[trustless://eil] Vitalik Buterin - Why Trustlessness Now?

Transcript

I'm not sure how many of you guys were here before, but it has been a wonderful day. And I get the honor of introducing the man that needs no introduction, Vitalic Buterine. Okay. So, great. Ask a question. That's uh Okay. So, I see my first slide down there. I don't see it up there. Um let's wait for this to but the

Okay, perfect. So, why trustlessness now? I mean, I see I think I see we can only see the you know the middle 80% of the slide. Um I don't know whether to classify this as a trust problem but uh that's um you know so software correctness is one type of trust as I am going to talk about over the next 20 minutes or so but uh why trust also now right is the uh topic that I wanted to talk about right so I look at this the whole day you know like we're talking about this uh theme of uh building applications that you know do not have choke points that can block you, that can steal from you, that can rug you and

you know like what is the point of this right and particularly what is the point of uh this year as a time to actually really start doing something about it. So first of all, you know, oh I see they fixed a bug. Uh why trust? So trustlessness is not the only thing. Trustlessness is everything, right? So I'll let the meme speak for itself. uh but uh you know basically we often you know like in DeFi we think about percentages like you know like 3.6% 2.4%. And uh of course if you don't care about DeFi then you're going to get 0% on your ETH but then uh the thing that the thing that matters the most in finance is not the difference between 3% and 2% and

it's uh the diff it's making sure you don't get 100%. Right? So in Argentina there was this event about 25 years ago at the Coralito where basically the uh peso previously had a onetoone exchange rate with the dollar and then overnight they made this decision where they basically force converted everyone's dollars to pesos and then immediately down uh devalued the peso uh and uh it dropped by about 70% over the course of two years. And so if you had your money in a centralized bank account, effectively during those two years, your APY would average minus 46.53%.

Who here remembers Mount Gaus? So if you had your coins back in Mount Gox, then Mount Gaus of course got hacked in 2014. If you had your coins in Mount Gaus, then like I think you lost something like 80% of your money. And it's like different for I think for different people and like finally they only managed to like really start doing proper repayments over the last couple of years and then actually out of that I think they discovered like some of some of the coins and then there was like all kinds of fancy legal wrangling and like actually the lawyers got a whole bunch of it. Um, and then uh down at the bottom, right, if uh the wrong people discover how much money you have or if

the wrong people discover who you donate to and if they discover where you live, then your APY is well, you're losing something more important than your money. So trustlessness is uh very very important. Um, also who here remembers FTX? So basically, yeah, you know, you know, I love this sentence, right? It like it goes on all of the ads. Like I think the FTX ads are like this perfect, it's like if you take like trustless colon values and like you, you know, invert them 180 degrees, right? Like imagine if you know you you turn them into like LLM LLM embeddings like swap the pluses and the

minuses and then you like take them back out, right? Like this is basically what you get, right? It's uh like one single person who uh voluntarily kind of put his face up to cover like 70% of every ad and uh basically saying I want to do good but for the good to happen you have to trust me and for my your money to not disappear suddenly you have to trust me and of course the money all uh ended up being uh abused to do some uh le leverage speculation effectively that ended up going wrong and a whole bunch of people lost their money. So this is what we're trying to do better then in Ethereum. Trustlessness is eroded very

easily, right? So trustlessness often does not disappear in one uh magic swoop. Trustlessness is something that easily degrades over time. And uh one of the analogies that uh I like that uh I know a lot of people use for this is email. So email is theoretically a permissionless and decentralized protocol. Anyone can make an email server. Who here has run an email server? who here has had the experience of sending email from your own email server and then realizing that like you're immediately put into a spam filter by basically every other corporate email provider.

