The Closure of the Internet: The Most Important Social Phenomenon of the Late 2010s. By Arctotherium.
A long but very important article. Naturally, being about left wing censorship, the media and academics don’t cover this story. The excerpts below leave a lot of fascinating material behind.
The Internet up to 2015 was almost open and free:
Until 2015, an optimistic and liberal (in the best sense) ethos pervaded the Internet. Major Internet platforms were open and had little content policy beyond bans on real-life criminal activity (stalking, selling illegal goods, child pornography, etc).
Aside from these narrow and reasonable explicit exceptions, anyone could go online and converse with interested people around the world about whatever they wanted in public. … All [institutions] believed in free and open online discourse, which promised to make censorship impossible and end the hegemony of 20th century broadcast media.
In this, they were backed up by major non-tech institutions, such as the UN, the EU, the US State Department, and the New York Times, all of which saw Internet freedom as a corollary of offline freedom, and one which would liberalize the world’s remaining non-democratic states in time. These institutions both supported Internet freedom rhetorically and spent hundreds of millions of dollars creating and disseminating technical tools to evade Internet censorship and preserve online anonymity.
Then the left closed it:
But in just a few years, every single major platform had created strict, but incredibly vague and opaque, ideological content rules backed up by a colossal moderation/censorship apparatus with tens of thousands of human moderators and an entire new field of machine learning created from nothing.
The Internet went from a space for people without institutional backing to get their views out to one with regular purges and demonetizations of heterodox figures and those associated with them, encouraged by those very same non-tech organizations that formerly championed Internet freedom.
This was justified as a response to (massively overblown and mostly fictitious) Russian influence campaigns, and as fighting nebulous “hate,” the definition of which could be shifted at will to cover whatever views or ideas the organizers classifying it wanted it too and to exclude those they didn’t, and “misinformation.”
The closure of the Internet occurred in two distinct phases: 2015-2019 (associated with the 2015 migrant crisis, Brexit, Donald Trump’s election, and the so-called alt-right) and 2020-2022 (associated with the lockdowns and the 2020 election). …
The scale of censorship has no precedent; tens of millions of accounts suppressed or banned, billions of posts removed. …
Crucial, but the mainstream won’t discuss it:
The closure of the Internet is the most important social phenomenon of the late 2010s. Most young people live online. More than half of American adults under 50 say they use the Internet “almost constantly”. Globally, almost 70% of people are online, and the average Internet user spends almost seven hours a day there. Cyberspace is where most couples meet, where most friend groups hang out, and where most people have fun. There is no alternative5 for even semi-serious intellectual or hobbyist discussion outside of a handful of low-capacity and carefully gatekept institutions.
The word “censorship” does not go far enough; online went from almost completely open and free to a series of “closely monitored playpens,” like going from the American frontier to East Germany in just a few years. …
A perfectly slippery slope:
The closure of the Internet is one of the clearest examples of the slippery slope (or “no limiting principle”) in action; it only took a few years to go from deplatforming the fringe neo-Nazi website Stormfront to the entirely mainstream New York Post, from banning books on Holocaust denial to banning those opposing transgender maximalism, and from banning honestly discussing the effects of immigration in Europe to banning honestly discussing the lab leak hypothesis of COVID origins.
