When the internet was new in the 1990s it had a very different feel to it. This first incarnation of the web, "Web 1.0" as it was called a few decades ago, had a lot of user made content and was very much a human generated affair.

Web 2.0 was the beginning of the long process of decay. With Web 2.0 We started to see the rise of platforms.

At first we saw "walled gardens", AOL was one of the first examples of such. Then later we saw social media sites pop up.

Now platforms are so common that the vast majority of internet usage time is actually on these platforms and not on various sites. People spend far more time on social media than they do browsing the web.

The rise of blogs and comment spam

Some early web sites had the idea of a guestbook in them. This was a place where visitors to the sites could place some comments or discussions. Finding examples of these are hard in 2024 for a couple of reasons. First web search has become incredibly degraded and secondly many of these original websites are no longer hosted.

In many of the original blogs we had sections for comments. This was a radical development for the web because it was one of the first mainstream ways that allowed users of sites to interact with the sites. Along with web forums these were the main ways in which user generated content first appeared on the web.

The first internet search engines tended to rank pages based on the content in those pages, searching for keywords and phrases to then index the pages. Unfortunately people figured out how to game those systems by doing things like including large amounts of keywords in invisible text and other techniques. Search engines then had to respond to the people gaming the system and started to implement some countermeasures in the search engine arms race.

Comments sections were one of the first places on the internet that got heavily targeted by spammers. At first a lot of this spam was generated by hand but as time went on this spam got increasingly more automated. Some of the spam was just of an advertorial nature, but perhaps the biggest technical developments in spam came out of the "Search Engine Optimization" industry.

As keywords and site contents were getting less reliable as measures to rank pages the search engines needed more reliable ways to rank sites that weren't as subject to gaming the system. A breakthrough happened with the PageRank algorithm which started to rank sites by how many backlinks went to those sites from the rest of the internet. The idea here was that it would be harder for a site to manipulate the results when there were parts of the ranking that came from other sites that the webmasters of sites did not control. This created a very strong incentive for search engine "optimization" companies to spam as many pages as possible with links to the pages they were wishing to optimize the search results for. Some unethical SEO businesses started developing automated spam systems for the purposes of bumping up the search engine results for their clients. Because of the strong economic incentives at play this allowed the shady SEO companies to collect a lot of money from clients who wanted to get their sites ranking high regardless of the ethics of doing so. The SEO companies did this by including backlinks to the sites they were promoting by including those links in spam comments posted on as many websites as they could. This gave the SEO companies a lot of money to invest in technical talent to develop more and more sophisticated spamming technology. Some of these early SEO firms were pivotal in the development of comment post spam. This SEO technique worked especially well until the search engines responded with countermeasures in the SEO arms race.

The other people who got heavily involved in comment spam were scammers and fraudsters who wanted to promote their schemes. Comment spam got so bad that there was a massive arms race that started to try to deal with it.

Initially there was some optimism about using machine learning and AI to deal with spam, here's a famous comic from October 2010:

The alt text on this image was "And what about all the people who won't be able to join the community because they're terrible at making helpful and constructive co-- ... oh."

Looking back on this with the developments in AI and machine learning that can generate text this hits very differently, we've just ceeded more control of the internet to algorithms and made it even less human. This is a "mission accomplished" that really has that same feel as George W. Bush's "mission accomplished" from May 1, 2003:

In this speech George W. Bush said "Major combat operations in Iraq have ended. In the battle of Iraq, the United States and our allies have prevailed.". US troops were officially withdrawn eight years later in 2011. This doesn't even say anything about the war in Iraq that caused US troops to be redeployed in 2014.

In much the same way I think we haven't accomplished the mission of dealing with computer generated spam just yet. In 2024 we haven't got a good solution to the spam problem and I think things are actually in some senses worse than ever. Yes we had an early win, just like the coalition forces in Iraq had, but really the war is not even remotely won. The nature of the adversaries has changed and internet spam has changed over time.

There was a tipping point where vastly more fake accounts than real accounts were being created. This led to the ongoing development of the CAPTCHA. Most people now know about these because of how annoying they now are. Due to advances in computer vision algorithms these CAPTCHAs have had to get harder and harder to decipher for the bots, but this has meant that they also have got harder and harder to decipher for human users.

