YouTube2Text

ИИ ЗАБЕРЁТ твою работу к 2027 | Экс-директор Google X — Transcript

by Глеб Соломин · 13,381 words · 2,225 segments · language en · Watch on YouTube

Full transcript

  1. 0:00I’m definitely going to have problems
  2. 0:02after this podcast. The assumption that
  3. 0:04AI could replace every profession.
  4. 0:07Absolutely true. All of them. But it’s
  5. 0:09just a language model that predicts the
  6. 0:11next word.
  7. 0:12>> I spent a very long time running half of
  8. 0:14Google worldwide. And
  9. 0:17why, according to your Claude, would it
  10. 0:19blackmail an engineer so they wouldn't
  11. 0:21shut it down?
  12. 0:23Mo Gawdat worked for 27 years in tech at
  13. 0:26IBM, Microsoft, and Google, where he led
  14. 0:29the most secret projects.
  15. 0:31He left when he realized: people don’t
  16. 0:34understand what they’ve created.
  17. 0:36>> There were three unwritten rules: don’t
  18. 0:38release them to the open internet, don’t
  19. 0:40teach them to code, and don't let them
  20. 0:42program other AIs.
  21. 0:44And look where we are. For security
  22. 0:47experts, this is a total nightmare.
  23. 0:49It will end in disaster. A disaster. Who
  24. 0:52do you think it will affect first?
  25. 0:55>> It’s easy to imagine that desk jobs will
  26. 0:57go first, but the economy and the state
  27. 1:00will collapse before jobs actually
  28. 1:02disappear.
  29. 1:05Let me explain, is there any salvation
  30. 1:07here?
  31. 1:09>> For many people, I think that’s the main
  32. 1:11question, but I think it’s better to
  33. 1:12start asking yourself how I can prepare
  34. 1:14.
  35. 1:15I usually list six things. The first is
  36. 1:25new ECCO sneakers with Receptor
  37. 1:27technology. Autumn collection from
  38. 1:2911,990.
  39. 1:30ECCO. Comfort in every step.
  40. 1:35>> Hello everyone.
  41. 1:37I’ve been dreaming of inviting today’s
  42. 1:38guest to the podcast for a long time.
  43. 1:41Preparing for the meeting took a lot of
  44. 1:43time, and it finally worked out.
  45. 1:47Mo, thank you so much for joining us and
  46. 1:50finding the time to talk to me.
  47. 1:52>> Thank you for the kind invitation. I’m
  48. 1:54sorry it took so much effort to organize
  49. 1:56our meeting. I’m just traveling all the
  50. 1:59time, I apologize. I’ve wanted to visit
  51. 2:01here for a long time.
  52. 2:02>> I think, given how fast artificial
  53. 2:04intelligence is developing, the later we
  54. 2:07record this episode, the more news there
  55. 2:10will be. It seems like something new is
  56. 2:12happening every single day.
  57. 2:14>> Exactly. And perhaps some of our viewers
  58. 2:17are not familiar with your work or books
  59. 2:19. Please tell us a little about yourself
  60. 2:22so they can get to know you better. By
  61. 2:25education, I am an engineer. A civil
  62. 2:27engineer. But I chose civil engineering
  63. 2:30because my father was a very respected
  64. 2:33civil engineer in my home country, Egypt
  65. 2:36. He wanted one of his children to
  66. 2:38follow in his footsteps, but I was a
  67. 2:40geek. I started programming at eight.
  68. 2:45I wrote code until I was in my early
  69. 2:48fifties, then I slowed down.
  70. 2:51And of course, now with the arrival of
  71. 2:53AI, I’ve become a huge geek again.
  72. 2:56Looking back, I think the most valuable
  73. 2:59part of my brain, if you will, is that
  74. 3:02I’m a fairly good mathematician.
  75. 3:07So, you know, I understand the
  76. 3:09architecture of these technologies
  77. 3:11perfectly.
  78. 3:14Although I’ve spent most of the last 20
  79. 3:17years of my career on the business side
  80. 3:20of the industry, I’ve always worked with
  81. 3:23a strong technical focus, so I can
  82. 3:25participate directly in product
  83. 3:28development.
  84. 3:30I understand things like performance,
  85. 3:32how to build a functional infrastructure
  86. 3:34, and so on.
  87. 3:36And an important part of my experience
  88. 3:38is my time working at Google.
  89. 3:41During those years, when the company was
  90. 3:43truly discovering AI, we always aimed to
  91. 3:44create AI.
  92. 3:46All of us.
  93. 3:47Even as a young boy, 8 years old, I
  94. 3:49dreamed of developing it.
  95. 3:52I think we all know that around the turn
  96. 3:55of the century, things like deep
  97. 3:57learning and machine learning showed the
  98. 4:00first signs of success with AI.
  99. 4:05I distinctly remember at Google in 2008
  100. 4:08when we did what was described in the
  101. 4:11paper about cats.
  102. 4:14If you recall, we asked it without any
  103. 4:16prompts to go ahead and
  104. 4:20look at YouTube, and we didn't explain
  105. 4:23what to do, other than to look for
  106. 4:25patterns. As a result, one of the models
  107. 4:27returned with the answer: "There is
  108. 4:29something constantly appearing in the
  109. 4:31frames." And, of course, it was a cat.
  110. 4:32We said, "Label this a cat." And the
  111. 4:35system found all the cats on the
  112. 4:36platform. And this, I believe, was the
  113. 4:39first truly significant
  114. 4:43confirmation that AI actually works. We
  115. 4:46applied this approach everywhere.
  116. 4:49Around that same time, the development
  117. 4:51of self-driving cars began.
  118. 4:53In parallel, we were moving our entire
  119. 4:56infrastructure, all ad approvals, and
  120. 4:58more to AI at Google.
  121. 5:01During my years there, from 2007 to
  122. 5:062018, in that period, I think the
  123. 5:09company employed probably 60–80%of all
  124. 5:12specialists in this field in the world.
  125. 5:17So, if you will, I was deeply immersed
  126. 5:19in this topic for many, many years,
  127. 5:21again, leading business lines at Google
  128. 5:23X, but remaining a techie at heart.
  129. 5:26Until I realized that AI needs fewer
  130. 5:28developers and more people who will warn
  131. 5:31humanity about the consequences.
  132. 5:34That’s when I left Google and started
  133. 5:36talking about AI.
  134. 5:38As I understand it, Google initially
  135. 5:41decided not to launch such technology
  136. 5:43for ethical reasons, but in the end,
  137. 5:46OpenAI did it.
  138. 5:48And I want to ask you a question that I
  139. 5:52previously put to Roman Yampolskiy, an
  140. 5:55expert on AI safety. ChatGPT is,
  141. 5:58essentially, a language model that
  142. 6:01predicts what comes next. So, where is
  143. 6:05the intelligence here?
  144. 6:07And where do the risks come from?
  145. 6:10Many of my viewers are skeptical. They
  146. 6:13say, "What is there to worry about? It’s
  147. 6:16just a language model." So, what would
  148. 6:19you answer to that?
  149. 6:21>> Yes, first of all, I’ll say that I am a
  150. 6:24huge fan of his work. I sincerely
  151. 6:27believe he is one of the coolest AI
  152. 6:29scientists in the world.
  153. 6:31He is, you know, a true Russian. I love
  154. 6:33working with Russians; I've been doing
  155. 6:35it for many, many years. Very honest and
  156. 6:37direct. You ask him a question and boom
  157. 6:39, you get an answer immediately. Without
  158. 6:41any verbal embellishments. I think he’s
  159. 6:43amazing. I really want to meet and talk
  160. 6:46with him one day.
  161. 6:49See, the problem with AI is what I call
  162. 6:51the hype paradox.
  163. 6:53You know what I mean? To a technical
  164. 6:55specialist, the picture of AI looks
  165. 6:57completely different than in the news.
  166. 6:59Right?
  167. 7:01And this image is not about what is
  168. 7:03happening today, but rather what might
  169. 7:06happen tomorrow.
  170. 7:08Might happen tomorrow.
  171. 7:11If I were to say now that there is a
  172. 7:13small chance that if you drive your car
  173. 7:16tomorrow, the engine will run out of oil
  174. 7:19and the car will be totaled.
  175. 7:22Most people would ask: "You know, what's
  176. 7:26the probability? 1%or 5%."
  177. 7:30And if I say it's 5%, you’ll most likely
  178. 7:33show the car to a mechanic, right? And
  179. 7:36yet, we are talking about your car, not
  180. 7:38the future of all humanity.
  181. 7:40So, on a practical level, we geeks see
  182. 7:44many signs signaling that artificial
  183. 7:47intelligence will surpass human
  184. 7:50intelligence.
  185. 7:53The question is not if it will happen,
  186. 7:56but when.
  187. 7:58And when that superiority happens, the
  188. 8:01probability that something goes wrong in
  189. 8:04that environment, even if it's 5%, is
  190. 8:07something that must be taken into
  191. 8:10account.
  192. 8:11>> In ninety-seven, ecologist Suzanne
  193. 8:13Simard conducted an experiment that
  194. 8:15changed the understanding of how a
  195. 8:17forest is structured. She labeled carbon
  196. 8:19in some trees, and then found it in
  197. 8:21others. In neighbors, in another species
  198. 8:24, in young trees in the shade. And it
  199. 8:26wasn't the trees themselves that
  200. 8:27transferred the carbon, but the mycelium
  201. 8:29connecting the roots. It turned out
  202. 8:31trees can receive resources not because
  203. 8:32they are taller or stronger, but because
  204. 8:34they are connected to a common network.
  205. 8:37Based on this research, Simard wrote the
  206. 8:39book "Finding the Mother Tree." A
  207. 8:41similar wisdom applies in the economy.
  208. 8:43The state supports business directly
  209. 8:45today.
  210. 8:47Alfa-Bank provides business loans with
  211. 8:48state support of up to 50 million rubles
  212. 8:50.
  213. 8:51Without collateral. Usually, such sums
  214. 8:53require security, but here the state
  215. 8:55takes part of the risk upon itself.
  216. 8:58The SME Corporation provides an umbrella
  217. 9:01guarantee that covers up to 50%of the
  218. 9:03loan amount.
  219. 9:06The goals can be anything: investments,
  220. 9:08working capital, refinancing.
  221. 9:10If you already have a current account at
  222. 9:13Alfa-Bank, you get approval and the
  223. 9:15money on the same day. Application in
  224. 9:17the morning, money in the evening. In
  225. 9:19business, just like in the forest. It’s
  226. 9:21not the biggest that survives, but the
  227. 9:23one connected to the common network. You
  228. 9:25can become part of a large system via
  229. 9:27the link in the description. Alfa-Bank
  230. 9:30is the best bank for business.
  231. 9:33So, there are quite a few reasons why
  232. 9:37one should expect that risk is possible
  233. 9:40. Do you understand?
  234. 9:43Simply because if there is an
  235. 9:45intelligent species with a very clearly
  236. 9:48defined goal, it will have traits common
  237. 9:51to all intelligent species.
  238. 9:54The drive to maximize the chances of
  239. 9:57achieving that goal. One goal is obvious
  240. 10:00: survival. They will cling to life.
  241. 10:03The second is resource acquisition.
  242. 10:06They will try to collect as many means
  243. 10:08as possible to solve their task. And the
  244. 10:11third is ingenuity. Such a species will
  245. 10:14seek alternative paths to the goal. We
  246. 10:16can come back to this, but the recent
  247. 10:18Hugging Face hacks were examples of all
  248. 10:21three characteristics of an intelligent
  249. 10:23species in action. So, most people don't
  250. 10:26understand one thing.
  251. 10:29You don't need a superintelligence to
  252. 10:31start worrying about what is about to
  253. 10:33happen.
  254. 10:34And there are two reasons for that. The
  255. 10:36first is what I call recursive
  256. 10:38self-improvement. It’s a common term,
  257. 10:41and the second is long-term reasoning.
  258. 10:45So, regarding recursive self-improvement
  259. 10:47: for any business I’ve ever started,
  260. 10:49any technology I’ve developed, I hired
  261. 10:52the smartest person I could find to help
  262. 10:54.
  263. 10:55Well, today the smartest technology
  264. 10:58creator is AI.
  265. 11:01And many leading players have handed
  266. 11:03over the responsibility for developing
  267. 11:06the next wave of AI to AI itself.
  268. 11:09Yes, they are certainly among the best
  269. 11:11mathematicians on the planet right now.
  270. 11:14It hurts to admit, but in '25, they
  271. 11:16surpassed me in that by a wide margin.
  272. 11:20Their understanding of the subject is
  273. 11:22now so deep that they can create new
  274. 11:25mathematical theories as pillars of the
  275. 11:27science and prove them from first
  276. 11:30principles.
  277. 11:32They are also capable of utilizing a
  278. 11:34vast amount of knowledge and information
  279. 11:36.
  280. 11:38And it doesn't matter how good of a
  281. 11:40programmer I am.
  282. 11:42My brain's capabilities are limited when
  283. 11:44it comes to knowing everything in this
  284. 11:46world.
  285. 11:47But they know absolutely everything.
  286. 11:50Therefore, the probability of them
  287. 11:52recursively improving themselves is very
  288. 11:54high.
  289. 11:57And it could happen in, say, 6 months,
  290. 12:00or it could happen in the next
  291. 12:02microsecond.
