Ethereum is a good example. It is facing high levels of congestion, so moved a lot of activity off chain onto “layer 2” protocols. But these are fairly concentrated, eliminating some of the perceived advantages of Ethereum and introducing new counter-party risks.
The blockchain hype is mostly just hype. The same centralizing incentives operate there as in the broader internet. The most important blockchain by far is Bitcoin, and Bitcoin mining (i.e. ultimate decision-making power) is extremely concentrated. This has been the case back in 2016, when https://urbit.org/blog/the-dao-as-a-lesson-in-decentralized-governance, an essential post on the topic of decentralized computing, was written, and this situation has persisted ever since. However, Bitcoin is still alive and functioning as intended despite the lack of decentralization.
A few reactions to Arnold's thought-provoking Theorem:
1. The impossibility theorem sounds like a problem for market economics to solve wherever it is allowed to do so. Different firms offer various solutions to the decentralization-vs.-computing-efficiency tradeoff. Internet users and consumers decide for themselves which solutions serve their needs most effectively, all things considered (e.g. price and quality). The most profitable solutions will win out, or at least I hope so.
2. Two Nobel Prize winners - Ronald Coase and Oliver Williamson - thought carefully about how different organizational structures would resolve the dilemmas between transactions costs and organizational complexity in general, if not specifically with respect to the internet. (The internet did not exist when they wrote their hypotheses.)
We have a terminology problem. A decentralized society is barbaric, almost by definition (and a centralized one is tyrannical, likewise). A distributed society is the task for culture, and engineering solutions should seek to support it for mutual long term success. Distribution involves cooperation (decentralization atomization). Blockchain is more properly considered as distributed, just like cash, not decentralized. This is both the point, and the source of a lot of confusion. Distributed architectures, especially robust ones, have cooperation mechanisms that (de)centralized ones fundamentally don't. That the internet has evolved into a hybrid distributed architecture with some worrying aspects is no argument against agent-based blockchain technologies. The issue is cost trends in critical functionality and scalability. An internet/blockchain comparison is apples and oranges since the constraints of the former are primarily in the hardware while the latter is in the software application itself.
When we start looking at particular verticals in/to which blockchains can be, and are being, applied, like gaming and virtual worlds, what's clear is that, for anything like the kind of imagined future we're seeing from founders and their investors, the increase in computing and energy is orders of magnitude greater than what's currently there. For example, the ramblings about "the metaverse" being a persistent, real-time virtual world in which individuals at reputed Meta level scale, requires an incredible amount of compute. We currently can support a couple hundred concurrent players. The point here is: where's the compute coming from? a third of the internet runs on AWS. I don't know the exact number but, from what I've read, nearly 25-30% of Ethereum runs on AWS. I don't think everything part of the stack will be decentralized; seems unlikely that energy, compute and hardware, OS-layer. It's obvious to me that the future is a hybrid of centralized and decentralized.
My sense is the real hope for this technology, maybe misplaced, is that sufficiently powerful engineering can unstick the cultural barriers. A less sticky internet, where it's easier to automate cooperation (smart contracts), to incentivize distributed ownership and processing (blockchain and associated currencies), and to maintain transparency (blockchain ledgers), could lower the cost of creating alternative systems, and raise the cost of pursuing rents, monopolies, info asymmetries, etc. Maybe a change in degree could yield a change in kind.
Ethereum is a good example. It is facing high levels of congestion, so moved a lot of activity off chain onto “layer 2” protocols. But these are fairly concentrated, eliminating some of the perceived advantages of Ethereum and introducing new counter-party risks.
The blockchain hype is mostly just hype. The same centralizing incentives operate there as in the broader internet. The most important blockchain by far is Bitcoin, and Bitcoin mining (i.e. ultimate decision-making power) is extremely concentrated. This has been the case back in 2016, when https://urbit.org/blog/the-dao-as-a-lesson-in-decentralized-governance, an essential post on the topic of decentralized computing, was written, and this situation has persisted ever since. However, Bitcoin is still alive and functioning as intended despite the lack of decentralization.
Excellent link. Thanks!
De-centralization isn't really possible in social systems either. See the Iron Law of Oligarchy.
I'm skeptical that culture can do what 'engineering' can't – engineering is just another form of culture after all.
A few reactions to Arnold's thought-provoking Theorem:
1. The impossibility theorem sounds like a problem for market economics to solve wherever it is allowed to do so. Different firms offer various solutions to the decentralization-vs.-computing-efficiency tradeoff. Internet users and consumers decide for themselves which solutions serve their needs most effectively, all things considered (e.g. price and quality). The most profitable solutions will win out, or at least I hope so.
2. Two Nobel Prize winners - Ronald Coase and Oliver Williamson - thought carefully about how different organizational structures would resolve the dilemmas between transactions costs and organizational complexity in general, if not specifically with respect to the internet. (The internet did not exist when they wrote their hypotheses.)
We have a terminology problem. A decentralized society is barbaric, almost by definition (and a centralized one is tyrannical, likewise). A distributed society is the task for culture, and engineering solutions should seek to support it for mutual long term success. Distribution involves cooperation (decentralization atomization). Blockchain is more properly considered as distributed, just like cash, not decentralized. This is both the point, and the source of a lot of confusion. Distributed architectures, especially robust ones, have cooperation mechanisms that (de)centralized ones fundamentally don't. That the internet has evolved into a hybrid distributed architecture with some worrying aspects is no argument against agent-based blockchain technologies. The issue is cost trends in critical functionality and scalability. An internet/blockchain comparison is apples and oranges since the constraints of the former are primarily in the hardware while the latter is in the software application itself.
When we start looking at particular verticals in/to which blockchains can be, and are being, applied, like gaming and virtual worlds, what's clear is that, for anything like the kind of imagined future we're seeing from founders and their investors, the increase in computing and energy is orders of magnitude greater than what's currently there. For example, the ramblings about "the metaverse" being a persistent, real-time virtual world in which individuals at reputed Meta level scale, requires an incredible amount of compute. We currently can support a couple hundred concurrent players. The point here is: where's the compute coming from? a third of the internet runs on AWS. I don't know the exact number but, from what I've read, nearly 25-30% of Ethereum runs on AWS. I don't think everything part of the stack will be decentralized; seems unlikely that energy, compute and hardware, OS-layer. It's obvious to me that the future is a hybrid of centralized and decentralized.
This was very helpful. Thanks.
My sense is the real hope for this technology, maybe misplaced, is that sufficiently powerful engineering can unstick the cultural barriers. A less sticky internet, where it's easier to automate cooperation (smart contracts), to incentivize distributed ownership and processing (blockchain and associated currencies), and to maintain transparency (blockchain ledgers), could lower the cost of creating alternative systems, and raise the cost of pursuing rents, monopolies, info asymmetries, etc. Maybe a change in degree could yield a change in kind.