Péter Szilágyi - Decentralization: origins vs. reality

So, uh, couple months ago, I was asked to do a talk here and I was kind of, um, in a bit of a pickle because I'm a technical person. I figured that doing a very heavy technical talk might not be ideal. And then it kind of dawned on me that decentralization is something that is extremely close to my heart. Yet I kind of feel that it is extremely misinterpreted all over the place. So I figured that amazing perfect thing to talk about.

Now if you want to talk about decentralization we kind of need to go way back to the roots about 80 years ago when there was this guy this Paul Baran published a really a milestone paper. Um what you need to know is we're in ' 64. It's post World War II. Everything was about war. So what we need really wanted to do is make systems that survive a war.

And up until this point, everybody knew that a centralized system is bad. You can just shoot it off. And then people kind of tended to create those that middle picture that middle picture designed their network topologies that way because I mean it's hierarchical. You can shoot bits and bobs off it, but I mean that's the best we can do. And this guy came along and started actually analyzing it that how many bombs does it take to destroy that network and the entire paper is about uh discussing the number of bombs the number of volleys it takes to to shoot the network down.

It's a really interesting paper and at the end of the paper he kind of ends up with a different uh suggestion that instead of creating these hierarchical structures why don't we just allow all our nodes to talk to everybody that's close by. And there's a really interesting analysis of how many thousands of bombs you need to take one of these networks offline. Now, if you remove this the war aspect from this paper and you you move it into a peacetime world, what uh the really nice observation in the paper is that if the network is tolerant to being blown up, then it's also tolerant of simply failing by itself. And why is that good? This really really good because in general in in any computing environment any engineering environment if you want to make something more reliable generally tends to be tends to cost significantly more.

So the more reliable something is the more expensive it gets. And the idea was that wait, if we have such a fancy mesh network where we tolerate it being blown up or at least parts of it being blown up, then uh what if we don't strive for 100% reliability anymore? What if we just say that okay, let's make it, you know, 50% reliable, good enough. Maybe half of the network will fail, but if it's going to be significantly cheaper, then we kind of win. And the second really observation is that wait if we made it cheaper significant let's made it 10 times cheaper then we can maybe reinvest some of those money to you know make it denser and by making it denser you actually raise the reliability and then you enter in this really interesting circle where you make things uh cheaper then you have more of them then you can make them even cheaper and then you have more of them and this really interesting phenomenon happens and um essentially this I at the end of the paper described the internet before the internet existed.

It's a it's really amazing paper. Now um this is kind of what decentralization was kind of defined. But how does this look like in our little world now? Ethereum launched almost 10 years ago. In two days it will be the 10 year anniversary of the Ethereum launch.

That's from the launch party discussion. That's Jeff. And uh my in my opinion, Ethereum was pretty damn close to being perfectly decentralized when it launched. You had uh a lot of nodes. Everybody who ran a node was kind of like an equal opportunity uh runner.

You had the full state history. You you basically did a full sync. Everything was full. If you wanted to run create blocks, you wanted to run a minor. It was completely free to everybody.

You could CPU mine, GPU mine. You did whatever you wanted. If you wanted to transact, you just shot your transaction into a transaction pool and everybody got it. Everybody operated on it. It was perfectly equal.

And the best thing about it was that uh it ran on everything, ran on anything. And when I say that, literally when Ethereum was launched, most people didn't really want to buy expensive miners. So, they just bought a beer crate from their supermarket and just dumped a minor an Ethereum node into that beer crate. And um I think the the other two pictures are my own systems that I made is uh one of them is a Raspberry Pi computer which is probably one of the least powerful Raspberry Pies ever made and it could run an Ethereum full node and also it's my first smartphone ever that I had. It also managed to run a full Ethereum node and um why was this good?

Well, that was in 2017. There were literally tens of thousands of nodes in Ethereum scattered all across the globe. And all of these nodes were mostly equal. Sure, some were more powerful, some were a little crappier. But uh now imagine if you were a state in a state of war and you just try to shoot that thing down.