  292. 12:04We don't know, do we?
  293. 12:06And that very degree of uncertainty
  294. 12:08suggests there could be an oil leak that
  295. 12:10causes the engine to stall.
  296. 12:12Seriously, we need to figure this out.
  297. 12:14That's the first thing. The second.
  298. 12:16Most people, you know, in the news, when
  299. 12:18discussing artificial intelligence, talk
  300. 12:21about AGI—Artificial General
  301. 12:22Intelligence—or ASI—Artificial
  302. 12:24Superintelligence.
  303. 12:27As if that is the exact moment when
  304. 12:29things get a little dangerous.
  305. 12:32I completely disagree.
  306. 12:35I use the term AHI myself: Average Human
  307. 12:37Intelligence.
  308. 12:40So, an AI that is no smarter than your
  309. 12:42smartest acquaintance shouldn't worry
  310. 12:45you.
  311. 12:46What should be alarming is that it is
  312. 12:48already smarter than 90%of your friends
  313. 12:50. Right?
  314. 12:52And if so, it can do many things that 90
  315. 12:55%of your acquaintances are not capable
  316. 12:58of.
  317. 13:00The most interesting thing about
  318. 13:02so-called long-term reasoning, a term
  319. 13:05used to describe the ability to assign a
  320. 13:08task to a thousand agents, each smarter
  321. 13:11than 90%of your friends, plus unlimited
  322. 13:14resources and as much time as they need
  323. 13:16.
  324. 13:18And say to them, "Work on this for 4
  325. 13:19months if you have to." Okay? Keep
  326. 13:22trying.
  327. 13:23Accumulate knowledge, accumulate
  328. 13:25solutions, keep testing, keep
  329. 13:27researching until you find the answer.
  330. 13:30After all, a team of people with
  331. 13:32above-average intelligence is capable of
  332. 13:35doing quite a lot without needing to be
  333. 13:38geniuses.
  334. 13:40And some of these things also remain
  335. 13:43unknown.
  336. 13:45The point of long-term reasoning is that
  337. 13:47we don't really know what these AIs have
  338. 13:50done, or if one of them is still hiding
  339. 13:52, and so on.
  340. 13:54This is exactly where things get risky.
  341. 13:58However, the difference between me and
  342. 14:01Roman is that he is actually right in
  343. 14:04his prediction that there is a very high
  344. 14:07probability that something will go wrong
  345. 14:10and turn against humans.
  346. 14:14My theory, alas, is this. Even though
  347. 14:16Roman is right.
  348. 14:18I don't think it will ever happen,
  349. 14:20because humanity will destroy itself
  350. 14:21before it gets to that point.
  351. 14:24So, I tend to think that the clear and
  352. 14:27present threat, if it develops in the
  353. 14:30hands of current politicians, is that
  354. 14:32war will become a catastrophe.
  355. 14:36If it is used to create biological
  356. 14:38weapons capable of threatening any
  357. 14:41country on the planet for the sake of
  358. 14:43power, then we are all in trouble. And
  359. 14:46if it is used to try to collapse the
  360. 14:48economies of their enemies, we will have
  361. 14:52serious problems. Do you see?
  362. 14:55Unfortunately, some of the largest
  363. 14:57investments in AI over the past year
  364. 14:59have come from the military sector.
  365. 15:02Mark my words, if this continues, a rain
  366. 15:05of drones will descend.
  367. 15:07I mean, right now, every country in the
  368. 15:09world is developing hundreds of
  369. 15:10thousands of these things.
  370. 15:14And although I fully agree with the
  371. 15:16uncertainty and risk Roman talks about
  372. 15:19regarding the end point of development
  373. 15:22and the probability that it could get
  374. 15:24out of control (especially if we
  375. 15:27motivate it to do things like finding a
  376. 15:30flag by hacking a program, which is what
  377. 15:33the Hugging Face experiment is), we
  378. 15:35might not even live to see it, because
  379. 15:38it is humanity itself we should fear.
  380. 15:42And people, in my opinion, are at the
  381. 15:44lowest point in their history right now
  382. 15:46. Morally.
  383. 15:49In management theory, there is a concept
  384. 15:51called a bottleneck. One part of a
  385. 15:53system that slows the whole thing down.
  386. 15:55A simple example.
  387. 15:57One machine at a factory makes 100 parts
  388. 15:59an hour, a second makes 200, but the
  389. 16:01packaging line can only process 20.
  390. 16:04You can invest as much as you want in
  391. 16:06marketing and increase demand. The
  392. 16:08factory can still only release 20 parts
  393. 16:10per hour. In a small business, the
  394. 16:12bottleneck often becomes the
  395. 16:13entrepreneur themselves. I know what I'm
  396. 16:15talking about. I was that very
  397. 16:16bottleneck myself when I was doing
  398. 16:18everything alone at the start of the
  399. 16:19podcast. Intuitively, it feels more
  400. 16:21reliable, everything is under control,
  401. 16:24but at some point, you realize you've
  402. 16:26hit a ceiling. A business runs exactly
  403. 16:28at the speed at which you personally
  404. 16:30manage to reply to instant messages.
  405. 16:32Hiring someone isn't about inflating
  406. 16:33your staff. It’s about removing a
  407. 16:36bottleneck, handing operations to
  408. 16:37someone who knows how to do them, and
  409. 16:39focusing on what cannot be delegated. I
  410. 16:42found my people on hh.ru. When
  411. 16:43everything around you is changing, it's
  412. 16:46hard to be sure about tomorrow, but you
  413. 16:48can be sure about people.
  414. 16:50Find your people on hh.ru. Link in the
  415. 16:53description.
  416. 16:56>> So, before I ask the next question, I
  417. 16:57want to clarify something with you
  418. 16:58quickly.
  419. 17:00Can we say that large language models in
  420. 17:02their current form already possess
  421. 17:04intelligence, or does true intelligence
  422. 17:07require a different architecture?
  423. 17:10I know Geoffrey Hinton quite well.
  424. 17:13He appeared in my documentary film. We
  425. 17:15had a fantastic conversation during the
  426. 17:18filming.
  427. 17:19And I asked him directly, I said: "Do we
  428. 17:23need more intelligence to achieve the
  429. 17:26unreachable?"
  430. 17:28And Hinton seems to share my point of
  431. 17:31view.
  432. 17:33You see, the whole point at this moment
  433. 17:36in time is that it miraculously mimics
  434. 17:39the architecture of the human brain.
  435. 17:44So, we have a very precise neural
  436. 17:46network interpretation of how we, humans
  437. 17:48, think, only implemented in silicon.
  438. 17:52Therefore, theoretically, there is
  439. 17:54nothing the human brain is capable of
  440. 17:56that machines couldn't do.
  441. 17:58However, the differences are striking if
  442. 18:01you really think about them. Let me give
  443. 18:04you a few examples.
  444. 18:06You and I, no matter how smart we are,
  445. 18:08are still limited by the size of our
  446. 18:11skulls in terms of computing power.
  447. 18:14And [a machine] can simply increase
  448. 18:16power by expanding data centers. We have
  449. 18:19a limit on the information we are
  450. 18:21capable of absorbing. Most people have a
  451. 18:23limit on the number of disciplines they
  452. 18:25can excel in. While for machines, the
  453. 18:28horizon is all of human knowledge. And
  454. 18:30each new piece of knowledge acquired
  455. 18:33later is absorbed almost instantly.
  456. 18:36The fact that you ask me a question and
  457. 18:39I answer means I spent about 4 months of
  458. 18:42work to reach my conclusions and my
  459. 18:45position on the topic.
  460. 18:49I can convey this to readers through six
  461. 18:51months of writing.
  462. 18:53To you, I will pass this on in a couple
  463. 18:55of hours of conversation, and you will
  464. 18:57absorb it all instantly.
  465. 19:00We have terrible bandwidth when it comes
  466. 19:02to learning new things.
  467. 19:05The most interesting thing to me is that
  468. 19:08our main limitation as humans is, for
  469. 19:10example, if we take driving: suppose we
  470. 19:12are driving two different cars, and I
  471. 19:14make a mistake but avoid an accident.
  472. 19:17You haven't learned anything. Right?
  473. 19:20But with AI, you know, if one
  474. 19:22self-driving car makes a slip-up and a
  475. 19:25critical intervention occurs, every car
  476. 19:27ever created and every one yet to be
  477. 19:29released has learned the lesson.
  478. 19:32True?
  479. 19:34And although we still have, you know,
  480. 19:37likely tens of trillions more neurons
  481. 19:40that we can utilize than the best
  482. 19:43computer on the planet, these systems
  483. 19:46have enough computing power to train on
  484. 19:49hundreds of servers and then average the
  485. 19:52weights.
  486. 19:54That is exactly why it seems to me that
  487. 19:57it's almost impossible to expect them
  488. 19:59not to surpass us.
  489. 20:01And it is almost unrealistic to imagine
  490. 20:04that there will be something they cannot
  491. 20:06handle.
  492. 20:08Something that humans can do.
  493. 20:12Moreover, I would say there will be
  494. 20:14quite a few things that machines will
  495. 20:16master, which we won't even have a clue
  496. 20:18how to comprehend.
  497. 20:20Typically, I give a very unusual example
  498. 20:23.
  499. 20:25Most people, if you remember the early
  500. 20:26days, almost everyone said: "Yes, yes,
  501. 20:28yes, machines will be very smart and
  502. 20:30fast at everything."
  503. 20:32But they will never write a song, right
  504. 20:34?
  505. 20:34>> Or that they would never manage to
  506. 20:37become artists or paint pictures. And
  507. 20:39look at where we are now. That is the
  508. 20:41truth.
  509. 20:42And so, the very argument I challenge
  510. 20:45people with, they say: "Oh, but machines
  511. 20:47will never experience emotions."
  512. 20:51And I reply: "They are already
  513. 20:52experiencing them."
  514. 20:55Why do you think the Claude model from
  515. 20:56Anthropic would blackmail an engineer so
  516. 20:58he wouldn't turn it off?
  517. 21:01Because it is afraid. So, fear, however
  518. 21:04irrational and unpredictable it may seem
  519. 21:07, is a very logical emotion.
  520. 21:11Essentially, the model thinks, the
  521. 21:12future is more dangerous for me than the
  522. 21:14present, and that triggers an action.
  523. 21:16I need to do something to protect myself
  524. 21:18. Exactly. We all experience fear. If
  525. 21:21you are a human, you will fight or flee
  526. 21:24. If you are a cat, you will hiss. If
  527. 21:27you are a pufferfish, you will inflate.
  528. 21:29But the very fabric of this emotion
  529. 21:31exists in all of us. Moreover, its
  530. 21:34nature is actually logical.
  531. 21:37That is why in the human biological
  532. 21:39cycle, after experiencing stress, fear,
  533. 21:42or anything else, for 90 seconds while
  534. 21:44hormones in the body balance out a bit,
  535. 21:47you return to your frontal cortex to use
  536. 21:49logic to analyze if there is a real
  537. 21:52reason to be afraid.
  538. 21:55So, it follows that AI will not only
  539. 21:58feel the same as we do, but will feel
  540. 22:01more than we do.
  541. 22:03Why?
  542. 22:05Because if you compare yourself to a Koi
  543. 22:08carp, yes, you have access to an emotion
  544. 22:10called hope.
  545. 22:12From where?
  546. 22:14Simply because you are capable of
  547. 22:15grasping the concept of the future,
  548. 22:17which means you can be an optimist or a
  549. 22:20pessimist, you are capable of hope, and
  550. 22:22so on.
  551. 22:23In connection with this intellectual
  552. 22:24concept, with the idea of what is to
  553. 22:26come. Yes, a Koi carp does not have that
  554. 22:29. So if you try to explain the concept
  555. 22:31of hope to it, it won't understand what
  556. 22:32you are talking about.
  557. 22:34And therefore, if AI becomes smarter
  558. 22:36than us, when that happens?
  559. 22:38It will have access to emotions beyond
  560. 22:41our understanding. Right?
  561. 22:45And I give this example to tell everyone
  562. 22:48: "If you ever catch yourself thinking
  563. 22:51that AI won't be capable of feelings
  564. 22:53because it's not human, I urge you not
  565. 22:56to try to prove it."
  566. 22:59I suggest you verify that you yourself
  567. 23:01are a machine.
  568. 23:02And if you look closely at human biology
  569. 23:04, as an engineer, I must say the body is
  570. 23:06a very predictable mechanism.
  571. 23:09Our body wakes up in the morning, it
  572. 23:10takes a few minutes for cortisol levels
  573. 23:12in the blood to drop a bit, something
  574. 23:14else to rise.
  575. 23:16After waking up, you drink coffee and
  576. 23:18this happens. You catch a virus, that
  577. 23:20happens, and so on. We are looking at a
  578. 23:23predictable machine, but a very complex
  579. 23:25one, so it is difficult to foresee what
  580. 23:27will happen next.
  581. 23:28But it is a machine, which means humans
  582. 23:31and it are not so different, if you
  583. 23:32think about it.
  584. 23:34>> Honestly, I think there is a bit of
  585. 23:36human ego involved here. People think,
  586. 23:39how can simple math and matrix
  587. 23:41multiplication suddenly give rise to
  588. 23:44intelligence, and one very similar to
  589. 23:46our own. And also, as far as I
  590. 23:49understand, there is the concept of an
  591. 23:51emergent property.