This is basically what happens, right? And so email is a protocol that was very egalitarian, very decentralized, very permissionless in theory. But then in practice there was this kind of de facto add-on of spam filtering that got added. And then in practice the individual providers I mean like started uh making trust relationships with each other and then in practice uh it's uh all of these things started kind of piling on top of each other and at the end of the day basically email ended up uh not really being actually permissionless in uh practice. So trust is eroded very easily. So if you use a trust dependent bridge to move your funds from chain A

to chain B then the bridge can steal from you. Now of course one of the things that you can do is while you can you know take your money and then you can split it into a thousand pieces and then you can send each piece a bit at a time and then if the bridge uh tries to rug you at most you're charged effectively a 0.1% transaction fee. But then of course the problem is like you have to write that software and then you have to like your entire trade takes like a bunch of hours and then your onchange transaction fees go up by a factor of a thousand. And so if you have often like you do actually wants to move a big chunk of your net worth from one place to another and if you want to do that safely then like we need a

trustless bridge. Um a lot of D5 protocols depend on onchain governance. Onchain governance is something that can can be hacked. It is something that has been hacked and it is something that will be hacked again. Um, onchain governance can get corrupted by uh people bribing coin voters in all kinds of offiscated ways. It can get corrupted because 95% of people don't even vote and then uh it's very easy to overpower everyone else that does. Uh it can uh get corrupted in all kinds of ways. Um third party dependencies in uh software.

Actually, this one also applies in hardware, right? Like basically our software and our hardware, the problem isn't even that it's made by one guy and you have to trust the one guy, right? Like you know there's this uh famous kind of problem in econ puzzle in economics that actually shows that there's a certain type of decentralization which is actually bad, right? And it's like the kind of you know like the end kings on a river problem, right? Basically, you know, if you have one king on a river, then the king charges the toll and the toll is going to be a high toll and then uh if like obviously the toll is going to be pretty high and so it'll deter a bunch of people from using the river at all and you know this causes a bunch of damage to the economy. But if you have

like 10 kingdoms on the same river and you have to get like 10 out of 10 permission to cross the river then or you know to s to sail from one end of the river to the other then like each one of them charges a separate toll and then when each one of them chooses their toll like basically they only pay 10% of the cost of setting a toll that's uh to so high that it deters people. And so effectively like you get 10 people who are all price gouching you to the point where most people are not even going to go through the river and then none of them really get much revenue, right? So it's like parallel versus serial, right?

It's like parallel decentralization is good. Serial decentralization where you need to like you need per like acceptance from node of n like that's actually bad, right? And so what we have in our software is like we have 14 different dependencies like a bunch of different software libraries and people are trained to just like casually import things by default and then if one of them breaks then uh you know you can't even install your software like I'm who here has had that experience. It's like you know you download a software package and it has like 20 dependencies and it turns out one of them like doesn't work on your particular version of Linux or whatever. Then if any one of them has a bug then uh you know you end up uh bas uh having a security vulnerability in

the whole thing. If any one of them changes in a particular way then uh you know the entire software has to rewrite. Uh then if any one of these dependencies is something that involves talking to a server then uh you know these are the kings on a river and any of them can sort of impose extra conditions on you and a lot of them will right and so and uh often you have like these weird kind of cross attacks right where you have systems that get sort of formally proven correct secure within one model but then there's this dependency in some totally different part of the stack and there's some like flaw that's not even foreseen And uh then with that flaw that's uh not because of that flaw like actually the

assumptions behind the security proof are not even correct anymore and there's like some weird combination attack that combines those two things. Right? So third party dependencies in a in a server security holes and the problem often isn't even that we have the that there's one actor you trust. If there was that would be a much better world. often you have like 15 actors that you need to trust and then actually each of them are trusting other people and then there's this whole like crazy chain um user interfaces on a server right so safe got hacked this year and uh of course safe uh the contract system was fine but the user interface it was on a server and so if you can hack the server

you replace the user interface with something else and uh if like by default users don't really have any defense against this and they lost over a billion dollars. Um, a centralized API in a DAP interface. So, this is a privacy leak, right? A lot of DAPs today, they uh access not just the RPC, they access a whole bunch of different servers. And uh if you have a DAP that accesses a whole bunch of different servers, then each and every one of those servers knows when you're online, they know your access patterns.

