Full episode with Vitalik Buterin (Mar 2020): https://www.youtube.com/watch?v=3x1b_S6Qp2Q
Clips channel (Lex Clips): https://www.youtube.com/lexclips
Main channel (Lex Fridman): https://www.youtube.com/lexfridman
(more links below)

Podcast full episodes playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4

Podcasts clips playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41

Podcast website:
https://lexfridman.com/ai

Podcast on Apple Podcasts (iTunes):
https://apple.co/2lwqZIr

Podcast on Spotify:
https://spoti.fi/2nEwCF8

Podcast RSS:
https://lexfridman.com/category/ai/feed/

Vitalik Buterin is co-creator of and author of the white paper that launched Ethereum and ether, which is a cryptocurrency that is currently the second-largest digital currency after bitcoin. Ethereum has a lot of interesting technical ideas that are defining the future of blockchain technology, and Vitalik is one of the most brilliant people innovating this space today.

Subscribe to this YouTube channel or connect on:
- Twitter: https://twitter.com/lexfridman
- LinkedIn: https://www.linkedin.com/in/lexfridman
- Facebook: https://www.facebook.com/lexfridman
- Instagram: https://www.instagram.com/lexfridman
- Medium: https://medium.com/@lexfridman
- Support on Patreon: https://www.patreon.com/lexfridman

So then what is the origin story, maybe the human side, but also the technical side of Ethereum?

So I joined the Bitcoin community in 2011, and I started by just writing.

I first wrote for this sort of online thing called Bitcoin Weekly.

Then I started writing for Bitcoin Magazine.

You have this funny kind of story, true or not, is that you were disillusioned by the downsides of centralized control from your experience with, wow, World of Warcraft.

The fact that that's the reason I do decentralization is witty.

Have you always had a skepticism of centralized control?

Has that feeling evolved over time or has that just always been a core feeling that decentralized control is the future of human society.

It's definitely been something that felt very attractive to me ever since I could have learned that such a thing is possible.

So You joined the Bitcoin community in 2011, you said you began writing.

Started writing, moved from high school to university halfway in between that and spent a year in university.

Then at the end of that year, I dropped out to do Bitcoin things full time.

And this was a combination of continuing to write Bitcoin magazine, but also increasingly work on software projects.

And I traveled around the world for about 6 months and just going to different Bitcoin communities.

Like I went to first in New Hampshire, then Spain, other European places, Israel, and then San Francisco.

And along the way, I've met a lot of other people that are working on different Bitcoin projects.

And when I was in Israel, there were some kind of very smart teams there that were working on ideas that people were starting to kind of call Bitcoin 2.0.

So 1 of these was covered coins, which is basically saying that, hey, let's not just use the blockchain for Bitcoin, but let's also issue other kinds of assets on it.

And then there was a protocol called Mastercoin that supported issuing assets, but also supported many other things like financial contracts, domain name registration, a lot of different things together.

And I spent some time working with these teams and I quickly kind of realized that this Mastercoin protocol could be improved by kind of generalizing it more, right?

So the analogy I use is that the MasterCoin protocol was like the Swiss Army knife, you have 25 different transaction types for 25 different applications.

But what I realized is that you could replace a bunch of them with things that are more general purpose.

So 1 of them was that you could replace like 3 transaction types for 3 types of financial contracts with a generic transaction type for a financial contract that just lets you specify a mathematical formula for kind of how much money each side gets.

You say financial contract, just the terminology.

This is just generally an agreement where either 1 or 2 parties put collateral in.

And then depending on certain conditions, like this could involve prices of assets, this could involve the actions of the 2 parties, it could involve other things.

They get different amounts of assets out that just depend on things that happened.

So a contract is really a financial contract is at the core, It's the core interactive element of a financial system.

Yeah, there's many different kinds of financial contracts.

There's things like options where you give someone the right to buy a thing that you have for some specific price for some period of time.

There's contracts for difference where you basically are kind of making a bet that says like for every dollar this thing goes up, I'll give you $7 or for every dollar that thing goes down, you give me $7 or something like that.

But the main idea that these contracts have to be enforced and trusted.

You have to trust that they will work out in a system where nobody can be trusted.

This is such a beautiful complicated system.

Okay, So you were seeking to kind of generalize this basic framework of contracts.

So what technically are the steps to creating Ethereum?

So I guess just to kind of continue a bit with this Mastercoin story.

So started by kind of giving ideas for how to generalize the thing.

