> I also think it's telling that even though Blockchain has been this hyped thing for 6+ years at this point, we haven't really seen it actually be used for anything outside of cryptocurrency, which in and of itself isn't used for much outside of speculation. On the other hand, machine learning is used in everything now and makes a lot of stuff better.
Did you know that the Neural Network has been around since 1958 [1]? Machine learning is not a technology that is just 6 years old. The latest AI trend is also not the first or second time that AI has been through a massive hype cycle.
The problem with the cryptocurrency space is that it's financial innovation. And just like financial innovation on Wall Street, this tends to draw out the slimiest people in society, because if you get someone to believe in your product they may well leverage their mortgage and throw their life savings at you. It's crushing to see people do this, especially because pretty much only the malicious projects get hyped up that much.
But that doesn't mean that there isn't any truly groundbreaking innovation out there. Cryptocurrency changes the fundamental scalability of society. A key bottleneck for human society is trust - at some point a system gets large and corrupt, and it becomes difficult to keep bad actors from imparting a large amount of negative influence. But cryptocurrency allows us to design systems that don't require any trust at all. They _cant'_ be corrupted, because a combination of incentives and cryptography keep everyone safe.
As this blog post shows, there are still a lot of rough edges out there, but the technology is innovating rapidly. I do think the hype is probably 5-10 years ahead of the technology, but in the grand scheme of technology (think of how long it too Arpanet to mature, or Neural Networks to mature) that is not much time at all!
which discusses economic limitations to the incentives for blockchain mining. (One part is that if a cryptocurrency gets too valuable, the value of a double-spend attack could exceed some models of the reward for honest mining. Another part is that if you have unregulated derivatives, you can own a negative amount of cryptocurrency, which means that your financial incentive can be to actively damage the cryptocurrency rather than helping it. Satoshi's paper seems to assume that you can only own a positive or zero amount of bitcoin rather than a negative amount, when arguing why miners are incentivized to be honest.)
(This is also true for the ability to short, or insure, any asset -- you can be financially incentivized to damage it -- but elsewhere this incentive is partly countered by law enforcement investigations of some trades and insurance claims where people profited significantly from accidents, disasters, or scandals. Smart contracts on blockchains let us build insurance and derivatives markets where you can bet against things without identifying yourself. In fact the whole underlying discussion here is about how the person who claimed this particular asset in Ethereum is anonymous and probably can't be punished for doing so, even if we believed that the claimant wasn't entitled to make this claim. That could be equally true if the person were collecting an insurance contract payout. That's potentially fine if contracts can't create new incentives to cause harm, but maybe not so awesome if they can.)
The paper is 2y old - which is a bit of time in blockchain space - and yet, no one has performed this attack successfully in the wild. So, either everyone is just loving Bitcoin and wouldn't harm it for profit, or it's not so easy (hint: you cannot short Bitcoin for hundreds of millions, it's not liquid enough).
Also, this is strictly about PoW, the ETH roadmap (and this is where we are coming from in this submission) is moving to PoS where attacks are potentially way more expensive.
> you cannot short Bitcoin for hundreds of millions, it's not liquid enough
That seems like a great explanation, but the paper's argument is definitely not that cryptocurrencies can't exist or can't work, just that they have a limited range of levels of adoption where the incentives will continue to point in the right direction. In the paper's model it seems that Bitcoin has just not reached that level, right? If you could double-spend or short enough value in it, the incentives would reverse.
> PoS where attacks are potentially way more expensive
I don't think Eric Budish agrees that PoS verification is categorically immune to this. The very last sentence of the paper mentions that it "will be interesting to watch [PoS] research develop, and see whether or not it constitutes a valid response to the critique in this paper".
This is a reasonable response - I definitely think blockchain is interesting but like I said, it's in need of some major additional refinement before it becomes practical.
>But cryptocurrency allows us to design systems that don't require any trust at all. They _cant'_ be corrupted, because a combination of incentives and cryptography keep everyone safe.
This article definitely doesn't describe a system that lives up to that ideal at all. Which is why it's so scary - when you remove manual oversight you're essentially saying "Hey, if you can hack this, you win!"
Did you know that the Neural Network has been around since 1958 [1]? Machine learning is not a technology that is just 6 years old. The latest AI trend is also not the first or second time that AI has been through a massive hype cycle.
The problem with the cryptocurrency space is that it's financial innovation. And just like financial innovation on Wall Street, this tends to draw out the slimiest people in society, because if you get someone to believe in your product they may well leverage their mortgage and throw their life savings at you. It's crushing to see people do this, especially because pretty much only the malicious projects get hyped up that much.
But that doesn't mean that there isn't any truly groundbreaking innovation out there. Cryptocurrency changes the fundamental scalability of society. A key bottleneck for human society is trust - at some point a system gets large and corrupt, and it becomes difficult to keep bad actors from imparting a large amount of negative influence. But cryptocurrency allows us to design systems that don't require any trust at all. They _cant'_ be corrupted, because a combination of incentives and cryptography keep everyone safe.
As this blog post shows, there are still a lot of rough edges out there, but the technology is innovating rapidly. I do think the hype is probably 5-10 years ahead of the technology, but in the grand scheme of technology (think of how long it too Arpanet to mature, or Neural Networks to mature) that is not much time at all!
[1]: https://www.computerworld.com/article/2591759/artificial-neu...