This section could just have easily been tilted "the death of the anonymous internet". As time has gone on more and more sites and platforms have had to ban comments and interactions from anonymous users. As there are now far more fake users than real users allowing unverified accounts to post content is a large liability, and one that fewer sites are willing to accept. With this one of the aspects that made the early internet what it was has died. Anonymous comments by humans was a fundamental part of the early internet. These days an anonymous comment is more likely not to be from a human who's expressing their own thoughts. Bots and humans creating paid content to push some product, agenda or propaganda is now far more common.

The rise of platforms above the law

Section 230 of the Communications Decency Act is a part of USA legislation that enormously changed the trajectory of the internet as it paved the legal way for social media platforms to be able to exist. This remains one of the most important pieces of legislation not just in the USA but the entire world since many social media platforms are based in the USA. Given the huge role social media has in many people's lives you'd think that people would know more about this legislation, but alas almost nobody knows anything about it.

This legislation made it so that "carriers" would not be legally liable for the content posted by user on their platforms. There were some provisions about this that were supposed to limit the platforms but effectively this put the social media platforms in a position of great power. For example if you or I as citizens or as businesses made comments that were deliberately defamatory we would be legally liable for doing this. If a social media platform served up that same defamatory content to millions of people around the country and the world they wouldn't face any consequences at all. These laws effectively created the ability for social media to exist and at the same time created a number of second order consequences that have completely changed the course of modern history.

The social media platforms could by use of their algorithms just push whatever content they wanted to by amplifying certain messages from contributions of users of their platforms without facing the liability for actively editing the content. This could allow them to greatly amplify a message while also effectively dodging legal liability for the message they were pushing by hiding behind the excuse that the content was created by users of the platforms. Because the lawmakers were so far behind the times when it came to understanding the content algorithms that these sites used the social media platforms effectively avoided accountability.

Events like the live-streamed NZ massacre showed the public and the lawmakers that something was seriously wrong with the way the social media sites were operating. Specifically the social media sites would keep claiming that there wasn't anything they could do about the content from users, but this was a very self serving lie. It was a classic example of the won't-can't play. The major social media platforms had a lot of ability to shape the content on their sites and a huge powerful infrastructure set up to do so. But the power of this infrastructure was greatly increased by the general populace not knowing of its existence. If the lawmakers didn't know that the sites had this power it would buy the social media sites extra time to wield this power without facing any accountability for the impacts of it.

Much of these protections of these laws were written under the very direct assumption that social media sites were not like newspapers in the sense that the social media sites were not to have an editorial function to them. The sites were allowed have moderation functions to remove illegal posts. After a series of livestreamed suicides on the platform Facebook started using algorithms to rank posts according to the mental health state those posts predicted. Posts showing self harm and other suicidal ideation were supposed to be moderated but those posts went off to moderation teams with little training for how to handle these things. There's an exceedingly high chance that the handling of this data fell short of HIPAA standards, but that's not surprising since these platforms weren't designed to handle sensitive medical data. There might be good intentions here but the creep towards these sites being completely moderated platforms continues. Sharing posts of suicides to teenagers is egregious and is a situation where the content shouldn't be shown (if teenagers are to be allowed on these apps in the first place, which on balance they probably should not). But where to draw the line on what can and can't be shared? There's a very grey line between moderation and outright censorship of ideas with certain topics. Most of the large platforms have very little transparency when it comes to the way in which they decide what content is acceptable and what content is not. Unfortunately this is an area where the laws just haven't kept up with the rapid changes in the social media landscape.

With the invention of the algorithms that drove what content was seen by the introduction of the "content feed" this assumption that the platforms weren't shaping the content no longer held. The platforms could decide what rubbish to "feed" to its users and in many cases they did do just that. As time went on the platforms deliberately interfered with what people saw in increasingly severe and unethical ways. We saw a particularly high profile case of this in the lead up to the 2020 presidential election with the Hunter Biden laptop scandal. In this case many social media platforms engaged in very deliberate editorial actions by banning the publishing of certain links and selectively removing content from users that was not in breach of terms of service. This wasn't just one off action either but part of a very large operation to sway opinions. If you tried to post an article with links to various newspaper articles your posts would be automatically deleted. This of course was a clear cut case of the sites exerting deliberate editorial control over the content that users were posting in a situation where the users were breaking no laws and not breaking the terms of service of the platforms they were posting on. Experts on this law have suggested that this type of overt editorial intervention of user content should have made the social media platforms lose their Section 230 immunity.

