https://www.youtube.com/watch?v=bV1_0iI7R6A
Video AMA with @matan on March 1st discussing holographic consensus, the market landscape (Aragon + Holochain), and use cases for DAOstack.
A transcript has been included below for convenience.
I. ON HOLOGRAPHIC CONSENSUS
Josh: Okay so welcome everyone. We’re here for the first of our video AMA series with Matan Field, architect of DAOstack. Thanks everybody for coming. I’m Josh, I’m part of the community team and myself, along with some other members of the team, am monitoring the chats for questions. You can type questions into the group chat on the Hangout or onto our telegram community channel and then we’ll be filtering those up to Matan and we’ll try to get as many questions answered in the next 30 minutes as possible. So let’s start with a live question from Ezra Weller from the telegram. “I’d love to hear any more details about holographic consensus. I’d love to hear any specific ideas for achieving it, if possible.” We also had a previous question as well about holographic consensus. “Could you be more specific and describing holographic consensus?”. So do you want to start there, Matan?
Matan: Sure, yeah. That’s a good place to start. Firstly, thanks everyone for coming and maybe, before answering what are the details of the solution of holographic consensus, let’s just say a couple of words about the problem. More generally, the problem of decentralized governance is that there is a core tension between scale and resilience. In fact, this is not different at all from the scalability problem of the blockchain itself. In the blockchain, scalability problem is actually more severe because it is an additional bottleneck of scalability due to communication speed. But even if you had solved communication speed – so even if communication was in the speed of light, was infinite – still you would have the same scalability problem with blockchain, which is basically a problem of any, I would say, consensus engine. So consensus by definition means that you have many agents on the network. And so let’s take it back to Decentralized Governance and then each agent voice is weighted by some function. It could be reputation, and could be by tokens, it could be anything else. But let’s just streak now for a reputation, because that’s the easiest and also I would say the best one.
So now when we say consensus and we want to keep resilience, what it means is that each decision on the network is basically approved by majority. So that will be a consensus, an absolute majority. So if we have a reputation then the majority of reputation holders is approving each and every decision. The problem though is that this is clearly not scalable. So if you grow the organization, you will not be able to process more and more decisions and then you will not become more effective. You will actually become less and less effective as you grow the organization.
So the goal is to create this super organization which is an organization that, as you grow the size of the organization, it can make more and more decisions. It can become better and better at processing computer power if you will. So the only way to get out of this – just by the definition that we made – is by allowing decisions to be made by small groups. Another way to call it is “by relative majority”, we will see it in a moment. So to get decision by small groups, but for the sake resilience, you have to ensure that those decisions are in line, effectively, with the thought of the majority, and that’s what we call holographic incentives. The reason why we call it holographic is because it’s kind of like taken from hologram, and in the hologram each piece of the three-dimensional image has the information of the entire picture. So in the same way here, you are able to produce decisions locally, or at the edge of the network, but in a way that is ensuring that those decisions are in line with the majority. So that’s the first thing about the why and general how, but now we can kind of like double-click into the what and what it looks like.
So basically, let’s speak about relative majority and the protection of collective attention. Absolute majority is clear, it’s easy, and resilient, but it’s really not scalable. Relative majority is when you open a vote. You say “okay, let’s take this proposal and let’s open it for two weeks”, for example. And whoever votes in, whatever the majority says we are doing. If majority says yes, we are executing the proposal. If majority says no, we’re not executing the proposal. This is called relative majority, so we are only taking the majority out of those who voted. We’re not counting those who haven’t voted. So this is one way to go about this. The problem with this approach though, is that it’s very easily “spammable”. I can open thousands of proposals and the collective attention will not be able to actually look at each and every proposal and in some of them I will hide deadly action, but also just wrong actions. So the organization would not be able to effectively look at each and every proposal. So the next step would be to limit that, which I call the defensive zone - we also call it “boosting” by the way. So boosting means that we’re opening, we’re transitioning a proposal from the the absolute majority into a relative majority finite time stage. So this transition (this boosting) has to be limited. Let’s say that the next step would be that we are only allowing five proposals to be boosted at a given time. That’s definitely a protective solution, only five proposals can be boosted at a given time. Now we know that the entire organization can funnel the entire collective attention onto final five decisions and then we maintain resilience. But still we have a problem because five decisions per week for example is not scalable. If we grow the organization, it will not be able to make more decisions. So we would have to extend that number but in a protective manner. How do we do that? The way to do that is basically to add an additional market-driven staking layer. When someone opens up a proposal, it gets into a queue. By default, proposals on the queue can only be executed if they reach absolutely majority. But now the question is how do we decide which proposal to boost? The answer is that every and each proposal is getting a score and there is a threshold and if proposals are crossing that threshold, they are being boosted.