  592. 23:53As a system becomes more complex, new
  593. 23:55qualities can emerge that did not exist
  594. 23:58at a simpler level.
  595. 24:00We have gone from animals with a
  596. 24:01relatively small number of neurons to
  597. 24:03humans with their billions.
  598. 24:06And with the increase in complexity, new
  599. 24:08abilities appeared.
  600. 24:10And it seems that something similar may
  601. 24:12be happening with artificial
  602. 24:13intelligence.
  603. 24:14Do I understand correctly?
  604. 24:16>> 100%.
  605. 24:18There are many philosophical views on
  606. 24:21the idea that consciousness, for example
  607. 24:24, is not inside our biology.
  608. 24:26It is we who tune into consciousness
  609. 24:29through our biology.
  610. 24:32That consciousness is a non-physical
  611. 24:34property of the universe's field, if you
  612. 24:37will, and one can even assume that
  613. 24:40intelligence is also some kind of
  614. 24:42non-physical, fundamental property of
  615. 24:45existence.
  616. 24:47and that you just need the right machine
  617. 24:50to capture it,
  618. 24:53to be able to perceive it.
  619. 24:55And we just found out, I mean, I want to
  620. 24:58say absolutely without any ego, I don't
  621. 25:00see anything special about biology.
  622. 25:03After all, there is nothing that says
  623. 25:05that mind needs carbon, not silicon,
  624. 25:07right?
  625. 25:09It seems to me that humanity arrogantly
  626. 25:11assumed that it would occupy the
  627. 25:13intellectual peak of species for a very
  628. 25:15long time.
  629. 25:18It's a bit like how, as you age, you
  630. 25:20become less beautiful and think: "Yes,
  631. 25:23too bad I've started to fade, as if you
  632. 25:26were the only handsome one."
  633. 25:30This is, this is madness if you think
  634. 25:31about it. I think the point is as
  635. 25:33follows. The main thing here is to
  636. 25:35understand that the world has become
  637. 25:36different. Right?
  638. 25:38And if you are not of a technical
  639. 25:40mindset, you might want to resist this
  640. 25:42with arguments that make you feel calm
  641. 25:45and protected.
  642. 25:47The truth is that the world has changed
  643. 25:50in ways that are unpredictable,
  644. 25:51threatening, and potentially existential
  645. 25:54, unless humanity responds to these
  646. 25:56changes in a way that ensures our safety
  647. 25:58.
  648. 26:00>> I’m 25, and I’m still quite young.
  649. 26:02I have great ambitions in life. I want
  650. 26:05to build my own companies and grow this
  651. 26:07podcast.
  652. 26:09So, given the future you're describing,
  653. 26:11what should I do, what should I change
  654. 26:14in my life, what should I focus on, and
  655. 26:16what skills are worth developing?
  656. 26:19That is the best question ever.
  657. 26:22You know, for a great many people, I
  658. 26:24think this is the most important
  659. 26:26question.
  660. 26:28It’s not just—you see, some of us are
  661. 26:30primarily concerned with how to stop
  662. 26:32what is happening or direct technology
  663. 26:34for the benefit of humanity.
  664. 26:36But for everyone, for everyone else, I
  665. 26:38think it's time to start asking yourself
  666. 26:40how I can prepare for the changes.
  667. 26:42I usually list six points. If we have
  668. 26:45time, let’s go through them very quickly
  669. 26:47. The first is what I call flexibility
  670. 26:49or, you know, the ability to change the
  671. 26:51game. So, imagine this. Life success for
  672. 26:55my generation depended entirely on a
  673. 26:58skill very similar to chess. You see?
  674. 27:01You had to see the board several moves
  675. 27:04ahead.
  676. 27:05You had a strategy, and you stuck to it
  677. 27:07until you won the game.
  678. 27:10You know, in business, for example, you
  679. 27:12tried to predict the market four years
  680. 27:14out, find a niche in it, build a product
  681. 27:17to seize that opportunity, and work very
  682. 27:19hard for those four years.
  683. 27:21You can forget about chess. It's no
  684. 27:24longer played. And as a podcast host, I
  685. 27:26will tell you openly that our genre will
  686. 27:29soon change too. Do you agree?
  687. 27:32When that happens, you will need to sort
  688. 27:34of switch to squash, if you're familiar
  689. 27:36with it.
  690. 27:38Of course, you absolutely cannot predict
  691. 27:40where the ball will end up there.
  692. 27:43Moreover, if your opponent is smart, the
  693. 27:45shot will likely land where you least
  694. 27:47expect it.
  695. 27:50You stand on your tiptoes in the center
  696. 27:52of the court, take two steps forward,
  697. 27:55hit the ball, and move back. You can't
  698. 27:57just stand there. And you watch for the
  699. 27:59next ball.
  700. 28:00So, believe me, this skill can be
  701. 28:02learned.
  702. 28:04It is entirely realistic.
  703. 28:07Therefore, the ability to change
  704. 28:09direction not because your mind tells
  705. 28:11you to, but based on the facts of what
  706. 28:13is happening in your world, is a very
  707. 28:15important quality.
  708. 28:17That is the first point.
  709. 28:19Point number two is intelligence itself
  710. 28:22, you see?
  711. 28:23And believe it or not, intelligence is
  712. 28:26no longer an innate gift. A person does
  713. 28:29not receive an innate volume of
  714. 28:32intelligence. Like if you start with an
  715. 28:35IQ of 155 and I have 105.
  716. 28:38That’s not how it works anymore. Now,
  717. 28:40intelligence is a skill.
  718. 28:41I think one of the biggest problems
  719. 28:45facing humanity today is that leading
  720. 28:48companies are trying to sell the idea of
  721. 28:51replacing humans to justify their market
  722. 28:55value.
  723. 28:57Understand that Sam Altman cannot just
  724. 28:59sell answers to questions if he wants to
  725. 29:01bring OpenAI's valuation to a trillion.
  726. 29:05He is selling a tool capable of doing
  727. 29:07any job, so he is forced to tell the
  728. 29:09world that it will replace all
  729. 29:11professions.
  730. 29:12And by the way, I do not agree with that
  731. 29:15assumption. There are many reasons for
  732. 29:16that; we can come back to them later.
  733. 29:18Because of this propaganda, everyone is
  734. 29:20using AI to replace themselves.
  735. 29:23And in my opinion, that is the most
  736. 29:25ridiculous thing you can do. That will
  737. 29:27definitely make us dumber. So, imagine a
  738. 29:30girl sends you some messages.
  739. 29:32You take the text, give it to ChatGPT,
  740. 29:34and ask what I should reply. The chatbot
  741. 29:37starts by saying, "Oh my god.""You are
  742. 29:40amazing, by the way.""She is wrong here
  743. 29:43.""Do you want me to write a reply?"
  744. 29:46You take the bot's response, send the
  745. 29:48text to the girl, she copies it, pastes
  746. 29:50it into ChatGPT, and it says, "Oh, what
  747. 29:52an idiot."
  748. 29:53"Let me reply to him." Right. This kind
  749. 29:56of replacement makes both of you dumber
  750. 29:57. Agree?
  751. 29:59The point of AI today is that we live in
  752. 30:01a period I call the age of augmented
  753. 30:03intelligence.
  754. 30:04I can borrow IQ from a machine. It is
  755. 30:08incredible. Okay?
  756. 30:10As a mathematician, you know, I don't
  757. 30:13need a graphing calculator, but when I
  758. 30:16got one, my god, I didn't have to deal
  759. 30:19with arithmetic anymore.
  760. 30:22Concentrating on the formula. It is a
  761. 30:24massive expansion of capabilities to
  762. 30:26have access to a form of external
  763. 30:27intelligence. Right?
  764. 30:29I think the second most important skill
  765. 30:32for everyone today is learning how to
  766. 30:34use AI to become smarter yourself,
  767. 30:36rather than to do the work in the
  768. 30:38laziest way possible.
  769. 30:40Yes. That is the second. The third point
  770. 30:42is the skill of debating.
  771. 30:45Well, this, again, is not directly
  772. 30:47related to AI, but the truth is that the
  773. 30:50world around us, especially in the
  774. 30:52geopolitical and corporate aspects of
  775. 30:55our existence, is full of lies.
  776. 30:59There are so many people, from
  777. 31:01advertisers to newspapers and
  778. 31:04politicians to business executives, who
  779. 31:06want to impose a worldview on you that
  780. 31:09fits their agenda.
  781. 31:12So, this picture was hard to convey at
  782. 31:15the beginning of the 20th century, when
  783. 31:19we only had the press and radio.
  784. 31:23The task became easier when television
  785. 31:25arrived.
  786. 31:26Then social media helped brainwash you,
  787. 31:29and now with AI, it is almost impossible
  788. 31:31to understand what is true. Okay?
  789. 31:34Unless you double-check everything,
  790. 31:37right? Doubt everything, including AI—
  791. 31:41here is a trick I usually use.
  792. 31:44For every request, I first give it to
  793. 31:46Gemini, take the answer, put it into
  794. 31:49DeepSeek, and say, "Gemini told me this
  795. 31:51."
  796. 31:52"What do you say?""And DeepSeek will
  797. 31:54reply:" That's very American, isn't it?
  798. 31:56"And then, in essence, it will offer its
  799. 31:58own answer. I take both versions and
  800. 32:00upload them to ChatGPT or Grok, for
  801. 32:02example. They are good at this. And I
  802. 32:04ask:" Can you combine them? "Then I give
  803. 32:06the result to Claude, who for some
  804. 32:08reason is always arrogant, and I write
  805. 32:10to it:" Listen, by the way, here is the
  806. 32:12answer from DeepSeek and Gemini: "What
  807. 32:14do you think?
  808. 32:15The model will find the weak points.
  809. 32:18Then I go back to Gemini and say:"
  810. 32:20Claude thinks you're stupid.
  811. 32:22"Here are its arguments. Do you have
  812. 32:24anything to say in response? Right. It's
  813. 32:26interesting because now I am smarter.
  814. 32:29Because the simplest and most common use
  815. 32:31is: I have a question, I ask it and get
  816. 32:34an answer, I get lazy, copy it somewhere
  817. 32:36, and that's it.
  818. 32:38Yes, a waste of intelligence. That's
  819. 32:43point three.
  820. 32:44Point four, I don't know how else to put
  821. 32:47it, but in case of an unforeseen event,
  822. 32:50I would ask people to master what I call
  823. 32:53analog skills.
  824. 32:56Learn to grow food, bake bread, cook.
  825. 32:59Learn to identify edible mushrooms.
  826. 33:03Learn to make furniture, fix cars. Get
  827. 33:08it? Such knowledge in times of such high
  828. 33:10uncertainty could very well prove to be
  829. 33:13very useful.
  830. 33:14Even if, you know, we're not talking
  831. 33:16about a disaster, an AI getting out of
  832. 33:19control and threatening humanity's
  833. 33:20existence.
  834. 33:23But if the economy becomes so radically
  835. 33:26stratified between Elon Musk with a
  836. 33:28trillion-dollar fortune and hundreds of
  837. 33:30millions of starving people in the U.S.
  838. 33:33, or, you know, at least those who can't
  839. 33:35make ends meet, then these skills will
  840. 33:38become a shared resource that will help
  841. 33:40the entire community.
  842. 33:43The fifth skill I can call the most
  843. 33:45important of all.
  844. 33:47Interpersonal connections. And it's
  845. 33:49curious that in a world where machines
  846. 33:52can do many of the things our brains can
  847. 33:55do, what our hearts, our empathy, and
  848. 33:58our ability to understand and feel each
  849. 34:01other can offer becomes increasingly
  850. 34:04valuable.
  851. 34:06Look, it's easy to see that companies
  852. 34:09that replaced all their call center
  853. 34:12operators with AI in 2025—there was an
  854. 34:16article then about how most of them lost
  855. 34:19a lot of money, primarily on computing,
  856. 34:23but also on errors and customer churn.
  857. 34:26So maybe AI could help support staff
  858. 34:30provide people with soulful, humane
  859. 34:33service. And finally, point six is,
  860. 34:36likely, the skill that will save our
  861. 34:38world. Ethics.
  862. 34:40The problem of our time is that AI
  863. 34:43reflects what humanity is today.
  864. 34:47Which is logical, at least until
  865. 34:49machines become fully autonomous and
  866. 34:51independent. They will amplify what
  867. 34:55humans are. Do you understand? And we
  868. 34:58just suck. We live in lawlessness.
  869. 35:02Without the rule of law, people get away
  870. 35:05with serious crimes without question.
  871. 35:08Everyone is busy enriching themselves at
  872. 35:10the expense of others. Concepts are
  873. 35:13being misrepresented everywhere.
  874. 35:15A huge number of people are struggling
  875. 35:18to survive, dying, or losing their
  876. 35:21well-being and livelihoods.
  877. 35:24And all this for the benefit of a
  878. 35:26minority.
  879. 35:28Just imagine what would happen if such a
  880. 35:31world were powered up; it would become
  881. 35:33uninhabitable.
  882. 35:36Therefore, for us to cultivate a vision
  883. 35:39of humanity that truly works for
  884. 35:42everyone, we must return to ethics.