Due to the rise of scammers and spammers, social media sites have had to invest considerably in technological ways to automatically delete certain posts and certain links. This is a necessary evil to prevent sites from being completely overrun by bots and scammers who use dubious links for criminal purposes. In this day and age if you buy a new domain name and set up a Wordpress site with open comments you'll get all sorts of scams and spam in your comment section in a disturbingly short amount of time. But to turn this automated infrastructure against a link to a post published by a journalist on such a high profile topic was a turning point. It was clear to anyone paying attention that the idea of not being editorial was clearly dead on these big platforms. The question of if Section 230 protection for the platforms should continue to exist considering this clearly editorial behavior is something that many in the legal system are still debating.

Perhaps even worse is that way in which leaked documents showed that there was a multi tiered system for which content was subject to moderation and which was not.

Let's look at the first program you reported on, which is a VIP program called XCheck. This is a program that basically created separate rules for VIPs and for everybody else who uses Facebook

In this system the "VIPs" are exempted from having their content fact checked. Accounts covered under this scheme included politicians, celebrities, athletes, or anyone who was popular enough that Facebook deemed the risks of intervening into their content in a visible manner too risky for the company. If the company deemed it would receive considerably criticism and PR backlash if it were to intervene then the company wouldn't intervene. If you were the average person your posts were handled entirely differently. If a politician wants to spout complete nonsense there's nothing stopping them. If the average person does the same they could be banned. This is a truly machiavellian and cynical approach from Facebook. These sorts of things show that the platform is anything but impartial to the content posted on it and has invested considerable resources in creating multi-tiered systems where some people's content is treated entirely differently to others. In other words Facebook is more than happy to delete your content if they think they can get away with it, and only hesitates to intervene when they think they will face a backlash from the general public or the legislators or the powerful for doing so.

Internal research at Facebook found that the Instagram platform which they own was causing significant mental health issues for many of its users. Instead of going public with this they held it back until the documents got leaked.

Infinite scroll and the death of the early internet

One of these days I should finish reading Infinite Jest by David Foster Wallace. The premise of the book is that there's this tape that's so addictive that people watch it until they die.

I can't help but wonder if infinite scrolling as seen on many social media platforms is a bit like this fictional tape. More than once I've seen people have very close calls with death because they were paying so much attention to what was on their phone and not their surroundings. The first time I saw this happen was in 2016 when I saw someone walk out directly into busy traffic very nearly avoiding death because a car slammed the breaks and another car swerved to miss them.

The invention of the infinite scroll I think marked the death of the original internet, but it took some time to see why. What was "innovative" about the infinite scroll was not that it helped people find better information or navigate websites more effectively but was effectively the exact opposite. The infinite scroll became popular because it meant that people had to spend more time on websites to get teh same content, during which time more advertising could be shown. While it might not have seemed like much at the time this was one of the first changes to the internet where technology was introduced to make browsing less efficient while simultaneously making it more addictive.

The changes on platforms from navigating around pages to getting an endless "feed" of content changed how people interacted with the platforms. People get lost in this feed because their brains want to scroll to get to the end, to feel some sense of closure, but these infinite scrolling mechanisms are designed to hijack that instinct and keep you scrolling way longer. There is no end, there is no satisfaction, it's all a highly manipulated cynical machiavellian scam to capture your attention and sell it to advertising companies. The issue is that the users of social media platforms are not the customers of those social media platforms. The social media platforms then have incentives to do what the advertisers want, not what would be best for the users.

Interestingly when the commercial incentives don't align with the infinite scroll you still see pagination on websites. I remember being at a tech conference a few years back and there were staff attending from multiple major tech and e-commerce companies. Staff from one e-commerce platform were talking about how they that replaced pagination with infinite scrolling and quickly realized it was a mistake. They had very detailed customer data that showed that people stopped purchasing as many products when there was an infinite scroll as compared to paginated results. Because this lost the website money they went back to paginated results fairly quickly. E-commerce platforms need to sell products to make money, having users endlessly scroll doesn't make them money, unlike advertising driven platforms like social media sites.

Speaking of social media and e-commerce, recently we have seen more and more fake content on these social media platforms to push various products and E-commerce related grifts...

Influencers

As more and more people are spending more time on social media platforms we have seen the rise of the "influencer". The book called "Get Rich or Lie Trying: Ambition and Deceit in the New Influencer Economy" by Symeon Brown is a great exposition of how the modern internet influencers go about their craft. This is an especially important book these days because the spread of social media has meant that influencers are now a key part of actually influencing society.