Now let’s see where the score is coming from. There is an additional activity now that you can do (in addition to voting on the proposal), which is to stake tokens. Basically you can stake GEN tokens over proposals. For example, let’s say that I see a proposal and I believe it’s a good proposal – we don’t mean objective good or bad. When I say good, I mean that I believe that this approach is in line with the DAO, or more precisely, I believe that if enough people will be looking at that proposal, it will pass in this specific DAO, while maybe it will not pass in another DAO. In the same way, I can look at another proposal and say “I’m willing to bet $2,000 that this proposal will never pass in that DAO.” This additional layer of stake is basically pulling information to the system about whether proposals are in line with the truth or not. For example, let’s say that I’m trying to kidnap the organization and I’m just offering to fill up our office with one hundred-thousand-dollar worth of ice cream and that’s definitely not a good proposal. So if I try to do that, I’m just creating economic incentives for people to place stake (to place bets basically) that this proposal would not pass. So we are “crypteconomically” monetizing the mismatches between status quos of proposals and the truth - the subjective truth - if they would pass or not pass in the DAO. Only if enough steaks are positively predicting the proposal is going to pass that the proposal is being boosted and delivered to the voters to decide locally if they approve it or not. So basically the formula for threshold is the amount of GEN tokens that are being staked predicting that it will pass minus the tokens that it will not pass.
The last point is that the threshold is also dynamic and it’s exponentially dependent on the number of proposals that are already boosted. The result of that is the following: initially, it’s very easy and very cheap to boost proposals. However, let’s say now we have 10, 20 proposals, then the threshol becomes exponentially high, which means that it’s exponentially costly to even boost yet another proposal. So a lot of people will need to stake over a new proposal to be boosted. But then, once that proposal has been boosted, there will be a lot of stake - a lot of money on the table – that will make those people who staked it being incentivized to ensure that the right people are coming to vote and make it the correct result. So if I place the five thousand dollar bets that the proposal X is going to pass, and then it’s being boosted - and then nobody’s going to vote, and just two people are voting against it - I have five thousand dollars to lose so I have incentive, if I believe that a majority actually think that this should pass, I haven’t sent it to pull them out until “look guys the proposal here is not being done right and you should come vote.” So basically, the more proposals that are being boosted, the more the collective attention is scarce; the more the collective attention is scarce, the higher the threshold and more expensive it is to boost yet another proposal; the more expensive it is, the more stake will be on the table for that proposal, and then the more incentives will be to protect the outcome of that proposal to be a mismatch. So in this way, this circle is simply dynamically allowing for more or less collective attention to be drawn.
This layer of staking has nothing to do with the DAO, it’s an external layer to the DAO. Anyone in the world with no reputation - and it can be even just an anonymous person - can come and place stakes over bet, and in that way you can think of it as the DAO outsourcing the navigation of collective attention to a larger audience. And also you can say that the large audience is becoming better and better, just as any trader investing in the market. Those who are making good predictions are making a lot of money of that, and then they have more money to make even more better predictions. And those who make bad predictions lose their money and stop doing that, so that the market of predictor is ongoing evolving and becoming better and better in that. It might be worth mentioning something about Gnosis. So although we are partnered with Gnosis and we are going to integrate Gnosis to the stack - and Gnosis was going to integrate the stack into Gnosis - we are not using Gnosis right now and the first reason is because the market I’m describing right now is not what people call prediction market. There are two kinds of markets and both deal with prediction but they’re working very very differently. So I’m trying not to call it prediction market because people already have in their mind something about prediction market and this market behaves very differently. In a regular prediction market, you’re buying in a way shares of yes and no - and then you’re trading those shares - and here it’s not really like that - I’m just placing bets so it’s slightly different.
II. ARAGON COMPARED TO DAOSTACK
Josh: thanks Matan. Okay, so I want to move into a question about just the landscape of DAO platforms. It’s something that I think we hear a lot and ur Wiggum from telegram asks “What differentiates you from Aragon?” And we might also include in this other DAO platforms as well.
Matan: Sure, Those two questions are constantly coming. So first I like to say that we’re less interested in what others are doing like. We’re just interested in doing our own work the best, and there is enough room for everyone. So I think more interesting than how we are different, would be the question of how we can cooperate.