And eventually this turned into a much more kind of fully fledged proposal that just says, hey, how about you scrap all your futures and instead you just put in this programming language.

And I gave this idea to them and their response was something like, hey, this is great, but this seems complicated and this seems like something that's we're not going to be able to put onto our roadmap for a while.

And my response to this was like, wait, do you not realize how revolutionary this is?

What was the name of the programming language?

So then I kind of went through a couple more rounds of iteration and then the idea for Ethereum itself started to form.

The idea here is that you just have a blockchain where the core unit of the thing is what we call contracts.

It's these accounts that can hold assets and they have their own internal memory, but that are controlled by a piece of code.

And so if I send some Ether to a contract, the only thing that can determine where that Ether, the currency inside Ethereum goes after that is the code of that contract itself.

And so basically, you're kind of sending assets to computer programs becomes this kind of paradigm for creating these sort of self-executing agreements.

It's so cool that code is sort of part of this contract.

So that's what's meant by smart contracts.

So how hard was it to build this kind of thing?

I mean, originally, I actually thought that this would be a thing that I would kind of casually work on for a couple of months, publish, and then go back to university.

Then I released it, and a bunch of people, or I released the white paper.

A whole bunch of people came in offering to help, a huge number of people and have expressed interest And this was something I was totally not expecting.

And then I realized that this would be something that's much bigger than I had ever thought that it would be.

And then we started on this much longer development slog of making something that lives up to this sort of much higher level of expectations.

Is it fundamentally like software engineering challenges?

So what are the biggest interesting challenges that you've learned about human civilization and in software engineering through this process?

So I guess 1 of the challenges for me is that I'm 1 of the apparently unusual geeks who was never treated with anything but kindness in school.

And so when I got into crypto, I expected everyone would just be the same, kind of altruistic and nice in that same way.

But the algorithm that I used for finding co-founders for this thing was not very good.

It was literally what computer scientists call the greedy algorithm.

It's the first 15 people who replied back offering to help kind of are the co-founders.

Oh, you mean like literally the people that will form to be the founders, co-founders of the community, the algorithm.

And so what happened was that these, like, especially as the project got really big, like there started to be a lot of this kind of infighting and there are a lot of like, I wanted the thing to be a non-profit and some of them wanted to be a for-profit.

And then there started to be people who were just kind of totally unable to work with each other.

There were people that were kind of trying to get an advantage for themselves in a lot of different ways.

And this just about 6 months later led to this big governance crisis.

And then we kind of reshuffled leadership a bit.

Then 9 months later, there was another governance crisis.

And then there was a third governance crisis.

So is there a way to, if you're looking at the human side of things, is there a way to optimize this aspect of the cryptocurrency world?

It seems that there is, from my perspective, there's a lot of different characters and personalities and egos.

And like you said, I don't know if, you know, I also like to think that most of the world, most of the people in the world are well intentioned, But the way those intentions are realized may perhaps come off as negative.

Is there a hopeful message here about creating a governance structure for cryptocurrency where everyone gets along?

After about 4 rounds of reshuffling, I think we've actually come up with something that seems to be pretty stable and happy.

I think, I mean, I definitely do think that most people are well-intentioned.

I just think that 1 of the reasons why I like decentralization is just because there's this thing about power where power attracts people with egos.

And so that just allows a very small percentage of people to just ruin so many things.

Within the Ethereum research team, I feel like we've found also a lot of very good people that are just primarily interested in things for the technology and things seem to generally be going quite well.

The technology side, like what have you learned?

What have been the biggest challenges of bringing Ethereum to life on the technology side?

So I think first of all, just, you know, there's like the first law of software development, which is that when someone gives you a timetable, I'm going to switch the unit of time to the next largest unit of time and add 1.

And so instead of taking 3 months, it ended up taking 20 months to launch the thing.

And that was just, I think, underestimating the sheer technical complexity of the thing.

For example, 1 of the things that we've been saying from the start that we would do, 1 is a switch from a proof of work to a proof of stake, where proof of stake is this alternative consensus mechanism where instead of having to waste a lot of computing power on solving these mathematical puzzles that don't mean anything, you kind of prove that you have access to coins inside of the system.

And this kind of gives you some level of participation in the consensus.

Can you maybe elaborate on that a little bit?

I know that a lot of people say that the idea of proof of stake is really appealing.

So basically the idea is like, if I kind of lock up 100 coins, then I turn that into a kind of quote virtual miner.

