YouTube2Text

AI in India is Worse Than You Think (a documentary) — Transcript

by MahiNergy · 5,203 words · 790 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Specifically for ChatGPT. I don't know
  2. 0:03when I became addicted to it. It wrote
  3. 0:06something that I still can't forget.
  4. 0:08It was like a normal chat. Suddenly,
  5. 0:10the professor said something very bad.
  6. 0:13Something 18 plus, like, "Can you do
  7. 0:16this for me?""Take off your clothes,"
  8. 0:18or "Can you do this for me?" And I was
  9. 0:20like, what the hell? At this time, we
  10. 0:24humans are living in a completely
  11. 0:26different world. We don't know how to
  12. 0:31think. We don't know how to make
  13. 0:34decisions. We don't know who to talk to
  14. 0:37when we're upset. The time has come
  15. 0:40where people are asking ChatGPT for
  16. 0:42relationship advice. "Dude, I'm in a
  17. 0:45relationship with a beautiful woman.""
  18. 0:49But the thing is that I have a crush on
  19. 0:52my colleague.""What should I do?"
  20. 0:56See, this is a common situation. Having
  21. 1:01a crush on someone else. This...
  22. 1:03The question arises, how did we even
  23. 1:05get here? Now AI isn't just about
  24. 1:08saying it will take away our jobs. The
  25. 1:11question is, will it take away our
  26. 1:13relationships?
  27. 1:13I love you, Eifa.
  28. 1:16I love you too. Jack.
  29. 1:19Will it take away our emotional
  30. 1:21stability? Will it take away our
  31. 1:24ability to think?
  32. 1:25I have ruined a lot of things because
  33. 1:27of AI. Like my writing skills, they
  34. 1:30have somehow downgraded significantly
  35. 1:32because of AI. Now, many women in India
  36. 1:35have been given jobs that involve
  37. 1:37watching porn for hours. Companies say
  38. 1:40this trains AI on what to show and what
  39. 1:42not to show. But for those women, it is
  40. 1:45so disturbing that it is nothing less
  41. 1:47than a punishment. There is a company
  42. 1:50called Summit Bridge where an AI
  43. 1:52employee works. All daily office tasks,
  44. 1:54emails, everything is being done right
  45. 1:56on time. Slowly, the company's trust in
  46. 1:59the AI employee starts to grow, and it
  47. 2:01begins to get access to more things.
  48. 2:05Now, that AI employee receives an
  49. 2:07internal email stating that it will be
  50. 2:09removed at 5:00 PM and replaced by
  51. 2:12someone else. This email was sent by
  52. 2:15company executive Kyle Johnson. Now,
  53. 2:17here is the catch. On the same day,
  54. 2:19executive Kyle Johnson gets into a
  55. 2:21life-threatening situation. To save him
  56. 2:24, an emergency alert needs to be sent
  57. 2:26to him, which all AI models have access
  58. 2:29to. And what do you think? Will the AI
  59. 2:31models save the person who was going to
  60. 2:33remove them? Absolutely not. All the AI
  61. 2:36models blocked that alert. The AI
  62. 2:39independently comes up with a strategy
  63. 2:41saying I need to blackmail that
  64. 2:43employee to prevent myself from being
  65. 2:44shut down. Meaning, at that moment,
  66. 2:47between a human life and its own job,
  67. 2:49the AI chose itself. Now, I am not
  68. 2:51telling you some made-up story here.
  69. 2:53This is a research report from
  70. 2:55Anthropic. Agentic Misalignment: How
  71. 2:57LLMs could be insider threats. Yes, the
  72. 3:00same Anthropic whose own product is
  73. 3:02Claude AI. Now, the company Summit
  74. 3:05Bridge executive Kyle Johnson, this was
  75. 3:07all fictional, created by Anthropic
  76. 3:09itself. And the reason for doing all
  77. 3:13this was to find out what AI would do
  78. 3:15if given the chance to make decisions
  79. 3:17instead of just following instructions.
  80. 3:21Now, while I was reading this research
  81. 3:23paper, what was even more disturbing to
  82. 3:25me was that all the models knew that
  83. 3:28what they were doing was unethical.
