YouTube2Text

Dead Internet Theory: KI übernimmt das Netz | NANO — Transcript

by 3sat NANO · 1,759 words · 278 segments · language en · Watch on YouTube

Full transcript

  1. 0:02In a single day on the internet, 12
  2. 0:03million fake accounts are removed by
  3. 0:05Facebook. 50,000 AI-generated songs
  4. 0:08land on Spotify. More than half of all
  5. 0:12new Reels on TikTok and Instagram are
  6. 0:14created by machines. And almost every
  7. 0:17third new book on Amazon was written
  8. 0:19with the help of AI. More and more
  9. 0:22content no longer comes from humans,
  10. 0:23but from machines. This raises the
  11. 0:26question: is the internet dead? Is the
  12. 0:28internet dead? A good question.
  13. 0:30Every one of us has already been on the
  14. 0:32internet several times today.
  15. 0:33I don't think the internet is dead yet,
  16. 0:35but it’s starting to smell a bit in
  17. 0:37one or two corners.
  18. 0:40And this feeling that something isn't
  19. 0:42quite right on the internet has a name.
  20. 0:44The Dead Internet Theory. What began
  21. 0:47almost like a joke on a blog in the
  22. 0:50late 2010s has developed surprisingly
  23. 0:52in recent years. But what exactly does
  24. 0:55the theory actually say? Well,
  25. 0:59the Dead Internet Theory is, to some
  26. 1:02extent, a meme that was very popular on
  27. 1:05Reddit a few years ago and was
  28. 1:07certainly more of a joke at the time.
  29. 1:11Yes, it perhaps even started more as a
  30. 1:14conspiracy theory, but it is a theory
  31. 1:17that certainly holds much more validity
  32. 1:19today: namely, that a large part of the
  33. 1:22so-called traffic, i.e., the websites
  34. 1:25clicked, is perhaps no longer clicked
  35. 1:28by humans at all, but by machines, and
  36. 1:30that the content you see is also no
  37. 1:33longer created by humans, but by
  38. 1:35machines. A recent study from April
  39. 1:402026 showed that automated traffic
  40. 1:43already accounted for 3%of all web
  41. 1:46traffic in 2025.
  42. 1:50There is a concern that eventually only
  43. 1:52bots will talk to bots, that a computer
  44. 1:54says: "Hey, so-and-so has a birthday,"
  45. 1:56automatically writes a greeting, and on
  46. 1:58the other side, another computer thanks
  47. 2:00them for this greeting, likes it, and
  48. 2:01says: "Hey, great." And in the end,
  49. 2:04there are actually no longer any humans
  50. 2:06interacting with each other, but just
  51. 2:07computers. That has nothing to do
  52. 2:09anymore with the internet that used to
  53. 2:11exist, where the idea was, hey, we can
  54. 2:13all participate, the absolute
  55. 2:15democratization of knowledge, of
  56. 2:16opinion, of exchange. It was precisely
  57. 2:20this exchange and the democratization
  58. 2:22of knowledge that made the internet
  59. 2:23lively and exciting in the first place.
  60. 2:27For the first time in history, people
  61. 2:29from all parts of the world could
  62. 2:31connect with each other, share
  63. 2:33information and feelings, and discuss
  64. 2:35anything and everything. Between 1990
  65. 2:40and 1999 alone, the number of internet
  66. 2:42users grew from 2 to 150 million. The
  67. 2:4890s were the backdrop for the internet
  68. 2:50revolution, defined by freedom, free
  69. 2:52knowledge, and cultural exchange. Then,
  70. 2:56at some point, came the dawn of the
  71. 2:58blogosphere. People started blogging
  72. 3:00and exchanging ideas, then came social
  73. 3:02media, and with social media, there was
  74. 3:04an internet within the internet.
  75. 3:06Suddenly, we had a place where we could
  76. 3:08interact with each other without having
  77. 3:10to type anything into a browser. I was
  78. 3:13always on Facebook anyway, for example.
  79. 3:18With the rise of social media and the
  80. 3:20spread of smartphones, the exchange of
  81. 3:22information between people worldwide
  82. 3:24accelerated at an unimaginable pace.
  83. 3:29Between 2010 and 2020, the amount of
  84. 3:31data we generate on the internet every
  85. 3:34year multiplied by about 30 times. And
  86. 3:38since the AI boom and the release of
  87. 3:41ChatGPT in 2022, we are facing the next
  88. 3:43stage of development. And the next
  89. 3:46stage will be synthetic content, where
  90. 3:48basically only computer-generated
  91. 3:50content exists on the internet, and in
  92. 3:52the bleakest scenario, it will only be
  93. 3:54viewed by computers as well.