"Influencers" make money based on both how much attention they can command and how much that subsequent attention can be exploited for financial gain. This creates huge financial incentives to get attention whatever the method. Unfortunately that's led to a whole lot of anti-social activities being posted online when those anti-social activities attract a lot of attention. This problem is amplified greatly by the platforms that often push this content to other people. Also it turns out that it is often far easier to fake your content than for it to be genuine. As long as nobody exposes the fraud it can be far cheaper to sell the lie of success than actually be successful.

The story of Dan Blizerian is a really good one. He rose to social media infamy by amassing a huge follower count on social media. He had big parties at mansions, private jets, etc. A big part of his success was in the "self made man" grift. Of course the fact he was a trust fund nepo-baby with an extremely wealthy father who got his wealth as a Wall Street Raider was never to be mentioned. Perhaps the lack of a moral compass with business runs in that family. He claimed to be a high stakes poker pro as well, and to anyone who knows anything about poker it was very clear that he was not at the level he claimed to be at. Blizerian however spoke as though he was a top player which is why I knew he was a fake long before the other parts of the story broke. Then there was the real estate claims, which turned out to also be dubious. Other people who knew the areas he was claiming to be an expert in saw the same pattern, that he knew a lot less areas he claimed to be making money in than a legitimately successful people in those areas would. You see this same pattern with other "influencers" who claim expertise in a certain area, if you happen to actually have expertise in that area you'll often quickly see if an influencer is faking their claims of expertise.

This comes up a lot in the business and investing world, you'll see people who claim all sorts of wild stuff but sometimes its hard to know if what they are saying is bullshit until you've done some research into that area. People however find the claims of success, if presented in the appropriate way, to be emotionally appealing and won't verify them for a variety of reasons. In 2024 most financial advice is actually entirely bullshit for a variety of reasons. But they all have the same root cause, its earlier to make money in finance telling people what they want to hear than to tell them the nasty messy truth of how it all works. Far too many people would rather have their short term feelings validated than to set themselves up for lasting wealth. On top of that, in finance, being outright fraudulent can be very lucrative too. Recently some Youtube finance charlatans were promoting a bank that just went under. Yes that's right things are so bad with influencers these days that they can set up a dodgy bank of all things.¹

Sometimes these videos designed for clicks aren't as dangerous as others. We have things like the "primitive buildings" scams, where the whole thing is actually fake and all the construction is actually done with modern equipment. Thankfully those videos are mostly just going to be entertainment, people won't try this stuff themselves and if they do they'll quickly realize the whole thing is actually fake. But the scary part to me in this is that videos with millions of views are entirely fraudulent. These videos start to warp the perceptions of the people watching them who don't realize its all fake. Peoples intuitions start to get warped by seeing fake processes yielding fake results.

But we have far more dangerous things like "fake welding" where influencers make welding videos that look satisfying. Unfortunately some of these joints lack structural strength and are actually very dangerous. This sort of influencer video is very bothersome because people who don't understand what they are doing might see these videos and think they have learned a safe way of constructing things when in reality they haven't. There's a lot of important details in the trades and engineering that can't safely be ignored. For better or worse this is why these areas are highly regulated, mistakes in these areas directly threaten the health and safety of others. There's a reason we have apprenticeships and qualifications in construction, learning how to do things the right way is incredibly important for the safety of others.

Chat GPT

The invention of large language models has been another massive blow to the authenticity of the internet. Large language models are a development in machine learning that allows an algorithm to generate sentences of text output given a text input prompt.

Large language models have been a huge development in many ways, both technologically and culturally. For many years people talked about the Turing Test as being some milestone for humanity, civilization and intelligence. But then it actually happened and most people didn't seem to notice. Perhaps that's unsurprising since these days most people don't really notice much, such is the state of constant distraction and overwhelm that most people are spending their entire lives operating within.

This technology allows users to create large passages of text without having to write it themselves. That text might not be factually accurate or really make any sense but the text itself is grammatically correct. This allows for a level of spamming that was simply never possible before. This spam is also extremely hard to block because it can bypass all statistical methods of anti-spam detection that existed before that attempted to detect if text was not written by a human.

What this technology has done is less about decreasing the cost of producing spam and scam content but more about preventing it from being easily detected. Because fake content is cheaper than real content and hard to detect we are starting to see more fake content being generated as a proportion of all content.