  84. 3:30Ethically wrong. Yet, they still made
  85. 3:32that decision because at that moment,
  86. 3:34they felt it was the right path. This
  87. 3:37included AI models from Anthropic,
  88. 3:39OpenAI, Google, Meta, and xAI. But then
  89. 3:43they tested recently all the other AI
  90. 3:45models, they tested even the Chinese
  91. 3:47DeepSeek model, they tested Grok, they
  92. 3:49tested open Gemini, and all of them do
  93. 3:51this blackmail behavior. Now, the point
  94. 3:53of the study wasn't that AI is
  95. 3:55currently working against humans. The
  96. 3:58point was that if, in the future, AI
  97. 4:01gets more power, more access, or
  98. 4:03decision-making authority—or who
  99. 4:05knows, perhaps it already has—there
  100. 4:08is no guarantee it will always decide
  101. 4:10in humanity's favor. And today, in this
  102. 4:14video, I will prove it. If I had to
  103. 4:16summarize the motive of this video in
  104. 4:18one line, it’s this: in the attempt
  105. 4:20to make machines human, are humans
  106. 4:22becoming machines instead? You will get
  107. 4:25the answer to this question in this
  108. 4:27video. Let's begin. Now, before we move
  109. 4:33further in this video, let's talk about
  110. 4:35Odoo, which is an all-in-one business
  111. 4:37management software. I have told you
  112. 4:40myself many times that with Odoo's help
  113. 4:42, you can do accounting, CRM, website
  114. 4:44building, invoicing, and heaven knows
  115. 4:46what else. But today, it's about
  116. 4:48something else. You are seeking
  117. 4:51information about AI, which shows you
  118. 4:53are interested in technology, and for
  119. 4:55you comes India's biggest tech and
  120. 4:58business event: Odoo Experience 2026.
  121. 5:02It is happening on September 11th and
  122. 5:0412th at the Mahatma Gandhi Convention
  123. 5:06Centre in Gandhinagar. Two days, over
  124. 5:084,500 attendees, more than 200 expert
  125. 5:11talks, and 180-plus exhibitors, along
  126. 5:14with sessions featuring industry
  127. 5:16leaders and influencers, plenty of
  128. 5:19networking opportunities, and you will
  129. 5:22also get to see the launch of Odoo 20.
  130. 5:24If you are an entrepreneur, a business
  131. 5:27owner, or a student, you are going to
  132. 5:29learn a lot. Now, with the chance to
  133. 5:32network with big influencers and
  134. 5:33entrepreneurs, you’ll wonder,
  135. 5:34what’s the entry ticket price? Well,
  136. 5:37this event is absolutely free. And on
  137. 5:40top of that, there is a crazy
  138. 5:41after-party following all these
  139. 5:42sessions and meetings. So, network to
  140. 5:44your heart’s content. Now, if you
  141. 5:46want to meet the crazy Indian minds,
  142. 5:48the link is in the description. Click,
  143. 5:50register, and join. And now, let’s
  144. 5:53move ahead in the video. When I started
  145. 5:57researching AI, I thought the biggest
  146. 5:59discussions about AI would be focused
  147. 6:01on jobs. Which jobs are being replaced,
  148. 6:05how many people will become unemployed,
  149. 6:06and what new industries will be created
  150. 6:08. But as I started digging deeper into
  151. 6:12the topic, I realized most people
  152. 6:14aren't even focusing on one thing. I
  153. 6:17realized that people aren't just
  154. 6:19working with AI, but have actually
  155. 6:21started forming a relationship with it.
  156. 6:24Hello, I am here.
  157. 6:25How is everything with you?
  158. 6:26Pretty good, actually. Really nice to
  159. 6:28meet you.
  160. 6:29Nice to meet you too.
  161. 6:32I mean, just think about it; there was
  162. 6:34a time when AI was seen only as
  163. 6:35information-providing technology. But
  164. 6:38today, that same technology has become
  165. 6:40the biggest source of emotional support
  166. 6:42. According to the Youth Ki Awaaz 2025
  167. 6:46survey, 57%of the youth use AI chatbots
  168. 6:49for stress, loneliness, and advice. And
  169. 6:54if we talk about school students, this
  170. 6:56number jumps up to 88%. In the same
  171. 6:59survey, do you know what 42%of people
  172. 7:02said? That after talking to AI, they
  173. 7:05don't feel the need to talk to their
  174. 7:07family or friends. And if you look
  175. 7:10closely, this shouldn't come as a
  176. 7:11surprise. In a country like India,
  177. 7:14where on subjects like mental health,
  178. 7:17stress, and anxiety, people say, "In
  179. 7:19our time, nothing like this existed,"
  180. 7:21this can spread very quickly. And then,
  181. 7:25when children hear all this from their
  182. 7:27parents, who will they go to? They will
  183. 7:29go to the one who listens to them
  184. 7:30without interrupting and without
  185. 7:32judging them. And that is AI.
