Back to page

📄 posts/
#lang pollen

◊(define-meta title "The Purpose of a Decentralized Internet")
◊(define-meta published "2021-07-12")
◊(define-meta topics "Society, Tech, Themelio")

◊section{Why decentralize?}

We often hear that Internet technology is far too centralized. Many crucial everyday services are dominated by one or a few massive providers (Google, Facebook, Amazon, etc.) that don't necessarily merit our trust. In response, interest in alternative decentralized technologies has greatly increased. In fact, this perception that the current centralized Internet ecosystem is "broken" is key to the astonishing rise of blockchain-based technologies, like cryptocurrencies and decentralized finance (DeFi).
Many even hope for technological advances to bring about a new "Web3" of fully decentralized applications, to ◊link[""]{"make the world work better for all"}, to ◊link[""]{"fundamentally change what we can do online, how we do it, and who can participate"}...

But why do we want to decentralize the Internet? What exactly is wrong with technological "centralization"? Commonly cited faults of centralized systems include lack of privacy, poor security, and unreliability. Decentralization, in turn, is often marketed as the solution to this host of issues plaguing centralized systems. However, the only convincing arguments for decentralization all boil down to one thing: resisting pathological social centralization.

Let's take teleconferencing as a case study. Here is a pattern often criticized as dangerously centralized: most people use one of a few major teleconferencing platforms, like Zoom or Google Meet. Why exactly is this bad?

Are these centralized platforms unreliable? It is true that they might one day go out of business and we'd have to reschedule all our meetings, but these companies want to make money, and that requires providing extremely good reliability. And even if you're particularly paranoid of a centralized service going down during some critical event --- maybe you're navigating a rocket over Zoom --- I'm pretty sure a custom insurance policy to cover that would be cheaper than setting up a high-reliability decentralized teleconferencing system. Unreliability, then, is probably not the source of criticism.

The problem isn't even privacy, ◊em{per se}. Most people really do have virtually "nothing to hide"; for the remainder, having their darkest secrets locked away in a datacenter, jealously guarded from competitors, and used to target Google Ads isn't exactly a catastrophe.

Instead, the really scary scenario here is rather obviously ◊em{state-sponsored mass surveillance}. We find such a teleconferencing arrangement disquietingly "centralized" precisely because it gives a disproportionate amount of power to the sovereign state, the very pinnacle of social centralization.

This investigation is not restricted to teleconferencing platforms: apply it to every other Internet centralization nightmare, and you'll find the same social centralization monster at the bottom. Popular centralized services on the free market are more often good at what they do, but the risk of abuse is inherent in their centralized infrastructure. What makes this risk a reality is social centralization.

When we decry the terrible centralization of the CA-based PKI of the encrypted Web, we should remember that its worst failures (e.g. hacked certificates used for mass surveillance in Iran or untrustworthy CAs run by government-linked Chinese entities.) are fueled by social centralization. When we blame the social media companies for the excesses of so-called surveillance capitalism, we must realize that such practices are impossible without our highly centralized social order. After all, Facebook's scandalous affair with Cambridge Analytica only becomes a nightmare in a world so centralized that a single ad campaign affects the rulers of hundreds of millions.
◊; Iran: diginotar incident. Chinese: ICANN, WoSign

The interest in decentralized technology, then, really reflects the problem of extreme social centralization in our world. We want a trustless PKI not because Verisign & co suck at running their business, but because we want to prevent mass surveillance from demolishing Internet security. We want decentralized cryptocurrencies not because banks charge too much fees (they really don't), but because we need to defend against the social manipulation and legalized expropriation of centralized fiat currency. The point of decentralized technology is to provide tools to counteract the pervasive social centralization of our modern world. 