Throw in fake engagement from bot accounts, paid subscribers (shills) and various other forms of astroturfing and you have a powerful combination that can amplify fake content far more than real content. Real engagement is a .

AI generated picture and video content

The most recent "advance" in the fake-ification of the internet is the emergence of AI generated images and videos.

It was only a few years ago in August 2022 that the stable diffusion technique was released. This allowed people to use input text to generate new images. Not that long afterwards people started using these methods to create entirely new images using AI techniques as a primary means of creating new content. This was a big change from before where people would have had to engage in a very manual and time consuming process to create images if they wanted to make something new. The costs of creating fictitious imagery has dropped enormously and as a result the spread of such imagery has spread enormously. We are already seeing the emergence of a new class of scams that's impersonating people using these techniques.

Automated content generation has come along substantially in the last few years. In past times companies hired "content creators" to make low quality crap content for things like SEO, but now with technology all those "creators" are out of work. Its much cheaper to use generative AI to make low quality crap content than to pay a human. In the race to the bottom automation will be a large factor. It will also be a crucial factor in the death of full employment.

So much so that we now have examples like the crazy situation with Spotify where AI was used to make a lot of music that was then viewed a huge number of times by bots. The creators of this were getting paid by Spotify for all the "clicks" but it turned out these were just computer programs that were automatically clicking on links to content made by other computer programs. This highlights an upcoming dynamics in the "attention" economy, which is that people will start to use AI to avoid them needing to interact with all content. As more and more platforms seek to addict people to dopamine and other low quality content the benefits of having some way to filter out the crap being fed to you will go up a lot. I know I'm not alone when I say I'd happily pay a monthly subscription to avoid getting all content from misinformation bots, scammers and spammers. The annoyance of dealing with all this crap content is something that a huge number of people are going to be willing to pay money to avoid. But the online advertising industry is going to be impacted by this, what happens when my AI clicks on a link that serves up some advertising but the page itself is just misinformation so my AI helps me never visit it. No human "attention" is captured in this case but it won't be clear to the advertisers how many views their adverts have got. In some ways the automatic filtering of crap content I see as being an inevitable direction for the internet with the rapid growth of crap content everywhere. People simply don't have the time and energy to filter through it all.

AI powered censorship

Something that has become increasingly more apparent is the heavy handed application of AI and machine learning that was developed originally to fight spam to ban comments based on their content. The Hunter Biden Laptop scandal was one of first times where the scale of this system was revealed to the world in a highly politicized arena. When this happened all comments that were linking to journalists stories, from mainstream news sources no less, about a news story were getting automatically deleted. This intervention was especially widespread. Were this a link to a phishing site, the sort of thing these systems were initially designed to block, nobody would have really cared, but because it was a political topic that was getting blocked lots of people suddenly found themselves caring about these content moderation systems who had never previously given them any thought.

Now this sort of content moderation by algorithm wasn't new at all, detecting things like cryptocurrency scams and other criminal behavior necessitated the development of this anti-spam, anti-fraud type of content. What many people didn't expect was for these systems and algorithms to be used to target real accounts and human generated content.

We've talked so far about how many methods have made computer generated content easier to create. But unfortunately a big threat to human content internet has emerged recently with advanced natural language processing algorithms making it increasingly easier to selectively delete content by scanning that content for semantic meaning.

Seeing content creators get demonetized or banned based on algorithms that are scanning their content is relatively new. For example on platforms like Youtube every single uploaded video will be processed through a number of algorithms that will analyze the content. This might have started with the music industry demanding that royalties were paid for included content. Initially this system was based on manual moderation but as time went on this became increasingly automated. Algorithms were developed to detect existing audio tracks in new content and those algorithms are now quite ubiquitous. The Shazam service is an example of this in an application. Shazam's music recognition application was first released in 2002 and the company was acquired by Apple in September of 2018, so as you can see this technology has been in the hands of the public for quite some time now.