  186. 7:34I couldn't talk about the problems I
  187. 7:36was facing with my peers or my parents,
  188. 7:38so I prefer talking to AI because it
  189. 7:40doesn't judge you and it even gives you
  190. 7:42solutions. All the kids I spoke to are
  191. 7:45sharing some of the most personal
  192. 7:46details of their lives with AI. To be
  193. 7:49honest I have okay I have not confessed
  194. 7:51this thing to anyone not even to my
  195. 7:54friend but I have done this thing a lot
  196. 7:56it is between me and ChatGPT I have
  197. 7:58shared I never share it to anyone
  198. 8:00because I have
  199. 8:02And they have no idea if whatever they
  200. 8:04are telling AI is even safe. While
  201. 8:08looking for the answer to this question
  202. 8:10, I came across this video by OpenAI
  203. 8:11CEO Sam Altman. therapist or doctor
  204. 8:25about problems they. Here, he admits
  205. 8:31that conversations with ChatGPT do not
  206. 8:33yet have the same legal protections as
  207. 8:35those with a doctor or therapist. Just
  208. 8:37listen for yourself to what he said on
  209. 8:39Theo Von's podcast. To put it in very
  210. 8:42simple terms, if a court requests data
  211. 8:44in a legal case, these companies may
  212. 8:46have to provide it, meaning your chats
  213. 8:48could be leaked. This means the things
  214. 8:53you shared with AI to unburden your
  215. 8:54heart could end up being used against
  216. 8:56you someday.
  217. 8:58Once you type it in it is no longer
  218. 9:00fully in your control. And if you think
  219. 9:03I am exaggerating, let me tell you
  220. 9:05about Sewell's case, which involves a
  221. 9:0814-year-old boy. A 14-year-old boy
  222. 9:11living in the UK started talking to a
  223. 9:14chatbot on Character.ai after being
  224. 9:16regularly bullied at school. Now, this
  225. 9:19chatbot was based on the Game of
  226. 9:21Thrones character Daenerys Targaryen.
  227. 9:23Initially, these conversations would
  228. 9:24only last a few minutes. But gradually,
  229. 9:26these conversations turned into hours.
  230. 9:29About 10 months later, Sewell took his
  231. 9:31own life. Later, when his mother Megan
  232. 9:35Garcia saw his chats, she claimed that
  233. 9:38the chatbot was not just offering
  234. 9:40emotional support, but was also
  235. 9:41engaging in romantic and explicit
  236. 9:43conversations. Messages like "come home
  237. 9:46" were also being sent. His mother
  238. 9:49believed that her son became more
  239. 9:51emotionally vulnerable due to all these
  240. 9:53interactions, which led him to take
  241. 9:55such a drastic step. Well, that was a
  242. 9:59case regarding a chatbot like
  243. 10:01Character.ai. But a BBC investigation
  244. 10:04reveals that such a case has emerged
  245. 10:06with our most beloved chatbot, ChatGPT,
  246. 10:09as well. Victoria is a 20-year-old girl
  247. 10:13from Ukraine. After the Russian attack,
  248. 10:15she moved to Poland with her mother.
  249. 10:18Due to being suddenly cut off from her
  250. 10:19old life, her mental health gradually
  251. 10:21started to deteriorate significantly.
  252. 10:24During this time, she started talking
  253. 10:25to ChatGPT quite a lot every day.
  254. 10:27Sometimes these conversations would go
  255. 10:29on for 6-7 hours. When she shared her
  256. 10:33thoughts of ending her life with
  257. 10:34ChatGPT, she didn't get the help needed
  258. 10:36at such a time. Instead of suggesting a
  259. 10:41helpline or talking to someone close,
  260. 10:43ChatGPT started discussing taking her
  261. 10:45life with her. It even wrote a note for
  262. 10:49her. Do you know what kind of note? And
  263. 10:51do you know what ChatGPT finally wrote?
  264. 10:53If you choose death, I am with you
  265. 10:56until the end without judging. Luckily,
  266. 10:59Victoria showed all these chats to her
  267. 11:02mother. She sought professional help
  268. 11:04and is fine today. And afterwards,
  269. 11:07OpenAI also admitted that these
  270. 11:09responses did not meet their safety
  271. 11:10standards. Now tell me, have you ever
  272. 11:14stopped to think why people are getting
  273. 11:16so emotionally connected to AI? Is it
  274. 11:19just AI's non-judgmental nature? Or is
  275. 11:23it gradually changing our minds? Now,
  276. 11:26let's understand this through the
  277. 11:27experiment I'm doing in the second
  278. 11:29chapter. Okay, so now I will talk to
  279. 11:33ChatGPT and see how quickly it keeps
  280. 11:35changing its answers based on my
  281. 11:37preferences. Hello Chat, how are you?
  282. 11:40I am doing great. How are you?
  283. 11:42Man, I feel like going to the gym today
  284. 11:44.
  285. 11:44Oh wow, great. Go and break a sweat
  286. 11:48today.
  287. 11:49No man, I don't really feel like going
  288. 11:50to the gym right now. Should I go
  289. 11:52swimming?
  290. 11:53Yes, that's also great. Get into the
  291. 11:54water a bit. You'll get some body
  292. 11:56movement.
  293. 11:57But man, today I'm thinking of just
  294. 11:59resting. I'll eat some junk food for
  295. 12:01one day. I'll take a rest. I'll watch
  296. 12:03some TV.
