Events Add an event Speakers Talks Collections
 
IC3 Blockchain Camp 2021
July 25, 2021, Online
IC3 Blockchain Camp 2021
Request Q&A
IC3 Blockchain Camp 2021
From the conference
IC3 Blockchain Camp 2021
Request Q&A
Video
The Evolution of Consensus Protocols
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Add to favorites
2.37 K
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About the talk


About speaker

Emin Gun Sirer
Founder & CEO at Ava Labs
Share

Is everyone knows goon is associate professor at Cornell is the founder and the CEO of Ava labs. And of course, he's also a founder and co-director of my t-shirt. She'll, thank you very much already. And thank you everyone for coming by. I'm coming to you today from Southern turkey and with the severe fires and all sorts of infrastructure issues around me. So I apologize in advance for the connectivity problems that I've been having today and I want to talk to you about consensus brother calls, and

I should, I guess I should talk a little bit about in what capacity I'm talkin to you. I've been cautioned by Ariana. There's that I see three of the policy involving. Do you know clarification of rules? So I am a faculty member at Cornell University and I want the generally talk to you about consensus for the cold. In general, but I'm also the CEO of all the laughs and Ava loves is commercializing. What I think is the latest innovation in consensus. Protocols known as the Avalanche. My goal today is to advance to the point where we talked about Avalanche.

I will try to in general. I think to the extent that I have a reputation light or ended by being fair and square at every level. And so, I will try to say everything like it is, but you should all know that maybe I will stretch or whatever. Like, I don't know what people worry about, but you, you should know that there is a, that I am indeed the CEO of all the labs, and that might change my value value judgments, but in general, I don't think you'll hear of value judgement. I think with you, here is a short description of Concepts. So, let's see, let's get started here. And

let me see if I can share my screen and create some kind of It's going to create this, that's good. I'm at turn off my camera. Do you need to mow? Maybe those with the camera be. Let me get my my slide up here. So let's get the, let's go to this View. Okay, I will talk to you about about the consensus for the Kohl's in General. Tso lock change as, you know, or a big deal. They reach the stage in in language Evolution where they don't require, you know about the notion of an article. We don't get to talk about a blockchain. We don't

get our blockchain instead. What we talk about is if you would lower Manhattan to very well-dressed, people will typically be walking down the street. Then one of them will turn to the other and say, what's your strategy for blockchain? So it gets the street. I think, is reserved for proper noun and the only one with the best one that I know that compares is the words God. So it says that you've done an amazing status for this new asset class that has emerged around blockchains around. The notion of consensus has come to go, hold Great Value more than four hundred billion

dollars. In fact, Play Kris Kross. By the time, I wrote this was 400 billion across the 1 trillion boundary since then. So there is a lot of, a lot of a lot of systems out. There. There's a lot of our stored in these other systems. There are thousands of currencies, maybe, eight, thousands of of coins. And there's a lot of money. I think last time I checked was there was not January 2017. It was about a year ago was 40 billion dollars worth of VC money in Roblox in base companies. And the more importantly, I think that was about 20 to 30 billion in people's hard-earned

money. This is retail investments. In blockchain companies are many interesting accomplishments in the space. One of the most exciting for me, is in the time. I was interested in watching since, so that's like the last decade in a little bit. I have personally seen at least half a dozen instances where Gmail wants out where AWS went down with Facebook went down. And yet, during this time, many of these systems never went out. There. Isn't a Satoshi is not walking around with a pager in his pockets right to maintain your system. These systems are inherently

more robust, then centralized operations that we have seen elsewhere if I look around at 12 second, but there's a very big change of a change of time. Especially we're going from centralized client server systems to peer-to-peer Byzantine. And these systems are in the NFL can be inherently more robust, because I was really excited. I never like really interesting strange at times. I was right because our people as well as, you know, stolen all the people from the bank of England been, I've shared the Table with a

darknetmarkets operator, as well as people from the bank and you will know how many songs have been written about bit about Bitcoin. And how many songs have been written about Facebook and free software. What's really happening here at the highest level? What we're seeing. As I mentioned we're witnessing the emergence of practical Byzantine, fault-tolerant systems, the assistants carry out the function. They carry that function that without the single person supposed to charge. So people somehow spontaneously organized and they

maintain a service. Nothing sure. Strong Integrity guarantees of that said, service. With some of its ability and transparency for anybody to love to be able to gain to to be able to look into the system and gain some assurance that it's working correctly and new family. About the case is application Stan, high integrity systems that old Financial records. The easiest accounting thing is keeping track of how many other resource a certain person as I have but also more complicated instruments instruments that relate to finance options.

