EthCC 2: Matan Field

so how everyone until I'm being told to start how everyone are feeling okay yeah can start okay great so my name is Madan from Dow stock and I will tell you tell you about holographic consensus and how it can be well useful for scalable dowse and what skill the scalable dowser anyway okay so just a quick quick reminding about what do we mean by dows on the blockchain so basically those are composed of two elements the ownership structure and then the big Sandra's governance structure or senders governance system to handle that ownership so by ownership we mean like there are four types of ownership so external assets internal assets external rights and internal rights so external right external asset would be something like the Dow holding its wallet theorem ether or EC 20 tokens and etc internal asset would be the Dow managing his own its own tokens right so the Dow can can own a token contract and mean tokens on the go external rights would be for example it Dow holding voting power in another Dow or perhaps the Dow having permission to change the - X platform just as the DX now is going to do to have in about a month and a half from now and internal rights of course is the reputation system the interpretation system of Dow itself so what what is the governing system then think about the machinery then there is a pipeline proposals running through that machinery okay you can propose the dal to do anything maybe do allocate funds maybe to vote in another Dao maybe to check to token due to whitelist tokens or a dot X plus form and so on so forth and then people vote so the ages of the Dao so the reputation holders will vote yes or no that puzzle this is the basic scheme where proposal are yes/no proposals you can have other scheme where the proposed are not just a yes/no but yes/no is the simpler scenario so we'll start with that notice that different votes have different ways we call that reputation or voting power so the votes are awaited differently if you don't have reputation you can still make proposal but you cannot vote and then if majority of rotation says no roughly you'd think of that as the Dow saying no and if majority of reputation saying yes you would think of that as the Dow saying yes so what what do we now buy scalable does and why should we care what why should we worry about them so firstly scalable in what sense so scalable in basically all dimensions so firstly should they should be scalable in the number of voting agents also known as voters so many many influencers or if you wish a wide distribution of power why do this efficient of reputation you know then they should be scaled by the number of proposing agents so not only you have maybe thousands of people voting but also thousands of people proposing so it's not just a single person making proposal once a week and then a lot of people are voting about it but also many people are proposing and that's possible because anyone can make a proposal or anyone can make any proposal so many scaling the scalability number of proposing agents and finally scalable number of proposals so that would be the frequency or the throughput of the DAO so how many decisions in the unit of time how many decision every week the Dow is making that's the throughput that's the real bottleneck and by the way that's not different we'll talk about it but the problem we'll talk about a problem soon and it will not be different from the problem of scalability of blockchain itself so now before we dive in to do more technical details let's hold for a moment and try to imagine and I know it's not it's not easy it's not trivial you can you can close your eyes if you need try to imagine five thousand people collaboratively engage in a Dow making five thousand proposals every week ok so many ideas or every month let's say every month so every you know on average an agent would have an idea once a month and we want that Dow to come from run screen through all of those ideas and say if they are good or bad if they are the best ones and then implement the best ones of them ok so that try to imagine that how how powerful that would be how many of you feel that you came roughly imagine that scenario is completely esoteric or roughly okay just a few okay now if you if you if you were able to imagine it very well deeply you had to encounter two things firstly you need to see that it's very different for any structure that you've seen before so it's not like company it's very much more lively and much more harvesting the collective intelligence okay so be a very powerful structure that we haven't seen yet the second thing that you had to come up with is that there is a big problem in that really big problem and in fact I would say the Beatles biggest problem the biggest bottleneck for scalable dowse which is the scalability problem so let's go over the scalability problem so the scalability problem is an inherent tension between three characteristics in tension between the centralization scale and resilience and it has nothing to do with you know with a specific topic it's a broad a problem in fact it's not different from the scalability problem of the blockchain itself so on the face of it it means that at least naively the centrist garner system cannot be scalable and rosy at the same time but let's let's go into details and see what we mean by that and why so firstly the standardization the centralization means wide distribution of voting power also known as reputation in the Dow so let's say we have five thousand people and reputation is widely distributed such that no less than 1000 people can achieve a majority okay so it's not five thousand people and then you know majority of population health by hell by ten it's actually at least one thousand people are needed to compose a majority in the down so this will be scalable or it's very decentralized ow then scalability is about the governance throughput so if you wish the amount of deposit every every week or every unit of time or the frequency of decisions that are made in that town okay so for example in a very scalable Dow you'd run through hundreds of decision every week or every or thousands you know every week and then finally resiliency is the fact if is its representativity so it's the fact that all those decisions actually represent the majority of them okay so although we are running so many decisions in every week still each and each of those decision should be representative and in line with the global majority of reputation in the Dow now of course if you put it all together so why distribution of power a lot of many people and scalability many decisions and resiliency representativity you come to the conclusion that at least naively you cannot have why distribution of power deciding very frequently in a very representative manner