ฝัง
- เผยแพร่เมื่อ 31 พ.ค. 2024
- 0:00 Intro
0:31 The byzantine generals problem
3:51 First solution
10:13 Importance
12:00 Blockchain-based solution
15:57 Wrap-up
Support us on Patreon: / polylog
We solve the Byzantine Generals problem, a fundamental problem of distributed computing. Then, in classic Polylog fashion, we spend the latter half of the video discussing why it matters in the broader context, specifically for distributed databases and blockchains.
Blog post: vasekrozhon.wordpress.com/202...
Code for the animations: github.com/polylog-cs/byzanti...
Filip Hlásek: Animator, script editor, beta reviewer
Richard Hladík: Script editor, beta reviewer
Václav Rozhoň: Writer, animator
Václav Volhejn: Narrator, animator, script editor
Thank you to our beta testers: Aashna, Coralie, Frido, Gabi, Jan (hi dad!), Julia, Martino
Animations: manim, a Python library docs.manim.community/en/stable/
Color palette: Solarized ethanschoonover.com/solarized/
Music: Thannoid by Blue Dot Sessions
Pictures: Wikipedia, DALL-E 3, Internet
![โอ้มายก๊อด! ไม่ถึงชั่วโมงหมดเป็นแสน 😅 [cc] แดนแพทตี้ SS2 | EP.42 |](http://i.ytimg.com/vi/OH2ShXfO6tA/mqdefault.jpg)
![[TH] VCT Masters Shanghai Playoffs DAY 1 // 100T vs FUT | FNC vs GEN](http://i.ytimg.com/vi/erQ7pGChzxA/mqdefault.jpg)
See the blog post for more related interesting stuff!
vasekrozhon.wordpress.com/2024/03/29/why-arguing-generals-matter-for-the-internet/
Byzantine video, byzantine haircut.
This was a great video.
It reacquainted me with concepts I hadn't needed to use for a while.
You explained it at a rate faster than I could recall it - it was like learning it for the first time.
Thank you for a great start to my day :)
I've watched the 3B1B video and it's great. It only talks about proof of work, though. Can anyone recommend an explanation of proof of stake? I've searched for one but all I can find is crypto bros talking about how supposedly cool it is, and the general idea, but not actually talking about how it works under the hood.
Can't wait for the next video. Thank you for your time and effort put in this video.
Love your explanations ❤ and visuals!! And sound design!! ❤
Fire video as always! Excited for the next one :)
Such such great content❤ Vis of algos are so so gooooooood🎉
The problem that makes all blockchain protocols fundamentally insecure is that there is no cryptographic proof of being the leader, so there can be disagreement on how the chain continues due to spoofing (note even when messages are signed a fake leader can send different messages for their own transactions, e.g. sending the same coin to different people). This chain split has happened multiple times in the past. It would be virtually impossible in a cryptographically secure protocol. But actual countermeasures rely on a computational advantage for the honest participants to succeed. A cryptographically secure system would require an astronomical computational advantage for the dishonest parties to succeed, not just a slight one.
Huh? Generally it's always a leaderless protocol so ofc theres no proof of whos the leader? The whole point is to make it decentralized.
Very cool video, but question about 13:40: can't quantum computing theoretically break modern encryption like quantum computing, or is that far off enough from feasibility to not be of concern here?
Software is moving away from quantum breakable algorithms (like RSA). Signal (secure messaging platform) for example started using both a traditional and quantum proof asymmetric encryption algorithms. And even if we don't invent quantum computers for another 100 years, organizations already store high value encrypted messages to decrypt once we have the technology. This is called store-now-decrypt-later.
THIS WAS BRILLIANT!!!
DUDE CONGRATS ON THE ACX THING
Pure gold!
13:20 Felt cute might delete later 😂
amazing
My initial solution is very simple, but it meets the criteria for any number of traitors.
Each general asks all other generals for their opinion. Then, if all generals respond "yes" AND the asking general wants "yes", then agree to "yes". Otherwise, agree to "no".