uh they uh potentially, you know, get information from the de from the DAP, especially if there's a software library attached from your mouse movements, they could probably identify you like that. They learn a lot of things. Um if a price oracle gives the wrong data, then uh you're going to make a trade on an AMM where you set the limit price to be yeah, you know, 2,000 instead of 3,000. And then your transaction might get shipped over to an MEV bot. you get sandwiched and you sold someone your ETH at uh at at $2,000 and you lost a third of your money. Um if an RPC node gives you the wrong data, then putting the wrong data in transactions enables all kinds of theft and DeFi protocols. If

your RPC node learns your access patterns, that's a privacy leak. Uh basically, uh you know, your RPC knows what you did last summer. And uh it would be nice if uh we uh had a world where that was not true. So it's not even just about the core. It's about all of these different pieces across the entire infra uh infrastructure stack where you're basically you're putting a huge amount of trust in like an entire set of different actors and often you don't even know who those actors are and then often it's not just the people you trust. It's like you trust someone who trusts someone you trust someone else and uh the whole thing is just incredibly brittle.

Trustlessness is not something you should think about as a feature. Trustlessness is hygiene. Who are you allowing to see you digitally naked, right? Who are you allowing trusting to not take your money? Who are you trusting to not permanently freeze your money um or your or your documents or your calendar or your messages um or simply by walking away, right? So often you know there's a lot of uh projects on the uh like websites on the internet where just the existence of that website depends on a server. Five or 10 years later the server uh disappears because the company goes bankrupt or they get bored of running that particular service

and a whole bunch of links stop working a whole bunch of data stops uh stops being accessible. So think of it as hygiene, right? Hygiene, you know, it's not about adding a feature. It's uh it's it's about subtracting dirtiness, right? It's about subtracting risk. And I think uh this is the kind of spirit to which uh with which you want to approach trustlessness as well. Trust dependencies are a problem far beyond crypto. So this is not just an internal crypto topic. This is a 21st century global society topic. Uh there's this book uh the uses and abuses of weaponized interdependence by Henry Ferrell and a couple of other people

which basically talks about how the uh the centralized traditional financial system originally started as this thing that was meant to help bring together a global world and give people freedom. But then because it had centralized choke points, it got turned into a tool by which powerful people were are able to censor and exclude and uh effectively project power all the way across the world. Um this also it's important to talk about hardware right I think uh a lot of us uh you know we think about trustlessness in software we sort of take hardware for granted but hardware also a lot of uh opportunities for back doors there's a

recent back door by which uh a whole bunch of uh like RFID smart cards to get into rooms that were uh vulnerable to being cloned uh so lots of these things happen right You're like in 2015 the slogan is not your keys not your coins. I fully believe in 2030 the slogan will be not your silicon not your keys. Um brain computer interfaces. So these are starting to actually exist. They're starting to be a thing. And uh at the same time, of course, we have LLMs that can, you know, read our facial expressions and uh guess uh probably how many hours you've been awake, if you've been happy, if you've uh if you've been

sad, if you're in the type of mood that today you might be especially vulnerable to being scammed. Um so that's uh information is power. Whoever has information about you has power over you. And BCIs are something that is uh powerful for productivity, powerful for empowering the disabled. It's also something that I think will be necessary to continue to give, you know, humans a maintain our relevance in a world of AI becoming more and more powerful. But at the same time, we're literally talking about things made by people like Elon Musk and Mark Zuckerberg reading our minds. Who here wants like Elon Musk and

Mark Zuckerberg to read your mind? Okay, a few hands. Um, I don't know, maybe I don't know who who here wants uh I don't know like 10 cents to read your mind. Um, okay. Also, not many hands. Um, let's uh let's go around the world. Who here wants Nokia to read your mind? Okay. Um, so I see Elan and Zuck get a couple uh a couple more takers. That's interesting. But like I don't know the dominant opinion seems to be that like you know you don't want uh you know the big corpse to be able to read your mind.

Uh actually it's worse reading your mind. Um also these things are going to be able to write to your mind. Um, so, uh, I the the I think this, um, might have been at one of the previous DACK days, but there's like literally a thing that you can put on and it's like only a little bit more obtrusive than sunglasses and like basically the thing like almost gives you a seizure. Um, and so writing to your mind is like way easier than you might think. Um, and so this is also a form of a form of trust.