simply cannot and why so it's well it's so firstly that's that's a big problem so it's big problem for big house that's the biggest problem okay so this problem just say it's it's the very fact of scarcity the natural law of scarcity of attention and the reason is the reasoning is very simple so if you have lots of fully engage people means that you have lots of proposals if you have lots of repose that means that you have little attention put on each one of them and then if you have a little attention on each one of them generally speaking they would not be representative so we have we have 5000 people and only 10 percents or 250 people are voting on a proposal 50 people that combinely comprise let's say 1 percent of reputation in that Dow there is no reason to believe that that 1 percent will be representative of the whole Dow so if we just stop time for a moment we stop the flow of thousand of proposal for a moment we stop that and then we ask all 5000 people to react just this proposal and then we measure whether you know the the little decision was representative of the global majority or not and in general there is no reason to believe it will be representative so that's a problem that's the reason or if put in very simple words in small doubt everyone looks at everything well as in large doubt no one looks at anything and if maybe putting its in some some little picture so this would be a small Dow and now we're looking at one proposal and since there are just a few agents and just a few proposals then well most agents will look on each and every proposal and as a whole that would be pretty representative however if you look at a larger Dow and here is that these are not thousand these are just 20 nodes and we want to operate at a thousand level if you have more nodes and then also many more proposals then on average a proposal will just gain very little fraction of the reputation and in general as I said before will not be representative so that's the draw the little character of the problem so now the solution which is just a logical outcome of what I said so far it's just in a way it's logically the only solution and which could be called holographic consensus is that dows or large scaled out scalable dowse need to allow for local decisions and local in the sense that only requiring little tension to put on each of one of them so local decisions but such that or which are ensured to be globally representative so to be in line with a Dao majority of reputation so this is by definition right we need scalability thus we need to allow for the senior to be made locally by little amount of tension but we also want to receive resiliency which by definition means representativity of those decisions of the global majority so this is just logical outcome and in drawing it would be like that so we have many nodes and many decisions and indeed still every decision will just capture very little attention from the dao however there is there should be that magic that we call or graphic consensus which somehow makes sure that those little decisions over here actually reflect the global majority even if you didn't directly ask all of those people if you just stop time and ask them you discover that these little decisions or local decisions actually represent the majority so this will be the drawing floor graphic consensus and again these local decisions that are ensured to be globally representative and of course the big question is about how you do that it sounds almost impossible right so in a very very summarised form I'll just say that and I'll come that at the end I would say that holographic consensus achieved by turning a social arbitrage into economic arbitrage so socialite Raj is the situation where those active voters that place their vote do not reflect will the global majority that's the so we are assuming scalability of the doubt we are assuming large-scale doubt we are assuming the centralization Y distribution of power and now we want to ensure for for resiliency which means that we want to avoid the situations of social arbitrage avoiding the situation where some active voters in fact do not reflect well the globe majority so how do we avoid that situation where just translating that social arbitrage into economic arbitrage making that an opportunity making the opportunity for anyone to make profit from successfully identifying and staking gains that mismatch and then just letting the free market to take care for the rest and let's see how that works so this I will describe now the genesis V zero two point two a protocol which I would argue that's probably the MVP for her graphic incentives you can still you know complete complete I so much but this is this is the you know the most advanced version that I think is the minimum viable protocol such that it doesn't have you know collapse attacks or very severe attack collectors okay so there are a few roles in a Dow so two of them are the regular ones are proposers and voters so anyone can make any proposal you don't need to be reputation holder in the Dow it doesn't cost anything so we want to tap into the collective intelligence you have an idea for Dow great just submit your proposal then voters you can anyone well in principle you can vote but only if you got reputation your accounted because your vote is weighted by your reputation so in a way that's less of a less permissionless in network although reputation can be derived from anything so it could be married to erotic reputation so if you mean reputation that is that is given to you when you make contribution to the dow but you'll also be a more economic of reputation for example i Quadling so that you can make a pleasure system such that reputation is allocated to you if you're taking your doubt oaken's and locking them for two years okay so it'll be more like economic reputation derivation and then the new actor is the predictors so it's also permissionless net anyone can place predictions and make a prediction whether a certain proposal will be approved or rejected by the voters and notice that while the voters are being asked what should be done predictors are just predicting what would happen okay what they think the voters will say so there are independent sectors so placing prediction on what voters would decide is happening by staking gen tokens that these are the tokens of the dolsot project you can place them in two ways you can stake tokens in two ways you can up stake or down stake a proposal if you think it's going to pass or fail and then there is the game the prediction game and in that game the right predictors eventually win the gen stake of the wrong predictors so if you make a prediction and you found out to be correct you are going to make profit from those who make the wrong prediction but now what would give you the