◊section{Decentralized technology is incompatible with a centralized society}

In fact, it doesn't make sense to develop decentralized Internet technology without aiming to radically reject social centralization. Building secure, endogenously trustworthy decentralized systems often requires great sacrifices in areas like convenience, performance, and cost. Only great payoffs in social decentralization can justify these downsides.
◊; link to endogenous trust
Blockchains, for instance, have notoriously low transaction throughput --- Bitcoin processes at most around 7 transactions a second compared to Visa's 1700 or so. Decentralized anonymizing networks like Tor or I2P often add crippling latency to basic web browsing compared to ◊link[""]{centralized VPN services}. 
◊; There really is no reason for your average, centralization-embracing Joe to listen to the neighborhood cypherpunk, give up his high-performance, well-polished centralized toys, and switch to Monero and Matrix.
In fact, if a person didn't care about centralization and only wanted performance, why would they ever choose the decentralized alternatives? 
◊; link to Geph 

All these performance sacrifices are pointless if our central social paradigm can be trusted. If we believe in our centralized society and government, we could always work within the system and produce better centralized solutions to fix whatever's wrong with the status quo. For instance, if governments are abusing fiat currency, we could just advocate for monetary reform --- maybe even introduce radical ideas like ◊link[""]{NGDP targeting} or a gold standard. If social media companies are violating user privacy or censoring content, we could switch to "ethical" companies like MeWe, sue the abusive companies, or even pass ◊link[""]{GDPR}-style privacy laws and enforce them with the full power of our centralized sociopolitical apparatus. Decentralization in itself targets the problem of centralization, and it comes at a cost. There is no point in decentralizing technology without aiming at decentralizing society.

So what about the multitude of people who promote wide adoption of decentralized technology while mostly accepting the current social paradigm and supporting the likes of antitrust legislation as means to curb Internet centralization? To these people, technological decentralization is one tool out of many that can help fix our current problems. They hardly see any reason to altogether give up on our entire social paradigm.◊sidenote["BR"]{Even the dazzlingly optimistic visions of a decentralized Internet offered by the likes of ◊link[""]{◊em{Blockchain Revolution}} don't seem to require dismantling sovereign states to realize!}

◊figure["cardano-screenshot.jpg"]{A ◊link[""]{quintessential example} of "decentralized technology as a tool for gradualist social reform"}

At first glance, this seems like a reasonable and moderate position. A closer look, however, reveals that almost always, proposals for gradualist fixes to Internet centralization using decentralized technology without fundamental social changes are not what they seem. Either they misapply decentralized tech in situations where conventional, centralized technology would work better, or they---perhaps unintentionally---hide radical implications of their proposals that shake the very foundations of our centralized society.

A rather innocuous and politically correct example is blockchains for political elections. Using blockchains in elections seems to be a no-brainer: blockchains excel at providing strong guarantees of correctness and transparency, both of which are essential in elections. In fact, many major players, including governments like the ◊link[""]{Swiss canton of Zug} and the ◊link[""]{state of Alaska}, are considering using blockchains to implement voting systems.

Let's leave aside the whole debate on the merits of blockchain voting in general◊sidenote["BV"]{Vitalik Buterin has a ◊link[""]{good blogpost} for the "pro" side, while some people at MIT wrote a ◊link[""]{whole paper} arguing extensively for the "anti" side.} and focus one a critical, yet often neglected, design decision in any blockchain-for-elections scheme. That is, is the blockchain used as a tool by existing election authorities to improve election transparency and security, or is the blockchain trusted as a ◊em{fully independent, authoritative source of truth} for key facts like "who won the election"?

Unsurprisingly, real-world proposals to use blockchains in elections generally employ the former approach. They use blockchains as nothing more than an implementation strategy for secure electronic voting. Often, this involves so-called "private" or "consortium" blockchains run by traditional election authorities that bear only a passing resemblance to trustless public blockchains like Ethereum or Bitcoin. These systems want the desirable properties of blockchains without the radical new model of social trust, so they leave out anything providing endogenous trust◊sidenote["ET"]{Also known as "autonomy", as in "autonomous software". I discuss it at some length in my ◊link["/posts/against-blockchain-governance.html"]{previous blogpost on blockchain governance}.}. Without endogenous trust, however, there's no longer anything special about the correctness and transparency guarantees given by a blockchain.