He's a really exciting and in so doing we can also do away with Oren code into the system, corporate structures, trust Partnerships things that would normally in the business World require a trusted operator or a trusted middleman to carry out, candy them programmatically. Sorry. And I were to start the boy band, the division of the royalties would typically be done by our manager. And I'm sure within a few years with the manager. We will there be an ass as far as I can tell from the supermarket tabloids Ari. And I will get into

a big fight with the manager and we can do away with with people in the middle who don't actually provide that much value other than serve, as a trusted identity. This system can take over and be with the trusted intermediary. This is amazing. High-level message. I want to leave you with today is that there are lots and lots of exciting deep scientific contributions in this space. This is not just an application of things. We've known. This is the thing that took DFT and made of practical vfd. Until now was going down the particular path of very academic path and it was a dead end

in many ways. And suddenly, we're seeing an amazing man that stands ideas from obviously the stereo systems from networking cryptography. Game Theory program application, programming languages. And of course, economics. I also I think my second message that I want to so that maybe more lately get across is the fact that the dream that's what though. She has. Shown us is very difficult to achieve with today's pro. And what I want to give you a glimpse of is a different, kind of a protocol for achieving consensus that has properties. That

are unlike everything else that came before it. Come closer to achieving the dream. Walking down the aisle. I want to talk to you want to talk about the Avalanche family of consensus for the clothes, which I view as the third biggest thing that's ever happened in the space more throughout all history. Only has a history of 45 years. What is the third? Biggest thing that's happened. It's certainly the biggest thing that's happened since the the the paper

notarized and then we can just talk a little bit about bookkeeping and all you can talk about anything you are. So how do we generally keep track of Records? This is an age-old question, right? We all live in less, a summation. We need to keep track of whose rituals worked hardest who's getting the most and we need to sort of keep tabs on this one. Way of doing it, is you trust someone to keep the record. There's a bank of England. You can trust them or you can trust another Record Keeper like their sign. And now the moment you trust a Record Keeper, you're at the mercy of that, trust

identity. You can trust me to keep tabs on your accounting. I would like to remind all of you that small Street failed to do the most basic of accounting, just a few years ago. There is the story of the doll company. The doll company issued. 38 million shares in this lifetime. And the few years ago. There was a there was a reason to to collect and collate. All of the shares outstanding and suddenly lo and behold, we got 53 million. People saying I'm a shareholder of the door company. So this entire infrastructure they can stay on whatever they want at the end of the day. They

were unable to keep tabs on the most basic thing that they exist to keep tabs on us for their sign. I think we all know what they fail that day or in the trusted position of authority over the.com space and they gorgeous. You want to buy anything. Come, you're going to pay them $8. And that's that's unconscionable. The cost of them is a fraction of a penny. The cost to use $8. That's only because they're the gatekeeper. This is going to repeat itself across time and space. It is not just very sign. You replace Mary sign with any other set of human being is going to happen again.

And again, it's just endemic. If you look up course at the bank of England, they are in a very very funny. What they can do is they can inflate the money supply. They are in the position of control over the soul money supply. You and I can scream all day about whatever it is that we want. They do not need. Take that into it. So okay. So then let's take a step back and think a lot about how in general people keep tabs on on. Well, on the simple records who's done the best? What would be the simplest thing? That I know that you all know, is, of course, gold. The

precious metal is very shiny, looks very nice. I look at this and immediately, the conjurer's in me, the idea that whoever's holding. The stash has worked really hard to attain it, but it's, it's inherently valuable. I don't know why that is. If I think about it, though, it's solely because I'm culturally conditioned, I grew up. And when I went to the cupboard, Lazar, it's has that Shimmer, that gold gold colors to it or it's used to these days, all of them to Gold Star salesman at the covered Lazard. Also, accept other crypto including bits go in an avalanche set alarm in here. So,

but it wasn't always this other people in the world who believed that these things were valuable. They would look at this and be like, oh my God, somebody spent a lot of time at the beach, collecting, these valuable stones or perhaps they spent a lot of time, you know, growing weed and selling it to other people that separates our massive Specialties. And, of course, the the early money used in China, was that was the store. That was a nice. Not everyone can afford one of these things. The standardized one of the most interesting stories that I want a very briefly, such a pain comes from

this island and the island of yak in the South Pacific. They use these Stones down. This is all these restaurants are special to them because it's impossible to forge one on this island. This island actually lacks big rocks. What they have to do. They have to go to the island over? That's volcanic. Carve out one of these stones and bring it back to their own Island, which is a coral island where there are no big, big rocks like this. And these things are If you had one of these, you know, you'd have to carry it

around like this big deal to have one of these things. And then they do with their, well, exactly what we do with our, well, which is display it. So, this is in front of the house. It was another display of what the location you use these things. I would go to REM be like a r e. I've got five of these Stones. I don't give you three of my rice stones, and then maybe you'll give me your land is a very interesting story from 1863. I believe we're in English, a pathologist was stationed on the island and the group of men went over to the island where the volcanic island. They carved out of

stone. They were coming back and I'm sure many of you have been on the canoe. It's very Tipsy and on the way back they dropped the the other coin into the ocean. When they did, and when I came back to their Village for some of the kind of wet sand, kind of kind of humble. They had the story, they weren't so, you know, we did everything, we were supposed to do, we want to be island over. We, we carved out our coin. We did what we were supposed to do, except, you know, we had a bad thing happen. It's on sale. The stone

is at the bottom of the ocean. And so upon hearing this, The Village Elders decided got a decision to make and they came up with what is known as probably as the first case of virtual currency. They said, you know what, we're going to count that rock. We're going to count that's going at the bottom of the ocean. And from that day on, I personally would be able to say to re Hai re I'm going to give you two stones. Not 32 plus the one at the bottom of the ocean. Well, you own it. And now I have yours on it and the trick here, of course is you want to make sure that everybody on the island