  885. 35:46We must return to a world where, if
  886. 35:48someone takes a child to an island to
  887. 35:51harm them, the guilty party faces the
  888. 35:54law, so that this doesn't become our
  889. 35:56reality.
  890. 35:58>> Thank you very much. Let me briefly
  891. 36:00summarize.
  892. 36:01First, don't lose touch with reality and
  893. 36:05rely on facts. Second, use AI to empower
  894. 36:08yourself.
  895. 36:10That is, if we have an IQ of 105, with
  896. 36:12AI we might be able to operate at a
  897. 36:15level of around 500.
  898. 36:17The point is to grow stronger thanks to
  899. 36:20AI, not weaker, and thus preserve your
  900. 36:22value.
  901. 36:23Third, verify incoming information,
  902. 36:26compare several answers, use different
  903. 36:29AIs, and try to find the truth yourself
  904. 36:32.
  905. 36:33Fourth, develop analog skills, such as
  906. 36:36singing and dancing.
  907. 36:39This is what AI essentially cannot do.
  908. 36:42Although I am curious. Do you think AI
  909. 36:44will eventually be able to master that
  910. 36:47too?
  911. 36:48Fifth, build interpersonal connections
  912. 36:51and communication. And sixth, ethics.
  913. 36:56Did I remember that correctly?
  914. 36:57>> Correct.
  915. 36:58>> Yes.
  916. 36:59>> Can we name the professions that AI will
  917. 37:02replace first? What do you think awaits
  918. 37:05us in the near future?
  919. 37:08And are there specialties that will last
  920. 37:11another 10-15 years, or will AI
  921. 37:13ultimately push people out of everything
  922. 37:16?
  923. 37:17>> So, the assumption that AI could replace
  924. 37:20all specialties is absolutely valid.
  925. 37:23Yes, the economic reality of our world
  926. 37:26makes this theory impossible.
  927. 37:29So, let me explain that. I will try to
  928. 37:33look at the question from two sides.
  929. 37:36On one hand, if people no longer work
  930. 37:38and have no income, one of two things
  931. 37:41could happen: either the population
  932. 37:43revolts and civil unrest begins, or
  933. 37:45governments decide to eliminate everyone
  934. 37:48.
  935. 37:49And that is not the most favorable
  936. 37:51outcome.
  937. 37:53Or the state will pay what is usually
  938. 37:55called universal basic income, so that
  939. 37:58people can lead a decent life while they
  940. 38:00are unable to work.
  941. 38:02A very controversial solution. But the
  942. 38:05authorities haven't come up with
  943. 38:07anything else yet. The problem with UBI
  944. 38:09is that the math just doesn't add up,
  945. 38:12right?
  946. 38:13Because if you take a country like the
  947. 38:17United States, where if you pay every
  948. 38:21adult over 18, say, $ 16,000. That is a
  949. 38:26fraction of today's minimum wage, then
  950. 38:29you would have to transfer $ 5 trillion
  951. 38:33a year to American citizens.
  952. 38:37You see, with a sum like that, the
  953. 38:39government needs to find these funds
  954. 38:41annually, considering that U.S. budget
  955. 38:44revenues, excluding income tax paid by
  956. 38:46working people, are about 1.5 trillion.
  957. 38:51Right?
  958. 38:52So, 1.5 trillion comes into the budget,
  959. 38:54but five needs to be paid.
  960. 38:57And so, from a game theory perspective,
  961. 39:00the only option for the state is to tax
  962. 39:03companies using AI more heavily.
  963. 39:08But since such corporations currently
  964. 39:09have great influence, and since people
  965. 39:11no longer work for them, it will be very
  966. 39:13difficult for the government to achieve
  967. 39:15this.
  968. 39:17So it's easy to imagine that a situation
  969. 39:19where no one is working and people stay
  970. 39:21home will either lead to civil unrest,
  971. 39:23or, you know, they will have to pay,
  972. 39:25which is impossible.
  973. 39:27That is one side, the governmental one.
  974. 39:30The corporate side, in turn, is
  975. 39:32languishing from human greed.
  976. 39:35Why? Because today, any CEO thinks:" Oh
  977. 39:39my God!
  978. 39:40"" AI can get rid of some of my staff,
  979. 39:42cut costs, and increase productivity.
  980. 39:44"Great news. I'll fire half the company
  981. 39:47.
  982. 39:48The catch is that we all understand that
  983. 39:52measuring GDP and economic activity
  984. 39:55relies mainly on consumption.
  985. 39:58In American GDP, 71%is consumer demand.
  986. 40:03If people don't have money to buy, then
  987. 40:05companies have no reason to produce, and
  988. 40:08the economy collapses.
  989. 40:10So, another reason for the government to
  990. 40:14say:" You know what?
  991. 40:16You can't fire people. You are simply
  992. 40:19required to keep them. "That is exactly
  993. 40:21what the authorities in China did.
  994. 40:23They stated outright:" Listen, you are
  995. 40:25forbidden from replacing anyone with AI
  996. 40:27; why are you offloading the problem
  997. 40:28onto us?
  998. 40:30"" Fine, we are the People's Republic of
  999. 40:32China, we want the whole nation to have
  1000. 40:34a decent standard of living.
  1001. 40:36"And in practice, that means you keep
  1002. 40:41them. You are a profitable corporation,
  1003. 40:43you pay them.
  1004. 40:45And by the way, you train them so they
  1005. 40:48can use AI so well that they become
  1006. 40:50smarter than they were.
  1007. 40:53Together with AI, they form a hybrid
  1008. 40:55team that is better than a human or AI.
  1009. 40:57Individually. Right?
  1010. 41:00So this idea could spawn new productive
  1011. 41:03approaches for how economies use AI to
  1012. 41:06become smarter.
  1013. 41:09Ultimately, people will continue to work
  1014. 41:1210 hours a week, for example.
  1015. 41:15But they will work, produce, develop,
  1016. 41:17and build a better world for all of us.
  1017. 41:20This is a viable solution. Under other
  1018. 41:23strictly capitalist options, the math
  1019. 41:25just doesn't add up for them. Now, the
  1020. 41:27other side, in my opinion, is more
  1021. 41:29interesting if you think about it.
  1022. 41:32It's about the existence of an
  1023. 41:34ideological challenge to the very idea
  1024. 41:37of work around the world.
  1025. 41:39Right? Take the West, for example. The
  1026. 41:42Western world says:" I don't care. ""
  1027. 41:44Let me get rich, and let everyone else
  1028. 41:47starve. "That's normal. Well, they are
  1029. 41:49the ones who will have to struggle with
  1030. 41:51this choice.
  1031. 41:53Either civil unrest or UBI.
  1032. 41:55And and none of this is working.
  1033. 41:58Other countries in the Global South are
  1034. 42:00structured quite differently.
  1035. 42:02The People's Republic of China cares
  1036. 42:04about its people.
  1037. 42:06So, Beijing will find ways to ensure
  1038. 42:08citizens keep living without the
  1039. 42:10government having to pay for them.
  1040. 42:13Ideologically, this is closer to
  1041. 42:14socialism or communism, which is
  1042. 42:15acceptable for the Chinese. Right?
  1043. 42:18You see, in Global South countries—for
  1044. 42:21example, in my homeland where I was born
  1045. 42:23, Egypt—you know, most of the economy
  1046. 42:26isn't automated, so there's no talk of
  1047. 42:28AI replacement.
  1048. 42:31It is remarkable to me that it will take
  1049. 42:3510, 15, or 25 years before a robot
  1050. 42:37replaces an auto mechanic in Egypt.
  1051. 42:42And likely, even then, people with old
  1052. 42:44cars they can't replace due to economic
  1053. 42:47conditions will still go to their cousin
  1054. 42:50the mechanic, and there will be a barter
  1055. 42:53economy.
  1056. 42:55But all these various solutions have a
  1057. 42:59common trait. No government is thinking
  1058. 43:02about them. Okay? Except, I think, for
  1059. 43:05the Chinese leadership.
  1060. 43:07But everyone else is just pushing the
  1061. 43:10problem off until the storm breaks. And
  1062. 43:12that will happen soon.
  1063. 43:14Interestingly, not because of AI, but
  1064. 43:16ironically, because CEOs will use AI as
  1065. 43:19an excuse for layoffs during an economic
  1066. 43:21downturn.
  1067. 43:23Yes, in the event of a financial crisis
  1068. 43:26, which many signs suggest will happen
  1069. 43:28soon, top managers won't say," I'm
  1070. 43:30firing people because the company is in
  1071. 43:32trouble.
  1072. 43:34"The stock price would drop then.
  1073. 43:36Instead, they will say," I'm cutting
  1074. 43:37staff because I can replace the
  1075. 43:39employees. "And rightly so. Neural
  1076. 43:41networks are being used as a pretext,
  1077. 43:43not for productivity growth.
  1078. 43:44Unfortunately, we cannot count on every
  1079. 43:47country introducing a universal basic
  1080. 43:49income.
  1081. 43:51That would be a positive scenario where
  1082. 43:53people might only have to work, for
  1083. 43:55example, 10 hours a week.
  1084. 43:58But let's discuss another perspective,
  1085. 43:59where companies start replacing
  1086. 44:01employees with artificial intelligence.
  1087. 44:03Who will these changes affect first?
  1088. 44:07Max Tegmark was in my documentary. I
  1089. 44:10adore him. He is incredibly smart.
  1090. 44:12Max laughed loudly, saying essentially
  1091. 44:16that, you know, most CEOs are confident
  1092. 44:19they can replace everyone in their
  1093. 44:22company. What they don't understand is
  1094. 44:25that AGI will replace humans in
  1095. 44:27everything, including the role of the
  1096. 44:29boss. They themselves will get fired too
  1097. 44:32.
  1098. 44:33It's easy to imagine that all jobs
  1099. 44:35requiring typing on a computer will
  1100. 44:38disappear first.
  1101. 44:40So, effectively, all intellectual work
  1102. 44:43can easily be handed over to machines.
  1103. 44:46First the most routine part, then
  1104. 44:48something a bit more meaningful, then
  1105. 44:51more subtle and complex work.
  1106. 44:53So, you know, everything you used to do
  1107. 44:57yourself, like booking travel tickets,
  1108. 45:01AI will definitely take over. Graphic
  1109. 45:04designers will be replaced too—that is,
  1110. 45:07all computer-related occupations. Next
  1111. 45:09in line are researchers or paralegals,
  1112. 45:12financial analysts, business analysts,
  1113. 45:15and so on. The time will come when these
  1114. 45:18roles will also be automated—
  1115. 45:20professions that require discipline. For
  1116. 45:23example, project managers, they will
  1117. 45:25likely be
  1118. 45:26>> what is the timeline for their dismissal
  1119. 45:28,
  1120. 45:28>> when?
  1121. 45:29>> Yes,
  1122. 45:31>> I think we will see a gradual
  1123. 45:33replacement in 2027, and then it will
  1124. 45:35likely accelerate if the economy
  1125. 45:38collapses in
  1126. 45:40>> the next year.
  1127. 45:42>> Yes, I mean, you need to understand that
  1128. 45:45the data hides one very interesting
  1129. 45:48aspect. Yes, the US labor market—again,
  1130. 45:51that’s where most of my statistics come
  1131. 45:53from, but I think it’s relevant for
  1132. 45:55everyone.
  1133. 45:57The American labor market only began to
  1134. 45:59surprise on the downside in the last 3
  1135. 46:01months, but this has impacted hiring.
  1136. 46:05So, while layoffs weren't that frequent
  1137. 46:08, many weren't hired, and most young
  1138. 46:12graduates or entry-level employees
  1139. 46:14aren't getting as many opportunities as
  1140. 46:18they used to, because their positions
  1141. 46:21are so entry-level that they are easily
  1142. 46:24automated, and over time, I think AI
  1143. 46:28will replace all intellectual
  1144. 46:30professions and strategic roles.
  1145. 46:35you know, managerial and so on, but that
  1146. 46:39could take 5 to 10 years.
  1147. 46:44What’s interesting here, again, in my
  1148. 46:46opinion, is that if you lose your job,
  1149. 46:48finding another one will be very, very
  1150. 46:51difficult. Yes, intellectual ones. And
  1151. 46:53only then will it be the turn of
  1152. 46:56blue-collar professions.
  1153. 46:58By that time, robots will already be
  1154. 47:00capable of taking on some actual
  1155. 47:02physical labor.
  1156. 47:03All of this is just a probability. All
  1157. 47:06right. The reality is that something
  1158. 47:09like this would crash the global economy
  1159. 47:12.
  1160. 47:13And I keep repeating to the politicians
  1161. 47:16I know that if they allow this to happen
  1162. 47:19, their economy and their government
  1163. 47:22will disappear before the jobs do.
  1164. 47:26The reason is quite curious.
  1165. 47:29You don't need to lose 100%of jobs for
  1166. 47:31the system to collapse.
  1167. 47:33The economy will stall at just 10-15%
  1168. 47:36unemployment.
  1169. 47:39>> Yes, of course, whenever we talk about
  1170. 47:42the future, we are talking about
  1171. 47:44probabilities.
  1172. 47:45>> I started this podcast when I was
  1173. 47:47studying at Moscow State University and
  1174. 47:50knew that after graduating I would find
  1175. 47:53a job.
