| layout | title | permalink |
|---|---|---|
page |
Can Technology Decentralize Power? |
/decentralizedROUGHDRAFTSHORT |
What will be the long run effect of technological progress on the distribution of power? Will governments and big tech companies closely cooperate to strengthen their control over society? Or can the cypherpunk dream of technology as a tool of individual freedom and privacy be realized?
These questions are discussed in The Centralized Internet Is Inevitable by Samo Burja and in this podcast featuring Burja and Palladium editor and cofounder Wolf Tivy. Their position is that the Internet is an inherently centralizing technology.
<script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>The future is total surveillance for total compliance of every action, word, and thought. This is the world we are bringing into being with technology.
— Wolf Tivy (@wolftivy) November 20, 2019
The unavoidable telos of technological means is totalitarian power.
I'll describe below how technology has already decentralized power to a large degree and why we might expect a continuation of this trend.
Compare the information landscape in Western countries now to what it was in 1990. Today we have easier access to a wider variety of information and narratives, which are often more critical of large institutions and elites than the narratives people were exposed to in the past. This criticism has eroded the perceived legitimacy of institutions and elites, as described by Martin Gurri in The Revolt of the Public (podcast with transcript, ten minute video).
In the pre-Internet era everyone was deplatformed by default. There were a limited number of TV channels, radio stations, and newspapers making it necessary for information distributors to exercise heavy curation. This gave information distributors cover to curate content partly based on the its narrative. They wanted to distribute profitable content but they also wanted to trade favors with powerful people and institutions who didn't like criticism.
The small number of information channels meant there was no room for niche content -- it was most profitable to try to appeal to a broad swath of the population. The high costs of pre-Internet information distribution gave advertisers more leverage over content. Both of these factors limited anti-establishment narratives.
On the Internet almost everyone can have a platform, but it's still true that only a small percentage of people have a large platform. There's still a fixed amount of attention to go around. Content on the Internet is implicitly filtered (everyone only sees a minuscule fraction of all content produced) but this filtering is done mostly based on popularity. There is some narrative-based filtering but far less than exists in legacy media.
So much information is produced on the Internet that people no longer need to ever return to the fold of mainstream narratives. Before the Internet even if you were reading subversive books you still probably got much of your news from TV or newspapers. Today someone might get almost all of their understanding of what's going on in the world and what it means from Joe Rogan, Eric Weinstein, Chapo Trap House, Scott Alexander, and Reddit.
Governments, the media, corporations, and educational institutions can no longer control which narratives are taken seriously and this is a direct result of the Internet.
One might object that most anti-establishment narratives spread via YouTube, Facebook, and Twitter, which are all becoming increasingly powerful platforms. Isn't this just a manifestation of a tendency toward centralization?
Control of information is decentralizing despite this because in a world of with near-infinite information channels deplatforming someone is more conspicuous than in a world of few information channels. Traditional media platforms must always exercise judgment about who they broadcast because of limited space, and they can use these judgments to deplatform people for reasons that are not obvious.
On Internet platforms deplatforming is the exception and the act of doing it attracts attention and demands an explanation.
Tech platforms do impose some narrative filter on information. There have been coordinated deplatformings of Alex Jones and Richard Spencer. Twitter recently prevented sharing of a story by the New York Post (a decision Twitter later said they were wrong to do), and Twitter sometimes puts warnings on tweets (including Donald Trump's). Tech platforms have started instituting "soft" filters on some content by reducing the likelihood that it appears in users' feeds. This week YouTube said it would start deleting videos which claim that the 2020 US election results are fraudulent.
It's important to compare this narrative filtering with what existed before the Internet. The relevant question is whether the Internet is leading us toward more centralization than what existed before, not whether it's leading us to more centralization than we think is ideal.
Jones and Spencer would never have had their narratives broadcast on mainstream media pre-Internet. Even after his deplatforming, Jones (via his site InfoWars) has almost ten times greater reach than The Economist. Jones' most recent appearance on Rogan's podcast has over 15 million YouTube views, 60% more views than the top rated network TV news show (ABC Nightly News).
Network effects have lead to most Internet communication being mediated by a small number of companies. China has shown us that if these companies are captured by a government the effect on power can be strongly centralizing. However there may be no plausible path for this sort of capture to develop in the West.
When the Internet was new the Chinese Communist Party was in a position of higher information control than Western governments. The CCP made a deliberate choice to maintain high information control over the course of the Internet's rise and they've largely been successful. This took a lot of effort and planning.
Western countries made little effort to control the flood of information stemming from the Internet and as a result have suffered the crisis of institutional authority described previously.
Maybe the US government could have avoided this fate if it had acted sooner, but it is now in a situation of reduced legitimacy and little control of the narratives its citizens are exposed to. Without these things it's difficult to enact and enforce the far reaching policies (mandated encryption backdoors for instance) that would be needed to transition to high information control over its citizens.
Digital privacy advocates had tried and failed to get people to use end-to-end encryption for decades. Technical challenges prevented the existence of user-friendly options and people didn't care enough to go through the cumbersome process of using these tools.