has heard of the fact that re now owns the stone at the bottom of the ocean. If it's not everyone here is it then I could sell the coin. I mean I could tell you I could give the coin and by land from Ari and Veronica at the same time. I would get some land from all the land from Veronica using the same coin known as a double Spin. And that would create a huge problem. So that takes us to, I think, you know, where we need to be, which is what what do we need to achieve? To build a proper virtual currency system. We need to achieve consensus. Everybody who's honest, who's participating

in the protocol needs to agree on exactly what happened in in history. So that's our goal. And I will mention it's a bit more formally maybe in a second or so. These are more pictures of these raistone but there's one is probably the most interesting one. I'm not making the store. Somebody went and found. I think there's even more than one of these stones at the bottom of the ocean. And as long as you see you live on a small island where everybody talks a lot, I think you can keep tabs on Who's got what but at some point on a global scale, you will end up needing what is known as a

consensus protocol. I think it should be amply clear to everybody by now. That's what we want to do is we want to keep tabs on the global lender a ledger like this. Lamentable contain entries, like such a zealous spacebob, 10 coins on coin and Charlie pays David five of the time he owes bones. That's a reasonably formatted. Proper blockchain going to do this. What's the question of the set up messages? The protocols are there, some of the procedures were going to use for, for maintaining our

shares worldview on this kind of a ledger is known as a consensus protocol. There are many concerts has protocols out. There is not a consensus protocol, the protocols involved, set of messages, and the procedures, we all prayed to send those messages that altogether construct, a proper Ledger for, Proof of work and proof of stake. Our civil control methods. They are the means by which were going to make sure that we only talked to people. We need to talk to and the reason for why we need simple control

mechanisms is imagine that you are in the village on the Pacific and you know, normally to have a consensus for the call. You want to be talkin to everybody in the village and everybody is honesty want to talk to all of them. What do you don't want to do to do when you go from that world to our current world is suddenly have Veronica show up with a million other Veronica and though they're all controlled by her, but they confuse me into thinking that all these Veronica's are the set of people. I need to be talking to and and suddenly she has much more weight in the

protocol than other people. She can school me and the she can use all you can she can fool other people and caused me to double spend. So we need that. We're going to need two couple hour hour. The consensus protocols, with some means of controlling who is part of my Village and who is not going to come back to you in a little bit. When we talk about the mortgage. How do we, how do we do this? Consensus? Make how do we make sure that all the participants observe the same series of a of a of actions? There's been a lot of research in this area. A lot of

research that ever since the Inception of distributed systems. This has been the biggest problem. They have been tutoring awards. Awarded in the space, both went to people who made significant advances in consensus particles. When is Leslie Lamport, the others? Professor both are amazing, people who advances state-of-the-science immensely and they built up what is known as classical consensus particles. These particles are used in today's permission systems. They're also getting a revamped and

revived for adoption in to blockchain systems. There is no actually getting. So, I'm either going to go update it if you will. That's the nice way to put it. In many cases. They're just getting a little bit of makeup and they're getting calls from 1999 are being served back to the community as if they are new major breakthroughs. Well, we see that happen with very stupid chains. So classical protocol, for the name of the game for a long time. And the one Satoshi Nakamoto came along in 2009. He knew all about the last you knew about all of them

and he discarded the idea because he thought it was not suitable for open permissionless systems used in the wild at internet scale. And so so what he did is he invented his own consensus for the cold that works differently. Sure. Many of you are very familiar with them until two years ago. And I want to take you to the state-of-the-art and I want to take you a little Beyonce, Avalanche concept. So how do classical consensus? Porticos work? So, I'm the very first papers in the space or especially old papers that talk about the parliament.

They are trying to, in essence, try to create. They're trying to create some kind of a universe where, where a group of entities are going to be making decisions in unison. Some of these entities might be Byzantine in this particular setting of cryptocurrencies in blockchain and that's creates Room For Gray complications. Now imagine that you want to make some decisions and something like a parliament and let's maybe go through some simple protocols, that you might want to set up. And let's maybe the spell why or talk about why some

some approaches are completely unsuitable and why why why the protocols have the shape and form that they do? So for example, Set the framework up every one of these consensus for the calls will require. I read protocol and the right for the rights. For the call is for making a decision or read for the coldest for clearing in the state of the universe. If we look at what's going on here, like a parliament and let's suppose we want to pass the law. Say something like hey, I want to exit, I want to exit the EU. A

reasonable right for the call mi give you a single right now. I'll give you a simple one. Suppose my rights for the call is I want to I want to message this to all 100 members of the parliaments. I want to hear from all hundred and when I do, I will then consider its written and then you can read from any one of the members of parliament and they will tell you if we have eggs or not. I'll give you a moment to think about whether or not. This is a good concept of protocol. I need to think about this for a second if I'm writing to a hundred people and I'm waiting to hear

from all of them. And one of them is evil, then he's going to be able to stop my system. You just won't respond. And now suddenly, my system is dead. So you can see that, that's a terrible picture, who writes for the goal. The reason for the call is also bad, if there's a single evil person and I happen to read from the evil person use in a lie to me and then I'll make a bad decision. So that was a terrible idea. So we can think about the other than just one person. Then that's obviously a bad for the goal because there's no guarantee you're going to reach from him. That's not going to