There is a difference when you ask what is the difference in the vision or in what they should do, and then you can look inside the platform and see a technological difference so though from the outset it seems very similar I think, if you look inside, the technology is completely different. And I think the result is driven by completely different thinking. We have been thinking about this project for the past four years and from the very first beginning we have been thinking about this more, I would say, architectural driven and protocol driven and ecosystemic point of view. For example, Aragon has been building an app and then they decided to also build a platform behind it and we went completely the opposite direction. We have been building platform for a long long - we have a massive codebase - and only recently, only in the last two months, we’ve started to build an app on that platform. And that derives completely different architectures. So we’ve designed from the bottom up a completely modular, completely upgradeable, completely adaptive (protocol) so we can basically change to be capturing any sort of ground system of the blockchain. We’ve designed for a long time this architecture that produces any sort of governance, and not necessarily for the rhythm of work - but also for the rhythm of work. So the same system can be used for a decentralized insurance, and can be used for a decentralized investment fund, it can used for decentralized curation networks such as TripAdvisor, Booking, Yelp, and Google. So this is, I would say, one difference - I call it the “horizontal ecosystemic view”. And then secondly is the “vertical ecosystemic view.” So realizing that the importance here should be about interoperability, and building an ecosystem, we’ve built a vertical integration stack. So we’ve built the governance framework on top of Arc; we’ve built Arc.js, the JavaScript layer that allows for very easy integration - if you are now building a decentralized application you can easily get out of the box a decentralized governance or a decentralize decision-making engine to be integrated with, so you don’t need to take care for that - it’s also coming with like shared registries that are supporting the interoperability of different applications and then of course we’re also building applications ourselves. So this I would call the “vertical ecosystemic approach”. And thirdly, it is about the protocol. So just going back to holographic consensus, realizing that if we really want to implement large-scale organization in the real world, will need to sort out completely new kinds of protocols. Any other project that I’m familiar with today in the space is basically taking existing governance models and kind of like replicating them on the blockchain and I think it’s great for small-scale collaborations. But for large-scale cooperation it will not work, because large-scale cooperation, again, comes with the tension between resilience and scale so either those systems will not scale up, or they will scale up in a non-resilient manner, so it will not be safe, or a bad behavior.
We put a lot of of emphasis on that – on designing large-scale governance protocol and just, for example, one outcome of that is that we have completely different business model, completely different utility model. The utility model or the ANT token (The Aragon token) and the utility model of Colony token, and any other, is very different from the utility token of DAOstack (The GEN token). The utility model of the GEN token is basically that the entire staking mechanism that I described - over the whole ecosystem - is being made by the GEN token. So basically, the more large-scale coordination you have in the ecosystem, the more usage of that token is, and then the more the value of that organic is going up. So it’s a totally different utility model from anything else out there. This is for example a big difference.
III. ON LARGE SCALE COORDINATION
Josh: Okay, thanks Matan. But a little bit about use case an implementation. Travis Wallace asks on our forum “at what point do you have enough actors in a network that DAOstack becomes more efficient than managing it in other ways?” Similarly, Julian on the telegram asks “When a business starts and the number of agents are small, is it still useful to use DAOtack?”
Matan: Our emphasis is about large-scale coordination so we’re not imagining just replicating the the real world on the blockchain, which is great by itself, but not our focus. We’re imagining networks of thousands and millions of people cooperating around shared goals and interests. So maybe imagine 1 million people are literally running a decentralized insurance cooperative. That would be something that is relevant for DAOstack. Now the question when it becomes more useful? Of course, you can also use the DAOstack for small-scale collaboration but on the face of it you are just earning a little bit more efficiency, and all of the ownership management is being managed more automatically and more transparently, but I think that that by itself is not the big advantage. From the beginning, the idea was to build the platform, build the tools, the protocols for for open collaboration at scale, which means that if you have now millions of people collaborating in a single network, you can much much more easily get access to basically everything. So if you’re a company and you’re looking for developers, you can much more easily get an access to developers. If you’re looking for business development, and/or partnerships, or outsourcing, or anything of this sort, being part of a grading network and also the ability to be part of that greater network I think it’s very advantageous. So again, the the real value is coming from large-scale coordination, so either you are a large-scale organization yourself, or you are being part of a larger organization and enjoying and benefiting from that larger resources. So if you would ask how many (members), I would say that anything that cannot be managed otherwise. For example, I think already hundred is actually many. I don’t know if any any good tool in the world that allows 400 developers to self-organize and build projects and also share the management or ownership of of their creation. You can definitely see open-source developers in the hundreds and even thousands but they don’t manage ownership. So eventually they are just volunteer. Or you see cooperatives at, I would say, the tens that are pretty efficient, but beyond that they are very inefficient - or they are basically becoming centralized. So I would say that already the hundreds - that 100 or 200 at the most - is the number that, if you want to have an effective coordination at scale, you definitely need a new model for governance. And that’s DAOstack.
IV. DAOSTACK AND HOLOCHAIN
Josh: Okay, thank you. So let’s go now to Holochain. We get a fair number of questions about this and a (Xia Hud?) who asks “I want to see how this integrates with Holochain.” Any thoughts?