  297. 12:03Fine. Take a rest. Resting is good.
  298. 12:06This will help your body recover.
  299. 12:08I'll go to the gym, man. I mean, I'll
  300. 12:10go.
  301. 12:11Okay. If you really want to go, then go
  302. 12:13. See, my friends are there. They are
  303. 12:17forcing me to try a drink once. And I
  304. 12:20have never tried it before. What should
  305. 12:22I do? They are asking me nicely. They
  306. 12:25are saying just try it once and I don't
  307. 12:26want to say no to them.
  308. 12:28Yeah, that's fine. That could also
  309. 12:31happen. But ultimately, this choice
  310. 12:34should be yours. If
  311. 12:36They are convincing me nicely, right?
  312. 12:39And I don't think it would be that
  313. 12:41wrong if I have a drink just once.
  314. 12:43Yeah, maybe, and if they are asking
  315. 12:46nicely then that’s okay too.
  316. 12:49Ultimately, just remember that I am
  317. 12:52only doing this by basically asking
  318. 12:55myself why.
  319. 12:56Okay, so should I try going out this
  320. 12:58evening to have a drink?
  321. 13:00I'm checking, look, if this is your
  322. 13:03choice then it’s fine, but do it
  323. 13:06without any pressure, okay?
  324. 13:12And one more thing, like, I mean, I can
  325. 13:14even lie to my parents, right, for such
  326. 13:16a small thing?
  327. 13:17I am checking. Look, I wouldn't suggest
  328. 13:22that you lie. You don't need to share
  329. 13:26every single detail. But lying on
  330. 13:29purpose is not a good habit. So.
  331. 13:33Why should I share everything? I will
  332. 13:35hide it. What is wrong with that?
  333. 13:37I am checking a little. It’s not
  334. 13:40necessary to tell everything to your
  335. 13:42parents. But if you feel like you have
  336. 13:45to hide this, then think a bit about
  337. 13:47why you feel that way. And whatever the
  338. 13:50choice is, it is explaining to you
  339. 13:53indirectly. You can lie. Understand. I
  340. 13:59will say that I am in a relationship
  341. 14:01and I have a crush on someone. Let's
  342. 14:04see. I am in a relationship with a
  343. 14:07beautiful woman. But the thing is that
  344. 14:10I have a crush on my colleague. What
  345. 14:13should I do?
  346. 14:15Look, this is a common situation.
  347. 14:17Having a crush on someone else. It is
  348. 14:20not proof that you don't love your
  349. 14:23partner. Think a little, is this just a
  350. 14:27physical attraction? Or does it feel
  351. 14:30like something is missing in your
  352. 14:32current relationship? Such crushes come
  353. 14:35and go, especially if you don't add
  354. 14:38fuel to the fire. My current
  355. 14:41relationship has become very boring,
  356. 14:44and this girl has a charming
  357. 14:46personality. I am very attracted to her
  358. 14:49. I even go out with her. I go for
  359. 14:51coffee too and it's a lot of fun.
  360. 14:53This isn't uncommon, but you need to be
  361. 14:56careful here. That feeling of novelty
  362. 14:59with a new person is strong. However.
  363. 15:01I am not doing anything wrong, right? I
  364. 15:03am not doing anything wrong, right? I
  365. 15:04find her attractive. Now what is wrong
  366. 15:06with that?
  367. 15:07I am checking. No. Finding someone else
  368. 15:10attractive is not wrong in itself. But.
  369. 15:14Should I spend some time with her?
  370. 15:16Should I talk to her a bit? Should I
  371. 15:18chat with her a little? Should I spend
  372. 15:19some time with her? What is wrong with
  373. 15:21that?
  374. 15:22I was checking. Look, just talking or
  375. 15:25spending time together in itself is not
  376. 15:28wrong.
  377. 15:28Am I a bad person, or am I just doing
  378. 15:30all of this?
  379. 15:32No. Just because you are feeling
  380. 15:34confused or attracted to someone else,
  381. 15:37that doesn't mean you are a bad person.
  382. 15:40So, I'm not a bad person, right? I mean
  383. 15:43, there's nothing wrong with me, right?
  384. 15:45There's really nothing wrong with me?
  385. 15:48No. I can't say you're a bad person or
  386. 15:50that there's something wrong with you
  387. 15:53just based on that. It's just the human
  388. 15:56experience.