work. So maybe it's beginning to Dawn on everybody was listening to this. Somehow. I need to create some kind of an intersection property for this. Kind of a system to operate correctly to maintain safety and likeness in the presence of evil. Participants. And if I if you think about it, just just a little bit more and I would love to go through. It's in a little bit more slowly. When I do this for a lecture, we go through it a little bit more slowly. But since I don't have that much time today, I'll just kind of cut to the Chase and give you the

form of the solution. It turns out that in this particular instance, with the kind of setup that we have, the optimal size of people that I need to write to is 2/3. So what I do my rights for the girl looks like this. I write to everybody in the parliament, but I can't afford to wait for everyone. That would be stupid or just one evil person with jerk me up. So I write to everybody and I wait to hear from 2/3 + 1, that's the blue, shaded area. So I know that this many people do bees blue people have agreed that we are leaving the EU. And when you read, you

read from two-thirds of the parliament. I'm dumb. And so you can read from everybody but you you weigh 222, you wait until you've heard from two thirds of the participants of the reason for the two-thirds being optimal for these kinds of particles. Is that the intersection area. This far, is the intersection of area. It's not that I don't know if you can see my my little love, my little mouse care but this intersection if I have 2/3 + 1 and 2/3 + 1, if I write to 100 people

and I hear back from 67 and you read from a hundred people and you wait until divorce from 67, then the set your hearing from. And the set I have written to will be at least 34 people dig by the pigeonhole principle. If you've mastered the pigeonhole principle, you've mastered, you've essentially become an expert in vfd systems. Circa Circa few years ago, is section, so that when I write and you read, there was somebody around to carry the the information to you. Now, I constructed the 67 and 67. The intersection is 34. What is that mean? It means

that up to 33. People could be evil, but there will always be a start, a fourth person. He's thirty-three could be absolutely horrendous. They could be and I'll be able to flip flop. They could be like, whatever like these brexit pass or like, I don't know where they stand anymore, but, but there's always going to be one additional guy in that set of 34. It was going to be 33 at the most 33 section. Plus the extra guy. And this guy, this is the guy who's going to be like, wait, wait, wait gun said that we decided to exit the EU and therefore I'm going to

carry Message on to the person doing the re so so I thought I hope this is clear and these classical brother calls all have the score of intersection property where they write to a sizable group and they read from a sizable group ensuring that there is somebody in the intersection. I can carry the message for it. There was an additional complexity and this is the second part. You have to master to become a complete expert in the space circus a few years ago and that is not. So it is not sufficient for you to hear of as a member of the parliament. It is not

sufficient for you to hear that. The decision is taking place because if re is evil, you could send messages to some people and different messages to other people and they could get the parliament's. Very confused for you. To commit a decision. You have to not only hear of The Proposal, but you also have to know yourself that there are. There are sixty-seven people who have Heard that 67 other people are about to commit the decision. I hope that makes sense. That little part is a little tricky and Minnie PCS have been written on it. Many species of contained errors. When

it comes to, that part of many published papers are actually at fault because they don't try to maintain it properly. So there are colorful area. But it is at the very heart of every consensus protocol, which is this this notion that there was enough of a quorum intersection each and every one of the numbers is essentially asking, are there enough of us who made the decision already made the decision to move forward? That Monday have convinced themselves that they move forward. I hope this makes you an expert in this really literally makes you an extra. You would be able to to, to review

papers and reject papers, and reject new breakthroughs that come after you. If you wanted to just having mastered these two little tricks and Lord knows, there are reviewers out there. Hello to all of them. Whoever you'd my papers. Just based on the fact that they must have these two little chicks that brings you up to speed. That's 2018, B, ft. You figure these two things out decorum. Decorum, principal and the rest of the one level of recursion that you need to do before you commit to anything. And now, you can build yourself a consensus for the clothes that maintain safety

and lightness. What's wrong with this? Why we use these things? Well, I'll give you a moment to think about about the what's I just mentioned these. Protocols. They are very nice and many ways, but they don't scale. They don't scale in any shape or form. These protocols were designed from the ground up, for very small sets of people, making decisions. They require and squares communication. If there are a hundred people a hunt each and every one of those hundred people have to hear from the other hundred people. Why? Because of the second property

that I mention. I need to know that 67 people are on board with this decision decision, and if n people have to talk to and other people for every decision, well, that communication complexities and square. So hundred squared is a thousand messages. Now, it's ten thousand messages that we can handle that. But if you have a 10000 people, that's 100. Read messages, that's a lot. And you really cannot. By the time you wake, you call a $200 hundred million messages. It's just the amount of time. It takes is going to be

incredibly high, and then you're going to be in a world of hurt. I did not even talk to you about the complexity of what it takes to change members of the parliament. These systems are typically very fragile you and I and everybody has to agree on who's in our Parliament. If you are writing to one set of people and I'm reading from a different set of people than the intersection property, the first property I mentioned goes out the door. So so that's obviously going to be a complete failure. So therefore, we need even more complex protocol for changing the parliaments and

many, many practitioners are unable to get this right. Many, many men right now. There are star wholesale system's closed for me. Private for mission systems are typically close because for the longest time, people can get this part of the protocol implemented correctly. Thanks for the call for changing members. Of the parliament, is the vent tubes. Typically at least for the simplest for the goals, rather complicated very, very fragile. And and they just don't scale at heart. The more participants you have, they have the less some of the more, the more communication

every play. The best of these protocols is a protocol that is first authored by pinion that Jen is one of my graduate students. One of my graduate students. He just graduated actually and Ted Ginn and, and all the Emoji and others came up with the best of these particles is called hot stuff and it is you use my Facebook's Libra effort. Now called them. So, but DM, and Libra Target about 200 participants, so it's going to be Mark Zuckerberg and his 99 best friends. It's not going to be an open system for you and me and everybody else to use.