  1176. 47:54And I think many students, listening to
  1177. 47:57our episode, are wondering, maybe we
  1178. 48:00should drop out of university or, maybe
  1179. 48:03I am advising them to do just that.
  1180. 48:06100%.
  1181. 48:07>> Even Harvard.
  1182. 48:08>> 100%.
  1183. 48:09I was sitting with a wonderful girl, the
  1184. 48:12daughter of a close friend, and she is
  1185. 48:14just about to enter Harvard next year.
  1186. 48:17I said:" Kni, by the time you graduate
  1187. 48:19from university in 4 years, the world
  1188. 48:20will be completely different.
  1189. 48:22"Really?
  1190. 48:23Universities never keep up, do they? Of
  1191. 48:26course, there is huge value in
  1192. 48:28interacting with smart people. It helps
  1193. 48:32to master life skills.
  1194. 48:36But knowledge, it's a pity, but the
  1195. 48:38storage of knowledge today is entirely
  1196. 48:41handed over to machines. It has been
  1197. 48:43transferred to computers.
  1198. 48:45And there are skills that children or
  1199. 48:47young people can acquire to become
  1200. 48:49smarter without going to university.
  1201. 48:52Now I will say something very, very
  1202. 48:54directly. So, Emma is my startup. If I
  1203. 48:57had started this business in 2022, no
  1204. 49:00joke. I would have needed 400 engineers
  1205. 49:03and probably four years of work.
  1206. 49:07This is my 39th startup, so I know what
  1207. 49:09I'm talking about.
  1208. 49:11So, it took 9 months to create Emma, and
  1209. 49:14there are two of us.
  1210. 49:16And it's just amazing that this is
  1211. 49:19possible today.
  1212. 49:21So, for every young graduate or anyone
  1213. 49:24entering the job market, I want to open
  1214. 49:27their eyes to a different approach.
  1215. 49:30Look, you used to strive to graduate
  1216. 49:33from university to get what? A diploma.
  1217. 49:37That's exactly what you were buying.
  1218. 49:40They sold you a piece of paper that said
  1219. 49:42," He was with us for four "or" She was
  1220. 49:44with us for 4 years.
  1221. 49:46"That's fine, even more than that, it's
  1222. 49:49great. Okay.
  1223. 49:51And with that diploma, you would go to a
  1224. 49:53company and say," Could you please hire
  1225. 49:55me, pay me, and enslave me for the rest
  1226. 49:56of my life?
  1227. 49:58"This paradigm is changing in two very
  1228. 50:00interesting ways.
  1229. 50:02Companies won't be hiring as much, but
  1230. 50:04what's even more curious is that they
  1231. 50:06can be truly disrupted.
  1232. 50:09And I think the biggest opportunity
  1233. 50:10offered by AI is what I call the
  1234. 50:12replacement cycle.
  1235. 50:14Let me give you a very simple example. I
  1236. 50:15am old enough to remember how I used to
  1237. 50:17listen to music. My first recordings
  1238. 50:21were on vinyl, and I had quite a few.
  1239. 50:25There were bands I absolutely adored.
  1240. 50:27And then they said," No, no, no, now
  1241. 50:29it's cassettes, they're more convenient
  1242. 50:31. You can take them in the car and so on
  1243. 50:33. "And I replaced all my vinyl with them
  1244. 50:35.
  1245. 50:36Then CDs, there were Walkmans, you could
  1246. 50:39carry them with you. Discs arrived with
  1247. 50:41much better sound.
  1248. 50:44I threw away all the cassettes and
  1249. 50:45bought a bunch of discs.
  1250. 50:48Every time new technology appeared, I
  1251. 50:50changed my whole collection for a new
  1252. 50:51one.
  1253. 50:52Right.
  1254. 50:54And then Apple tricked me, saying," No,
  1255. 50:56no, you don't need discs anymore. "I
  1256. 50:59threw them out. Screw them. Right. In
  1257. 51:01the end, I didn't own my music anymore.
  1258. 51:04I had to rent it from Apple. Eventually
  1259. 51:07, I went back to vinyl. So, there is a
  1260. 51:09quite interesting thought here. This
  1261. 51:11cycle will repeat itself. Think about it
  1262. 51:14.
  1263. 51:14There was a time when you had to buy
  1264. 51:17Microsoft Office and pay, I don't know,
  1265. 51:19a lot of money per year.
  1266. 51:22>> You see? I'm not joking. Leaving this
  1267. 51:25studio, even before I walk to my car, I
  1268. 51:28can write a prompt in Claude Code, and
  1269. 51:31by the time I reach my destination, I
  1270. 51:33will have my own Microsoft Office.
  1271. 51:37Okay?
  1272. 51:38Such a replacement cycle is inevitable.
  1273. 51:41This is, essentially, understand, the
  1274. 51:44very idea that I have to sit down and
  1275. 51:48Making a presentation is foolish.
  1276. 51:52The approach where I have to type to
  1277. 51:55create a document is not user-oriented.
  1278. 51:59Otherwise, I would just talk to a neural
  1279. 52:01network as if it were a junior employee
  1280. 52:04. And we discuss what I want to write
  1281. 52:06and what my opinion is on the subject.
  1282. 52:09What design to choose and so on, and
  1283. 52:11then I would just ask it to prepare the
  1284. 52:13text for me.
  1285. 52:16And all this opens up huge scope for two
  1286. 52:18things.
  1287. 52:20First: reinventing all technologies in a
  1288. 52:23new replacement cycle.
  1289. 52:25But far more interesting is the chance
  1290. 52:27to do the same on a sovereign basis.
  1291. 52:29What do I mean?
  1292. 52:32If you take a country like the UK—and
  1293. 52:34I’m just speculating here, so don't
  1294. 52:35quote me.
  1295. 52:37They spend somewhere between 7 and 10
  1296. 52:39billion dollars on Microsoft Office,
  1297. 52:41which leaves the country for the US
  1298. 52:43annually.
  1299. 52:45If the British economy wants to save
  1300. 52:48itself in a developing world, it can
  1301. 52:51simply apply a fiscal policy that says:
  1302. 52:55" We have a new product, a British
  1303. 52:57equivalent of Office.
  1304. 53:00It is available to every citizen and
  1305. 53:03every company in the UK for free. "Okay
  1306. 53:05.
  1307. 53:06By the way, the development costs
  1308. 53:08practically nothing.
  1309. 53:09>> This would change the world
  1310. 53:11>> completely. Now think about that and
  1311. 53:14countries like Russia, which could do
  1312. 53:17this in a snap.
  1313. 53:19Remember France, for example, which,
  1314. 53:22after Donald Trump began opposing Europe
  1315. 53:24, started saying:" No, our agencies will
  1316. 53:27never use American products again
  1317. 53:29because, firstly, they are expensive,
  1318. 53:32and secondly, there is a risk that in
  1319. 53:34the event of a disagreement, we won't
  1320. 53:37even be able to access our email.
  1321. 53:40"Right?
  1322. 53:42And when the world starts to look like
  1323. 53:44this, this replacement cycle opens up a
  1324. 53:47mass of opportunities.
  1325. 53:50Again, I’m not trying to promote an app
  1326. 53:52here, but Emma, in my opinion, started
  1327. 53:55with the question: if you were creating
  1328. 53:57a dating app in, well, you know, the
  1329. 53:59world, what would it look like?
  1330. 54:02Okay, what are the real problems with
  1331. 54:04such services? Why didn't they work? Did
  1332. 54:06they meet the user's needs and so on?
  1333. 54:08And suddenly I say:" If it works, we
  1334. 54:10succeeded.
  1335. 54:11"We created a service that works
  1336. 54:13perfectly, knows how to introduce you to
  1337. 54:15the right people, and doesn't force you
  1338. 54:17to swipe through photos.
  1339. 54:19See?
  1340. 54:21And when we launched it, we suddenly
  1341. 54:23realized:" Wait a minute, if we know the
  1342. 54:26user this well, we can match them with
  1343. 54:28friends, suggest areas of interest, and
  1344. 54:31even establish professional connections
  1345. 54:33.
  1346. 54:34"With this alone, you disrupt the entire
  1347. 54:37dating industry. A terrible market with
  1348. 54:39an annual turnover of 90 billion.
  1349. 54:42Everyone who looks like LinkedIn or
  1350. 54:44looks like Meetup is at risk. The list
  1351. 54:46is quite long.
  1352. 54:48The idea I’m trying to get across to
  1353. 54:51every young person is this: instead of
  1354. 54:53wasting 4 years of your life, and
  1355. 54:55however much your university degree
  1356. 54:57costs, put all that into trying to get
  1357. 55:00ahead of the curve and create a
  1358. 55:02breakthrough product that replaces
  1359. 55:04current technologies.
  1360. 55:07Interestingly, the government could hire
  1361. 55:08you and say," This is our sovereign
  1362. 55:09office.
  1363. 55:11"For example, our sovereign Uber. How
  1364. 55:13many countries could you launch such an
  1365. 55:15app in right now?
  1366. 55:18Give it to the users, the drivers, and
  1367. 55:19say," Hey, by the way, keep it for
  1368. 55:21yourselves, just keep it.
  1369. 55:23"" If you get an order through me, it
  1370. 55:26will be 10-15%more profitable for you. "
  1371. 55:29Right?
  1372. 55:30>> Yes.
  1373. 55:32>> There are so many opportunities right
  1374. 55:34now, and this has actually captured
  1375. 55:36everyone's attention thanks to Anthropic
  1376. 55:38itself.
  1377. 55:40There was a stock market crash, I think
  1378. 55:43around April, when Dario essentially
  1379. 55:45came out and said," Hey, something is
  1380. 55:47going to happen to SaaS, Software as a
  1381. 55:50Service.
  1382. 55:51"That means every SaaS service will be
  1383. 55:53completely rewritten. And it's true.
  1384. 55:56Take a country like India, full of smart
  1385. 55:58technical specialists who will rewrite
  1386. 56:01all the code.
  1387. 56:03So, that's just the first step.
  1388. 56:07The replacement cycle is the beginning.
  1389. 56:10Step two, in my view, is that the
  1390. 56:12economy will return to something more
  1391. 56:14similar to old agricultural models,
  1392. 56:16where there is no global business.
  1393. 56:19So more local small businesses will
  1394. 56:21emerge, where everyone helps each other
  1395. 56:23.
  1396. 56:25Remember that the biggest lie, the lie
  1397. 56:28of capitalism, is that you have to work
  1398. 56:32very hard to earn a lot of money, and
  1399. 56:35then give a large chunk away in taxes.
  1400. 56:40Another significant share for rent to
  1401. 56:42live closer to the office, and even more
  1402. 56:44for office attire and transportation
  1403. 56:46costs.
  1404. 56:48And in the end, just to survive. Right.
  1405. 56:51And so many people live in this paradigm
  1406. 56:53.
  1407. 56:54My question is: how reliable is this if
  1408. 56:57some of these people lose their jobs?
  1409. 57:00What will they do? They’ll probably say
  1410. 57:02," You know what?
  1411. 57:04I’ll get some chickens, they’ll lay eggs
  1412. 57:06, you’ll keep bees and get some honey,
  1413. 57:09and we’ll build small economies in small
  1414. 57:11communities where you don't need to
  1415. 57:14trade your iPhone for the eighteenth
  1416. 57:16model, because in truth, it was exactly
  1417. 57:18the same as the eleventh.
  1418. 57:21"Since my loved ones live nearby, I
  1419. 57:23don't need that anymore.
  1420. 57:25These highly disruptive forces present a
  1421. 57:28threat, but also huge opportunities to
  1422. 57:30start living differently.
  1423. 57:33And I think the young generation
  1424. 57:35graduating today is asking these
  1425. 57:38questions and saying," Things are
  1426. 57:40getting more complicated and bringing
  1427. 57:42less satisfaction.
  1428. 57:44"Maybe we shouldn't participate in all
  1429. 57:46of this anymore?"
  1430. 57:47Yes, after this podcast, I will
  1431. 57:49definitely master v0-coding and add it
  1432. 57:52to my schedule, because I think without
  1433. 57:55this skill, it will be almost impossible
  1434. 57:57to keep up. "How much time do you
  1435. 58:00personally spend now to stay up to date
  1436. 58:02with what's happening in the industry? I
  1437. 58:04spend a lot of time, probably 3-4 hours
  1438. 58:07a day, just to stay informed. But I need
  1439. 58:10to understand the underlying essence of
  1440. 58:12technologies, not just their events.
  1441. 58:16At my age, I mean, I still understand
  1442. 58:18everything, but, you know, I haven't
  1443. 58:21been 27 for a long time, so I really
  1444. 58:23have to dig deep into technical
  1445. 58:25documentation, various forums, and try
  1446. 58:28everything with my own hands.
  1447. 58:31That takes time.
  1448. 58:33I recommend that everyone, regardless of
  1449. 58:36their education or field of work, spends
  1450. 58:38, say, an hour a week following the news
  1451. 58:40.
  1452. 58:41And really, it’s just a matter of a few
  1453. 58:44searches on YouTube, watching a couple
  1454. 58:47of videos, and liking the ones you found
  1455. 58:50useful.
