“I've never taken the dead internet theory all that seriously, but it seems like there are actually a lot of Twitter accounts run by LLMs these days,” Sam Altman, CEO of OpenAI, wrote on X last week in his typical lowercase style.
Altman, the CEO of the company that created ChatGPT, the world's most popular artificial intelligence text generator, sneered at X. “You're absolutely right! This observation isn't just smart – it shows that you're operating at a higher level.” wrote one user, imitating the flattering tone of ChatGPT text.
Altman was referring to an idea popularized by a 2021 post on the online forum Macintosh Cafe Agora Road: The Internet, once full of human life, is now dead, run entirely by bots and for bots. “The Internet seems empty and devoid of people” wrote IlluminatiPirate, the theory's pseudonymous author at the time. The promise of free exchange between people has disappeared. The Internet has been “hijacked by a handful of influencers.”
Conspiracy theory
In 2021, almost two years before the launch of ChatGPT, the idea that the Internet was run by robots sounded far-fetched, as did the explanation that “the US government is gaslighting the entire world population with artificial intelligence.” The Atlantic published an article about this theory under the title “The Dead Internet Theory Is Wrong, But It Seems True.”
Bots—automated computer scripts that rank websites for search engines and social media content for platforms—were part of the Internet, but they couldn't generate compelling content.
“We haven't had artificial intelligence operating at such a scale that we could actually have plausible artificial intelligence accounts running the Internet,” Adam Aleksic, linguist and author of the book Algospeak: how social networks are changing the future of languagetold TIME. The dead internet theory “used to be a fringe conspiracy theory, but now it looks much more real.”
Death of the Internet
The business model for online content creation is simple: Advertisers pay creators for the attention their content gets, which allows creators to keep creating and people to keep looking at what they like. Except that over the past few years, people have become surplus to requirements.
A March report from ad analytics company Adalytics found millions of instances since 2020 in which ads from brands from Pfizer to the NYPD were shown to web crawling bots rather than to real users, undermining advertisers' investments. In some kind of comic casesthe ads were served by Google's ad server to Google's own bots. According to Impervacyber security company. In 2024, it reached 51 percent, surpassing the share of internet traffic from people for the first time.
Even if some Internet denizens were bots, they were mostly passive observers. That changed in 2022, when Sam Altman's OpenAI kicked off the generative AI race. Since then, the amount of AI-generated content has increased dramatically. The percentage of websites in the top 20 Google search results that contain AI-generated content has increased by 400% since the release of ChatGPT. according to at Originality AI, a startup that creates content detectors using artificial intelligence.
“It’s in the platforms’ business interests to shove slop down our throats, because over time, if there are more AI accounts, they’ll have to pay human creators less,” Aleksic said.
Search engines such as Google have begun providing summaries of articles online using artificial intelligence. Instead of visiting content creators' pages, users can get a review without leaving the search engine. Fewer clicks on content meant less ad revenue flowing to creators.
As the complexity of AI-generated content increases, its reach has expanded beyond social media. In August Send reported that the stories published by “Margot Blanchard” in Wired and at least five other media outlets were shut down after the author was revealed to be an artificial intelligence. For scammers with a creative flair, artificial intelligence represents a new way to make money quickly.
As such, IlluminatiPirate's vision of a virtual wasteland created by and for bots is more plausible than ever.
Human cost
The rising tide of waste is creating problems for AI developers as well. LLMs or large language models such as ChatGPT are trained online. If AI-provided summaries continue to divert revenue from original content creators, high-quality content could dry up, leaving modelers with nothing to learn from. Article published in Nature in 2024 showed that AI models “collapse” when trained on data they themselves generated.
In response, some cloud service providers, such as Cloudflare, have proposed limiting access to the websites they host and making bots pay a fee to enter. This can help creators get back the income they need to continue creating. “My utopian vision is a world where people get content for free and robots have to pay a ton of money for it,” Matthew Prince, the company’s CEO, told TIME. previous interview.
The stakes are higher than just the Internet. People, like great language models, learn from what they read. “It’s not just that we’re surrounded by bots,” Aleksic said. “The thing is, we're starting to look more like bots.” In July Scientific American reported that words commonly used by ChatGPT, such as “get into” and “meticulous,” began appearing more frequently in conversational podcasts following the product's 2022 release. “Real people have caught the idiosyncrasies of LLM speech,” Altman noted in mail on Monday.
There is nothing wrong with changing the use of language. But the algorithms that control what we see online “already present reality differently than it actually exists” by promoting extreme content from human users. The AI generation could further weaken collective reality by reducing the human input that anchors online discourse to what real people think.
“There's a growing perception gap in America where people think other people's views are more extreme than they really are,” Aleksic said. “This is AI psychosis on a mass scale.”
