Summary:
- Information systems today are largely centralised – with Amazon, Facebook and nation state governments as examples.
- Centralised systems are more efficient than decentralised systems, although they are less robust and more vulnerable to security attacks.
- Mobile phones are an example of how technology has decentralised computing power.
- Ethereum is a technology that is attempting to decentralise information.
- Ethereum is highly inefficient today. The question is whether technology can take it to a point where efficiency is high enough to be irrelevant.
Mobile Phones
When computers first emerged, it wasn’t clear we would all, someday, have one in our pocket (and soon attached to our brain?).
The reason we all have mobile phones is not because mobile phones have become just as powerful as larger computers – it is still more cost effective, on a per calculation per second basis, to build a large computer. The reason is simply that mobile phone computers are fast enough that computing power is now just one among many attributes to consider in choosing a phone.
Restated another way: In the early days, only big computers made sense. Over time, technology improved so much that it is worth us each having a personal computer, even if operating a larger computer would be more efficient. Technology has allowed us to decentralise computing power
Technology may allow us to (re-) decentralise production of food, clothing and medicine
Extending this framework, consider now the farming of vegetables, the raising of livestock, the production of vaccines and the manufacturing of clothes. Today, all of these seem best outsourced to large centralised companies. However, we should ask the question of whether technology can get to a point where these become processes that are done in our homes. To be clear, I don’t mean a reversion to classical small-scale farming), but rather to high-tech versions such as home chemistry/pharma kits and home cultivated “lab” meat.
COVID driving robustness in systems instead of efficiency
With COVID we realised that many of our supply chains, particularly around urgent care, are concentrated in certain parts of the world. Whereas this has allowed efficiency, it has not allowed for robustness when travel stops. COVID is therefore a driver of decentralising processes, e.g. building vaccine production in each country (and maybe some day, in each home).
2019 was a narrative of efficiency through specialisation through centralisation. 2020 brought more of a narrative of generalisation through decentralisation.
Cyber Hacking as a Driver of Decentralisation
We have lately heard of cyber attacks on pipelines (US) and healthcare systems (Ireland). Systems containing sensitive information are always vulnerable. If information is valuable but not sensitive, then you can make backup copies. If information is valuable and sensitive, the best you can do is to defend against attacks. The only alternative is not to collect and store that information centrally in the first place, but this hurts efficiency!
For a company operating in healthcare, it is currently highly inefficient not to have the patient’s information on hand. The same goes for Facebook and Google – neither company would be valuable if they were to give up on their centralised storage of information. Today, it is simply inefficient for a government or company not to store lots of information centrally.
But what if it were possible to efficiently store and access information in a decentralised way?
We think decentralisation. is too inefficient to be possible
Today, we don’t believe that information can be decentralised because we think this kind of system is too inefficient. We believe monetary and payment and government systems will always have an advantage to systems that are decentralised.
These statements are probably correct, centralised systems will always be more efficient. However, what is missing here is that decentralised systems may get to a point where their relative lack of efficiency becomes unimportant, allowing other attributes like robustness and security to prevail.
Large computers are still better than small computers on the basis of calculation efficiency. However, small computers are now good enough that this doesn’t matter. Likewise, decentralised blockchains right now are inferior to centralised systems. The question is, with technology, will they improve to a point where their disadvantage no longer matters.
Ethereum as a Test Bed for Decentralisation
Far beneath the hype and price swings in crypto there is an effort to answer the deep question of whether technology can make possible the decentralisation of information. Ethereum is one embodiment of this effort.
Roughly speaking, there two challenges that Ethereum faces with respect to decentralisation – 1. the energy problem, and 2. the speed problem.
- The transaction validation problem (aka the energy problem)
Contrary to popular news, the problem of energy consumption in Ethereum mining (and Bitcoin) is the less interesting one. At its core, energy consumption is an answer to the question of who gets to say which transactions on Ethereum are valid and which are not (since there is no central bank or private company to make that call). In very rough language, Bitcoin first solved the transaction validation question by using a maths problem to dictate who deems a batch of transactions to be valid. This is called Proof of Work and is the same mechanism that is used on the Ethereum blockchain as of May 2021. Solving random maths problems takes computing energy, and this is why cryptos today are energy intensive.
