The Dead Internet Theory (DIT) is an idea that the internet is mostly fake. It claims the internet is basically a massive botnet, filled with “content” made by bots interacting with other bots. A place where real human activity is almost non-existent.
To someone who’s never heard of the theory, DIT might sound like a conspiracy theory. And honestly it started out like that. Surely, all those people on social media are real and product reviews on a website must be made by real people. Zuckerberg might be a lizard, but he’s not a computer (yet).
Though there are still visible parts of the internet made by people, there’s an enormous amount of bot-generated stuff hidden away. The more I use the internet, the more bot-generated stuff I see, and the more empty the internet feels. There are moments where I ask myself: “what am I reading? Did someone really write this?” And with AI inference becoming mainstream, I think the DIT has become a lot more convincing, and might inevitably become a model, rather than a theory.
Text articles
Sometimes I need to look for an article about something. Maybe it’s something a software related tutorial or a recipe for a dish, these are the two most noticeable topics. If I search for a software tutorial article, articles begin to repeat themselves after the 3rd search result or so. By this, I mean that the same article contents are posted, only with a difference website, layout, and claimed “author”. Sometimes, one article will literally be a clone of another. Some articles just sound like a bot wrote them, or give extremely odd context like “what is a programming language”, or “what is Linux”.
For recipe websites, I have to scroll 3 pages down to find the actual recipe. The first 3 pages are fluff about what a dish is about, and how the author’s grandmother used to prepare it. No one reads that junk; if you’re going to use ChatGPT, please tell it to be less verbose.
So it’s becoming kind of hard to find human-written text on the internet. Mainstream sites will all move to AI written posts and increase their output. This leaves only smaller websites run by a small team or individuals, which are harder to find.
Spam and botnets
About two months ago, I opened an email account using this domain to receive email. Within 10 days, I started getting spam. I was able to get rid of it, but the speed of bots being able to scrape my website was quite frightening.
And these emails weren’t coming from random domains someone bought to send spam, it was coming from Gmail. So someone is either making an army of spam accounts in Gmail, or people’s accounts got hacked and were put into a botnet. I’ve seen the botnet story happen to people I personally know, so I can only imagine the full scale of hacked email accounts out there.
Then there are botnets outside of email. Notable ones are the crypto comment or softcore porn bot threads on Youtube. A bot account will post a link to a Telegram room (lol) or Discord server (lmao) and tell you how to make millions off a new shitcoin. Then 10 bots will reply to that first bot to try and make it look more legitimate. Or a bot with a cute/hot girl as a profile pic posts a suspicious link and some ❤️ emojis to make it alluring.
And it’s not just Youtube that is filled with these botnets, they exist on every social media platform. Or even more generally, anywhere that has a decent amount of traffic. Search for “I hate texting” about Twitter, or search for the percentage of internet traffic generated by bots.
Bonus: The man who uploaded 2 million videos to Youtube
No, that header is not an exaggeration.
Roel Van de Paar is a “Youtuber” that uploaded over 2 million videos to Youtube. His “content” is tech support questions you would find on Stack Overflow. Here’s a little math to help understand this number.
- Youtube opened in mid 2005. We’ll assume it’s been 19 years since they operated
- Assume Roel Van de Paar uploaded videos at a constant rate since 2005 (he actually opened his channel at the end of 2012)
2 000 000 videos ÷ 19 years = ~ 105 000 videos/year = 288 videos/day = 12 videos/hour
This is clearly not the work of a person. No human can ever upload this many videos. Automation is clearly behind the work.
My theory on how this is possible is that Roel uses either some cheap home server or a few cheap VPSes to automatically make videos. This is probably doable with a script for web scraping and using FFmpeg or similar software. He then uploads a video through Youtube’s API, so no manual work is involved. His computer(s) upload when he’s at work, sleeping, and when he’s on the toilet.
If you look at his channel, the views per video ranges from a few thousand to low tens. But imagine that every video got a measly 5 views. That’s still 10 million views total, a significant number. Roel would probably get decent ad revenue from this, and it costs him almost nothing in return. Because spamming the internet is profitable, people keep doing it. Even though it sucks for everyone else.
So now what?
I don’t know. Soon enough, deepfakes will actually be good enough to fool people. A future LLM might pass the Turing test. Maybe the hypothetical technological singularity will be achieved. Who knows if the consumer economy will be able to exist when a lot of people can’t find income.
The only thing that seems immune (kind of) to all of this is the physical world. At least when I physically see someone outside, I can be certain that they are not a bot. Older generations that grew up before the internet and digital world tend to say the younger generation spends too much time online. Maybe they’re right, but in a different way. Not because screens damage our eyes and we’re too sedentary, but because the physical world is where we really exist.