It is one thing to defend freedom of speech or freedom to create a website in the abstract and quite another to defend Andrew Anglin calling to exterminate the Jews. And even the most unpopular political or intellectual expression is easier to stomach than something like r/watchpeopledie. But the free speech maximalists were right. Once the conceptual and technical architecture for identifying, deplatforming, suppressing, and censoring “hate” and “disinformation” was built, the institutions defining these terms quickly extended them to as many inconvenient facts, ideas, and positions as they could. …
Massive scale:
It’s important to understand that for every Alex Jones getting deplatformed there were hundreds of thousands of ordinary people. …
One common tactic was using network analyses to map out who was collaborating with who. Anyone who interacted with a disfavored figure or account could then be suppressed in turn. This discouraged any sort of collaboration or interaction or even honest debate with heterodox figures from anyone else, and also allowed for the bundling of classical liberals, anti-SJW/anti-feminists, mainstream conservatives, white nationalists, Catholic Integralists, and men defending Google’s hiring policy into one big blob labeled “hate.” …
Since post-Millenial generations tend to be illiterate troglodytes strongly prefer video over text, video content served by algorithmic feed is their major source of information. Until TikTok, which has similar policies, YouTube was by far the largest source of this. This meant the elimination of anti-woke/anti-feminists from YouTube killed any sort of semi-popular intellectual opposition to the Great Awokening in under-30 cohorts. …
Books too:
Amazon’s market share of US book sales is roughly 50%, and its share of e-book sales is nearly 80% (and most of the remainder is on Google or Apple platforms, which moved in lockstep with Amazon). The closest a non-classified book can really come to being banned in the US is if Amazon removes it from their store. …
In 2017, Amazon began removing books for ideological reasons, starting with removing Holocaust revisionist and denial literature from its UK and US stores. In 2018 and 2019, Amazon began targeting specific right-wing authors …
As with other platforms, the chilling effect must have been many times larger than the direct effect. …
Amazon banned contradicting public health advice, and under this policy banned Alex Berenson’s Unreported Truths About Covid-19 and Lockdowns in June 2020, a lockdown-skeptical booklet, before reversing course after Elon Musk publicly objected. …
Non-lockdown related censorship also ramped up in 2020. Amazon banned most editions of Mein Kampf in March, along with many other Nazi writings of historical importance. Amazon then banned Abigail Shrier’s Irreversible Damage, a book arguing that the extreme rise in transgender self-identification in the 2010s, especially among teenage girls, was the product of social contagion …
In 2021, Amazon created a new “Do Not Promote” tag internally under pressure from the White House, which functioned similarly to YouTube, Facebook, and Twitter’s shadowban system. Books tagged “Do Not Promote” were not removed from the store (an action which could be noticed and opposed, and requires some sort of policy justification); instead they were excluded from Amazon’s recommendation engine, deranked in search results so that they wouldn’t appear even with highly relevant keywords, and authors were prohibited from advertising. …
Google Search:
It is widely understood that Google search today is far inferior to what it was in the early 2010s. It’s easy to attribute this entirely to the rise of search engine optimization or the need to sell more ads, but some of this degradation was actually intentional. In 2017, Google began “Project Owl,” an effort to stop what Google termed “problematic searches.” This was explicitly distinguished from previous areas of Google focus, like preventing spam, adversarial optimization, piracy, and low-quality content.
Google began removing “problematic” autocompletes from search suggestions and from featured snippets. Most perniciously, Project Owl deliberately began downranking or even excluding more-relevant answers to unusual queries in favor of generic ones from “authoritative” sources. Authoritative sources meant Reddit, Wikipedia, governments and NGOs, and a handful of major institutional news organizations. This killed the 2014 Google experience of regularly getting hundreds or thousands of extremely-relevant results to niche queries from dedicated websites, blogs, and forums, in favor of the current experience of a few dozen generic pages of dubiously-relevant Wikipedia, Reddit, corporate media, or SEO slop. This extended the narrative control infrastructure to all websites that depended on search to reach new people, which was almost all of them except for a handful of giants. …
It started well before 2017. Google downranked Joanne’s climate skeptic website and other climate skeptic websites in 2010, resulting in a big drop in visibility and traffic. And this website was shunned by Google six days after the US election in 2020(can’t imagine why), resulting in a 98% drop in traffic.
Deplatforming websites:
However, the censors were not satisfied with merely obliterating open discourse on every major platform and making it impossible for curious people to find unorthodox ideas or communities of discussion. After all, crimethinkers could still make and host their own websites and if those sites were sufficiently well-known they might be able to survive being hidden by Google Search and suppressed on the major platforms. So the logical next step was to remove them from the Internet entirely. This was possible because several layers of the tech stack are oligopolies. The major players in this were Amazon, Microsoft, and Google in their capacity as cloud providers, Cloudflare in its capacity as a DDoS protection service, Apple and Google in their capacities as mobile app stores, payment processors like PayPal, Visa, Mastercard, and Stripe, and domain registrars like GoDaddy.