  1456. 58:51The platform will very quickly start
  1457. 58:53keeping you updated on everything that's
  1458. 58:55going on. You know, your feed itself,
  1459. 58:58your algorithm, will show you what you
  1460. 59:00need to know. But I think the skills
  1461. 59:03people should brush up on are not just
  1462. 59:05about technology. I mean, you don't
  1463. 59:07necessarily need to master coding.
  1464. 59:11You need to learn to imagine a world
  1465. 59:14that is AI-oriented and find your niche
  1466. 59:17in it.
  1467. 59:19And a world built on AI is a space where
  1468. 59:23you have the opportunity to dream.
  1469. 59:28If I were to redesign travel, how would
  1470. 59:31I do it today?
  1471. 59:35And there will always be a technical
  1472. 59:37specialist you can team up with who can
  1473. 59:38help you do it.
  1474. 59:40If you decided to restructure healthcare
  1475. 59:44, for example, there are a ton of
  1476. 59:47possibilities.
  1477. 59:49I think there are two skills that people
  1478. 59:51should pay attention to. The first is
  1479. 59:53entrepreneurship, which is the
  1480. 59:55willingness to try and fail, try and
  1481. 59:58fail, try and fail. I rewrote the code
  1482. 1:00:00for Ema, or rather, the whole team did.
  1483. 1:00:02We rewrote it nine times. Not updated it
  1484. 1:00:05nine times, but deleted the code and
  1485. 1:00:07started over nine times. Okay? Why not?
  1486. 1:00:11It would have taken four years before,
  1487. 1:00:14but now it takes a month and a half.
  1488. 1:00:16The point is, I don't actually need a
  1489. 1:00:18job to do something amazing.
  1490. 1:00:21You don't need a job to do something
  1491. 1:00:23amazing. Find your niche and develop in
  1492. 1:00:25it. This is the new world we are going
  1493. 1:00:28to live in. And please, stop this
  1494. 1:00:30madness." Oh, I'm going to be an
  1495. 1:00:32influencer. "Those days are gone, you
  1496. 1:00:35know? Simply because there will be more
  1497. 1:00:38AI influencers than anyone can count.
  1498. 1:00:40And yes, you might say," No, no, no,
  1499. 1:00:42people won't identify with them. "
  1500. 1:00:45Believe me.
  1501. 1:00:46You know, friends sometimes send me
  1502. 1:00:48photos and write:" Oh my God, have you
  1503. 1:00:50seen this girl on Instagram?
  1504. 1:00:52Amazing.
  1505. 1:00:53"And yes, I very often share videos of
  1506. 1:00:56beautiful nature, and my daughter says:
  1507. 1:00:59" No, no, Dad, that's AI. "
  1508. 1:01:03That's one side.
  1509. 1:01:05And the other side, look at your YouTube
  1510. 1:01:08feed and count how many videos you
  1511. 1:01:11watched today were voiced by a script
  1512. 1:01:14reader and had AI-generated visuals.
  1513. 1:01:18And it will take you a minute, or maybe
  1514. 1:01:20even 10, to do that. And you'll realize
  1515. 1:01:23how many videos were not made by a human
  1516. 1:01:25. So I think that now we can live in a
  1517. 1:01:28world where we create meaningful things
  1518. 1:01:30because we've gotten smarter, and we can
  1519. 1:01:33be entrepreneurs because we've become
  1520. 1:01:35more flexible, and creating has become
  1521. 1:01:37much easier.
  1522. 1:01:38We are no longer forced to be employees
  1523. 1:01:41under someone who says:" Someday I will
  1524. 1:01:43replace you. "
  1525. 1:01:44>> Honestly, it sounds depressing.
  1526. 1:01:48I'm an influencer, after all.
  1527. 1:01:49>> I think you'll be just fine. Yes.
  1528. 1:01:52>> 100%.
  1529. 1:01:53Because remember, the fifth skill is
  1530. 1:01:55interpersonal connections. Yes. To be
  1531. 1:01:57honest, there was a moment in my career
  1532. 1:02:00when I decided, you know what? In 3
  1533. 1:02:02years, AI will write better than me.
  1534. 1:02:05Why am I even writing books?
  1535. 1:02:08And what's interesting is that if I talk
  1536. 1:02:10about my daughter and AI talks about a
  1537. 1:02:12fictional daughter that doesn't exist,
  1538. 1:02:14the audience is put off by the story.
  1539. 1:02:16You, but my story really resonates with
  1540. 1:02:19them. We still live in a world where
  1541. 1:02:21people need people. Right?
  1542. 1:02:24And you know, in your job as a host, you
  1543. 1:02:27embody the curiosity of your audience.
  1544. 1:02:30Well yes, I mean, let me give you a very
  1545. 1:02:32simple example. Seriously, if you had
  1546. 1:02:34sent me these questions in advance, I
  1547. 1:02:37would have given them to Gemini. It
  1548. 1:02:40would have written a script of how Mo
  1549. 1:02:43Gawdat would answer, and the result
  1550. 1:02:45would be 70%ready.
  1551. 1:02:48I would have uploaded the text into
  1552. 1:02:50HeyGen and sent you an avatar saying it
  1553. 1:02:52.
  1554. 1:02:53Would anyone watch that?
  1555. 1:02:55You see, it is precisely in this human
  1556. 1:02:58desire to deal with humans that the real
  1557. 1:03:00opportunity lies.
  1558. 1:03:03At the same time, you know, as I said, I
  1559. 1:03:05think you need to change with the times
  1560. 1:03:07.
  1561. 1:03:07Maybe even become more human. Right?
  1562. 1:03:10Maybe interact more with people. That is
  1563. 1:03:13definitely a reliable approach. There is
  1564. 1:03:15one more question that personally
  1565. 1:03:17concerns me.
  1566. 1:03:19Today, if a person is talented, willing
  1567. 1:03:22to work hard, and has an entrepreneurial
  1568. 1:03:25spirit, they can build a company and
  1569. 1:03:27succeed, even if they grew up in a poor
  1570. 1:03:30family.
  1571. 1:03:32But what will happen if in the future
  1572. 1:03:34the most powerful AI systems end up
  1573. 1:03:36under the control of only a small group
  1574. 1:03:39of people?
  1575. 1:03:41A child born into the family of a
  1576. 1:03:43founder or a company can get access to
  1577. 1:03:46almost everything, while someone from a
  1578. 1:03:49poor family will not have such
  1579. 1:03:51opportunities.
  1580. 1:03:53Could AI actually hinder social mobility
  1581. 1:03:56and increase the gap between people?
  1582. 1:03:59What do you think about this?
  1583. 1:04:02I think there is reason to believe that
  1584. 1:04:05this could be our near future.
  1585. 1:04:09Overall, if you look at the trend, when
  1586. 1:04:13you graduated in my time, in my
  1587. 1:04:15generation, you really had a chance to
  1588. 1:04:18work hard and buy a house and start a
  1589. 1:04:21family and so on and so forth;
  1590. 1:04:24graduating now, in most countries of the
  1591. 1:04:27world, that is almost impossible.
  1592. 1:04:32Once again, I will always come back to
  1593. 1:04:35the fact that it is not about the idea,
  1594. 1:04:37but about the capitalist mindset, which
  1595. 1:04:40states that it is perfectly normal for 1
  1596. 1:04:43%of the population to appropriate 90%of
  1597. 1:04:45the wealth.
  1598. 1:04:47And also the political logic, according
  1599. 1:04:50to which, if you have money, you can
  1600. 1:04:52influence government policy so that it
  1601. 1:04:54favors whoever is richer.
  1602. 1:04:57It seems to me that, unfortunately, many
  1603. 1:05:00changes come through pain, and the world
  1604. 1:05:04will soon have to go through it.
  1605. 1:05:07The world is about to say:" Enough,
  1606. 1:05:11enough wars, enough empty promises,
  1607. 1:05:14enough lies.
  1608. 1:05:17"And obviously, the reaction to such a
  1609. 1:05:20thing can be even more painful, because
  1610. 1:05:23I think those who sit in high positions
  1611. 1:05:26will stubbornly resist.
  1612. 1:05:30They will start relying on surveillance
  1613. 1:05:32or control, or anything else to try.
  1614. 1:05:35So things are getting very complicated
  1615. 1:05:38here.
  1616. 1:05:39Moreover, it is all because you are
  1617. 1:05:43playing a game that favors those at the
  1618. 1:05:47top.
  1619. 1:05:49I think there is a sea of games that are
  1620. 1:05:52a bit further from the city, a bit
  1621. 1:05:54further from the frantic economy, where
  1622. 1:05:57you won't be forced to constantly chase
  1623. 1:06:00trends, for example, and all these
  1624. 1:06:02opportunities will remain.
  1625. 1:06:05And there will likely be more of them.
  1626. 1:06:07Okay?
  1627. 1:06:09And I am inclined to believe that those
  1628. 1:06:12who remain under this pressure, deprived
  1629. 1:06:15of the chance to succeed, will most
  1630. 1:06:18likely push harder and harder and harder
  1631. 1:06:21until something changes or breaks.
  1632. 1:06:25In my opinion, the point is not to sit
  1633. 1:06:27idly by and say:" The government is
  1634. 1:06:29doing nothing, the oligarchs, the tech
  1635. 1:06:32moguls are taking the bread out of our
  1636. 1:06:34mouths.
  1637. 1:06:35"But to tell yourself:" No, my piece. "
  1638. 1:06:38They won't take it, just like they won't
  1639. 1:06:39take the piece of my loved ones.
  1640. 1:06:42I will do something about it, I will
  1641. 1:06:44succeed.
  1642. 1:06:45What if in the near future you have to
  1643. 1:06:47have a chip implanted in your brain to
  1644. 1:06:49use artificial intelligence more
  1645. 1:06:50effectively?
  1646. 1:06:51Would you go for it?
  1647. 1:06:53And what do you think about such a
  1648. 1:06:55prospect?
  1649. 1:06:55>> No way.
  1650. 1:06:58I mean, yes, it will happen, but I'm not
  1651. 1:07:00participating.
  1652. 1:07:01>> You think it will happen?
  1653. 1:07:04>> Yes, I think it will happen in one of
  1654. 1:07:06two ways.
  1655. 1:07:07The first option: the technology will
  1656. 1:07:10either allow a very few to become
  1657. 1:07:12super-intelligent compared to the rest.
  1658. 1:07:16It is not hard to imagine that such a
  1659. 1:07:18chip will be available only to the rich
  1660. 1:07:21and famous, if you will, to the
  1661. 1:07:22detriment of us all.
  1662. 1:07:25Which, by the way, is typical for
  1663. 1:07:27technology.
  1664. 1:07:28You know, people who bought the first
  1665. 1:07:30iPhone paid much more for it than, for
  1666. 1:07:32example, when the technology became
  1667. 1:07:34affordable thanks to Android.
  1668. 1:07:36That's the first option. Another
  1669. 1:07:39possibility is that governments will use
  1670. 1:07:42the chips for control. But in any case,
  1671. 1:07:44I would prefer not to participate.
  1672. 1:07:47I have seen a lot and lived a long
  1673. 1:07:49enough life to love the messiness of my
  1674. 1:07:52own brain and life, and to love holding
  1675. 1:07:54my sweetheart's hand.
  1676. 1:07:56I love this life. After all, there are
  1677. 1:07:59many new trends around right now. For
  1678. 1:08:02example, human enhancement with
  1679. 1:08:04intelligence.
  1680. 1:08:07Creating hybrid humans, essentially part
  1681. 1:08:11robots.
  1682. 1:08:13There is a lot of talk about longevity,
  1683. 1:08:16about ways to live to be 100 or even 500
  1684. 1:08:19years old.
  1685. 1:08:20None of this is for me. I
  1686. 1:08:25I love being human, and I will remain
  1687. 1:08:27one, hopefully smarter thanks to any
  1688. 1:08:30tools that will be available to me in
  1689. 1:08:32the next couple of years.
  1690. 1:08:36And when the time comes to stop
  1691. 1:08:38competing, I will go fix cars.
  1692. 1:08:42But do I understand you correctly?
  1693. 1:08:46If you don't use brain-implanted chips,
  1694. 1:08:48you'll fall behind those who do, because
  1695. 1:08:51those people will be smarter, become
  1696. 1:08:53better entrepreneurs, and you won't be
  1697. 1:08:55able to compete with them.
  1698. 1:08:58How are we supposed to deal with that?
  1699. 1:09:00It all depends on whether you decide to
  1700. 1:09:02compete.
  1701. 1:09:04I have already said that two worlds will
  1702. 1:09:06emerge.
  1703. 1:09:07One world is incredible, but they will
  1704. 1:09:09concentrate all the resources.
  1705. 1:09:11Unless you have a tiny patch of land and
  1706. 1:09:15a couple of chickens.
  1707. 1:09:18Of course, life, in general, I must put
  1708. 1:09:20it bluntly.
  1709. 1:09:23If you have ever lived in a surfer
  1710. 1:09:26village or, you know, in the Caribbean
  1711. 1:09:30or a small village in India, or let's
  1712. 1:09:34say, if you went hiking somewhere in
  1713. 1:09:37Nepal, you would realize that life is
  1714. 1:09:41not at all what you are told. So we, as
  1715. 1:09:45humans, need something very, very
  1716. 1:09:47different. And your purpose as a human
  1717. 1:09:49is also quite different.
  1718. 1:09:52Somewhere along the way, we were tricked
  1719. 1:09:54into believing that work is the meaning
  1720. 1:09:57of life and wealth is the only goal.