Since then, another approach to transaction validation has emerged called Proof of Stake. With Proof of Stake, the validation of each batch of transactions is randomly assigned someone holding the Ethereum currency (called Ether). This approach – and perhaps there will be even better approaches still – does not require solving random maths problems, and so the energy requirements are a tiny fraction of the main blockchains today.
Though not yet implemented on Ethereum (it has been implemented on other blockchains like Celo.org), there are known solutions to the energy problem.
2. The decentralisation problem with Ethereum (aka the speed problem)
The more interesting problem in blockchain is how to keep the system both decentralised and quick to approve transactions.
A simple way to think about whether the system is decentralised is whether it is practical for owners of standard computers to participate in validating batches of transactions. Bitcoin took a solid step in this direction, but you still need to have specialised computers in order to help validate Bitcoin transactions. Ethereum 2.0 – coming soon – will allow normal computers to be used to support transaction validation. Indeed there are over 150,000 computers already set up and ready to host the new version of Ethereum when it emerges over the next months.
So, the core challenge for Ethereum is that the software has to be able to run on a standard computer. This means that the whole network is somewhat limited by what a personal computer can do. If you think this sounds a bit stupid then you are right – it is like trying to do mobile phones in 1970. Slow.
Right now, as of May 2021, so many people are using the Ethereum network, that the cost of doing transactions can be over $100, which is insane. Ethereum simply cannot keep up.
This brings us to two broad solutions to the speed problem:
i. Use bigger, less centralised, computers
If you use fancier computers (like big Amazon datacenters), you can get great speed. By just having a few computers do validation – rather than hundreds of thousands – information can propagate much faster. However, this is at the cost of decentralisation (people can no longer support the network with their personal laptops).
Networks like Binance (with heavy centralisation) and Celo.org (with more centralisation than Ethereum but much less than Binance), have taken this approach to solving the speed problem. They require more specialised hardware to support transaction validation, so the networks are less centralised but they are faster.
ii. Use better technology (software)
Ethereum is taking a different course and sticking to the goal of being supported by personal computers. To do this, Ethereum is relying on mathematical tricks allowing more information to be packed into the same amount of computer storage. There are various solutions here and you will find words such as “zksnarks”, “optimism” and “sharding”, all of which are ways to pack more information into less storage via mathematical tricks. As I see it, these mathematical tricks are real knowledge progress, in the same way that Einsteins General Relativity is real knowledge progress – and I say that with only some exaggeration.
Decentralisation beyond Efficiency
I’m reluctant to say that technology necessarily reduces centralisation. True, we now all have mobile phones – so computing power is more decentralised – but much of the information is stored centrally with Amazon and Google. More accurately, technology changes what is centralised. At first, technology centralises things (like garment making on factory looms) or the use of supercomputers in the 1970s, but maybe later technology improves so much that “everyone can have one” type decentralisation occurs.
In 2021, the centralisation of information is driven by a need for efficiency in information exchange. Decentralised systems for information are currently inefficient, just as small computers were in the 1970s. That may change. Just as computers went from factories to phones, technology may take us to a point where factors other than efficiency – such as security and robustness – result in the decentralisation of information. Ethereum is an early test case in this.
Footnotes: A Summary of What Drives Centralisation
Summarising, in no particular order, we have desires for:
- Efficiency – driving centralisation. (e.g. storing customer information on one database)
- Robustness – driving decentralisation. (e.g. every country having vaccine production facilities)
- Security – driving decentralisation (e.g. customer information being held only on their computers, not a central company or government network)
- Technology – first driving centralisation (via economies of scale or network effects) but in some cases (such as computation) reaching an inflection point driving decentralisation again (e.g. mobile phones).
Reading:
- A great piece from Vitalik Buterin on the the tradeoffs between efficiency and decentralisation on blockchains.