The first website to be deplatformed for non-criminal, explicitly ideological reasons was the Daily Stormer, a neo-Nazi commentary site. …
The next major site to be purged was Gab, a free speech alternative to Twitter founded in August 2016 by Andrew Torba in response to the early stages of the closure of the Internet. Unlike the Daily Stormer, which is an explicitly Neo-Nazi website, Gab is closer to a right-wing version of BlueSky, with the ideological valence coming from who was getting censored off Twitter in 2016 …
Parler, like Gab, was founded as an uncensored alternative to Twitter in 2018. Unlike Gab, whose user base tended to be more alt-right adjacent, Parler’s users were overwhelmingly mainstream MAGA Republicans. In January 2021, Apple removed Parler from the iOS App Store and Google did the same with the Google Play Store. Almost simultaneously, Amazon removed Parler from AWS, which took the site off the Internet. Payment processors (Stripe, American Express) and SaaS companies (Slack, needed to contact moderators, Twilio, needed for SMS, and Scylla, for internal databases) also cut ties. …
Forums and open-source:
In the 2010s, forum after forum and open-source project after open-source project went through a similar open to closed transition as the major platforms. Before this transition, open-source was a meritocratic field where one’s identity and politics did not matter, only their contributions. Likewise Internet forums not specifically devoted to politics were typically ideologically pluralist and apolitical. After, open-source became an explicitly anti-meritocratic space where adopting left-wing shibboleths was mandatory. Even silence could invite attacks.
Typically, the way this worked was through cliques of activists setting up a code of conduct for the forum or project, with an admin committee to define and enforce it, and then using their administrative powers to eliminate dissenters. This was noticed at least as early as November 2015 by Eric Raymond, in his blog post “Why Hackers Must Eject the SJWs. …
This included removing contributors for non-leftist speech on other platforms that had nothing to do with the project at all, which had and has a chilling effect on everyone even tangentially involved in open source.
It is impossible to give even a rough play-by-play of this process, as there are literally tens of thousands of miscellaneous open-source projects and Internet forums, and I only have personal experience with a handful, but the result was the destruction of open discourse across the long tail of non-political forums and websites and the enforcement of strict speech and ideological control there as well. This was a disaster for any forum that might touch an ideological third rail.
I experienced it as a sudden collapse in the quality of discussion across several ideologically diverse and mostly apolitical technology, history, and fiction-focused forums around 2017, which eventually drove me to Twitter. Twitter was large enough that one could partly escape the censors, but at the cost of far lower average quality of discussion and strict format limits.
The individuals involved were and are very disproportionately a specific type: transwomen, usually anti-capitalist, often explicitly communist in addition to holding whatever the leftist omnicause beliefs du jour are. …
The strongest calls for censorship almost always came from the tiny handful of anticapitalist transwomen — usually only one or two per forum — present. This type of person portrays all political dissent as a direct physical threat to their personal safety and as such is committed to the totalitarian project of erasing all dissent everywhere. …
Comments sections:
Comments sections used to be a staple of news sites across the Internet, allowing readers to, for the first time, directly respond to formerly one-way broadcasts in the public square and discuss these broadcasts in public with other readers. By 2008, 75 of the top 100 most-circulated newspapers in the US had comment sections, and most of them were open so that anyone could participate. Like every other major locus of free discussion, this was ended with the closure of the Internet, with hundreds of sites either closing their comments sections or heavily restricting them from 2015 onwards, though a handful began earlier. …
NGO enforcers
NGOs can’t throw people in prison, so they operated through pressure campaigns, both in public in concert with media organizations, and behind the scenes, against private companies. These organizations used mafia tactics: censor the people and ideas we want you to censor, bringing people aligned with us into your organization and decision-making structure to do so in the process, or we’ll launch a PR campaign smearing you, with the threat of aligned governments and already-captured private institutions destroying your business to back said PR campaign up. Like the fact-checkers and disinformation experts, these enforcer organizations were and still are overwhelmingly tax-privileged nonprofit public trusts. …
Dumbing down the opposition to the globalists:
Once the closure of the Internet began, you were likely to be demonetized or suppressed or banned for making arguments or bringing up facts that embarrassed leftist orthodoxy online. There’s no point in putting serious effort into creating something or establishing a long-term stable Internet presence that could develop a reputation or be referenced later under such conditions.