And so we do not want things like this to have untrusted 57 dependencies of dependencies uh corporate supply chain dystopias the way that we have become accustomed to with web 2. An important thing to think about today is not about uh cryptocurrency. Cryptocurrency is the warm-up. Um we have the technology to improve on this. So let's first of all know we'll talk about blockchains. Um so we have much more secure bridging protocols that uh people this week have been uh talking about. Um then we have uh governance minimization. Uh so there's a lot of work in building protocols that uh are

much less reliant on governance and where governance has a clearly defined clearly limited role. Uh where as many functions as possible are made governance free and those functions that cannot get time locked. research into capture resistance. I know there's a a group called anti-apture that's been analyzing, how to make Dowo governance more difficult to capture. Dependency minimization, uh so making your code not just like gratuitously depend on a thousand thou uh like 10,000 line dependencies if you only need one function. If you need one function, just like stick it in yourself. um UIs on uh IPFS. This would have uh you know

prevented the safe hack. Like I think we need a world where DAP UIs uh live on IPFS and their hashes are onchain and you actually need the DAO itself to approve um an upgrade of the uh of of the interface directly. Onchain version control also of uh software libraries. Um ZK snarks opening up a lot more frontiers uh for uh like privacy preserving applications. Privacy is a super important dimension of trustlessness. Uh private reads. So there is uh basically if you want to make your RPC not get all of your data.

There's these like two work streams that we're working on in Kohaku. One is the TE plus ORM approach. Basically, you know, you do still have to trust this trusted hardware thing. But the good news is that effectively instead of trusting an RPC provider, you're trusting like a two of two between the RPC provider and uh the hardware. And then there's this other approach P which is purely cryptographic. And you know this is like mathematically ideal. The problem with P of course is that it's like less efficient but also the efficiency has been improving quite a bit. So we can improve on a lot of these things.

Um Okay. Now, trustlessness goes beyond crypto. Local first software, including local first AI. So, who here has had a uh conversation through a messaging app in the last week? Okay, that's a that's a good baseline for uh just uh how how many people don't raise their hands for anything. Um, who here has had a conversation over a messaging app where as part of that conversation you have to like click on a message to autored.

Okay. Now, chances are the auto translate that you used was not local first and so even if it was on an encrypted app for the translation you're shipping everything off to Google. Um so there are very good local LLMs already right and this is something that we can actually use more that we can integrate more and uh for translation is one excellent use cases there's a lot of excellent use cases for this right I think uh there's a big open space for someone to really you know like make an operating system that like properly uses local first tools to do translation you know like including real time, voice translation, transcription, talk to the

computer, deep research, uh agent stuff, you know, code editing, all of these use cases. Do it open source, do it local. This is also a trustless technology. It is important secure communications. So, I think over the last two years, we've seen this uh amazing flip from Telegram to signal. Uh for how many people here is Signal their primary communications app? For how many people was that true two years ago? See, okay. Like still many, but probably about like half or even less than half as much, right? So, signal is really getting up, which is good. There are things that are being built that attempts to be even stronger than signal. So, session and simplex are two of them. There's also more secure

operating systems. How many people use graphine OS? So um also secure open- source hardware again not your silicon not your keys this will become an important theme um also a plug DAC day um coming up very soon and uh DAC day one of the things that we have here this is an air quality monitor uh where the newest version has some cryptography built in basically ZK and FHE stacked on top of each other so that you can collect data uh about local environments without compromising the privacy of uh any individual and so you can enable science and public health and

all of these good things without the downsides of basically a biopanopticon biopanoptic I think panopticons are a big deal right so uh one of the things that it's good to appreciate is like people you know people often focus on you know the flashy cameras and there's these like scary pictures and you have like you know like five different uh cameras going going around covering like a 360 by uh by uh by 360 view and then sometimes for like added irony value, you know, the camera is like right above a poster that's uh or like uh that says, you know, whoever gives up liberty for safety deserves neither liberty nor their safety and then you have like the camera right above it, right? Uh but

like that's only that's such a small part of it, right? Then like there was a paper uh this here where basically like literally Wi-Fi hotspots you can turn them into surveillance cameras. Then on the on the other side like audio is enough to determine the password of someone typing on a keyboard in an adjacent room. So that's been a thing for a while. Also if we go back to air sensors right um the CO2 level in a room it can tell you how many people are in the room. It can tell you when people entered the room, when people left the room, pot distinguish between uh different people. It could even tell you when people, you know, like started and