What do you do? It's just there's nothing to shoot down. So it was an amazing golden era for Ethereum and uh it's not only my opinion if you look at the price charts actually this was the time when Ethereum almost flipped Bitcoin. It was 7% I think 7% away from actually flipping uh Bitcoin in market cap. Unfortunately after this era came the era of it's not enough and uh basically pick your poison.

You can make say that it wasn't fast enough. It wasn't cheap enough. It wasn't quickly. It wasn't finalizing quickly enough. Basically, there were lots and lots and lots of different properties that you could just find dense on.

And yes, those were actually good arguments. So, yes, it wasn't fast enough. It wasn't cheap enough. It wasn't everything enough. But the question is how do you fix it?

And this is where things get complicated. Um, a decentralized system is a bit messy and fixing it is a bit messy. Whereas we know that if we centralize a bit, we know how to fix it. It's it's easy. It's it's just natural.

So we kind of started going down on a less than ideal um pathway in my opinion. For example, um in in one of my earlier slides, I I enumerated kind of four properties that I found super important. One of them was that every node was equal. The other one was that uh everybody could mine if you wanted. Uh everybody had access to transactions and everything was cheap.

In this new world, we started go having these little baby steps away. One of the baby steps was that block explorers appeared. This is kind of natural. No harm there. You could instead of running if instead of checking your balance on your own local node, you just checked it on ether scan.

Yeah, it's perfectly fine. There's absolutely no harm. Then people realize that wait so ethers scan is super nice for humans but what if you want to have something similar for DAPs where programs don't have to screw around with their local nodes they can just tap into something that's somebody else maintains and that's how infura and all the RPC providers uh uh came to be and eventually currently the the standard that we're striving for stateless validation where we say that okay let's move the entire state out into external operators and uh even validators should just use them. Okay, baby steps. Baby steps.

Uh different aspect was uh kind of me when it appeared every transaction was kind of public and then people started abusing them, sandwiching them, front running them, back running them, all kinds of things. And this what got in the end this was called me and everybody agreed it's bad. So what we did we made everybody do it. We made it so easy to do that now since everybody's doing it, it's less bad. And um this in my opinion was kind of a bad approach because instead of fixing a problem, we made it worse.

Now all of a sudden people didn't want to be sandwiched. So these private transaction pools appeared, private order flows where all of a sudden you as a minor, I mean you could mine except you didn't have access to the transactions anymore. So you were forced to use somebody else's services. And again this goes on uh from a mining perspective. Originally in when Ethereum was launched fairly quickly mining pools appeared.

These mining pools were kind of centralizing but they were still limited by physical world by GPU availability and electricity. After st switching over to proof of stake the physical limits kind of disappeared and staking pools took their place. The problem was here that all of a sudden you can really quickly move all the stake from one pool to the other. So it's kind of pools get a lot more power and then with liquid staking again it gets it compounds up until the point where nowadays if any researcher is talking about tweaking the Ethereum network's issuance number one problem will liido torpedo it number two problem what's best for the network and that's a really bad situation to be in. Lido, by the way, is a really nice player.

I don't want to put shade on them, but there is that possibility that people need to take care of. And of course, last but not least, uh due to technical reasons, um after we switched to proof of stake, the minimum mining, minimum staking amount was raised to 32 ETH, which is approximately $100,000 give or take based on price uh fluctuations. And uh this was a bad choice. I mean, bad choice. Everybody knew that it kind of makes people a lot it cuts off a lot of people from participating but the idea was that no worries we will fix it and then later well everybody who wants to be a minor has 100k so let's just go from there and then we can assume that you have a fiber network because I mean if you have 100k that's the least you could do and then nowadays vitalik is talking about having uh provers in the network which uh with a power requirement of 10 kilowatts because 10 kil kilowatt is kind of like an average home user quoting Vitalik uh in Romania if you live in an apartment uh you are kind of limited to 4 kilow if you live in a house and have a singlephase current by law you are limited to 9.