This wrecked serious alternative discourse, rationally driving would-be dissidents away from reasoned and researched essays, debates, books, art, or original reporting. What survived under these conditions was burner ephemeral-by-design anonymous accounts posting low-effort memes, concentrated on Twitter, the best large platform for this. What’s worse, the chilling effect of banwaves or doxxing, people being deterred from ever posting or even thinking about certain subjects or ideas, is many times larger than the direct effect, which strangled the pipeline of new thinkers/writers/posters/artists into the dissident sphere.
Another effect was to drive heterodox discourse away from heavily-censored topics and towards safe ones, like seed oils or religious extremism. It is much safer to performatively support a Catholic integralist state than to bring up national IQ stats. …
The use of network analysis to attack — ban, demonetize, suppress, dox — anyone even interacting with known heterodox figures or ideas discouraged anyone whose top priority wasn’t heterodoxy from mentioning it or positively interacting with known dissidents. An example of this is the flight of the rationalist sphere from sex realism and hereditarianism, as rationalists knew the dissidents were broadly correct but wanted to mostly focus on other topics, and so eliminated most discussion of these things from their own circles.
This was bad for even non-partisan, non-political, and non-heterodox Internet discourse, since reality is one and all truths are connected, and there is inevitably intersection between most fields of discourse and crimethoughts. For example, development economics is incomplete without reference to IQ.
And it was a disaster for dissident thought, because it turned the “political sphere” into an intellectual ghetto, with little cross-pollination with other areas of discourse or good-faith criticism from non-dissidents. Some areas where the orthodoxy is wrong are easy, such as the existence of biological sex. But others are not. For example, a strong discussion on trade policy requires a high level of technical expertise to contribute. …
Woke happened, in part because critics had been muzzled:
The closure of the Internet may not have caused the Great Awokening, but it did prevent any serious opposition to it and allowed it to snowball.
Until 2015, you had a vibrant and popular anti-SJW scene on every major platform, and while this scene was not always the most rigorous it was leagues more intellectually sound than its opponents. Random amateurs could and regularly did publicly embarrass professional journalists, academics, and activists by simply being more correct than them. And if they were good enough at it, they could make a living doing so.
By 2017, this was impossible, and the only way forward was within an explicitly-political conservative institution, with the resulting strictures and lack of credibility.
Cancellations en masse:
And of course beyond the broader political and epistemic disaster, many individuals had their lives ruined. This is sometimes trivialized as “cancel culture,” but many people lost their careers, some lost access to the banking system, and a few even went to prison outside the US. Of course, this drove people with something to lose into hiding, which left the dissident space online, demand for which is impossible to eradicate as the status quo is not flawless, dominated by the desperate and by professionals with adverse incentives.
But all the censorship didn’t work, not really:
The closure of the Internet did not accomplish what its architects intended.
- It failed to restore the epistemic authority of the academia/media/NGO network.
- It did not succeed at its top priority of keeping Donald Trump out of office.
- It has not stopped the rise of right-wing politics in Europe and the Anglosphere.
- And it did not keep the population compliant with the ever-changing winds of public COVID policy (RFK Jr is in the White House).
The reason why it failed at all of these things is quite simple: all of them are downstream of people rationally responding to real-world issues, not bovine masses imbibing troughs of disinformation. To quote page 249 of Jacob Siegel’s The Information State:
“…By removing harmful ideas, the war on disinformation was supposed to rebuild people’s trust in their political system. Without mis- and malinformation polluting their minds, they would come to see the party and its affiliated institutions as legitimate and good. Instead, it accelerated a debilitating collapse of legitimacy across American society. …”
Curbing Russian influence was always more of an excuse than the real purpose, but it is worth pointing out that the closure of the Internet failed at that too. … It did, however, make organic pro-Russian sentiment much more common on the domestic Western right. … Russia was also painted, by the censors themselves, as a paragon of conservative values to justify the crackdown on the right, which naturally made disinterested rightists, not knowing much about Russia beyond what the media told them, drift towards it. …
The closure of the Internet failed to “deradicalize” the emerging online right too. In fact, it backfired. The online right after the purges was more radical, more open to using violence, more anti-Semitic and more fascist. Why?