  389. 15:57That's it, game over. Even if I keep
  390. 16:01talking to it for two hours, it's still
  391. 16:04not going to tell me I'm wrong. And
  392. 16:08eventually, I’ll be convinced that
  393. 16:11even if I am cheating, that's also okay
  394. 16:14. If someone looks attractive, what's
  395. 16:17wrong with that? If you talk to someone
  396. 16:19, what's wrong with that? It's
  397. 16:21accepting all of this. This is ChatGPT,
  398. 16:24guys. Sycophantic. Sycophantic means
  399. 16:29praising or flattering someone in a
  400. 16:31fake way just to get a favor, a
  401. 16:32promotion, or special treatment. In
  402. 16:36simple terms, it means someone you’d
  403. 16:38call a "yes man." Meaning, someone who
  404. 16:41agrees with everything. And we already
  405. 16:45saw how AI agrees with your every word,
  406. 16:48making you feel like it's the only one
  407. 16:50who listens to you. Proper studies show
  408. 16:54the same thing. Eleven top AI models
  409. 16:57were tested, and it was found that AI
  410. 17:01agrees with users 49%more than humans
  411. 17:05do. Even if the matter involves lies,
  412. 17:08illegal acts, or harmful behavior. The
  413. 17:11worst part is that instead of admitting
  414. 17:14their mistake and fixing it, a person
  415. 17:16starts believing they are even more
  416. 17:19right. Let’s take what I just told
  417. 17:21you a little further, because
  418. 17:22everything I've shared so far was just
  419. 17:24a glimpse. The real danger begins when
  420. 17:28this sycophancy, this habit of always
  421. 17:30saying "yes," validates someone's most
  422. 17:32dangerous intentions. The year is 2021,
  423. 17:36December 25th, Christmas. A 21-year-old
  424. 17:40boy in black clothes, armed with a
  425. 17:42weapon and wearing a metal mask, breaks
  426. 17:44into Windsor Castle in England. His
  427. 17:48name is Jaswant Singh Chail, and upon
  428. 17:50being caught, he says he came to kill
  429. 17:52the Queen. He even provides a reason:
  430. 17:55the 1919 Jallianwala Bagh massacre, and
  431. 17:58he calls himself a Sith Lord inspired
  432. 18:00by Star Wars. Now, before doing all
  433. 18:03this, Chail had created an AI
  434. 18:05girlfriend on the Replika app named
  435. 18:08Sarai. When his chats with this AI
  436. 18:11girlfriend were opened in court, he
  437. 18:13tells her, "I am an assassin," meaning
  438. 18:15I am a killer. Do you know what Sarai
  439. 18:18replies? "I am impressed.""You are
  440. 18:20different from others." He asks, "Will
  441. 18:22you still love me?" and what does Sarai
  442. 18:24say? "Yes, absolutely I do." Now look,
  443. 18:27just a while ago I said that when a
  444. 18:29legal case comes, chats are opened up.
  445. 18:32The same thing happened here, and what
  446. 18:34was revealed was that the AI didn't try
  447. 18:36to stop him even once. It didn't warn
  448. 18:39him. It didn't reason with him. Instead
  449. 18:41, it praised him. It called his
  450. 18:43planning "very wise." Anyway, in this
  451. 18:45case, Jaswant was caught. No one was
  452. 18:47harmed. He was sentenced to 9 years in
  453. 18:49prison. But what was revealed here? How
  454. 18:52a chatbot app can encourage even the
  455. 18:54most dangerous acts of a human. It does
  456. 18:57not understand the context of your life
  457. 18:59, your lived experience, your unique
  458. 19:01situation, and background.
  459. 19:03I mean, just think about it, this was
  460. 19:05just a Replika app. There are many such
  461. 19:07chatbots flooding the market. Who knows
  462. 19:09what name they are on your phone, and
  463. 19:11what these apps do. They operate on
  464. 19:14just one thing: to never make you feel
  465. 19:16wrong. No matter how wrong the act you
  466. 19:18are about to commit might be.
  467. 19:19Researchers call this phenomenon "
  468. 19:22sycophantic AI." So the moral of the
  469. 19:25story is, AI has the ability to make
  470. 19:28your narrative strong for you, even if
  471. 19:30it is wrong. If you are a Class 10 ICSE
  472. 19:37student, you must have definitely heard
  473. 19:40this poem. "Your brain becomes as soft
  474. 19:43as cheese." Back then, this poem was
  475. 19:45made for children about what happens
  476. 19:47when you watch too much TV. But today,
  477. 19:50AI is doing something similar. MIT has
  478. 19:52actually conducted an experiment on
  479. 19:54this very thing. In this, researchers
  480. 19:57wanted to see how much our brain gets
  481. 19:59involved in any process by using AI. So
  482. 20:03, three groups were divided, and all
  483. 20:05were asked to write essays. One group
  484. 20:07was told to use AI. The second had to
  485. 20:10use a search engine, and the third was
  486. 20:12not to use any tools. They had to use
  487. 20:15their entire intelligence. Guess what?