I suspect there is a better solution that allows the honest generals to agree to the majority, assuming the majority is greater than the sum of the minority plus traitors. I am currently too tired to devise such a solution.
As far as I can tell in this protocol, the generals will always decide "no" unless everybody had the initial opinion of "yes". Under this protocol, traitors can force the answer to always be no if they want (although that wouldn't mean a failure of the protocol).
So this solution seems very similar practically to the trivial solution that was ruled out despite satisfying the new rule that was introduced to eliminate the trivial solution.
Awesome vid >:3
Wee making it out of bullying with dis one fam 🗣️🗣️🗣️🗣️🗣️🗣️🔥🔥🔥🔥🔥🔥🔥🔉🔉🔉🔉🔉
you still have to worry about situation where there is unknown numbers if tractors?
I love this explanation. Thank you so much.
nice video
As someone who works with bft algorithms on a daily basis this is refreshing clear and we need more of it.
Suggestions: Can talk about asynchronous bft algos (eg. HoneyBadger or Dumbo) or DAG based algs (eg. Narwhal and Tusk).
Got distracted from your cuteness 😅
why do you need 10 votes to choose the local opinion? If the difference between yes and no is greater than twice the number of traitors, they can't overturn the majority so 9 votes are enough.
In general, if you have n voters and x traitors, you must trust n - x votes and you can trust (n + 1)/2 + x votes. Setting both equal yields x = (n - 1)/4 as the max number of traitors this system can deal with, in which case you would have to trust (3n - 1)/4 votes. So I guess that's the ideal choice as it survives the most traitors. For n=12 it yields 8¾, so 9 (if you have ≥8¾ votes, you have ≥9 votes, and otherwise ≤8).
great video, but i can't lie that hair goes kinda crazy 😅
idk why this flopped so hard so far
ahoj ríšo 👋
AMONG US
Can’t wait for quantum computing to make RSA trivial to break and the consequences it has for the world.
bro you look like this guy from despicable me
Wait, is RSA still a thing? 😮
Can we please stop the Blockchain bs? There is an older solution that is much more flexible and more importantly *proven* secure. Read up universally composable security or secure multiparty computation. Both approaches allow you to create a virtual trustworthy central authority from a system of unreliable interconnected parties.
How is creating a central authority a solution to wanting a *decentralised* network based on consensus? The entire *point* of blockchains (and block lattices or whatever else is new) was to get away from centralised leadership.
Honestly, at this points it almost feels irresponsible to talk about blockchain and crypto from a purely technical point of view without discussing the absurd amount harm they cause...
I actually think it's wonderful to see a video focusing only on the *core mathematics* behind the blockchain algorithm, without actually talking about cryptocurrency as the only application for blockchains. I think in particular they did a good job of connecting directly to the idea of creating consensus without the cryptocurrency-specific details like block rewards etc. The many real life issues with crypto can go in another video imo!
What harm do they cause alone? The only harm I can recall always comes down to bad actors that intend to do the harm to begin with.
agreed - unfortunately blockchains are a politicised data structure, and talking about them neutrally legitimises them. the proof that they are unnecessary is that despite being around for decades now, blockchains have not been widely adopted at all for any usecase other than crypto and NFTs; and yet even with their niche usage, the energy and hardware consumption to physically power blockchains is staggering
this is like saying to talk about internet is irresponsible because of the harm is causing. you are also confusing crypto with cryptocurrencies which are very different.
@@smaza2the internet has been around for even more time without being “useful” to most people. if you take a look at the academic work that came out in 2023 you’d be surprised (eg. HotStuff 2, Shoal, Fin). I think it’s irresponsible to politicise the tech as you did without even understanding where it stands and where it’s going.
who tf unironically has a bowl haircut
polylog
Someone developed poor consensus on how to cut your hair. Sorry for the roast. But you might want to adjust where you go, or how you do it.
you look like toxi$