WEBVTT 00:00.976 --> 00:09.000 In a conversation, Sam Altman, the head of the AI company OpenAI, described his plans for a ChatGPT successor. 00:09.000 --> 00:15.679 This is supposed to be ready for use in the coming months. Altman said the following: 00:15.679 --> 00:25.048 "I believe we are close to a world in which you can have a successor to ChatGPT that watches your computer screen all the 00:25.048 --> 00:32.017 time, follows every meeting you attend, records every call, and thus has a perfect context 00:32.017 --> 00:37.279 of your entire life – of everything you see." 00:37.279 --> 00:42.840 Soon, the AI will therefore not only respond to requests but act proactively. 00:42.840 --> 00:51.161 It will suggest ideas, recognize potential problems, and support everyday life. What a brave new world! 00:51.161 --> 00:57.736 By comparison, classic dystopias like 1984 – "Big Brother is watching you" – seem like a harmless puppet show, 00:57.736 --> 01:02.178 because Orwell's observers were still sitting behind a wall. 01:02.178 --> 01:08.526 Altman's observers sit in your own smartphone or PC and soon in your own brain. 01:08.526 --> 01:12.650 They become constant companions, supposed friends. 01:12.650 --> 01:15.343 The bait of convenience is tempting. 01:15.343 --> 01:23.476 Who wouldn't want to outsource a few annoying appointments or be supported at work by a perfect authority? 01:23.476 --> 01:30.363 But as soon as people become dependent on the new AI supports, the AI will begin to reprimand them. 01:30.363 --> 01:37.520 The narrow line between "support" and "patronization" will then no longer be defined by the users but by algorithms. 01:37.520 --> 01:41.204 Thus, proactive help becomes patronizing intervention. 01:41.204 --> 01:47.020 Here are some examples of what that looks like in concrete terms: 01:47.020 --> 01:50.695 1. You want to write a critical comment about government policy? 01:50.695 --> 01:55.437 The AI, which monitors every keystroke, "corrects" the text in advance. 01:55.437 --> 02:00.780 It proposes more harmless formulations or does not publish the post at all because it 02:00.780 --> 02:04.040 is classified as "incorrect" or "polarizing". 02:04.040 --> 02:11.060 Google's AI is already blocking harmless reviews based on its own definitions. 02:11.060 --> 02:15.078 2. One searches for information on controversially discussed topics online? 02:15.078 --> 02:22.414 The AI filters the search results and only shows those sources that it considers "trustworthy" – that is, system-compliant. 02:22.414 --> 02:25.585 Shadow banning already exists in social media. 02:25.585 --> 02:34.269 The content is not blocked, but it is no longer displayed in search and thus no longer made available to subscribers. 02:35.769 --> 02:39.906 3. Reading and writing are no longer necessary, because AI summarizes demanding 02:39.906 --> 02:42.683 texts or writes entire articles and speeches. 02:42.683 --> 02:49.870 The publicist Tom-Oliver Regenauer recently wrote about this in a position paper: 02:49.870 --> 02:56.020 "Whoever outsources reading, thinking, and writing unlearns them – and very quickly at that. 02:56.020 --> 03:01.599 The brain is comparable to a muscle. If it is not trained, it atrophies... 03:01.599 --> 03:10.049 If one considers what fatal consequences the introduction of the smartphone since 2007 has produced, it does not take much 03:10.049 --> 03:16.005 imagination to extrapolate what effects AI will have on the cognitive abilities of our 03:16.005 --> 03:19.616 species in a disproportionately shorter time. 03:19.616 --> 03:26.343 AI is – like the smartphone – first and foremost military technology and an instrument of degeneration. 03:26.343 --> 03:32.294 A tool that the fewest can use responsibly and productively. 03:32.294 --> 03:38.755 The orientation-, will- and thoughtless "smombies," who already shape the cityscape, do 03:38.755 --> 03:43.840 not yet seem to serve as a deterrent example." 03:43.840 --> 03:49.367 4. After a very short time, the AI recognizes personal political views, reads private messages 03:49.367 --> 03:52.821 and knows the relationship and contact network. 03:52.821 --> 03:58.580 It does not use this knowledge to support people in their development, but to "optimize" them. 03:58.580 --> 04:02.526 In plain terms, this means: adaptation to the required norm. 04:02.526 --> 04:10.245 Nudging, i.e. subtle opinion control, was already successfully mass-tested during the Corona years. 04:10.245 --> 04:15.040 So how much power do we, the users, want to give ChatGPT and Co.? 04:15.040 --> 04:19.198 Tom-Oliver Regenauer makes a clear statement on this: 04:19.198 --> 04:27.392 "I refuse to train a system that permanently spies on me and already has the potential to kill me. 04:27.392 --> 04:36.490 Because that is exactly what Palantir, ChatGPT from OpenAI, Claude from Anthropic and Grok from SpaceXAI are already doing 04:36.490 --> 04:43.948 today in Gaza, Ukraine, Iran and other war zones, where autonomous weapons systems are controlled by 04:43.948 --> 04:52.424 precisely these systems – just like those mini-drones that, according to EUROPOL, will soon patrol our cities. 04:52.424 --> 05:00.370 Anyone who uses AI systems trains "SkyNet," a central database that will control the real "Terminator." 05:00.370 --> 05:06.784 And thanks to Boston Dynamics, Unitree and Tesla, it will soon be at our door."