And the system itself kind of automatically and randomly assigns that in a virtual miners are right to create blocks at particular intervals.

And then if someone else has 200 coins and they walk on the walk, there's 200 coins, then they get a kind of twice as big virtual miner, they'll be able to create blocks twice as often.

So it tries to do similar things to proof of work, except instead of the thing rate limiting your participation being your ability to crank out solutions to hash challenges, the thing that reallimits your participation is how much coins you're locking into this mechanism.

So that limited participation doesn't require you to run a lot of compute.

Does that mean that the richer you are, so rich people are more, like their identity's more stable,

verifiable, or whatever the right terminology is.

Right, and this is definitely a common critique.

I think my usual answer to this is that proof of work is even more of that kind of system.

I didn't mean it in that statement as a criticism.

The proof of work is the same kind of thing, but in the proof of work, you have to also use physical resources.

Yes, and burn computers and burn trees and all of that stuff.

Is there a way to mess with the system over the proof of stake?

There is, but you will once again need to have a very large portion of all the coins that are locked in the system to do anything bad.

Yeah, and just to that, maybe take a small tangent.

1 of the criticisms of cryptocurrencies is the fact that I guess for the proof-of-work mechanism you have to use so much energy in the world.

Is 1 of the motivations of proof-of-stake is to move away from this?

Like what's your sense of the, maybe I'm just under-informed.

Is there like legitimately environmental impact from this?

So the latest thing was that Bitcoin consumed as much energy as the country of Austria or something like that.

And then Ethereum is like right now, maybe only like half an order of magnitude smaller than Bitcoin.

the status of proof of stake as the mechanism that Ethereum moves towards?

you move to a different mechanism of consensus within a cryptocurrency?

So Ethereum 2.0 is a collection of major upgrades that we've wanted to do to Ethereum for quite some time.

The 2 big ones, 1 is a proof of stake and the other is what we call sharding.

Sharding solves another problem with blockchains, which is a scalability.

And what sharding does is it basically says, instead of every participant in the network having to personally download and verify every transaction, every participant in the network only downloads and verifies a small portion of transactions.

And then you kind of randomly distribute who gets how much work.

And because of how the distribution is random, it still has the property that you need a large portion of the entire network to corrupt what's going on inside of any shard, but the system is still in a very redundant and very secure.

How hard is that to implement and how hard is a proof of stake to implement?

Like on a technical level, software level.

Proof of stake and sharding are both challenging.

I'd say sharding is a bit more challenging.

The reason is that proof of stake is kind of just a change to the how the consensus layer works.

Sharding does both that, but it's also a change to the networking layer.

The reason is that sharding is kind of pointless if at the networking layer, you still do what you do today, which is you kind of gossip everything, which means that if someone publishes something, every other node in the client hears it from on the networking layer.

And so instead we have to have kind of sub networks and the ability to quickly switch between sub networks and other sub networks, talk to each other.

And this is all doable, but it's a more complex architecture.

And it's definitely the sort of thing that has not yet been done

So most most of the networking layer in cryptocurrency is you're shouting broadcast messages.

Smaller group, but do you have like a bunch of subnet?

And you have to switch between, oh man, I'd love to see the, so it's a beautiful idea.

So from a graph theoretic perspective, but just the software, like who's responsible?

Is the Ethereum project, like the people involved, would they be implementing, like what's the actual, you know, this is like legit software engineering.

How do people collaborate, build that kind of project?

Is this like almost like, is there a software engineering lead?

Is there, is it a legit, almost like large scale open source project.

So we have someone named Danny Ryan on our team who has just been brilliant and great all around.

And he is a kind of de facto kind of development coordinator, I guess.

It's like, you have to invent job titles for this stuff, right?

The reason is that, like, we also have this unique kind of organizational structure where the Ethereum Foundation itself kind of does research in-house, but then the actual implementation is done by independent teams that are separate companies and they're located all around the world and fun places like Australia.

And so you kind of just need a bunch of almost nonstop cat herding to just keep getting these people to talk to each other and kind of implement the spec, make sure that everyone agrees on what's going on and how to interpret different things.

So how far into the future are we from these 2 mechanisms in Ethereum 2.0?

Like what's your sense of the timeline, keeping in mind the previous comment you made about the sort of general curse of software projects?

So phase 0 just creates a proof of stake network.

And it's actually separate from the proof of work network at the beginning, just to give it time to grow and improve itself.