incentive to begin like if there is a really good proposal why should I make a petition it's good because nobody will think it's bad so for that reason to bootstrap the prediction game the Dow itself puts a down stake like predicting against each and every proposal okay so that's bootstrapping their prediction games that kind of act market making to the prediction game we call it also the Dow stake the Dow putting down stake in each and every puzzle okay so now what's the process the process is comprised of three steps there's regular proposals there is the boosting condition and it was the booster proposals let's see how it works out so every each and every newly submitted puzzle is by default a regular proposal regular proposal means that it requires absolutely of all reputation in order to approve that proposal which means that by definition it's resilient but of course it's not scalable and that lives for a long time let's say for a couple of months okay now boosted puzzle is a new title okay they know if you wish a new stage a new status and proposal get boosted getting that status if they stand in the boosting condition which we'll see in the next slide now this come this time is it's it's a shorter of your voting period let's say one week or three days depending on the use case and there is guaranteed resolution in the end of that voting period however in that voting period it requires just relative majority of reputation in order to approve that decision so let's say the out of 500 people just 50 have voted and let's say just for simplicity they're all equal reputation holder just 50 have voted then if 26 are saying or more saying yes the proposal will be approved and executed so what's the boosting condition we said before the doubt down stakes each every proposals and then the proposal are being ranked by what we call the predictors confidence so that's see the confidence is the ratio between the total amount of up stakes and the total amount of Dow stakes including the down stakes of the Dow now notice that this is actually the confidence level meaning that if it's four it means the predictor network the free market decide that occur motive Lee that there is a chance of four to one it's going to pass and it's really the confidence in the sense if you think that you have the chances are five to one you have an economic incentive statistically to actually place more up stake and if you think that the chances are actually three to one you have an economic incentive to put more down stake so this is the creative confidence level of the proposals and if that confidence is sufficiently high and we'll discuss in what sense sufficiently high for continuously so for some time so let's say for 24 hours it keeps high for continuously for all that time then and only then a proposal gets boosted so that threshold alpha to the power and be alpha is just the parameter in the doubt the DA can change the parameter but it's a parameter is the scaling difficulty parameter it's apparatus slightly larger than one for example 1.2 and NB is the number of currently boosted proposal so if you have more booster prosit right now it means that you are diluting more the collective attention which means that it needs to be yet harder to boost another decision and that's how it comes it's protecting it's very protective formula because it's exponential in and be it could have been more softer milder form this is a very protective which means that the resilience level will be very high in that Dow but the scale built it will be somewhat stretched okay you can make it you can this is a real tension you can play between scalability and resilience and finally resolution so regular proposals can well be resolved in three ways they can either be boosted in that timeframe or they can be resolved with absolute majority if I've some agility has voted yes or no it's being resolved or it can be dismissed if it times out then booster proposals are always resolved when the voting ends they resolve by relative majority so if relative majority says yes it's it's a yes if you say no it's a no given a quiet ending will come back that in two slides and then as we said with a prediction game right predictors gained the steak of wrong ones so if you made a wrong prediction you're losing your steak if you made a good prediction you're gaining the steak the lost lost steaks of the other ones and you're gaining it pro rata and then there is one more thing that I'll just skip is about reputation flow it's somewhat orthogonal to Howe graphic consensus but early represented voters gained some of the early non presenting voters it's not very important for that matter and let's see how representativity comes about because we the whole purpose of holographic consensus was to argue that we take it down it's decentralized it's scalable and one now want to build machinery that also makes it representative or resilient okay so let's argue why we got represent IVA T so as we said before a proposal can be decided by relative majority and thus the risk of non represent if they only if it get boosted so again if it's not boosted it means that you need absolute majority and by definition its representative it can only be decided by relative majority if it got boosted but then it only got boosted if you got sufficient up stakes predicted predicting that you're gonna pass at least relatively doubt due to the minimal basic down stake now once you have sufficient up stakes you have a big economic incentive for the free market to bet against it to down stake and now it means that you have four nonzero positive not clearly yes or no you'd have predictors staked on the two sides of its faith now the staked predictors have the incentive to ensure representativity so if they think that in the process of the boosting boosted voting period if they think that the result the outcome is not reflecting the the you know the the action globe majority they can call out to the voters they think will shift it to the other way and they are basically protecting the represent IVA T and the last thing is that quite ending requires the outcome of booster posle to not to shift at the last stage I won't get to it right now just in the quick drawing so that's there is a stream of puzzles in the queue of regular proposals and then people placing predictions on those proposals if there is a signal for proposal to be confident in its passing then it jumps into the boosting queue in the boosting queue proposed are being open for a finite a short time shorter time where people vote people actually can also vote before and then when it times out there is a decision etc so this is this is the whole process and tomorrow or I think 10 o'clock I'll also demo the the actual interface that does that so to wrap it up just against things the whole