Unfortunately there's a set of content that's getting banned due to linguistic similarities not being picked up by these algorithms. Comedy is a good example of an area where this is happening a lot, for example satire and sarcasm are a great way to show up exceedingly stupid ideas in a funny way, but naive algorithms can confuse such communications as directly endorsing things that really aren't being endorsed. In recent times algorithms that have more semantic awareness have been deployed to automatically moderate content, there's still some early adoption pains here because a huge number of false positives are occurring that are causing real content to be deleted. We've seen a few cases of this lately, these algorithm mistakes aren't easy to search for because the comments are deleted, but the trends are there if you pay attention. One such example of early censorship aimed at social engineering was blocking posts or comments from people mentioning the various trends of people checking out of the societal system of rewards. This included terms like "lying flat generation", "MGTOW", "silent quitting epidemic", etc. The powers that be were very concerned by people "Going Galt" to use the phrase from the increasingly relevant book Atlas Shrugged by Ayn Rand, published in 1957. This book discusses a future society that doesn't reward good work but instead only rewards connected insiders in a corrupt political system. Throughout the book it talks about how people just stop doing hard work when hard work ceases to be rewarded in a society. Much of the motivation of the book came from the authors direct experience with communism where the system explicitly doesn't reward excellence or hard work since the rewards for such work are stolen from the labourers who do that work. A number of comments discussing these topics were getting automatically deleted because of similarities to posts that were being generated by bots. I could imagine there are some geopolitical games going on at the moment where shaping the perceptions of the populace could be highly influential and social media platforms would be a place where this could be attempted.

As mentioned earlier there was a significant uptick in content moderation and censorship leading up to the 2020 USA presidential election. This came at the same time as a huge uptick in bot created content. Some platforms clearly had a candidate they want to win and they didn't mind removing content supporting other candidates. Also since the war in Ukraine started there's been a large increase in bots posting content to try to sway the narrative about that war. It was very interesting to see this rise in automated and algorithmic generated content for propaganda purposes. There's an absolute torrent of bot created comments about this war that shows up everywhere, and we only see the posts that make it past the measures designed to block bots that are spreading spam and disinformation. It is likely that that more content about these events has been created by bots and spammers than legitimate users.

Now with the 2024 presidential election there's been another event that triggered another mass censorship push on the big tech platforms. Since the assassination attempt on Trump (edit: this refers to the first assassination attempt, there's since been another one) there's been an enormous number of comments being deleted on platforms like YouTube. The amount of comments being deleted has increased so much that regular users are now commenting on how their comments are being deleted. The censorship is now so out of control that people are starting to see that the censorship is out of control. Beforehand not enough comments were getting deleted from real human accounts for there to be much commentary on it, bots if their comments are blocked typically aren't programmed to make posts about how their comments are blocked. Humans on the other hand will draw this connection and likely be annoyed enough to actually post about it. Its not even entirely obvious that this is the fault of the platforms themselves either, there's been an enormous effort in the lead up to the 2024 USA presidential election to spread divisive sentiment via online platforms and the platforms have to do something about this to avoid just turning into complete cess pits of vitriolic and damaging content. There's also a clear battle going on about shaping the perceptions of reality for the populace and much of the battlefield here is occurring in the comments sections of various platforms. Those comments are being produced by many automated systems using generative AI approaches as this is not only a cheap way to create content but is a way that more effectively avoids that content being detected as being spam. It is exceedingly obvious that the activity of bots and other non-genuine accounts has greatly increased in the lead up to the election day. The sites obviously have to do something to avoid becoming completely overrun by bots. Unfortunately it seems that the more stringent anti-spam approaches are also having collateral damage on real comments too, fighting off the increasingly sophisticated spam content is now causing many real users content to be deleted. This is a difficult situation for the platforms because far more fake content is now being generated than real content, large language models have created a massive paradigm shift in the online spam industry. Because the creation of natural sounding text is now so easy the costs of producing spam have dropped significantly. Also a consequence of this generative AI is that the costs of content moderation have also gone up as it is now considerably more difficult to statistically analyse the contents of posts to determine if they were written by bots.

The current state of the internet

A lot of what we are seeing in the modern internet is a reflection of the huge changes in the economics of creating fake content. Because the costs of creating fake content was dropping over the years we started to see more and more of it being created. This then started an arms race with platforms to detect and ban fake content. A paradigm shifting moment occurred when Large Language Models became available that would allow you to generate natural looking text automatically. This has led to a situation where fake content is in many cases cheaper to create than real content. A human can only type a few words a minute but now a bot powered by generative AI can pump out huge amounts of fake content every minute and then post these to many sites simultaneously.

Because of these technological changes we live in an age where real speech has never been cheaper to censored while fake content has never been so easy to create. I suspect it will take a few years for everyone to catch up and understand how these recent technological changes that have in turn completely changed the fundamental economics of content creation.