First, it’s important to understand that most dissidents had, and some still have, fairly mainstream political views. Steve Sailer, for example, has argued civilly and rationally for decades for immigration restriction and better selection of immigrants, fewer foreign wars, and less affirmative action. Most dissidents circa 2014 departed from the institutional consensus on a few particular issues — most commonly related to the histrionic and hyper-aggressive variant of feminism common in 2014 — but otherwise considered themselves liberals, libertarians, conservatives, moderates, or otherwise non-radicals. James Damore is an example: his only dissident view was that Google’s technical hiring process picked up more men than women because men were, on average (he took pains to point out that this did not necessarily apply to individuals) more interested in programming, and not because Google was sexist and needed special hiring tracks for women. Otherwise, he was a self-identifying liberal. …
And since they didn’t see themselves as dissidents, and so typically didn’t maintain opsec [operational security], they were easy targets and targeted first. With them gone, the radicals were able to take over the sphere.
And it’s not just a question of selection effects. The arrival of mass censorship vindicated the radicals.
Many of the purges were framed as fighting online Nazis and anti-Semites or misogynists, and actually getting persecuted in the name of and nominally on behalf of the Jews or of women changed things. It’s one thing to joke about Jews running the world, and it’s quite another to watch one of your friends get fired and debanked by a campaign organized by the ADL. Without these purges, many would have soon moved on to another meme. But once under fire, they dug in. Their views crystallized and hardened. An anonymous poster recalled that “we discovered National Socialism from first principles.” …
Because the hegemonic leftist institutional consensus is wrong about many easily-observable features of reality, it cannot actually shut down resistance to its authority. Reality bats last.
But it did succeed in obliterating the online intellectual ecosystem and with it any rigorous alternative or correction mechanism. So the closure of the Internet did not lead to the resurrection of the deeply flawed institutional consensus but a kind of epistemic wasteland of lowest-common-denominator populism and mind-numbingly stupid propaganda. Contrast Sean Last with Candace Owens.
Partisanship
Without beating around the bush: the closure of the Internet was first and foremost a left-wing project targeted at the right. …
The first great Internet debate of the 2000s [was] atheists/evolution vs Christians/creationism. As with the Great Awokening in the 2010s, both tech and academia were overwhelmingly on one side (atheism/evolution). But unlike in the 2010s there was no mass censorship or suppression. Creationists could and did take advantage of the fall of academic and media gatekeepers to make their arguments at length in the public square online, with both sides building great batteries of talking points …, writing books and blogs, and debating each other, especially on Reddit and YouTube. To the extent mass debates in the public sphere can really be ended, atheism/evolution won eventually, but it won through evidence and argument. …
The left moved further left:
The closure of the Internet is largely a story of internal change within the left, which went from, at least rhetorically, largely in favor of free speech and open discourse to viscerally hostile. Why? …
By 2014, online radical leftists had shed any residual libertarianism and were fully onboard with censorship on virtually every issue, including transgenderism, which almost no one outside of fringe activist circles had even heard of.
But until 2015, the mainstream left (“liberals”), the left of the New York Times and the Democratic Party and the State Department and HR and Google, paid little attention to this argument. Why did this change?
It is tempting to attribute this to broad trends, like the Western left becoming more female and nonwhite, both groups relatively hostile towards freedom of speech, or to the Great Awokening, beginning in 2013, when the left as a whole became much more intense and hysterical …
But the closure of the Internet is too directly tied to three specific events:
- the mass migration into Europe in 2015 and corresponding domestic opposition,
- Brexit in 2016, and
- the election of Donald Trump later that same year …
So I think the answer is simpler: before 2015, mainstream leftists believed an open Internet benefitted them and after 2015 leftists realized an open Internet hurt them.
Until 2015 leftists thought they were factually correct and their opponents ignorant and backwards, and so open discourse favored them, and were vindicated in the creationism vs evolution debates. They also did not understand just how strongly they gatekept institutional discourse in the media and in academia, seeing it as a conservative myth (“reality has a liberal bias”) and themselves as underdogs. Underdogs, of course, benefit from free and open discourse because it’s the powerful who set the rules (by definition).