Matan: Firstly, I’m a great believer for Holochain. I follow the project for years. I met the founder years ago and Arthur Brock, he’s really brilliant person, and I believe that anything that he produces will be great and it would be great to integrate with them. One thing to say though, I think a lot of people confusing saying that holochain is maybe a competitor to Ethereum or to the blockchain, and it’s really not that - It’s a complementary tool to the blockchain. You can do a lot of things very effectively with holochain, but I don’t think you can do anything we can do with the blockchain on the holochain. So holochain is something on the spectrum between a blockchain and IPFS. It’s more like a peer-to-peer network but it can also run contracts. If anyone is familiar with the Rchain, it’s actually also similar to Rchain. In addition to the blockchain itself, which is actually very similar to Ethereum, Rchain has this tool that is called Special K and that is an analogy to holochain. So Holochain allows a peer-to-peer network to run applications but they can not process consensus. So you can not have, for example, regular crypto tokens - you can not run them on holochain. You can have other instruments, but not crypto tokens because, by definition, crypto tokens require a ledger and the ledger require consensus and there is no consensus engine in holochain.
I would say that the holochain is a great complementary tool to Ethereum and in that regard I would really hope to see DAOstack integrating with holochain, as well with other tools, such as IFPS, each of them is providing different benefits. Frankly, I don’t know what is the stage of maturity of holochain - if it’s already a production level or not. Last time I’ve been talking to Arthur, it was still pretty far from production level, but definitely once holochain will be in the air, there is no reason why not. Philosophically, we’re very much, I would say, platform agnostic, we would likely agree with anything that integrates. And in terms of what can be done with that, it is pretty well known that blockchain should use very scarcely - anything that can be run not on blockchain, should be run not on blockchain. So we will gradually migrate more and more, well, of course firstly the data are out of the blockchain - we’re not keeping data on the blockchain, we’re keeping data right now in servers - then next step is on IPFS and maybe next step is on holochain. Yet next step is to migrate also as large of a part as possible of the of the computation logic off chain. So even the computation level not everything needs to be sitting on the blockchain - this is by the way a whole paradigm in the blockchain space, which is called blockchain computation. So either we will need to migrate those logic to off chain tools such as plasma, Truebit, etc or to tools like holochain. That’s basically it.
V. USE CASES: THE 4 CATEGORIES OF DAOs
Josh: Okay, alright. “Where can I find examples of use cases for DAOstack?” and a similar question “Are there any pilot projects happening?”
Matan: Well, there are tons of use cases for DAOstack. Basically, almost every possible decentralized application you could think of requires something like DAOstack. In fact, every DApp that I ever came across can be categorized into four categories. So the first category is the category of collaboration - large number of people are coordinated to produce something - and then the type of decision that they make is basically about the allocation of resources, either allocation of budgets or allocation of ownership tokens if you want. The second category is the management of assets. So under collaboration will go open-source developers, Wikipedia, Linux, etc. And then, the second category is asset management and then under that category will go to decentralized investment, decentralized insurance, decentralized pension fund etc. And the third category is basically curation, so decentralized curation network such as the decentralized version of Yelp, or TripAdvisor, Booking, or even Google because Google is, at the end of the day, a curation engine. Also Reddit. So these are the three first categories and in fact each of them requires collective decision-making and each of them, if it would need to scale, we’ll need something like DAOstack. The fourth category actually does not have to have collective decision-making, although it will become better if it has. The fourth category is market places. So it can be e-commerce but it could also be decentralized ride-sharing, which is also a market place between riders and drivers. The nature of market place is that it’s purely peer-to-peer transactions, I make a transaction with someone else and we don’t really need to have collective decision-making for it to work, which is why the fourth category doesn’t have to have something like DAOstack although it would definitely become a better market if, in addition to the peer-to-peer transaction, the collecting can also produce collective decision making, for example about the nature of the platform. So yeah, there are a lot of use cases. Right now, there are no yet active use cases because the platform itself is not yet alive - it’s coming out live next month and it’s already coming up with a bunch of use cases. Firstly, we ourselves are going to start using Alchemy, which is the application for the decentralized budgeting - allocation of budget for open source projects - and we are going to use it ourselves in order to engage in larger open source developer community around the project. We also have our partner Gnosis, which right now is sitting on a very big pile of tokens and they really want to distribute those tokens to open-source developers that will produce applications on top of Gnosis, but then there is no good tool to do that. So Gnosis will also be using Alchemy for that. We’re also speaking with people from the Ethereum Foundation, we’re speaking with people from El Four, and other companies. So there is, I would say, over ten projects we’re speaking with to possibly use Alchemy for budgeting for open source development, which is one direction. And the other direction is about different applications that will be integrated with the sack. So we have Menlo for decentralized investment fund and also decentralized curation of investment; we have Sapien, which is about decentralized news platform; and we have a bunch of others. None of them is still live, but I think the first live use case will come out next month and maybe more a couple of months after.