And so these systems incur the costs of using a blockchain without the benefits. On top of this, a large body of research already exists on the design of secure electronic voting systems with conventional, centralized trust assumptions. More often than not, these non-blockchain systems achieve goals out of the reach of most blockchain-based systems touted recently, including very good coercion resistance, anonymity, and performance. Even relatively radical institutional reforms are possible by, say, running the voting system on computer systems run by outside observers. Clearly, there is absolutely no need to drag in an over-engineered log-structured replicated database◊sidenote["HF"]{Like Hyperledger Fabric} that calls itself a blockchain.◊sidenote["PC"]{There's a very strong argument that ◊em{all} private/consortium blockchains are useless: all the supposed transparency, immutability, etc of blockchains goes out of the window when a trusted party --- or whoever hacks the trusted party --- gets to bend and break all the protocol rules. I can't think of anything that would benefit from using a private blockchain over a conventional replicated database, perhaps enhanced by a few blockchain-inspired changes like Byzantine fault-tolerant consensus.}

What about the second alternative? Vitalik Buterin, for example, implicitly assumes in his excellent ◊link[""]{blogpost on blockchain voting} that "blockchain voting" will happen on public blockchains like Ethereum:

◊blockquote{◊p{Public blockchains, on the other hand, have permissionless economic consensus mechanisms (proof of work or proof of stake) that anyone can participate in, and they have an existing diverse and highly incentivized infrastructure of block explorers, exchanges and other watching nodes to constantly verify in real time that nothing bad is going on.

A blockchain-based voting system where a trustless, public blockchain serves as the authoritative source of truth would indeed be the qualitative revolution in transparency, correctness, etc. that private-blockchain-based systems cannot be. Yet adoption of such a system necessarily comes with enormous implications. First of all, it implies that ◊em{conventional authorities are fundamentally untrustworthy}. 
In our present, centralized social order, we typically already trust political checks-and-balances like judicial review and competitive elections to "constantly verify that nothing bad is going on". Introducing a public blockchain for this task implies that not only our existing authorities suck at their job, there's nothing in the conventional, state-based political process we can use to fix the problem. It's extremely radical to imply that instead of trusting the integrity of, say, a multipartisan Election Committee appointed by the nation's best statesmen, we'd rather trust a faceless autonomous program living on a network of computers, much of which sit on the territory of enemy states. 

Even more consequentially, such a blockchain-based electoral system gravely wounds one of the founding myths of the sovereign state --- popular sovereignty. That is, the idea that the overpowering authority of the centralized state is justified because, in the famous words of Abraham Lincoln, it is "of the people, by the people, for the people"◊sidenote["NK"]{In fact, this is one of the few principles that nearly every modern government pays lip service to. Even blatantly tyrannical governments like that of the "Democratic People's Republic of Korea" claim that they represent and rule for the benefit of "the people".}. 
A fully blockchain-backed voting system would mean that ultimately, all legal authority and legitimacy instead derives from the result of some distributed on-chain computation.
Of course, this computation will almost certainly be some kind of weighted majoritarianism reminiscent of existing democratic voting schemes, but as is well known◊sidenote["ST"]{In an often misquoted passage from ◊em{The Memoirs of Stalin's Former Secretary}, even Stalin says "I consider it completely unimportant who in the party will vote, or how; but what is extraordinarily important is this—who will count the votes, and how."}, it doesn't really matter who votes, but how and by whom the votes are counted. 
◊; though the sheer flexibility of code facilitates very different variations such as quadratic voting or futarchy???

At least with a conventional voting system, enough civics classes and national founding myths can create a widespread perception that the meatspace voting system defined by the constitution concurs with the will of the people. It's much harder to do so with a rigidly formalized algorithm run by machines that are by definition outside any organized human control. It's no wonder that some people think blockchain voting ◊link[""]{gravely endangers Democracy™}. 

Similar analyses can be made for other efforts to make the world a better place through decentralized technology --- either they're better off not decentralizing, or their vision of a "better place" involves radical social decentralization. The very purpose of decentralized technology is to decentralize society.

◊section{Where to from here?}

However, knowing that decentralized technology is inherently against our current, hyper-centralized social order is not enough --- we should know what it's for.
What does a decentralized social order worth aiming for even look like? 

In an upcoming blogpost, I will discuss a model for a practical decentralized sociopolitical order that we can start building today in cyberspace and beyond. It will not be some anarchist thinker's utopian dream or an imaginary ◊link[""]{"ancapistan" with McNukes}. Instead, it originates from a seemingly unlikely source: the polycentric political order of Western Europe's High Middle Ages.