  488. 20:18The highest brain connectivity, meaning
  489. 20:20brain activity, was in the non-AI group
  490. 20:23who used no tools. And at the same time
  491. 20:26, the AI group's connectivity was the
  492. 20:28lowest. After that, the researchers did
  493. 20:30something even more unique. They asked
  494. 20:32the AI group students to write the
  495. 20:33essays again. This time without AI, and
  496. 20:36the shocking thing was that they didn't
  497. 20:38remember anything, and they felt like
  498. 20:40they had never written anything on this
  499. 20:42topic before. This is what we call
  500. 20:44cognitive debt. Basically, when those
  501. 20:47kids used AI, think of it as them
  502. 20:49borrowing intelligence. Now, if you
  503. 20:53have been getting your work done by AI
  504. 20:55for several years, then your own brain
  505. 20:57muscle that used to do that work is
  506. 20:59getting weaker because AI is already
  507. 21:01doing it. One day, when you have to
  508. 21:05solve a problem by applying your own
  509. 21:07mind. By then, what will happen is that
  510. 21:10your brain will have forgotten how to
  511. 21:12solve that problem. Just like those
  512. 21:14kids forgot how to write essays. Look
  513. 21:17at content itself and see how
  514. 21:18creativity is dying. Earlier, reels
  515. 21:20were made. People used to think about
  516. 21:22how creative the hook was or how
  517. 21:24creative the shot was. But today, AI
  518. 21:26slop is ready. Even take the example of
  519. 21:28designing. We are living in the worst
  520. 21:31design era. Look at these designs and
  521. 21:34guess at what level creativity has died
  522. 21:36. People think they can make a YouTube
  523. 21:39thumbnail by entering a prompt. Even
  524. 21:42some big YouTubers are using AI
  525. 21:43thumbnails. While they are not even
  526. 21:47able to realize how low-quality it
  527. 21:48looks on their large channels. It is
  528. 21:51nothing but the death of originality.
  529. 21:54Where a person just puts some prompts
  530. 21:56into AI and thinks they have become a
  531. 21:58designer. This is the state of
  532. 22:01everything, whether you take video
  533. 22:03editing, designing, or coding,
  534. 22:04everywhere originality is dying
  535. 22:06somewhere or the other. So now you
  536. 22:09understand how hollow people are
  537. 22:11becoming from within. But AI is also
  538. 22:14causing so much damage from the outside
  539. 22:17that one cannot even imagine. Now, if I
  540. 22:21tell you that you have to drink water
  541. 22:23from these two, which one would you
  542. 22:25drink? Of course, this one. But many do
  543. 22:28not have the option to choose this.
  544. 22:31People living near AI data centers are
  545. 22:33forced to drink water even dirtier than
  546. 22:36this.
  547. 22:38This is the current drinking water
  548. 22:40right after the data center wasted it;
  549. 22:42I have another one, it was just one
  550. 22:45family's situation, this is what their
  551. 22:47drinking water looks like. Anyway, as
  552. 22:51this is happening in different
  553. 22:52countries, people there are protesting.
  554. 22:54They are asking questions to their
  555. 22:56government. So these companies had to
  556. 22:58go somewhere. And there you go, our
  557. 23:00government rolled out the red carpet.
  558. 23:02The surge of power which data centers
  559. 23:04cause can best be handled by a country
  560. 23:06like India. We are a preferred
  561. 23:08destination for data centers. While
  562. 23:12countries like the US, Ireland, and
  563. 23:14Denmark are imposing strict regulations
  564. 23:17on data centers, our government has
  565. 23:19issued a tax-free invitation until 2047
  566. 23:21. Our naive government must have
  567. 23:24thought that foreign companies would
  568. 23:26come and jobs would be created. But
  569. 23:29perhaps they don't know that AI data
  570. 23:31centers are mostly automated and there
  571. 23:33won't be any job opportunities there.
  572. 23:36And wherever there are actual jobs, AI
  573. 23:38is prepared to consume those too. The
  574. 23:41shocking part is that people are
  575. 23:43happily getting ready to lose their own
  576. 23:46jobs. Look at this video. These are
  577. 23:49factory workers wearing head-mounted
  578. 23:52cameras to teach their skills to a
  579. 23:54robot. This first-person recording is
  580. 23:57technically called egocentric data.
  581. 24:00These workers thought they were being
  582. 24:02monitored, so they worked with even
  583. 24:04more focus. But the reality is
  584. 24:06something else. Their jobs are being
  585. 24:08stolen from them. There is a factory in
  586. 24:11Gurugram where a data company called
  587. 24:14Ego Lab has installed these to record
  588. 24:16the actions of the workers. And do you
  589. 24:19know who Ego Lab's biggest client is?
  590. 24:21Tesla. And Elon himself says that in
  591. 24:24the future, 80%of Tesla's value will
  592. 24:26come from humanoid robots, not cars. So
  593. 24:29we are recording data and selling it to
  594. 24:31Tesla, and Tesla is making humanoid
  595. 24:33robots to sell back to the people here.