graphic contest is achieved by turning a social arbitrage so the situation where active voters do not reflect well the global majority into an opportunity into an economic arbitrage where anyone can make a profit from successfully identifying and staking against that mismatch and then again the free market will take that opportunity my dad closing the economic arbitrage by that closing the social arbitrage and again with that let's wrap up just imagine 5,000 people collaboratively engage in a Dow proposing 5,000 ideas every week and meritocratic Elysee winning all of them funding and implementing the best ones thank you [Applause] do we have do a time for like a couple of questions yeah okay great hi thanks a lot for the presentation I just wanted to know a bit more precisely how do you attribute reputation in your system right yeah so this is this is a big big topic somewhat orthogonal to this and again our infrastructure our like governance framework is completely agnostic you can you can have modules different modules that are locating repetition differently so I can just give you a few example that we actually implemented and you can think of many other examples so one the simplest example is just meritocratic lis so when you make contribution to the Dow that thou gives you reputation when you make a valuable contribution the Dow auto cash you reputation for example one way to do that manually that you are you're just making a proposal to the Dow to get reputation so I'm contemplating a white paper or that idea or that meetup and I'm asking from the Dow to get hundred points of reputation and that votes yes or no if they vote yes you get that reputation okay so to be like manual meritocratic allocation another one which for example we've implemented for the DX Dow is more like for hotlink so you have some tokens you know maybe that's the Dow stock and maybe some other tokens and then if you are locking your tokens for let's say a year or two you're gaining reputation so by that you're not proving any meritocratic contribution you are proving your long-term alignment in a sense with the Dow and then you can have many different other modules which are more you know more economic or more meritocratic and then different dowse can play with different modules so we have a complete you know generic and modular framework I'll maybe remind that a little bit tomorrow but that's but that's a big big topic hello first I would like to say really awesome holographic consensus I think this is one of the best things that happen to governments in this space and I have two questions one thing that I've been wondering I don't have an answer at all is when we speak about scalability and big social consensus in essence have you thought about cultural influences cause probably decision-making processes are different in different cultures so that would be one question that I have and the other is how do you see how would you compare it there's a lot of talk here in the community us about liquid democracy so how would you compare it with holographic consensus great questions so culture democracy so let me tell you something about culture I think about the doll has a culture that should have a culture and by if you wish it or not it will have a culture so let me actually give you example and again I'll mention that tomorrow but we ran an experiment with hundred people are about hundred people in our kind of like our own little Dow which we call the Genesis de which was amazing and you could have seen in that experiment culture spontaneously emerging for example we have put no limit on how much reputation you can you can ask and we know usually people kind of like probably the last 120 score they hunt 180 at some point just average 200 and if you ask more than 100 people tell you know everyone got hundred if you ask more than hundred need to know so just completely spontaneous culture or initially people just ask for reputation they could connect their LinkedIn and say this is me and I would actually actually get a reputation and over time people say no we don't know that's you if we want to do that confirm you know verify your identity and people needed then to verify their identity to get that reputation so bottom line yes culture is a very important thing in a dhow you would have a culture endow you can when you are founding it now you have the opportunity to you know a by us the culture into one or another direction by creating a very strong Schilling point if you're saying this now is founded for that purpose this is our culture you know this is a code of behavior you cannot enforce that of course but you are highly biasing towards that because then if I think that you know those who are in line with that culture will probably buy in those who are not aligned will buy out and then also if I think that everyone think about that culture is the baseline I will more likely to behave according to that so that's that's a very important topic the second thing which is liquid democracy or delegation is a very important element and it's just a second layer module on top of that so on in addition that this is not the full picture of scalability another talking skill bit is compositionality or imagine just those made of thousands of dollars or a mesh network of dolls so I just described how to scale up a single agency but now the thing about the mesh network in Internet of agencies this really how that would look like and inside and the third element of scalability is of course delegation which is super important but it can just be a second layer module on top of that um thank you more time I have a clarifying question you mentioned that anyone can make a proposal but also that every proposal is automatically down staked doesn't that allow people to basically print reputation by just creating proposals that are automatically staked or is there something that I'm missing in here the fool if anyone can make a proposal without reputation and any proposal is automatically down staked so that Dow stack puts reputation against the proposal it's not it's not dowel stuck it's the Dow itself it's not putting reputation it's down staking with tokens so Jen tokens every proposal get down staked and then if it's well if it's passing then that Dow stake is lost to the predict good well the ups takers which means essentially it's eventually its governance cost so that that will lose some you know it know before it will lose some amount of Jen tokens for each successfully approved proposal and if it doesn't pass and there's no cost so it cannot it cannot pass without cost because there is real cost you write the cost of attention right but if it doesn't pass there's no cause someone can't use someone cannot make proposals to like take money out of this okay thank you very much [Music] [Applause]