stopped doing various physical activities in the room. So you might think CO2 is just a number, but actually CO2 reveals everything, right? So this is I think uh really you know like thinking about how do we get information about the physical world in a way that is not a total privacy dystopia is a super important topic and how do we do this in a way that we can actually trust. Um, so we can make the Ethereum ecosystem a standard bearer insisting on these goals, right? We as Ethereum can be an ecosystem that really adopts and takes in trustless and uh maximally trust minimized technology both uh tru

uh to protect people's assets from being stolen and to protect people's assets from being permanently frozen and to protect people's privacy and uh we can use these things across the Ethereum ecosystem and at the same time we can build and we can use tools that have value that go far beyond Ethereum and uh that can enable a much um a much more secure world and I think a much brighter uh 21st century and so I invite all of you to join us uh in this journey and work on trustless technology. Thank you.

All right, we have time for just two quick questions. Um, the first is going to be, what do you think of the state of trustless bridges? >> I think it's a I mean it's improving, right? because uh I remember like two years ago basically everything um was very trust dependent um and then you know like this year there's been all of this uh like progress um like there's I mean even like the intentbased stuff in the OAF and then now you know like we have EIL finally like announced and uh it's amazing work that I'm excited about. Um so the the level of technology is uh is really going up right and uh

there's like more and more concret uh like actual concrete guarantees that you can get out of the bridges that uh are both both the ones that are in alpha and even the ones that uh the ones that are in production which is good. We're definitely far from 100%. Right? So I mean to get to 100% right you also have to remember you know the bridges have to be trustless and then the other big thing is uh the uh L2s themselves right because like there's no point in having a super amazing super perfect uh bridge if the underlying L2s that you're moving between are still basically multisig right and so I see there's a lot of good progress uh continuing in uh L2s moving

to stage two I see a lot of good progress in uh like bridges progressively reducing uh reducing trust assumptions and like thinking about this you know like rigorously and like and not settling for sec for like for second best and really asking like how do we really get these assumptions down like basically get people to the point where the experience you have in the Ethereum L1 L2 world is like the same standard of security that you would have if you're just moving things around on 01. And so, no, I think it's, you know, it's a it's a process and like I've I've been seeing a lot of a lot of signs of improvement.

And so, I'm uh like I'm I'm optimistic that uh you know, like something will uh something that really satisfies our standards of trustlessness will be available quite soon and then will become look very mainstream over the couple of years that follow. And given the context of MEV, isn't the block proposer a trust dependency? >> It depends what you use it for, right? I think because uh the way you ask whether or not something is a trust dependency is uh you ask how much can you get hurt if if they hate you and uh basically what the block proposer can do is uh they can censor you for 12 seconds. And if you have like one actor that is the

builder that is or you have like two actors that are the builders that are making 99% of the blocks, then those two actors can censor you on average for 20 minutes, which is like potentially enough to liquidate you, which is quite bad. We're not currently at anywhere close to that like that level, but we need to really make sure that we're not going to get there. And so I think you know we really need to get fossil uh so that there's more actors in parallel who can include your transactions quickly. Um also another thing of course is that if you as a user are sending your transactions to a builder directly the builder learns your IP address. That's also a form of trust and I think this is also why me not just mempools but even me like combinations of memp pools and

mixets are important. Uh then uh the other thing of course is that for many types of transactions like you're not sending from an EOA account and there's today a lot of use cases that rely on relayers and relayers you know they often break and uh I mean they also they're also a privacy leak and and then and they also sometimes break as well and so that's you know like something else that needs to be improved. So blog proposers are uh like they're very very far from being custodians which is good but at the same time we don't want to be complacent and uh just assume that blog proposers are this kind of magic cloud

in the sky that's going to do the right thing. >> And will you humor me with one more fast short one from a little 10-year-old hacker? >> Sure. >> Do you know where Satoshi is? Fortunately, I I do not and I give him a great, you know, a massive round of applause for like actually successfully preserving his privacy for all of these 15 years. A lot of people have told me that that can't happen and it and it has. >> I appreciate it. What a wonderful talk. Big round of applause for The girl signing up.