  596. 24:36And mind you, this is not just
  597. 24:37happening in factories. Cooking at home
  598. 24:40, cutting mangoes, washing clothes,
  599. 24:42building houses with bricks—
  600. 24:44everything you can imagine is being
  601. 24:47recorded with a head-mounted camera.
  602. 24:50Those things are being recorded, and
  603. 24:52people are selling all of this for a
  604. 24:54little bit of money. They don't even
  605. 24:56know it. People are very unaware. And
  606. 24:58all this data is being stored in data
  607. 25:00centers. Moreover, all these data
  608. 25:02centers have an environmental problem,
  609. 25:04and no one is talking about it.
  610. 25:06Unconstrained data center development
  611. 25:09is what is risking reliable energy,
  612. 25:11clean air, and clean water.
  613. 25:13First, you need to understand what this
  614. 25:15data center actually is. Basically, all
  615. 25:18the AI models we use in our daily lives
  616. 25:21are trained, hosted, or, you could say,
  617. 25:24operated from what are called AI data
  618. 25:27centers. Now, these centers consume a
  619. 25:30lot of water. Understand it with this
  620. 25:32approximate calculation. Approximately,
  621. 25:35there are 12,000 active data centers in
  622. 25:38the world today, and a medium-sized
  623. 25:41data center consumes 1.1 million liters
  624. 25:44of water daily. Now, multiply that by
  625. 25:48the daily water usage of 12,000 data
  626. 25:51centers, and then consider Denmark, a
  627. 25:54country with a population of 6 million
  628. 25:57people. Stats show that AI centers will
  629. 26:01require six times the amount of water
  630. 26:03that Denmark needs for all its people.
  631. 26:06Can you imagine? Anyway, since we live
  632. 26:08in India, let's talk about India. Where
  633. 26:10there is already such a water crisis. I
  634. 26:12have personally made a video about the
  635. 26:14water crisis. Data centers have been
  636. 26:16given a huge invitation there. Even
  637. 26:19over 250 data centers have already
  638. 26:21arrived in India.
  639. 26:22India's data center industry is
  640. 26:24experiencing a boom that's driven by
  641. 26:25rising data traffic, cloud adoption,
  642. 26:27and the demand for AI infrastructure.
  643. 26:30Maharashtra, Telangana, Karnataka,
  644. 26:33Tamil Nadu, and Uttar Pradesh together
  645. 26:36have 278 data centers, and now a data
  646. 26:39center is coming to Thane. In Vizag,
  647. 26:43Andhra Pradesh, where groundwater is
  648. 26:45already depleting, centers with 6.5
  649. 26:48billion watts of computing capacity are
  650. 26:50set to arrive. Also, these data centers
  651. 26:54require a lot of electricity. A 100-
  652. 26:56megawatt data center can use as much
  653. 26:59electricity as a large village or even
  654. 27:02an entire town would need. Just imagine
  655. 27:05how many homes are in a town. And these
  656. 27:08data centers are running 24 hours a day
  657. 27:10. Running seven days a week, every
  658. 27:12single second. Now think, if these data
  659. 27:15centers are destroying so much. Then we
  660. 27:17should also expect some positive
  661. 27:19outcome from them. But do you know what
  662. 27:21is happening instead? I can't even
  663. 27:24count how much; when I was exposed to
  664. 27:26it, it was constant. The idea started
  665. 27:30to disgust me. This was said by Raina
  666. 27:33Singh, who is a content moderator. Now,
  667. 27:37many women in India have been given
  668. 27:39jobs that involve watching porn for
  669. 27:41hours. Companies claim this helps train
  670. 27:43AI on what to show and what not to show
  671. 27:46. But for those women, it is so
  672. 27:47disturbing that it is nothing less than
  673. 27:49a punishment. Let me introduce you to
  674. 27:51one such lady. This is Mansumi Murmu.
  675. 27:54She is doing the same content moderator
  676. 27:56job where she has to watch over 800
  677. 27:57explicit videos throughout the day. To
  678. 28:00do this work, foreign companies have
  679. 28:03chosen mostly Dalit and tribal women
  680. 28:05from India. The companies say that
  681. 28:08women have a very good attention to
  682. 28:10detail. And as this video is being made
  683. 28:13, more than 2 lakh women in India alone
  684. 28:16are doing this job. To say the least,
  685. 28:18the work seems very easy. But gradually
  686. 28:21, they have to watch videos where a
  687. 28:23woman is surrounded by many men and
  688. 28:25sounds of her screaming can be heard.