9 kilowatt yeah that number needs a little bit of love still uh and um with all these aspects what we are what the people are trying to do there's nobody's trying to be malicious here Everybody's trying to really improve the system, make it more efficient. But time and time again this little question arises that well we know that it's not the best for decentralization but you know can we improve it and then think about the problem and the idea is that we want to create a global settlement layer and we want to displace Swift because Swift is the evil banker settlement system and the problem is that I'm not sure how many of you here know this story arc of swift uh before swift existed In 70 banks were actually settling transactions via telegram not the messenger telegram you know the morse code ancient technology telegram and it was completely 100% decentralized you could a bank in New York could send a message to London a bank in London could settle something in Shanghai and it worked except it was a little bit shitty a little bit expensive a little bit slow a little bit of everything so came the brilliant people and said we should make it more efficient And thus the swift cooperative was born out of this purely good um ideal that we would just like to make something a bit more efficient and they were successful. They were brilliant people working on it and in about 20 years the swift system became the global settlement layer. Of course, the problems began when in u 92 the US government uh realized that oh wait this is a nice juicy honeypot of information and they asked permission to access it. Permission was instantly denied.

Swift is a European company. It's headquartered in Belgium. The United States has no jurisdiction over Swift. So no bug off. So the United States waited 10 more years until 911 happened and that's when the really the floodgates opened because uh they enacted this anti-terrorism law which stated they can do pretty much anything in the name of anti-terrorism and even though they don't have jurisdiction over swift swift did have a backup data center in Arizona and that was kind of game over and the US government just approached the Arizona data center and told them that well from now on we have access to your data and do not dare tell the world about it.

And that was for five entire years. The US government was kind of data mining the crap out of the global payment network without anybody's knowledge. In 2006, uh the European Union actually found out and didn't do anything. It turned out that they tried. They were really big.

It was a big chaos and the US government just steamrolled uh through Europe. people were saying I mean I think back then European Commission ministers or whatever they're called were saying that u uh the US ambassadors were like doorto- door doortodoor salesman they were going to everybody every day trying to convince them to no you're not allowed to remove our access and eventually they gave in and uh in 2012 the weaponization of swift was completed when the US government actually forced them to uh unbank Iran and uh being unbanked from the Swift network was equivalent to being uh economically nuked. And that's that's not my term that's actually somebody one of the operators term. And um kind of the moral of the story of the Swift network was that it started out or at least before Swift, we had a beautiful decentralized system and then gradually we started making it more efficient and it was there nobody ever wanted to make something that's capturable. Nobody ever wanted to, I don't know, create something bad.

They were legitimately really nice people, smart people working together across 200 countries, 11,000 institutions trying to create something that just works a tiny bit better. And it took them approximately 30 years to capture it and 40 years to weaponize it. Now, in Ethereum, we are early. That's a nice quote. It's it's still early.

We'll we'll um in about two days will be the 10 year anniversary, which means that um even though in my opinion we started going down a similar path, we are way way way very far away from having troubles. However, even if history does not repeat itself, it does tend to rhyme. And uh which kind of gets me to a quote from uh Joe Biden. He he made this speech last year uh for the 80th anniversary of D-Day. The speech was that uh democracy is never guaranteed.

Every generation must preserve it, defend it, and fight for it. And for me, decentralization is kind of eerily um similar to the whole notion of democracy versus authoritarian government that if you have a if you look at China, Dubai, it is very very efficient. So you could it everything works and if you look at the elections that were happening in Europe in the US over the past uh years or decade it's messy really messy it's it's like it's chaos but u still we somehow tend to prefer that chaos over the super efficient thing and for me I think this is the message that I really want to get through is that the same way with democracy Decentralization will never be as efficient as a as a more centralized system. It will never be as cheap, as fast as everything. But it's a trade-off that we need to make if we actually want to be able to be part of that system in the long ter run in 20 years, in 30s, in 40 years.

And um so I guess my my my the takeaway that I'd like you guys to to remain with is that while we are uh pursuing the number go up just make sure that um every single baby step that we make try to look at the system as a whole. Try to look at from where we started and where we ended up with instead of looking at where we are today and where we are going to be tomorrow. And maybe if we do have this mentality, then we can avoid actually being boiled alive by different players. Thank you very much. [Applause] Amazing.

Thank you so much, Peter. That that was um that was fantastic. So many things I didn't know. Um okay. Well, guys, it's time for questions again.