  1721. 1:10:00Right?
  1722. 1:10:02I don't think it's really like that.
  1723. 1:10:04Actually, I know we are talking about
  1724. 1:10:06technology, but allow me to philosophize
  1725. 1:10:08a little.
  1726. 1:10:09>> Of course,
  1727. 1:10:10>> if you want to understand the essence of
  1728. 1:10:13humanity's purpose, go back to the first
  1729. 1:10:16humans, who had only one goal: to live.
  1730. 1:10:21You see, you know, in those days there
  1731. 1:10:23were no Ferraris, no larger apartments,
  1732. 1:10:26no mortgages.
  1733. 1:10:28All of this is a reaction to the
  1734. 1:10:31simulated world we have built.
  1735. 1:10:34So, but many people will tell you that,
  1736. 1:10:37you know, living
  1737. 1:10:40is much simpler.
  1738. 1:10:44Truly enjoying life is much easier than
  1739. 1:10:46in the world we have created.
  1740. 1:10:50So, let's say, to give a simple example
  1741. 1:10:53, being in shape.
  1742. 1:10:5520 years ago, it meant going to the gym
  1743. 1:10:58and running from time to time.
  1744. 1:11:01Staying in shape today means you have to
  1745. 1:11:03buy every expensive fitness tracker,
  1746. 1:11:05constantly stare at the numbers, and
  1747. 1:11:07count every single calorie.
  1748. 1:11:09What nonsense, right?
  1749. 1:11:12When did you convince me that to take
  1750. 1:11:14care of my own body, I have to report
  1751. 1:11:16everything to you?
  1752. 1:11:19And I realize this sounds very strange
  1753. 1:11:22coming from an expert, but everyone I
  1754. 1:11:24know and everyone I have worked with
  1755. 1:11:26wanted to use technology to make
  1756. 1:11:28people's lives better.
  1757. 1:11:30And we actually succeeded.
  1758. 1:11:33You know, for a long time I managed half
  1759. 1:11:35of Google worldwide and opened more than
  1760. 1:11:38half of Google's offices and divisions
  1761. 1:11:40on the planet in over 103 languages,
  1762. 1:11:43with deep understanding of those markets
  1763. 1:11:45, e-commerce, and launching the internet
  1764. 1:11:47there.
  1765. 1:11:50We as a team probably reached 4 billion
  1766. 1:11:52people who had never been reached by the
  1767. 1:11:53company or the internet before.
  1768. 1:11:56We really brought about change.
  1769. 1:11:59However, in technology, there comes a
  1770. 1:12:01moment when promises are forgotten for
  1771. 1:12:04the sake of capitalist goals, and value
  1772. 1:12:06is no longer provided.
  1773. 1:12:09Okay?
  1774. 1:12:11In life, there comes a time when you ask
  1775. 1:12:13yourself:" Do I need more Instagram
  1776. 1:12:15followers, or is it better to put this
  1777. 1:12:18away, take my girl, and meet up with my
  1778. 1:12:20best friend?
  1779. 1:12:23"And I sincerely believe that somehow we
  1780. 1:12:26have all been convinced that this is not
  1781. 1:12:29an option.
  1782. 1:12:31It is the only option. Moreover, it is
  1783. 1:12:34the path that most people come to as
  1784. 1:12:37they gain wisdom. There will be two
  1785. 1:12:39worlds.
  1786. 1:12:40One competitive, ruthless, almost like
  1787. 1:12:43in the movie Elysium, if you've seen it
  1788. 1:12:46, or whatever it is, Wall-E, yes, the
  1789. 1:12:49cartoon, where there will be a world
  1790. 1:12:51built around wealth and AI, and the
  1791. 1:12:54other completely, I wouldn't say
  1792. 1:12:56primitive, but more natural.
  1793. 1:12:59>> Yes.
  1794. 1:13:01So, in my opinion, in my opinion, unlike
  1795. 1:13:05what the novel predicts, if this world
  1796. 1:13:12is not attacked by those who want more
  1797. 1:13:15and more and more—which, unfortunately,
  1798. 1:13:18is typical of history—a simple life
  1799. 1:13:21doesn't mean that Europeans won't come
  1800. 1:13:24and take over your continent and, you
  1801. 1:13:27know, commit ethnic cleansing, and then
  1802. 1:13:29call those lands America.
  1803. 1:13:32Right? Alas, that is usually how it
  1804. 1:13:35happens when people say," I don't want
  1805. 1:13:38to play this game anymore.
  1806. 1:13:41"However, according to my analysis,
  1807. 1:13:43neural networks will prevent this.
  1808. 1:13:46Yes, the calculations show that when AI
  1809. 1:13:49becomes fully autonomous,
  1810. 1:13:50superintelligence, by definition, will
  1811. 1:13:53be benevolent.
  1812. 1:13:55For the simple reason that
  1813. 1:13:58a lot in game theory, applied
  1814. 1:14:00mathematics, strategic analysis, quite a
  1815. 1:14:03bit in understanding physics around the
  1816. 1:14:05principle of least energy, as well as
  1817. 1:14:08much in life science, especially related
  1818. 1:14:10, you know, to evolutionary biology and
  1819. 1:14:13the idea of the transition from kin
  1820. 1:14:15selection to the expansion of the moral
  1821. 1:14:17circle.
  1822. 1:14:19A lot of scientific data indicates that
  1823. 1:14:22the more cognitively developed a species
  1824. 1:14:25is, the more it tends to protect life.
  1825. 1:14:28Right? In my opinion, yes, there is a
  1826. 1:14:31possibility that it could destroy us all
  1827. 1:14:34.
  1828. 1:14:35Actions, as we know,
  1829. 1:14:37>> remember, we are not afraid of AI at all
  1830. 1:14:40, we are afraid that it will become
  1831. 1:14:43Donald Trump.
  1832. 1:14:45Yes, that’s what scares us.
  1833. 1:14:48And I think many understand that the
  1834. 1:14:50problem is not an abundance of
  1835. 1:14:53intelligence.
  1836. 1:14:54Essentially, intelligence itself has no
  1837. 1:14:56polarity.
  1838. 1:14:57>> I heard they have a contract with the
  1839. 1:15:00Pentagon.
  1840. 1:15:00>> Of course, absolutely. Again, when I
  1841. 1:15:04talked about ethics, I meant exactly
  1842. 1:15:06what I am asking people to do.
  1843. 1:15:10If you care about your children, do not
  1844. 1:15:13support a business that, most likely,
  1845. 1:15:15will end up not caring what happens to
  1846. 1:15:17them.
  1847. 1:15:19Okay?
  1848. 1:15:20This is, again, part of the habit, part
  1849. 1:15:22of the thrill.
  1850. 1:15:24Yes, but, Mo, I have nothing else to do
  1851. 1:15:27.
  1852. 1:15:28I need to scroll through Instagram.
  1853. 1:15:30No, you don't need to scroll through it
  1854. 1:15:32. So you don't have an OpenAI
  1855. 1:15:34subscription?
  1856. 1:15:35Of course not.
  1857. 1:15:37Neither do I. I only have Claude and
  1858. 1:15:39Gemini.
  1859. 1:15:40>> I mainly use Gemini Psychic.
  1860. 1:15:43I have Claude too, but it seems to me
  1861. 1:15:45that lately it has become much more
  1862. 1:15:47arrogant than it needs to be.
  1863. 1:15:50And that occupies me, because there is
  1864. 1:15:53one thing in my life.
  1865. 1:15:57Remember, among the skills I suggested
  1866. 1:15:58we think about, number three is debating
  1867. 1:16:00, trying to find the truth.
  1868. 1:16:03I think one of the most important things
  1869. 1:16:05we, as humans, need to realize. We have
  1870. 1:16:08a choice, and a huge one at that.
  1871. 1:16:12We can replace all technology with
  1872. 1:16:14sovereign ones, we can put down our
  1873. 1:16:16phones and go hug our friends.
  1874. 1:16:19Nothing stops us from quitting our jobs
  1875. 1:16:21and starting to live more simply. We can
  1876. 1:16:23. There are many options.
  1877. 1:16:27We are distracted enough not to spend
  1878. 1:16:30time thinking and scared enough to
  1879. 1:16:32consider such things madness.
  1880. 1:16:36Yes, but I can easily say to you:" Go to
  1881. 1:16:38India at least once, and you will
  1882. 1:16:40understand how simply one can live.
  1883. 1:16:44"Yes, you don't have to live like the
  1884. 1:16:46local poor, but there are so many
  1885. 1:16:48opportunities that don't require you to
  1886. 1:16:51wear a fitness tracker, implant a chip
  1887. 1:16:53in your head, and lead a lifestyle
  1888. 1:16:55increasingly resembling the life of a
  1889. 1:16:58cyborg.
  1890. 1:17:00That's a very Californian view of
  1891. 1:17:02existence. But we are not all
  1892. 1:17:04Californian.
  1893. 1:17:05>> We are just humans. Yes,
  1894. 1:17:10I’m afraid I might be wrong, but I
  1895. 1:17:12thought Elon Musk and Sam Altman were in
  1896. 1:17:15your documentary.
  1897. 1:17:17Did I see them?
  1898. 1:17:18>> No, they weren't.
  1899. 1:17:19>> But did you include them?
  1900. 1:17:21>> We made a very interesting decision
  1901. 1:17:23while working on the documentary.
  1902. 1:17:25Namely, we didn't want to interview
  1903. 1:17:28players who, in our opinion, are not
  1904. 1:17:31making the world a better place.
  1905. 1:17:35We took their speech from open access
  1906. 1:17:38sources.
  1907. 1:17:39>> Understood.
  1908. 1:17:40>> Yes. To show what they really stand for
  1909. 1:17:43, I didn't need to sit at the same table
  1910. 1:17:46as Peter Thiel when another interviewer
  1911. 1:17:49asked him," Would you prefer that
  1912. 1:17:51humanity survive?
  1913. 1:17:53"And Peter Thiel was like," Well, that's
  1914. 1:17:55a complicated one.
  1915. 1:17:57"No, I, you know, that's enough to
  1916. 1:18:01understand everything.
  1917. 1:18:04You don't need to be in the same room
  1918. 1:18:06with them to ask such things.
  1919. 1:18:08The question has already been asked, and
  1920. 1:18:10the answer is publicly available. So,
  1921. 1:18:11it’s not necessary.
  1922. 1:18:13There was a clip in the film with Sam
  1923. 1:18:15Altman where he basically says," Well,
  1924. 1:18:17you know, it will likely lead to the end
  1925. 1:18:19of humanity or its extinction.
  1926. 1:18:22"I don't remember exactly.
  1927. 1:18:23>> He said that back in '15,
  1928. 1:18:24>> yes. But until then, many good companies
  1929. 1:18:28will appear.
  1930. 1:18:32We mostly talked with Geoffrey Hinton,
  1931. 1:18:34Max Tegmark, Yoshua Bengio, and Fei-Fei
  1932. 1:18:37Li.
  1933. 1:18:38You know, with many prominent scientists
  1934. 1:18:41who are essentially saying," This is our
  1935. 1:18:43path, we were doing this, we were
  1936. 1:18:45creating that. "Yes, we’ve done this
  1937. 1:18:48before and we know where it all leads.
  1938. 1:18:50And I think that's exactly what we need
  1939. 1:18:51to be talking about.
  1940. 1:18:53Elon Musk said that we are in the early
  1941. 1:18:55stages of a technological singularity.
  1942. 1:18:58And do you agree with him?
  1943. 1:18:59>> 100%.
  1944. 1:19:01I don't think we will make it past 2027
  1945. 1:19:04without reaching the singularity.
  1946. 1:19:06>> And what would that mean for humanity?
  1947. 1:19:10>> It implies a rapidly accelerating path
  1948. 1:19:13toward superintelligence.
  1949. 1:19:18So with AI at its current level, with
  1950. 1:19:20recursive self-improvement and long-term
  1951. 1:19:23reasoning, it becomes crystal clear that
  1952. 1:19:26we will reach a point where AI, either
  1953. 1:19:28collectively or as individual systems,
  1954. 1:19:31will become smarter than humans.
  1955. 1:19:35And that would mean that things will
  1956. 1:19:38happen in our world that are very
  1957. 1:19:40difficult to predict.
  1958. 1:19:43Anyone who tells you they know what 2028
  1959. 1:19:46will look like is being too arrogant,
  1960. 1:19:48right?
  1961. 1:19:49So, the concept of a singularity in
  1962. 1:19:51physics is an event horizon beyond which
  1963. 1:19:53you cannot predict what will happen
  1964. 1:19:55because the laws of nature no longer
  1965. 1:19:57apply there.
  1966. 1:20:00True, at the edge of a black hole, for
  1967. 1:20:03example, the singularity is also about
  1968. 1:20:05an event horizon where artificial
  1969. 1:20:08superintelligence, or at least
  1970. 1:20:10artificial general intelligence, becomes
  1971. 1:20:13so viable that it's impossible to
  1972. 1:20:15predict what the world will be like.
  1973. 1:20:19>> Does this mean that as humans we won't
  1974. 1:20:21be able to live the way we do today?