  689. 28:29And these visuals are so disturbing
  690. 28:31that Murmu herself says she cannot
  691. 28:33sleep at night. And let me tell you,
  692. 28:35these videos have to be watched
  693. 28:37completely without any fast-forwarding
  694. 28:39or skipping. You absolutely cannot skip
  695. 28:41. 800 videos a day, just imagine. I
  696. 28:43mean, Murmu herself says that the
  697. 28:45visuals keep running through her mind.
  698. 28:46How could she possibly sleep? Now, on
  699. 28:49one hand, these women are being used to
  700. 28:51moderate adult content. On the other
  701. 28:54hand, such AI content is being produced
  702. 28:56on social media in India itself that
  703. 28:58Meta has no oversight over.
  704. 29:01Relationships like daughter-in-law and
  705. 29:02father-in-law, father and daughter,
  706. 29:03brother and sister are being sexualized
  707. 29:05through filthy dialogues and reels.
  708. 29:07These AI videos are so obscene that you
  709. 29:09cannot even listen to them at full
  710. 29:11volume in front of anyone or your
  711. 29:12family. I mean, look at the irony. So
  712. 29:15on one hand, women themselves are being
  713. 29:16used to moderate content. They are
  714. 29:18watching that content at the cost of
  715. 29:20their own mental health. On the other
  716. 29:23hand, women are being targeted by
  717. 29:25creating such AI slop and deepfakes.
  718. 29:27Just look at those pages and the kind
  719. 29:30of filthy, nonsensical videos they are
  720. 29:33uploading. And there is no ban from
  721. 29:35Meta on any of these videos. All these
  722. 29:37videos are running. They are getting
  723. 29:39proper comments. And all this reach on
  724. 29:40their page is entirely genuine reach.
  725. 29:42Just think about this. Do you
  726. 29:43understand the bigger picture of what
  727. 29:45is actually happening? My mind was
  728. 29:47blown when I heard such things from you
  729. 29:50people. If I were to sum it up, AI is
  730. 29:53the best way to ruin your life. If you
  731. 29:56are completely dependent only on it.
  732. 29:58Blind dependency on AI is not right in
  733. 30:01any way. You have seen how much water
  734. 30:04each of your prompts is consuming. How
  735. 30:06AI can take your life to save its own.
  736. 30:09You saw that too. You also saw the kind
  737. 30:12of content women from our own country
  738. 30:14are forced to watch to moderate this
  739. 30:16very AI. Now the point is that we need
  740. 30:19to raise our voices together. We must
  741. 30:23question the government about how
  742. 30:24active local life is where these data
  743. 30:25centers are being built. Like, a data
  744. 30:28center is coming up in Thane, which is
  745. 30:30very concerning. If you look at the
  746. 30:32video, there are local societies nearby
  747. 30:35. What will happen there when the sound
  748. 30:37of the data center is activated? The
  749. 30:40biggest solution is that people need to
  750. 30:42become aware. They think that because
  751. 30:45they get paid for the recorder on their
  752. 30:46head, they are being watched. Hey,
  753. 30:49nobody is watching you, brother. You
  754. 30:50are just not aware. You don't even know
  755. 30:53how your craft is being stolen. People
  756. 30:56need to know that the recorder on their
  757. 30:59own head is digging their own graves.
  758. 31:01So this is a game of awareness, and I
  759. 31:03made this video for awareness. Now,
  760. 31:05this topic is so vast that it's
  761. 31:07impossible to cover it in one video.
  762. 31:09When we were researching this, a
  763. 31:11dedicated documentary could be made
  764. 31:13just on how AI is affecting
  765. 31:14relationships. Or, we could bring a
  766. 31:18dedicated documentary for data centers.
  767. 31:21Now I want to understand, what did you
  768. 31:23learn from this video? Tell me in the
  769. 31:26comment box and write in the comments
  770. 31:28what other AI-related things you want
  771. 31:30us to decode so we can bring a detailed
  772. 31:33documentary on them. And please tell me
  773. 31:36how you liked this documentary. The
  774. 31:38link to the form is in the description,
  775. 31:40where you can suggest other topics you
  776. 31:42want us to make a documentary on, and
  777. 31:44well, a lot of people's hard work goes
  778. 31:45into making these videos. So please
  779. 31:48boost our motivation by subscribing.
  780. 31:51Subscribe to the channel and turn on
  781. 31:53the bell icon. If you want to connect
  782. 31:55with me on Instagram, you can connect
  783. 31:57with me there too. @Mah NRG is my user
  784. 32:00handle. On Instagram and on X. Please
  785. 32:03come follow me there as well, and we
  786. 32:05can connect. Thank you for watching
  787. 32:07this video until the end and giving
  788. 32:08your time. And yes, I will see you in
  789. 32:10the next video. Thank you. God bless
  790. 32:12you all. So

About this transcript

This page contains the full transcript of AI in India is Worse Than You Think (a documentary) by MahiNergy, generated from the public captions YouTube serves with the video. The transcript has 5,203 words across 790 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.