Do people get to choose, sorry to interrupt, do people get to choose, I guess, which consensus model?

get to choose to move over if they want to.

Then phase 1 adds sharding, but it only adds sharding of data storage and not sharding of computation.

And then after that, there is the merger phase, which is where the accounts, smart contracts, all of the activity on the existing ETH1 system just gets cut and pasted into ETH2.

And then the Proof-of-Work chain gets forgotten.

And then all the things that were living there before just kind of continue living inside of the proof of stake system.

So for timelines, phase 0 has been almost fully implemented.

And now it's just a matter of a whole bunch of security auditing and testing.

My own experience is that right now it feels like we're at about a phase comparable to when we were doing the original Ethereum launch when we were maybe about 4 months away from launch.

So how, you know, it took it took like over a decade for people to move from Python 2 to Python 3.

How do you see the move from like this phase 0 for different consensus mechanism?

Do you see there being a drastic phase shift in people just kind of jumping to this better mechanism.

So in phase 0, I don't expect too many people to do much because in phase 0 and phase 1, the new chain, they get it deliberately and if it doesn't have too much functionality turned on, it's there just like if you want to be a proof of stake validator, you can get things started.

If you want to store data for other blockchain applications, you can get started.

But existing applications will largely keep living on each 1.

And then when the merger happens, then the merger is a operation that happens all at once.

So that's 1 of the benefits of a consensus system that on the 1 hand, you have to coordinate the upgrade, but on the other hand, the upgrade can be coordinated.

Casper FFG is the consensus algorithm that we are using for the proof of stake.

Is there something interesting, specific about CasperFFG?

Like some beautiful aspect of it that's worth mentioning?

So CasperFFG combines together kind of 2 different schools of a consent algorithm design.

So the general 2 different schools of the design are right.

1 is a 50% fault tolerant, but dependent on network synchrony.

So 50% fault tolerant, but it didn't tolerate up to 50% of faults, but not more.

But it depends on an assumption that all of the nodes can talk to each other within some of a limited period of time.

Like if I send a message, you'll receive it within a few seconds.

And the second school is 33% fault-tolerant, but safe under asynchrony, which means that if we agree on something, then that thing is finalized.

And even if the network goes horribly wonky the second after that thing is finalized, there's no way to revert that thing.

That's fascinating how you would make that happen.

If you just search like archive, as in like ARXIV and Casper FFG, it's right there.

That's an archive, the paper's an archive.

This idea of just putting out white papers and papers and putting them on archive and just putting them publicly, is that at the core?

Is that a necessary component of cryptocurrency?

Is that the tradition started with Satoshi Nakamoto?

What do you make of the future of that kind of sharing of ideas?

And it's definitely something that's kind of mandatory for crypto because crypto is all about making systems where you don't have to trust the operators to trust that the thing works.

And so if anything behind our system works is closed sourced, then that kind of kills the point.

And so there is a sense in which the fundamental properties of the category of the thing we're trying to build just forces openness.

But also openness just has proven to be a really great way to collaborate.

And then there's actually a lot of innovation and academic collaboration that's just kind of happened ad hoc in the crypto space the last few years.

So like, for example, we have this forum called ETH Research, that's like E-T-H-R-E-S-E-A-R and then dot C-H.

And there we publish kind of just ideas in a form that's kind of half formal, like it's halfway in between.

It's a kind of a text write up and then you can have math in it, but It's often much shorter than a paper.

It turns out that the great majority of new ideas, they're just fairly small nuggets that you can explain in 5 to 10 lines.

They don't need the whole formality of a paper.

They don't require the kind of like 10 pages of filler.

And so introduction and conclusion is not needed.

And so instead you just kind of publish the idea and then people can go comments on it.

No, so just Casper FFG is just kind of combines together these 2 schools.

And so basically it creates this system where If you have more than 50% that are honest, and you have a network synchrony, then the thing kind of goes as a chain.

But then if network synchrony fails, then kind of the last few blocks in the chain might get replaced.

But anything that was finalized by this more asynchronous process gets, like, can't be reverted.

And so you essentially get a kind of best of both worlds between those 2 models.

I'm going to be reading the Casper FFG paper.

Apologize for the romanticized question, but what to you are some or the most beautiful idea in the world of Ethereum?

Just Something surprising, something beautiful, something powerful.

See all Lex Fridman transcripts on Youtube

Ethereum Basics (Vitalik Buterin) | AI Podcast Clips