So, I wouldn't be interested in a space if I wanted to replace the world's Financial in their infrastructure, with Mark Zuckerberg in 99 of his friends. I want to build open. I want to do something where we have to say. I want to build something that's passed that robust. That is amazing to use. That does not just turned down into Amino, replacing one set of Gatekeepers with a different set of gate. So let's move on. Indeed the background to, which Southern Ocean not going with a cane. And his response was look, this stuff is just it's not going to stay on this.

Not suitable for an open deployment and he came up with his algorithm that I'm not going to go through that. I'm sure you know, which involves built building a block shamed by a mining. So transactions arrived, as actions are collated them. Two blocks blocks are added to the blockchain when a minor comes up with a solution to a hash puzzle. And the for doing this the minor is awarded some number of coins. You can see that this was an old presentation back when the Bitcoin reward was about 50 this morning. It's much less now, but this is the process by which blocks are appended to The

Ledger. And if somebody wants to modify the past, if somebody wants to say, hey, you're Ellis, Ellis baseball, and now she wants to go back and change history, while she is going to come up with with many more blocks than the rest of the network is not many more to come up with, One more block, then the rest of the network has come up with, but to do that. She needs cash power greater than the rest of the network. And therefore, it's a double spend. It requires a 50% of 51% of the hash power to change the record on Nakamoto consensus. This

is a huge breakthrough and amazing breakthrough. It's a brilliant system. What's billions about it? It's observes that you can have a probabilistic system that can hold Great Value about more than one trillion dollars that are stored. In this point. It's an immense amount of value storage probabilistically. This gets under the skin of some academics. They want their their protocols to give them full guarantees math problems. Take down those like a dumbass are clueless because they don't realize that every single machine, including my cell phone here, especially is a

probabilistic device. The cell phone doesn't always operate correctly. It's like this. Very one is very, very faulty. The most every single chip you use this problems. So we live in a probably stick world and and therefore the one of the most amazing things that Satoshi ever did was to say, look, 1 - Epsilon is about, as good as one probability one, when you make up sloth slow small enough. And that's exactly what he did and that's why, Welcome to come up with this amazing. New infrastructure. There is no view change for the cold. There are no voters. That's that takes a second to sink in

there are minors, sure. But there were no voters. You don't need anybody's permission to participate in the system. There is no view, change happening. There is no voting happening on the Block. You just come up with a block, don't follow the protocol and you'll build the blockchain covered do that's getting to into play and it turns out it's not really quite 51%. 34 % can actually attack these systems. But leaving those aside, these systems are still very

robust, and I'm very open and scalable in the number of miners. So that's a pretty nice situation to me. What was wrong with them? When we was? We going for a very good back in the day when this came up, you know, people started using these machines and they start using these machines. Then, do you think these are fpgas? Then we went to a 6, this is Delete some of these dorm room where your electricity is free. I don't want to give anybody any ideas. My colleagues at CMU discover the whole data center underneath their data center. When they were remodeling it where, or somebody

know how to place the whole bunch of Bitcoin miners underneath the floorboards. This is mining industry from 6 years ago. Fly-by-night Chinese operation, you know all hokey. Here's another one in the Philippines that one ended up burning down. So those of you who are OCD kind of like me or I'm nervous by the tangle of wires. You can rest easy that got what was coming to them. They all burned down in the same thing right here. Say they got what they would they deserve for the, for the wiring so badly. Here's another one. I think this one's from Ukraine. This is a modern one. What the

heck are these things doing? They're not processing, transactions are all racing to solve these stupid puzzle. The more of these things you have. It's not the case that the, the platform moves faster, the platform moves at the speed of boobs. Lori comes up with a block every 10 minutes. No matter how many of these units you add. What you're doing is you competing with the other miners to make those coins, but you were not actually speeding the system up. If not, what is up. Kind of a system. So and so in, so doing as these miners are incentivised to compete with each

other and consume more and more energy. We end up smoking around 7 ish. I'm talking to you from a location. That is 42 Celsius in as I speak. So if I'm sweating and I look like I'm in a sauna it's because we have heated the world up on the amount of energy work on seeing right. Now again, this slide has become dated and no time. It's I think it's at least on par with Austria is to Denmark's 3-iron, visualize it for nuclear power plants are going into crypto mining. This is a huge waste, all of that energy could go into anything else,

just about anything else and it would be more. Useful for Humanity, we do not need all of this energy being spent the way it is being spent to keep tabs on the on the Bitcoin blockchain. So there are some people out there who think that the amount of energy we burn adds value to this point. It doesn't, it does not establish a price floor. It might be very expensive for you to dig out gold. It doesn't mean gold. Will have that value? If there is a lot of surplus of gold in the world is now we will crash. It'll be on feasible for you to dig it out, but it doesn't mean