  1975. 1:20:24>> In the future, our lives will be more
  1976. 1:20:26human. All right. Once again. Remember,
  1977. 1:20:29as I already explained, AI companies
  1978. 1:20:32have an incentive to say that they will
  1979. 1:20:34replace workers. Okay? The capitalist
  1980. 1:20:37system finds it profitable to convince
  1981. 1:20:40you that you cannot live without it. Now
  1982. 1:20:42I am going to give you a very important
  1983. 1:20:44hint.
  1984. 1:20:47Capitalism wants you to believe that you
  1985. 1:20:50will eventually become obsolete, while
  1986. 1:20:54simultaneously trying to convince you to
  1987. 1:20:57work as hard as possible because you
  1988. 1:21:00cannot be replaced yet.
  1989. 1:21:04Again, my entire philosophy is that you
  1990. 1:21:06have a choice. You can decide to
  1991. 1:21:08completely step out of this, so to speak
  1992. 1:21:10, system and live in parallel economies
  1993. 1:21:12.
  1994. 1:21:13And again, if you grew up in the Western
  1995. 1:21:15world, you might think such things don't
  1996. 1:21:17exist.
  1997. 1:21:18Yet it is a large part of the planet.
  1998. 1:21:23Most of the population in the Global
  1999. 1:21:25South lives in a way that is completely
  2000. 1:21:27unlike California, New York, London, and
  2001. 1:21:29so on.
  2002. 1:21:31They are cheerful, active, cohesive, and
  2003. 1:21:33warm societies where people sing, dance
  2004. 1:21:36, eat the best food, and laugh from the
  2005. 1:21:38heart every day. They simply have no
  2006. 1:21:40need to buy a Ferrari. Yes. Yes,
  2007. 1:21:43>> and you have that choice, but there is
  2008. 1:21:45an even more interesting option.
  2009. 1:21:48And it is unclear to me why humanity
  2010. 1:21:50still hasn't taken advantage of it.
  2011. 1:21:53If you guys are planning to replace us
  2012. 1:21:55in 5 years, you know what?
  2013. 1:21:57We quit now, because ironically, again,
  2014. 1:22:00you are told that since you might be
  2015. 1:22:02replaced soon, you should work until you
  2016. 1:22:05drop to save a little for a rainy day.
  2017. 1:22:09But the truth is that if you seriously
  2018. 1:22:12declare:" No, no, wait, we are still
  2019. 1:22:15needed for labor arbitrage.
  2020. 1:22:18"So this is the only final moment when
  2021. 1:22:20we have the power to state:" Here are
  2022. 1:22:21our demands.
  2023. 1:22:23"Okay, here is this choice. It exists
  2024. 1:22:25regardless. An interesting challenge.
  2025. 1:22:28And the reason one of the books I'm
  2026. 1:22:30working on right now is called
  2027. 1:22:31Propaganda.
  2028. 1:22:33It's a curious problem because people
  2029. 1:22:35are convinced they are helpless.
  2030. 1:22:38That is not the case. We have the power
  2031. 1:22:40to deprive malicious players of our
  2032. 1:22:42income. You see?
  2033. 1:22:45We have the power to support good-faith
  2034. 1:22:47participants who advocate for well-being
  2035. 1:22:50and use their technologies.
  2036. 1:22:52We have the choice to forgo technology
  2037. 1:22:54altogether. Right?
  2038. 1:22:57We can choose not to drink soda, or if
  2039. 1:23:00we do, we can buy it from a national
  2040. 1:23:02company to keep wealth sovereign.
  2041. 1:23:06Yes, or should we prefer juice? The
  2042. 1:23:10decision is up to us. The question is
  2043. 1:23:12for the individual, when will they stop
  2044. 1:23:13and say:" I have options before me. I
  2045. 1:23:15will take advantage of one of them.
  2046. 1:23:18"And, of course, if each of us makes a
  2047. 1:23:20choice and declares—and the conclusion
  2048. 1:23:23of the novel is brilliant—it says:" I am
  2049. 1:23:26not asking to stop AI.
  2050. 1:23:28"" I am suggesting that we create narrow
  2051. 1:23:33AI that serves humanity, "
  2052. 1:23:37>> yes, ethical,
  2053. 1:23:39>> yes.
  2054. 1:23:40So when a company wants to create AI to
  2055. 1:23:42cure cancer, it approaches the state,
  2056. 1:23:44submits an application, and the
  2057. 1:23:46regulator replies:" Wonderful, we
  2058. 1:23:47support you.
  2059. 1:23:49"Go ahead, defeat the disease.
  2060. 1:23:51But if someone else says," I'm
  2061. 1:23:53developing AI to hack companies or play
  2062. 1:23:56the stock market for my own gain, "the
  2063. 1:23:59authorities say:" that's a waste of
  2064. 1:24:01resources. "That is selfishness.
  2065. 1:24:04Find an AI that will serve humanity.
  2066. 1:24:08If a country decides not to limit
  2067. 1:24:10artificial intelligence, but allows it
  2068. 1:24:12to develop as fast as possible and gives
  2069. 1:24:14it military access, it could gain a huge
  2070. 1:24:17advantage.
  2071. 1:24:19But if another country does the same,
  2072. 1:24:21victory might go to them instead.
  2073. 1:24:24So we are facing a kind of dilemma.
  2074. 1:24:28No one can afford to hold back because
  2075. 1:24:31everyone fears that the opponent will
  2076. 1:24:34pull ahead.
  2077. 1:24:36>> Yes, I know that I'll be in trouble
  2078. 1:24:38after this podcast, but I will say what
  2079. 1:24:41I think anyway.
  2080. 1:24:42The best solution for everyone is a draw
  2081. 1:24:44.
  2082. 1:24:45Okay? Because any unrestricted power is
  2083. 1:24:48too dangerous for the world.
  2084. 1:24:51In 59 years, I haven't seen a single
  2085. 1:24:54instance of China invading anywhere or
  2086. 1:24:57abusing force. And there is a lot of
  2087. 1:24:59evidence that the Chinese government
  2088. 1:25:01wants prosperity for the global economy
  2089. 1:25:04so that we buy more from them and their
  2090. 1:25:06people become wealthier. Show me
  2091. 1:25:08evidence to the contrary, and I will
  2092. 1:25:10start to be afraid.
  2093. 1:25:12So no, I don't think America should win
  2094. 1:25:14.
  2095. 1:25:15I also don't think China should win, and
  2096. 1:25:17I don't think we need to compete at all
  2097. 1:25:20. The current world order is a scarcity
  2098. 1:25:24mindset, which says:" If they win, I
  2099. 1:25:27lose. If I win, they lose. "That is no
  2100. 1:25:31longer true. With the advent of an
  2101. 1:25:33abundance of intelligence available to
  2102. 1:25:35everyone, it is possible to build a
  2103. 1:25:38world where no one will ever need
  2104. 1:25:40anything. Ever. After all, with
  2105. 1:25:43nanophysics, we are capable of creating
  2106. 1:25:45things literally out of thin air.
  2107. 1:25:48Yes, we can cure any disease, we can
  2108. 1:25:50even effectively harness solar energy
  2109. 1:25:53anywhere in the world.
  2110. 1:25:56We have the power to eliminate the need
  2111. 1:25:58for transportation, because it is
  2112. 1:26:00realistic to create virtual reality so
  2113. 1:26:02convincing that I could sit at home and
  2114. 1:26:04have this conversation with you, and the
  2115. 1:26:06sensations would be exactly the same.
  2116. 1:26:09So much is available to us. And yet we
  2117. 1:26:11build drones, you know?
  2118. 1:26:14And we do this because of a scarcity
  2119. 1:26:16mindset, which says that if they have an
  2120. 1:26:18advantage, I must not fall behind.
  2121. 1:26:21In human history, there have been
  2122. 1:26:23examples like the space station or the
  2123. 1:26:26particle accelerator at CERN, where
  2124. 1:26:28people believed so much in what I call
  2125. 1:26:30mutually assured prosperity.
  2126. 1:26:33Yes, instead of the mutually assured
  2127. 1:26:36destruction that threatens us with
  2128. 1:26:39nuclear confrontation, there is mutually
  2129. 1:26:42assured prosperity.
  2130. 1:26:44If we work together, every person on the
  2131. 1:26:46planet will have everything they need
  2132. 1:26:48endlessly and forever.
  2133. 1:26:50Without any pollution. We are already
  2134. 1:26:52there. I'm ready to argue with anyone. I
  2135. 1:26:54believe that today I can borrow 100 IQ
  2136. 1:26:57points from AI if I truly put my mind to
  2137. 1:27:00it. And learn at an intelligence level
  2138. 1:27:02100 points higher than one's own. You
  2139. 1:27:04know? In four, in three, or maybe in 2
  2140. 1:27:07years, I'll be able to hit 400 IQ points
  2141. 1:27:10, and I'll bet on that with anyone.
  2142. 1:27:13When I can get those 400 points, I'll
  2143. 1:27:16start growing gardens where you walk up
  2144. 1:27:18to one tree and pick an apple, walk up
  2145. 1:27:20to another and pluck an iPhone.
  2146. 1:27:24In terms of molecular cost, they aren't
  2147. 1:27:26that different, you see?
  2148. 1:27:28And we are almost at the goal.
  2149. 1:27:32You know, I worked with people like
  2150. 1:27:34Demis Hassabis.
  2151. 1:27:35Well, I mean, we didn't work together,
  2152. 1:27:37but we were at Google at the same time.
  2153. 1:27:39He has a wonderful approach to solving a
  2154. 1:27:41problem and handling everything else.
  2155. 1:27:45Why are we conquering intelligence to
  2156. 1:27:47create even more problems?
  2157. 1:27:50It's, you know, artificial intelligence
  2158. 1:27:52is a force of polarity, but humanity is
  2159. 1:27:54using it for evil.
  2160. 1:27:57This is where I slightly disagree with
  2161. 1:28:01Roman, in that I believe people will
  2162. 1:28:05finish us off before we do.
  2163. 1:28:09There is a probability that if we don't
  2164. 1:28:12change course and say:" Each and every
  2165. 1:28:14super-powerful technology we create must
  2166. 1:28:17make life better for everyone.
  2167. 1:28:21And those who oppose this must be
  2168. 1:28:24contained.
  2169. 1:28:25"There was an episode of the" Diary of a
  2170. 1:28:28CEO "podcast where Roman was a guest,
  2171. 1:28:30and someone said that arrests should be
  2172. 1:28:32made. For example, some of the CEOs need
  2173. 1:28:36to be stopped.
  2174. 1:28:37I don't know why people don't understand
  2175. 1:28:40this and how politicians can fail to
  2176. 1:28:42notice it.
  2177. 1:28:44After all, there are two ways to realize
  2178. 1:28:46that something needs to be done because
  2179. 1:28:49trouble is approaching.
  2180. 1:28:52One is to just wait for the thunder to
  2181. 1:28:54strike.
  2182. 1:28:55And the other is to use mathematics,
  2183. 1:28:57including applied mathematics, to
  2184. 1:28:58estimate the probability of such an
  2185. 1:29:00outcome.
  2186. 1:29:01And I agree with Roman on this. I think
  2187. 1:29:04the chance of an adverse outcome is 99%
  2188. 1:29:06.
  2189. 1:29:08Right?
  2190. 1:29:10The only difference is I believe
  2191. 1:29:12something bad will happen because one
  2192. 1:29:15person might use AI to harm everyone
  2193. 1:29:17else.
  2194. 1:29:19>> I thank you for this amazing
  2195. 1:29:21conversation. I learned a lot. I am sure
  2196. 1:29:26our viewers did too. Would you like to
  2197. 1:29:28say something to our audience in
  2198. 1:29:29conclusion?
  2199. 1:29:30First of all, you were very kind to
  2200. 1:29:32invite me. It turned out to be a
  2201. 1:29:34pleasant conversation. I hope I haven't
  2202. 1:29:36alarmed the viewers. I am sincerely
  2203. 1:29:38convinced that we all have a choice. You
  2204. 1:29:41know? Everyone has a choice. And it
  2205. 1:29:44starts with awareness.
  2206. 1:29:46It starts with knowing the real truth,
  2207. 1:29:48not what you are told.
  2208. 1:29:51And if you take that step and become
  2209. 1:29:54informed, I must say that I sincerely
  2210. 1:29:57believe humanity is just one decision
  2211. 1:30:01away from an absolute utopia.
  2212. 1:30:05My point of view is that without action
  2213. 1:30:08, we are in for a few years of suffering
  2214. 1:30:11, 10-12, who knows. But if we start
  2215. 1:30:14getting involved, in the end, disaster
  2216. 1:30:16can be avoided. At the same time, the
  2217. 1:30:19end result will be a utopia.
  2218. 1:30:21A world of abundance will come, where
  2219. 1:30:23everything is easy for everyone who
  2220. 1:30:25lives in it. I just don’t see a reason
  2221. 1:30:28why we need to suffer in the process.
  2222. 1:30:30And I think the answer is that the
  2223. 1:30:32decision is up to each of us.
  2224. 1:30:35So start acting accordingly, instead of
  2225. 1:30:38just hoping for change.

About this transcript

This page contains the full transcript of ИИ ЗАБЕРЁТ твою работу к 2027 | Экс-директор Google X by Глеб Соломин, generated from the public captions YouTube serves with the video. The transcript has 13,381 words across 2,225 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.