow down stake I can earn $50 by putting $20 and say no this isn't this is this is an attack it's not going to pass and and get that money so once you have sufficient amount of up stakes you create economic incentive for people to put down stake but then once you have those two sides of the staking so you have ups takers and douse takers and now they vote they the relative vote short vote begins now they have the incentive the predictors have the incentive to call out for voters if they think the process is not representative so let's say that I I've predicted for it I think that is going to pass and now the vote goes on and now it seems to converge to know and then I'm going to lose a lot of money so I'm calling a I'm calling Alex that has a lot of reputation in that Dow and I just noticed that he hasn't voted and I know that he will approve it so I'm calling I've said Alex there is an attack over here the little group here is making a decision that is not reflective of the majority please cast your vote against that attack okay so we have people stakes on the two side of the equation to protect for A+ 'ti and that's exactly the outsourcing that the Dow is doing so at the end of the day the the day the predictors have three roles in the Dow the first row is fishermen they need to fish out the good proposals there's a huge queue someone is stupid look for those opposing and the Dow is paying for that okay you can also promote them in your social media but eventually they are the fishermen so the first role of the predictors is to find out the good put the good proposals and in amplify them by staking the second role is to protect and signal that there are mismatches so I see that you're up staking something I think you're either attacking or just not understanding that you know the culture of that Dow and I'm going to bend against that so I'm kind of like signaling that there is a mismatch and even then if it's going to be boosted my third role will be since I'm already staked would be to call out the voters I think we'll put the influence where it should be and that's protecting for representativity yeah quite ending just kind of like help that as well I've just said all that and it looks in the picture it looks like that there is like a stream this is like the shooting of thousand four paws are coming into the Dow those this is the prediction game people stake up and down and suddenly we have a good singer over there the something is going to pass these jumps the boosted stage there is a clock ticking we put placing votes at the end of the of the timer we have a decision and so on so forth and then we resolve the prediction game that's how it looks like now we can come up and wrap it all back together and try to put that back into this statement or graphic consensus the fact that you get small local decisions made in a way that in a high frequency but in a way that ensures representative the large majority we achieve that by creating this crypt economic game that essentially translates a social arbitrage in the opportunity the economic arbitrage and the free market of course will take care of the rest and with that you can imagine I hope that you can imagine now five days five thousand fear and developers they are working on increasing the the scalability solution of etherium and they are working all together and come managing a fund by the way it really could imagine that you could have the ability in the film itself so imagine that you know you could have increased 1% of the etherium economy distributed to those who will solve a scalability problem who thinks that the token the ether will go 1% up at least if we saw scalability only that oh my god I think it will go like hundred 100x up but so if we believe that the the ether will go at least 1% up if we saw scalability and if ether is 10 billion dollars means that we are essentially managing a fan of 100 million door and he made it we can actually manage that fun we print 1% of ether to those who solve scalability we can now you know 100,000 people can co manage and 100 million doors and basically reward those who will give the best implement the best idea for scalability and really build up a huge funding machine so for that we'll need a whole graphic on cells become need to run you know thousands or hundreds of thousands of decisions every months here and so forth so try to imagine that thank you [Applause] any questions yeah Thanks okay yeah very good so first email just one comment it's actually I would rather not try not to put it as a category inside future a future key future key because future keys in situation where the predictors are eventually also making a decision and that's part of the reason why it's actually somewhat when in portable it's a long long conversation future key may work for some use cases but not in a generic one so I'm trying to enough to separate holographic incentive from future key so Jesse's Dao yeah Jesse's that was launched in August 2018 we launched it with 50 people with this model with a fully working interface on a minute so that basically about two hundred and thirty proposal will have submitted about half of them were accepted about half of those who were accepted were asking for reputation so newcomers that wants to play the game which was completely organic we didn't do anything with that this grew from 50 people to about 130 or 120 and another half about 60 proposal we're asking for funding the Dow was managing about $15,000 I think every month oh sorry maybe somewhat less I don't remember exactly but something like that poor Park we you know they've been building so the initial definition were really to build they you know just any contribution to this project with the ecosystem in large and so on so forth so people build two interfaces with that funding we will people organize five meetups without funding to books content videos and a portrait handli handling paint portrait of italic pottery true and and so that's what so far it was pretty loosely defined okay so you could also see their contribution pretty loosely achieved now the next stage so we know it's down where we are upgrading we're in about a couple of weeks we'll have a life again the second version of the apps which will be the beta of the app is called alchemy it's an interface firdaus will have alchemy beta a much faster better protocol much better design and so on so forth and somewhat sometime after we will launch Genesis beta so a new version of the Dow that is now aimed with much grander goals as you know per your questions so the goal that we see for that there would be to [Music] basically take us you know the next big step towards decentralization of the project so we from the beginning of that project we had in mind that this project should be you know eating its own dog food and we are trying to reach that's that stage as soon as possible with advancing of the technology so with the way we think about it we think about it as in like grand alliance it eventually is owned and managed by the peddlers you know clients and holders supporters of the project so those who you know either either of those sectors will be part of the decision makers want to be the very grand out that then fund those initiatives fund either in ether or die or in gen tokens in focus of the project itself actually locked token because we want to fund people to you know to be stakeholders for the long term and basically support contributors of all kind that that are building the growth of this ecosystem and whether it is develop infrastructure or communicate or organize events or build bridges to other technologies such like build a bridge between the stack and Aragon or the second kleros and we are very much encouraging these kind of beaches so these are like the thing I think would suit best these things [Applause]