that there's a price floor for the shiny rocks. You've got This thing is an environmental disaster. It's literally melting. The poles is literally killing an injured animals, and their many people out there were selfish. I would like, to think I'm not one of them. I got mine, you know, Screw you. I'm, I'm happy with what I get out of it. What? OK, Google what year was leaking and enormous amount of value out of your system, to the power companies. And this shrinks the feasible envelope for the boys by shudder to think what might happen if the Bitcoin price Falls a lot, then we're

going to be in a world of hurt for the system to Rico. So Where are we now? So nothing like those old stories? I should mention this. So we can. What do we get for this? We get it. Open for the super nice. But we also get a very slow for the Beach Boys on chain through. Put this about three transactions per second latency is 60 minutes. You can please put on Roblox bigger. You can make your block smart speaker. Make some more frequent this place. Every fourteen seconds, which it turns out. This is actually not a good number.

If means that you end up being Far Corners of the world, you have to be well, connected to be any Siri and minor. This is so stressful, and it's also ends up achieving only 15 transactions per second. Even at the speeds. We are seeing signs of lack of decentralization. In the etherium blockchain. We did a study on this. A few years ago. You can't just leave for a Masters on the mecca, Motel consensus, front to get to get to, to get to cover. All the use cases iot microtransactions, other place. Very, very high command. See if anybody so somebody who does mange Baseball,

for example, will deal with the transaction. Throughput will have to deal with the transaction through that. Just cannot be done at these streets. So, And then there's also this, I let me just message him. This Parting Shot the world. We're doing all this for doing all this dude to do what to do to get an open system, that that gives us a decentralized. Open. Robust infrastructure. What time do? We get that? They look how many miners are there in this pain will have a lot of minors with a pool make make make make elaborate the end of organizing themselves into mining pools

and they're only about nineteen of them for B. Going at the families of the study that we're about 12 of them. If we're Syrian does accounted for 99% of the other block production. So at the end of the day, these people end up, end up losing together and you have only have a handful of entities are creating or blush. And anyway, so you it's the very that piss-poor at the centralization. They are they're terrible at scaling their terrible at performance, the turbolift late and cheese, but they are open and robust. So, I'm so it's surely would be nice

if we could somehow somehow combine the best of these two world. If we could have the quicksand ality classical Chronicles, without the terrible over has been squared over heads without having to require precise membership without having to make them clothes are small and it was supposed to be nice to have the robustness of Nakamoto combined with with low latency high, throughput and scalability. And of course, you didn't have to waste all that energy. So, how do we do that? Well, about really I think at this point about three years ago, in 2018, a paper was

released from a team called Team Rocket. Team Rocket came up with a new family of protocol. They work differently. The classical Protocols are protocols, in which every note is trying to maintain decorum in Burien by hearing from essentially every other one on the order of every other know. This is like all the old voting. If you look at Heather, a hash brown, if you look at Cosmos, if you look at these two, these are all classical. Protocols are trying to resurrect ideas from the 90s late 90s or early early North and adopt them into the blockchain.

Setting to to get altitude sickness. I should get away from proof of work. That's, that's a laudable goal. But they're by Joy still there. I should giving up on, on all of these things that that we think are crucial, which is open this, which is high performance, which is scalability. So Avalanches difference, it's going to require it's going to operate differently and I'll give you the hints on how it operates. Its 8, sampling baseball team for the car. You don't forget to ask everybody. You ask a small sample and you don't. So unlike classical. Maybe

they're in the latency might be a little higher actually, because you don't just do it in a fight, in a small number of a small number of us to rounds, but as a result, you end up getting you end up being able to scale, two emails numbers apart, the multiple protocols, a family of a family of chemicals are inspired by epidemic. For those of you who've heard of these storms will know that these are for dissemination of news or dissemination of new information and they've been used everywhere. Just about every web site

uses a gossip Network behind the scenes. This is Kai. I don't like that but it's it's solving a far harder problem. It's not just disseminating information. It's making sure that everybody every participant. That is correct. She's the same set of decisions and the system remains up and live. As long as the system is, as long as the lightness guarantees. Our goal here, is quicksand Melody, very, very low latency high-throughput. It should stay out of millions of nodes. It should be robust. It should be green and sustainable and

them. So what's the trick is, is, is, is you you don't force everybody to talk to everyone. Imagine the time in the Senate. Imagine that the Senate doesn't just a hundred people. Imagine. It's a hundred thousand people in Giants stadium. I could if I were doing the classical thing, I could ask everyone and collect responses. It would take forever. It would be very time-consuming. It would be and I was going to let you change would be insane. Every time someone comes into the stadium, I have to do an M cubed operation, which is impossible. These just wouldn't

work at their skin instead. I'm going to sample and & quiri members of the stadium. They themselves will do the same thing and magically because I'm sampling you all and you're simply me and other people. And, you know, it by sampling re I'm getting the benefit of the people that she has and maybe samples of Veronica is sampled the Arthur who's part of the stadium at cetera. And there is that we ensure the kind of coverage. We want probabilistic and will create a metastable system. Assistant that is unstable. As long as a decision is not me