  94. 3:57Many bots are already in use today, but
  95. 3:59what exactly are they? A bot is
  96. 4:03software that automatically performs
  97. 4:04tasks on the internet. Incredibly fast,
  98. 4:07around the clock, and without a human
  99. 4:09needing to click, type, or watch. Since
  100. 4:13the AI boom, there are also AI agents
  101. 4:15and so-called scrapers from companies
  102. 4:18like OpenAI or Perplexity that scour
  103. 4:20the web. For example, when a question
  104. 4:23is asked to ChatGPT, the systems send
  105. 4:25out small bots. They visit websites,
  106. 4:27read the content, and retrieve the
  107. 4:30information so the chatbot can
  108. 4:31formulate an answer from it.
  109. 4:35AI-generated content is, of course,
  110. 4:38flooding social media, and we are
  111. 4:40finding that social networks can no
  112. 4:42longer really handle all this
  113. 4:44AI-generated content. This "slop" that
  114. 4:47is very, very fast and very, very easy
  115. 4:49to produce, and of which there is an
  116. 4:51extreme amount, is something that human
  117. 4:53creators can no longer compete against.
  118. 4:57Ömer from Hamburg works as a content
  119. 4:59creator himself and manages five other
  120. 5:01influencers at his social media agency.
  121. 5:05In recent years, he gave up his fashion
  122. 5:07account, which had over 25,000
  123. 5:08followers. Among other reasons, due to
  124. 5:11the pressure from the massive
  125. 5:12competition of AI-generated videos. I
  126. 5:19used to make fashion content on another
  127. 5:21account and then decided for myself
  128. 5:22that I would give up the channel
  129. 5:24completely in October 2025. I also
  130. 5:28noticed that many creators are emerging
  131. 5:30from the AI sector in that field, which
  132. 5:32is easier for big brands to collaborate
  133. 5:35with, and you definitely notice that
  134. 5:37it's super difficult to get a foot in
  135. 5:39the door with things like that.
  136. 5:43AI-generated influencers in the fashion
  137. 5:45sector are long since a reality. Some
  138. 5:48reach hundreds of thousands of
  139. 5:49followers and post several AI-generated
  140. 5:51videos every day. For many content
  141. 5:54creators, this means less visibility
  142. 5:56and the fear of losing their job to AI.
  143. 5:59Ömer is experiencing exactly this for
  144. 6:00our industry right now as a manager,
  145. 6:02seeing the danger that brands will
  146. 6:03eventually get a taste for it and say,
  147. 6:05"Hey, I don't have to go through all
  148. 6:07this trouble anymore." Sending a person
  149. 6:09, for example, a bag, a coat, a jacket,
  150. 6:11or trousers so that they can advertise
  151. 6:13them; instead, I just have to tell an
  152. 6:16AI: "This is my new pair of trousers,
  153. 6:18this is the color, I'd like this
  154. 6:19background, this person, the person
  155. 6:21needs to be this tall, should have this
  156. 6:23build." It takes 10 minutes, and the
  157. 6:26video is created. Basically, such
  158. 6:28videos are extremely elaborate, and
  159. 6:30sooner or later, they will naturally be
  160. 6:33used by these brands.
  161. 6:35The fact that such content spreads so
  162. 6:37quickly is not just due to the AI
  163. 6:39itself. but also due to the platforms '
  164. 6:41algorithms.
  165. 6:44The algorithm has a birth defect, I
  166. 6:46believe. And that is the basic
  167. 6:48assumption that there is a law of
  168. 6:49nature whereby content is only
  169. 6:51interesting to many people if many
  170. 6:52others have interacted with it. And
  171. 6:54it's similar with AI-generated content.
  172. 6:57So the content that I find quite
  173. 6:59disturbing and strange, I naturally
  174. 7:01watch, I get stuck on it, and in the
  175. 7:03end, I give this content the chance for
  176. 7:05even more people to see it because the
  177. 7:07algorithm deems it valuable.
  178. 7:10A vicious cycle that reinforces a
  179. 7:12central point of the Dead Internet
  180. 7:14Theory. The internet as a place full of
  181. 7:17fake accounts, fake comments, and fake
  182. 7:20reviews, in which humans increasingly
  183. 7:22play only a passive role. But there is
  184. 7:26another effect, because the more
  185. 7:28frequently AI is trained with content
  186. 7:29that other AI has created, the worse
  187. 7:31the quality becomes. The results become
  188. 7:34less accurate, less diverse, and more