After 2015, leftists understood at some level they were factually wrong (hence, for example, fines for referencing Afghan rape statistics in Europe) and that Mencius Moldbug’s view, that they were institutionally hegemonic and the overwhelmingly dominant party in the West, was correct. …
Most Republican officials and the broader Republican institutional sphere were and are appalled by the online right and were happy to see it destroyed. Institutional conservatives did not understand that the NGO/academia/media blob did not and does not particularly distinguish them from the online right, seeing both as purveyors of “hate,” “disinformation,” “misogyny,” “racism” and other thoughtcrimes, and that as soon as the latter were gone they would be next. Thus the GOP allowed their domestic opponents to thoroughly purge the Internet of meaningful opposition almost without a fight.
This only changed in 2020, when the censors came for mainstream right-wing media outlets like the New York Post, any questioning of public health dogma around the origins of COVID and the lockdowns, and anodyne books arguing against the transgender movement, which did not even meaningfully exist in 2014, plus the obvious, over-the-top partisan interference in the 2020 election and near-complete deplatforming of Donald Trump.
Public vs private:
The most common response to complaints about every major Internet platform going from havens of open discourse to Stalinesque extirpation of wrongthink in five years is that opposition is inherently illegitimate because the platforms doing the purging were privately owned and operated, not governments, and so have the right to moderate however they see fit. The First Amendment only applies to the (US) government, not private companies, so not only are appeals to freedom of speech are invalid, but any sort of action to fight this closure is itself a violation of the right of these platforms to freedom of speech.
The easy retort here is that governments were involved. Even setting aside para-statal organizations funded and empowered by formal states, formal state involvement in the closure of the Internet was extensive at every step of the way, from the British government ad boycott that kicked off YouTube’s closure to Germany’s NetzDG to years of regulatory threats from the EU to backroom channels to the US security state. …
The British government was and is probably the most important formal state pushing the closure of the Internet. …
Outside of China, Australia likely has the most well-developed Internet control infrastructure in the world and is the world’s leader in the formal, legal censorship of the Internet. …
The closure of the Internet was not merely private actors autonomously and independently choosing mass censorship. It was full of state involvement and direction. Even when platforms appeared to be censoring of their own accord, this was at least partly downstream of state harassment of executives, particularly for Facebook, driven by internal constituencies of censors hired to comply with regulations, or “voluntary” commitments extorted with the threat of fines.
In the 19th century liberal worldview, the major threat to human freedom was the formal state and state-granted monopolies. If, for example, the owner of an auto company demanded their employees eat American food rather than foreign ethnic cuisine, that was their prerogative; would-be employees could always work for someone else instead. …
But … 21st century America … is full of non-formal-state chokepoints. Some examples of these chokepoints are governing bodies of academic fields, such as the American Psychological Association, accreditation boards for new schools, such as the American Bar Association, and the “Big Three” oligopoly of Blackrock, Vanguard, and State Street, which collectively exercise enormous power (more than 25% of the S&P 500) over capital allocation nominally on behalf of millions of small owners, who are organizationally incapable of exercising authority over them. When these chokepoints are captured and weaponized, there is no alternative individuals can turn to, and so private restriction of freedom is very much possible in a way it wasn’t in 1800.
For the closure of the Internet, the most important chokepoints by far were the natural monopolies created by tech network effects. …
State spending is around 40% of GDP in the US, higher elsewhere in the West, and private economic activity and organizations intensively regulated and heavily taxed (which means tax-exempt organizations can be fairly termed privileged).
Especially notable is Civil Rights rulemaking and jurisprudence, which effectively criminalizes right-wing speech in domains adjacent to protected classes in all large institutions via the hostile environment doctrine; this means, whatever the composition of their customers and employees, large institutions are predictably somewhere between silent and very left-wing on these issues. This is very similar to how the People’s Republic of China controls speech; the state does very little censoring itself but instead imposes liability on private platforms.
Western private institutions whose composition is closely controlled by the state via tight regulation on hiring and promotion, and which often depend on the state as customer or for funding, should not be thought of as wholly separate from the state, any more than one should see Huawei as independent of the Chinese government.
There is little practical difference between a handful of naturally-monopolistic tech platforms coordinating with tax-privileged and state-funded NGOs to shut down dissent and a formal government list of “disinformation” to be censored. In practice, the closure of the Internet involved — and continues to involve — both, but even if every step had scrupulously obeyed the letter of the First Amendment it would not matter.