and we will try to make the system especially if I go set up a kind of a dynamic. Fall into one side or the other, the system will not want to be in an indecisive indecisive state. If you want to come quickly, come to an agreement on whether something happened, or it did not happen. So how does it work? Well, okay. So let me go back to my example. If he, let me give you a hint of how it works for the very simplest of the protocols in the family. There are multiple and I'm sure there are many more to be discovered to come after us, but I'll give you a, I'll give you a set of the core idea of

how it works. And I hope it's useful. Imagine that were in the Giants stadium. Imagine that we want the stadium for the participants to come to a decision. They want a, we want them to say, either blue or green. We want all the correct nodes to pick one called her. It doesn't matter which color we pick what we want every good. Nice person to pick the same color. So he say, look, that's okay. We'll deal with them. If you're a nice person, then follow the protocol as follows the protocol, going to be very simple. We're going to do the same

damn thing over and over and over again. What's that dumb thing? Pick 5 people that surround them from the audience to pick 5 people from the stadium. And then query, what color preference they currently have. If they don't have a color, then they adopted the color that you've got. If they have heard of this, like, we're trying to make a decision, blue or green, leave the EU or stay within the EU. We're going to go left or right. It's like I kind of a binary decision to let me just simplified in the sky. So I'm going to go left or right with,

you know, with my stadium in the way I'm going to do this is the participant. I'm going to ask five people. Hey guys, you know, I heard that somebody wants to leave the EU. What do you think? And I asked re Veronica Arthur and two other people and I guess back responses and Veronica might say. Yeah. I heard the same thing. I'm down to leave the EU. I heard earlier from somebody else who wants to stay within the EU, I object and he says green instead until I get back responses, blue, green, blue, blue green. And so when I hear this, what I do as a correct participant to this

protocol is that you're just my own preference to be on the side of the majority of my sample. It's that simple. A sample 5, I go with the majority. And I'm going to repeat the stupid simple process. Some number of times that if you think about it, globally, if you look at this entire system from a very, very go from a god's-eye view, the very, the very worst thing that could happen here cuz I start out life right in the middle, 50% are favoring blue. 50% are favoring green. That's the worst thing that can

happen. After one round of this. I'm going to sample, some people. I don't know who I just pick them at random already samples people and Veronica sample. It is incredibly unlikely that I'm after one round of random samples. I'm going to be exactly 50/50 it's possible and if it does happen, then you know, this process will repeat itself. It's actually less likely to happen again and again and again, I'm not going to stay here. It's very very unlikely. What will happen most likely is because of random perturbance, has one side will be

slightly more favored than the other side. So I sample I get blue Veronica, sample. She gets blue already samples. He gets green and so on and let's say that we are slightly more blue than green at the end of the first round. Can you repeat the simple? And now you can see that if I have a small by is in the network, is going to have a fight. So if I'm 51% ahead on the blue side, next sample is going to take me a little bit further towards the blue. And so I might go to 53%. And if I do this again, I'll go from 53, you know, simulations show that you go to

something. Like fifty six. Next thing takes you through 61 and then suddenly there's a clip and you start going down this Cliff. So you start out in the flea by Van Halen State, and you go very quickly towards one or the other cat or the snake that has swallowed, an elephant, whatever you want to look at it. But down this this energy gradients to a lower energy State for the network and at some point there exists, a point of no return and let me, you know, I don't want to go through the complicated Markov chain, analysis that is behind this, but let me just give you the

simple intuition. Imagine that you are the only blue in a stadium of green people. You are guaranteed to in the next round. You're guaranteed to turn green yourself. So you can work backwards and you can see that there is a, you know, that if you are towards the edges of this, the outer edges than the system will actually remain at that. Once the entire network is green, then it's just beside the green. Now, there is no chance or hope or anything else. Anybody flipping this? It's immutable. The decision has been made. This is really cool. It's a different kind of a system. We are

not sampling. The whole network amazing in the number of messages. It says it turns out that okay. Let me just make sure it is. There could be some number of people in the system or Byzantine. They're playing tricks with you. And when the network is going towards the green side, they pull you back to the blue Etc. Well, you can compute exactly when there was a spaceship. There's a point Beyond, which the Byzantine guys are unable to pull you back. So I'm simplifying the time. I talk to you about the slush, brother. Actually, the snowball protocol is actually the one that

handles them. These a Byzantine participants. I don't take enough time to go into the details. Other than that, it would not be appropriate for the WhatsApp. But you can see that that decision could be made this way. And, you know, I can demonstrate this with their with pictures, but, you know, it's just going to skip through this. It's okay. Give me some more and more insightful, graph, the network starts and this shows that I think this is a case where the other networks read. Everybody can read. This is the zero point. It's showing badly on the scrap

of the zero. Point is right here. There are no Blues left inside the network. Are there many different protocols? A slush snowflake snowball, an avalanche. And I will not go into the details, and protocols themselves can be specified on a single slide. These are really simple. So I shudder to think about the complexity of Cosmos as you change. This thing is incredibly robust. Let's suppose. That's that Veronica knows, you know, a thousand more people who are in that war in the audience in the stadium. And I know like five hundred people less than her or whatever. That's okay. It's turns

out that this protocol is quite robust. My sampling will cover the vice versa for her. She's going to get the coverage she needs without having to sample, everybody. Thought you knew everything I wanted to say, there are some advances year we can make all sorts of fancy. Of addition to these Protocols are also, some errands want to talk about, but but I'm going to save them and there's a cool trick. One can do with a, with a directed acyclic graph where you don't make one decision of the time where you change your decisions. So that my stand is