If anyone has uh want to put the QR code or indeed has their hand up right over here, microphone coming to you, sir. Hi, Peter. Uh, great talk. I'm I'm wondering, do you know any kind of systems that actually stood the test of time and managed to survive or do you feel like this is a slow fight and a slow loss and what we can only do is like just slowly lose and maybe if we do better we we just increase the time but eventually we will lose.

That's a brilliant question. Uh I'm optimistic. Yes. Uh I think every system has a tendency towards centralization. That's the natural progression of things.

So unless you actively push back against it, you end up in a state that where you don't want to be. That's in my opinion the default. However, uh if you look at the most popular decentralized system nowadays, I would be that would be the internet where yes, of course, you have Google in a bubble, you have Facebook in a bubble, but all of these are kind of bubbles within the internet. But the internet itself, you cannot really say that it is uh centralized. And in my my opinion, the the biggest um difference between the internet and say Ethereum or other systems is that the internet can scale down.

So if I want to run a web server, I can take the tiniest little potato computer and it's able to run a web server. It's able to participate in the internet not just a not just a client but act actually as a server as an active participant of the internet. Whereas with the system that we are creating here, I think it's this is one of the aspects that we're kind of tend to give up. We're trying to we tend to say that well we don't really want toasters. We don't really want potato computers.

But I think if you we start focusing a bit more on that that yes we should have toasters. We should have potato computers. They will be super lame. They will be able to do very little. But if you somehow permit them to still be part of the network even with their little capacity, then it means that it doesn't matter how many big bubbles you create within the system, you could always create your you could always carve out your little niche where where you do you and uh you kind of stay outside of these uh influence zones.

So the answer is I think the tendency is to centralize. I think there is an example a living example of a system which works and is not centralized and it's kind of up to every each of us to somehow try to push try to nudge everything towards the towards the design of the internet rather than the design of swift.

Awesome. Thank you for that. Any other questions? Ah the man in the baseball cap right there.

Thank you. Do you think Ethereum should become an um all use case solver in the future or do you see much more um staying as it is right now without avoiding all these concerns you you presented right now and um take advantage of um use case specific layers. So usually um I think with Ethereum one of the big problems that I see is that it's uh it wants to be an atomic state machine. It wants to be a central thing conceptually where everything has access to everything instantaneously. And the problem with this one is you cannot really scale such a system unless you try to cut off all the little weird dangling edges.

In my opinion, um, Ethereum could become everything. It could it can achieve its goals, but somehow it needs to give up this this um this notion that it wants to be a shared state for the entire world together. Rather, it should, in my opinion, go towards the this design where it's a collaborative state for the world. So, it's perfectly fine for a bit of Ethereum in Asia to cooperate with a bit of Ethereum in the US. But if they don't have to be the exact same thing, then they can somehow live separately, just collaborate.

Whereas, if you want them to be the same thing, then it's it's much harder to achieve.

One more over here. uh if you could go back in time with all the knowledge and experience what happened in the last 10 years, what were the maybe the major things that you would change with how Ethereum developed over those 10 years to maybe prevent this or just otherwise just some technical aspects or anything like that?

It's kind of hard to the problem is it's very very hard to prevent these kinds of things when uh when there's enough money on the line. Um I think what we need to do and it doesn't really matter whether we did did it 10 years ago or whether we should do it now it's rather instead of um I think we we have a tendency of looking towards less hard solutions and even though everybody's saying that Ethereum is tackling the hard solutions headon true and not true it is true in that Ethereum has amazing people who are working on really really hard solutions and solving them. The part where I say it is not true is that many people are too afraid of competition and because of the competition there's this um time pressure on people to it's not enough to solve the problem you need to solve it today and I think that's one of the problems where if you're not willing to sacrif I mean Ethereum is kind of number two blockchain forever and if you're always living in the fear that oh my god I'm going to lose that status I must never lose that status then you're always going to make decisions where it's it's not the best solution, but it's the best we have now and just roll it out and then we'll just fix it later. And I think this is the the mentality that I I would change a bit. I think it's better to fall down to number three, number four, number fifth place, but at the end end up with a system that can climb back on the ladder.