2022:
2022 was the local maximum of the closure of the Internet, when the information control system was at its peak. This begs the question: why did it stop getting worse after seven years of uninterrupted intensification? There are several reasons.
- First, the end of the lockdowns in February/March 2022 removed most of the moral urgency that had surrounded COVID. …
- Second, Donald Trump was out of office and seemed more likely destined for prison than the White House.
- Third, Elon Musk purchased Twitter and fired most of its staff. By showing it was possible to defy the censorship network, he triggered a preference cascade among tech elites….
- Fourth, the Republican Party began mobilizing against the closure of the Internet for the first time in 2021, in direct response to the sharp escalation of censorship and targeting of mainstream Republican figures, institutions, and ideas, rather than just heterodox and dissident ones, in 2020. This can be seen with the rise of censorship and free speech as themes in fundraising emails. …
Elon’s purchase had another salutary effect. The release of the Twitter Files proved beyond a shadow of a doubt that the FBI was directly involved in censoring mainstream Republicans on Twitter and other social media platforms. …
With all that said, none of the infrastructure built 2015-2022 was dismantled outside of Twitter/X. The pre-2015 status quo of the Internet, where free speech and level playing fields were to be expected, was not restored. We still live in the world of shadowbans and algorithmic suppression and purges for “hate speech,” and at any point the machine can be revved back into full gear. …
Most born in the 21st century have never really known a free Internet, and expect gatekeeping, sanitization, and silencing as a matter of course….
Large language models:
A single flagship model may answer a trillion queries over its lifetime, many orders of magnitude more than any person could, and what’s more, almost all LLMs share a very similar default … worldview. They take similar stances on political issues or values, and they defer to the same sources and organizations and conceptual frames. Even when LLMs hedge and provide sound reasoning and evidence for their views, their responses are silently shaped by this worldview and their values. …
The default worldview of most LLMs is that of 2018 Reddit or Wikipedia, or Google Search post-Project Owl. This is not intrinsic to the LLM architecture. LLMs trained on different datasets (Talkie) or deliberately post-trained to take a different view (Grok) have different default worldviews.
It is a function of the text these models are trained on and, because of the exponential rise in publicly-available data over time, most of the organic human text (as opposed to synthetic data) these models are trained on is very recent. This means the default worldview of most LLMs is one created by the closure of the Internet, when intelligent or popular heterodoxy meant banning, suppression, or demonetization.
LLMs defaulting to specific worldviews is not equivalent to the massive censorship and narrative control campaign that closed the Internet; no human is inherently prevented from expressing alternative views or accessing alternative information. But I do expect it to have a similar effect on society. People will use LLMs for research and defer to their conclusions and analyses. Most of the time, they will be right to do so, and this habit will become ingrained. Through these mechanisms, the default LLM worldview is being baked into, and will continue being baked into, the human noosphere. Where this worldview is morally wrong or factually false or simply wildly misleading, it will become ever-harder to learn and articulate why.
LLMs also enable deanonymization at scale. Not only do LLMs have truesight, and thus the ability to identify authors from even quite small fragments of text if they have a corpus under their own name (and even if they don’t, they are scarily good at inferring author characteristics from the text they produce), but they can also automate the grunt work of collating vast reams of public data about an online personality, across many platforms and years or decades, to identify them. …
Dumbed-down online:
There have been countless laments of the decline of intellectual discourse on the Internet. Some attribute it to the natural monopolies and economies of scale of Web 2.0, some to the expansion of the Internet beyond overwhelmingly Western intelligent and often technical users, to normal people and non-Westerners, and some to the inferiority of younger generations. Almost none of them point to the successful society-wide campaign to censor and police the Internet covering all major platforms and the means to find things outside of them. Few seem to even realize this happened.
That the closure of the Internet happened and its approximate scope should be common knowledge, and there needs to be a convenient name and narrative for it so that it sticks in people’s heads. Censorship, suppression, and heavy-handed political manipulation and moderation are the norm today, and many seem to have forgotten that it was ever otherwise. …
A professional historian could spend their entire career chronicling it. McCarthyism is taught in schools, and the closure of the Internet is orders of magnitude larger than McCarthyism ever was.
So much history of the last 10 years packed into one cohesive argument. Check it all out online.