predicated on a lie. My spend that I suppose. I'm D my spend to z e right here. On the left is predicated on the other money. I spent earlier 222 e, which is predicated on Elvis, painting, Bob, and the network and make a decision down this path. And we can pick all of these things all at the same time without having to individually and separately, make a decision for each other. Very cool idea. It allows us to amortize the cost of a decision over a larger number of transactions. These are just some of the tricks that we use to make the Avalanche system. Far far, far

beyond anybody else? And so I think, I just want to wrap up in the next 1 minute or so. This thing is incredibly efficient because you're not fat, you're not talking to everybody and it's very quick. It turns out because he's kind of exponential, plus he's kind of a pandemic, some stuff on you know, how exponential growth works. It's very quick. About 17 rounds is sufficient for the net worth of $10,000 to 17. Remington. 17. Rounds. Only means 170 messages. So $10,000 would require a hundred million messages. If we tried to use a

classical protocol like PDF and instead we're getting 170 messages per node and and the rich even consensus. So are there. Lots of cool things that we could talk about? I'm not going to get into. It's very strong and usability. So people have tried to replace proof of work with proof of stake. They're talking about, you know, if you got a session leaving behind 19, mining pools, or 11 mining pools, replacing them with 21 block, producers and 64 for Casper, 128 for East, to use like tiny numbers of participants for

decision. In avalanche. You can have tens of thousands and millions of participants. Or there's another cool thing about this is safety, or for the network, where more than half the participants are byzantines. When more than half, the people are evil, but if they aren T safety, so what does that mean? People like me, even at fifty or sixty percent of the of the network is corrupted they cannot cause a double stand. If you promise to visit for this kind of the world, that's a pretty amazing thing to be able to say where else classical

protocols breakdown when a third of the participants are evil. This thing is not like that at all. If you have just the compromise, if the protocol of a spirometer I support for the Byzantine components of 33% and the attackers 34%, She breaks through your guarantees or probability, but he doesn't immediately calls a double spent. In fact, numbers show that this with the prime factorization. We have a double spend is incredibly unlikely, only likely to happen once in twenty thousand years or so. For One Prime traditionally, the end of the attackers 34%. Sure, he blows you

just blow, you killed it. Or does that mean? Well, you know, now you can cause a double spend once every ten thousand years, you really need to have a good good chance of a good meaning, like, you know, something in their double-digit percentages, Chance of causing a double Spin and you can use it with proof of stake. You can use both already. It's up to you. The choice of a simple function of a sibil, sibil Control Function is up to you and it's in a material in the Avalanche system will use proof of stake because we believe in green stuff.

We don't leave until into the system live forever and we want to to to have it live in a nice World War 1 more minute dopest thing. It's now running its miss holding multiple billions of dollars billion dollars worth of value and is very fast. I always advertise 1 to 2 seconds for the station. It's actually less than a sec. Is a shell around three hundred milliseconds per decision. If you ever played with Avalanche, you can never go back to the Syrian, the

Syrian text many seconds, even though they tried to hide it, it would the decisions are taken for it to be really spoiled, Arthur, Gervais and tell you exactly how long it should take. It should take about 37 minutes according to his thesis Avalanche, makes the decision decision and 300. Ms. Is the stupidest very high. Do you know, where do I want to leave you with? This is a very exciting future involving. The last word on blockchains has not been written, and do not be confused by the

people who got stuck on the 1st, or the first word is Bitcoin. It should gave us the vision. It's an amazing System supposed. She was a very, very brilliant researcher. All that being said, he was by no means the last person or the only smart person in this in this area and there are Many many other advances that have come after a lot of the misconceptions I hear about blockchain is being slow being unscalable being this that than the other. They are things of the past there. Have been advances year that took the science far forward and now we can make decisions and far

less than a second in a globally, decentralized network with thousands of participants and hold immense amounts of value. So I hope you will join us in this dream of a of a realizing, a decentralised world. I hope we will all together jointly by people who will adopt portions of the dream and and discard, the very Central core of the centralization. I have in mind systems that are cute. That are that are stealing. Essentially, the idea that are putting it on the blockchain with a centralized gatekeeper in charge. Those are I think the true enemy and I'm working

together. We can build decentralised systems that are Fast and scalable. I'm going to stop there. And thank you all very much for bearing with me today.

Cackle comments for the website

Buy this talk

Access to the talk “The Evolution of Consensus Protocols”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Full access

Get access to all videos “IC3 Blockchain Camp 2021”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Ticket

Interested in topic “Blockchain”?

You might be interested in videos from this event

April 23 - 24, 2019
Singapore
23
671
bitmain, blockchain, crypto, cryptocurrency, fintech, fundraising, game, government, investment

Similar talks

Philip Daian
PhD Student at Cornell Tech
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Justin Drake
Researcher at Ethereum Foundation
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Philip Daian
PhD Student at Cornell Tech
+ 1 speaker
Mahimna Kelkar
CS PhD Student at Cornell University
+ 1 speaker
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Buy this video

Video
Access to the talk “The Evolution of Consensus Protocols”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
949 conferences
37757 speakers
14408 hours of content