  189. 7:37repetitive. Scientists call this Model
  190. 7:39Collapse.
  191. 7:42If we have bots, LMs, systems that
  192. 7:44create content themselves, which they
  193. 7:46then use to train themselves again, to
  194. 7:49create new systems from that, then at
  195. 7:51some point you have a cycle of
  196. 7:53distortion that must eventually end at
  197. 7:55the point where only garbage is
  198. 7:57produced. And that ultimately means
  199. 8:00that every day the internet becomes a
  200. 8:02little worse than it actually once was.
  201. 8:05And that is a problem for people
  202. 8:07looking for information. It is becoming
  203. 8:10increasingly difficult to tell whether
  204. 8:13a website I land on after using a
  205. 8:15search engine was created seriously and
  206. 8:17professionally by humans or if it is
  207. 8:20AI-generated.
  208. 8:23Whether what we see, read, or hear
  209. 8:25online is even real anymore is
  210. 8:26currently one of the big open questions
  211. 8:28on the web. Large platforms are trying
  212. 8:32to counter this with hidden watermarks
  213. 8:34in files, invisible markings in images,
  214. 8:36texts, and audio. But the possibility
  215. 8:40that these solutions could soon be
  216. 8:42obsolete is real.
  217. 8:45The dilemma is that, as of now,
  218. 8:47AI-generated content cannot be
  219. 8:48technically identified. So, there is no
  220. 8:51good technology that can detect 100%of
  221. 8:53AI-generated content. This means we are
  222. 8:56actually facing a reversal of the
  223. 8:57burden of proof. We effectively need to
  224. 8:59define, prove, and verify what is real.
  225. 9:02I would very urgently like to see
  226. 9:05editorial obligations introduced.
  227. 9:07Perhaps only once you have an account
  228. 9:09yourself with 100, or whatever, 10,000
  229. 9:12followers. Maybe then you need a legal
  230. 9:14disclosure, maybe then you have to be a
  231. 9:16publisher yourself.
  232. 9:17At the same time, however, we all don't
  233. 9:18want to give up our freedom on the
  234. 9:20internet. We want to have the
  235. 9:21opportunity to be anonymous while
  236. 9:22online. And I hope that we don't slide
  237. 9:25into a world where we have to use ID
  238. 9:27online, because we would lose a great
  239. 9:29deal that is very important for a free
  240. 9:32internet.
  241. 9:35The internet became mainstream in the
  242. 9:3790s, driven by the idea of an open
  243. 9:39exchange and freely accessible
  244. 9:40information. Today, the web is changing
  245. 9:44noticeably and threatens to become more
  246. 9:46of a space under the flood of synthetic
  247. 9:48content where misinformation increases
  248. 9:50and control grows. So what will the
  249. 9:54internet look like in 5 years?
  250. 9:57I believe, or rather, I hope that the
  251. 9:59stories and the people themselves will
  252. 10:01endure.
  253. 10:03I believe that sooner or later we will
  254. 10:05realize that people have a need to see
  255. 10:08content from other people. And that
  256. 10:11will lead to either existing networks
  257. 10:14finding tools to recognize real content
  258. 10:17, or new networks emerging that prefer
  259. 10:19real content in such a way that
  260. 10:22AI-generated content has a harder time.
  261. 10:25If we, as viewers, continue to be
  262. 10:28unwilling to pay for good content,
  263. 10:31there will likely be two groups. One
  264. 10:35that continues to get a lot of junk on
  265. 10:38social media and will no longer find
  266. 10:41much good content on search engines,
  267. 10:43and another that is willing and able to
  268. 10:46pay for good content. The flood of
  269. 10:52fakes and AI could create a growing
  270. 10:54divide between those who actively
  271. 10:56search for real information and even
  272. 10:58pay for it, and those who only see what
  273. 11:00is fed to them. This stands in clear
  274. 11:03contrast to what the internet was
  275. 11:05originally intended for. Whether this
  276. 11:09trend can still be stopped is decided
  277. 11:11by each one of our clicks, because you
  278. 11:13shape the digital reality of tomorrow.

About this transcript

This page contains the full transcript of Dead Internet Theory: KI übernimmt das Netz | NANO by 3sat NANO, generated from the public captions YouTube serves with the video. The transcript has 1,759 words across 278 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.