Whereas if you just try to always compete with the systems that are a lot more centralized, then I mean time is against you. And I think that's that's something that we should we should accept that some things might take longer. We should accept that some things might be less ideal in the short term if we can make it better on the long term. So that would be my the thing the mentality change I would do.

Fantastic. any other ah our blazerw wearing fellow over there

thanks uh we don't want governments to weaponize financial systems like swift but at the same time we don't like in the name of terrorism fight against terrorism as you said but at the same time we don't want terrorists to weaponize decentralized systems for criminality purposes where is the middle ground to preserve both because we need both to to fight against both we need somewhere a middle ground where is that middle ground.

I think the middle ground is in a social layer. So the same way that um I think my uh Emma from a previous talk, I'm not sure if he's still here, mentioned that um um a lot of tools that we build today can be used for both for good and bad. And it's kind of boils down to what you do socially. How do you do you punish or do you um if if you sees that something's going wrong, do you uh just step in and try to try to change course and I think um if we are trying to say that we don't want governments to be able to censor but we also don't want terrorism to be able to use it. I think the the answer is in the social layer that you need to create a decentralized system that is actually protected from governments.

But when you see that something bad is happening, there needs to be an appropriate global um attempt to actually root it out. And it doesn't need to be a technical solution. It doesn't have to be a solution part of Ethereum itself. It can be a solution part of the social layer built around Ethereum.

Thanks, Brill. Any other questions from the crowd? If not, we'll go to the board. So, Peter, it was great to hear your thoughts on what you would have done differently before, but indeed, what would you do did differently in terms of the long-term vision for Ethereum and decentralized system? Where should we focus our research?

Yeah. So, here um the prevalent thought is we don't want Ethereum to run on toasters. My prevalent thought is I want Ethereum to run on toasters. So that's uh I think that is an unlock that prevents a lot of problems from happening. So just allow anyone anywhere with the crappiest little machine to be part of this and then the rest kind of in my opinion will sort it sort itself out.

I love that this this talk should have been called Ethereum on toasters actually.

That's great. Um, what do you think of ETFs, Black Rockck, and all institutions jumping on Ethereum from uh Guay Guru 2011?

I mean, I don't know. It's uh I think any any system that becomes valuable enough eventually people want to be part of it probably because they think they can make a lot of money off of it and that isn't necessarily a bad thing. So them joining I think it's inevitable. If you if you make something successful enough then you will have all kinds of players and um I think being a platform I'm not sure whether Amazon had this thought at some point or not that if you're a platform you you shouldn't really compete with your customers. So the idea is that if Ethereum is a platform then anybody should just come and use it and it's not really our place to to like or not like them.

makes a lot of sense. Um, last question we have time for here. Is the Ethereum Foundation a centralizing force?

Oh boy, that's a loaded question. Um, yes and no. That's my perfect answer. I always like to evade answering the questions. Uh, the F is doing a lot of is attempting to do a lot of nonentralizing things.

So for example, I've in the past years I've had a number of projects that I wanted to do in the EF and these projects were denied based on the sole reasoning that if the project becomes successful, it's just too centralizing within the EF. So the EF at the government's level, they are legitimately trying to not be a central point of failure or a central point of control. Now from in my opinion from a research perspective from the directing uh or guiding the Ethereum protocol I think the EF is a bit too powerful in that usually you have these little research ideas that are debated within and you have some directions and then after they get published it's very very hard from an outsider to push back against it and that has two reasons. One of them is that they are very very high-profile people working there. The other reason is that they actually do collaborate with the ecosystem.

So they do check with big players. So by the time something bubbles out, it's generally done or at least the general concept is done. So it's again it's not really a black and white. It's somewhere a little gray area in between. Personally, I think it would be nice if uh if Ethereum research, if Ethereum direction was a bit more decentralized, if a bit more organization kind of participated in defining what the lower level layers do and uh and not so if you talk about EIP whatever for the EVM changes, there are thousands of companies that have infinite money that they everybody is just pushing and nudging and everything.

I think that works pretty well. But for the lower protocol, I kind of feel that there's not enough external contribution and feedback currently.