This changed in 2010 with the release of TextSecure, which evolved into Signal. Signal uses end-to-end encryption, produces very little metadata, is open source, and is as easy to use as any other messaging app. Signal has become popular among protesters, the European Commission, and the UK's ruling Conservative party.
WhatsApp, the world's most popular messaging app, has integrated Signal's technology and now two billion WhatApp users have end-to-end encryption enabled by default. Zoom, the world's most used video conferencing software, just released their own end-to-end encryption feature in October 2020.
We are in the midst of a massive global rollout of end-to-end encryption. For the first time in history normal people are regularly communicating remotely in a way that governments and companies lack the power to listen in on.
Thousands of software developers and researchers are now working on another problem: creating a private, uncensorable form of digital money that's easy to use and can support billions of users. If they succeed in separating money from the state it will make it much more difficult for authoritarian governments to control their citizens.
Cryptocurrency isn't yet widespread 11 years after its creation and there are many unsolved problems, but progress is steadily being made and 11 years isn't that long. It was 25 years between the release of PGP (difficult to use encryption software) and when end-to-end encryption became widely used. It was at least 20 years between the invention of the Internet and when its decentralizing effects on the flow of information became significant.
Western governments have tried for decades to ban end-to-end encryption and have mostly failed, with the partial exception of Australia's recent anti-encryption law. This law has been controversial and has been blamed for reducing the competitiveness of Australia's tech industry.
I call this law a partial exception because Australians can still use end-to-end encryption via Signal and WhatsApp. It's not technically possible for the Signal team to comply with the law, so the Australian government's options involve blocking Internet traffic to and from Signal servers or pressuring Apple and Google to remove Signal from their Australian app stores. They've done neither of these so far, and Signal's blog post about this legislation reveals how to easily bypass both of these measures.
The situation in Australia is worth watching but we won't get a good sense for whether Western countries could actually ban end-to-end encryption unless the Australian government makes a more serious effort to criminalize its use.
The US government has worked with tech companies to secretly get access to citizens' cloud data and they have probably secretly backdoored some hardware but there are no US laws mandating that computer hardware contain government backdoors.
The most effective technique for governments trying to access data on an iPhone in their possession is trying every passcode combination. Users can protect themselves against this with long passcodes (10 digits will take about 12 years). There are many instances of the FBI and NSA failing to get into iPhones.
Governments may be able to hack into your phone remotely using sophisticated spyware if they can get you to click on a link, but probably not if you have the latest iPhone OS.
Given that the Internet has already stripped Western governments of much of their perceived authority, it's far from clear that citizens will hand these governments the additional power they need to access all of our data. We are not yet living in an authoritarian state -- Western politicians still fear the consequences of enacting and enforcing unpopular laws.
When humans were hunter-gatherers power was much more decentralized than it is today. One group could not rule over that many people both because the state of technology meant that the ability to inflict violence was proportional to the size of a group, and because the lack of permanent settlements made it easy for subgroups to leave.
The invention of agriculture allowed permanent settlements, caused the accumulation of large stores of wealth that were hard to move, and created economic network effects around physical locations. Whoever controlled a location could extract resources from those wanting to benefit from being near it. Our modern system of nations is a continuation of this trend.
The Internet is bringing us back to a situation where it's less important to physically be in any one place. We can now get close to the center of economic and intellectual activity from any physical location. This increases competitive pressure on governments and tilts power toward individuals.
<script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>And other multibillion dollar tech companies have gone full remote.
— balajis.com (@balajis) December 12, 2020
We’re entering the next phase, what we’ve been building towards over the last few decades: a global two-sided marketplace of technologists and policymakers.
Go where we’re welcome, leave where we’re not. https://t.co/gy2f0nNQOn
Balaji dives deeper into the power of exit in this talk.
Even though the US government knows a lot about its citizens, its low information control and low legitimacy constrain the authoritarian actions it can take. For the US government to become significantly more powerful it would need to establish higher information control, which it could only achieve with the tech industry's help.
However it's not in the interest of tech companies to help the US government become authoritarian. Such a system would lock tech companies into a subservient role, subject to the whims of the ruling party. Zuckerberg, Bezos, and Page do not want to be in the position of Jack Ma, who was likely forced out of Alibaba by the CCP.
In the US the balance of power between tech companies, government, and citizens is more evenly distributed than in China leading to more of a (temporary) stalemate. In the struggle between tech platforms and the US government the support of the public is critical, leading to caution from both groups about abusing their power.
Most technologies increase power across the full spectrum between large organizations and individuals, but they do so unevenly.
Artificial intelligence becomes more effective the more data and computing power you have, so we should expect it to increase the power of large organizations more than smaller ones. The crossbow was an example of a technology that helped individuals more than large organizations -- it enabled untrained peasants to effectively wield deadly force. Encryption and cryptocurrency also help to level the playing field between large organizations and individuals.
Whether power centralizes in the future depends on the specific details of current and future technology, the existing structure of power when this technology is developed, and many other difficult to analyze factors. We're living through a period of rapid technological progress so it's hard to know these details, but we've seen many factors pushing in the direction of decentralization. For those of us who prefer to avoid an authoritarian future it would be a mistake to prematurely accept that fate.
I welcome offers to bet with readers who disagree with any claims I've made.
Follow me on twitter to be notified of my future writing.