YouTube2Text

Братья Либерманы. Почему ИИ сделает одних сверхбогатыми, а других – нищими? — Transcript

by Forbes · 12,346 words · 2,030 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Why is it so much harder for today's
  2. 0:02young people to get rich than it was
  3. 0:04for their grandparents? What's more
  4. 0:06valuable: creating an atomic bomb or
  5. 0:07artificial intelligence? How is it that
  6. 0:10everyone benefits from human knowledge,
  7. 0:12while a few companies in the US and
  8. 0:15China profit from it? And what will we
  9. 0:18do when robots come to take our jobs?
  10. 0:21Hello everyone. My name is Dmitry
  11. 0:22Yakovenko, and this is a special
  12. 0:24edition of Forbes Highlights. A month
  13. 0:26ago, we released a video on our channel
  14. 0:28in which we discussed how corporations
  15. 0:29and governments stole the internet from
  16. 0:31us and turned every user into a
  17. 0:33commodity. If you haven't seen that
  18. 0:35episode yet, the link is in the
  19. 0:37description. We asked brothers Daniil
  20. 0:39and David Libermanov to serve as
  21. 0:41experts for this episode. You may have
  22. 0:43seen them in these episodes. They are
  23. 0:45serial entrepreneurs, originally from
  24. 0:47Russia, who have been living in Cream
  25. 0:49Valley for several years. During this
  26. 0:51time, they have managed to attract
  27. 0:53investment from Elon Musk, sold their
  28. 0:56company to Snapchat, launched several
  29. 0:58startups, and are now trying to spark a
  30. 1:00revolution. A key characteristic of the
  31. 1:03Liebermans is that they're not only
  32. 1:04incredibly interesting to listen to,
  33. 1:06they also have their own unique
  34. 1:08perspective on the modern internet
  35. 1:09economy. We spoke with them for three
  36. 1:12hours, but only about 30 minutes of
  37. 1:14that conversation made it into this
  38. 1:16episode, so we decided to present our
  39. 1:18conversation as a classic interview.
  40. 1:20And one last thing. Social media
  41. 1:22algorithms are merciless, so if you
  42. 1:24like the content we produce, we kindly
  43. 1:26ask you to help us a little by liking
  44. 1:29it, subscribing, and even writing a few
  45. 1:31sentences explaining your agreement and
  46. 1:33disagreement. Thank you very much in
  47. 1:36advance. No other economic sector in
  48. 1:44the world probably has such a natural
  49. 1:47tendency toward monopolies. One search
  50. 1:51engine, one marketplace, one streaming
  51. 1:53service, two operating systems. Why, in
  52. 1:55your opinion, is the industry
  53. 1:56developing in this particular direction
  54. 1:58?
  55. 1:58Fifteen years ago, we wrote an essay
  56. 2:02called "copy." It was about our
  57. 2:06internal research. It centered on the
  58. 2:09phenomenon of copied products. A
  59. 2:11first-generation
  60. 2:12product is one in which, after the
  61. 2:16first copy is created, the cost of
  62. 2:20creating subsequent copies is
  63. 2:23negligible. So, you could spend
  64. 2:26100 years thinking about and inventing
  65. 2:29electricity, but once it's invented,
  66. 2:31everyone can copy it. It costs a
  67. 2:33lot of resources to train a model. But
  68. 2:37even the largest models now take up
  69. 2:41about 10 TB. That means they can fit on
  70. 2:45a single flash drive and can be copied
  71. 2:47billions of times at a very low cost.
  72. 2:50Everything we're used to thinking about
  73. 2:54the market, yes,
  74. 2:55doesn't work. For example, that very
  75. 2:58famous Supply and Demand graph that you
  76. 3:00learn in, I don't know, your first
  77. 3:03economics course, doesn't work for
  78. 3:05these products. And the very nature of
  79. 3:08the market, it turns out, is different.
  80. 3:10Monopolies in this particular area are
  81. 3:13very difficult to regulate. For example
  82. 3:17, in the United States, antitrust laws
  83. 3:19are very clearly formulated. But all
  84. 3:22these years, a monopoly has been
  85. 3:25defined by the fact that you can use
  86. 3:28your market position to raise prices
  87. 3:32for consumers. And all the tech
  88. 3:35monopolies have found a loophole in
  89. 3:37this, saying, "But we're giving away
  90. 3:39the product for free." How can we be
  91. 3:42blamed for offering Facebook or
  92. 3:45Instagram for free,
  93. 3:48but they're simply making more and more
  94. 3:50money from each user on advertising, an
  95. 3:52indirect product. In these products,
  96. 3:54like a social network, it's convenient
  97. 3:56that there's only one. Well, because
  98. 3:58all of our social networks, our social
  99. 4:02connections, happen in the same space.
  100. 4:05It's convenient that there's only one
  101. 4:06AIL. It's convenient that there's only
  102. 4:08one domain space. It's also interesting
  103. 4:11that if you look at this in copied
  104. 4:13products, the greatest value is
  105. 4:15achieved when everyone uses the copied
  106. 4:18product. Imagine, for example, that you
  107. 4:21have the first copy, and the greatest
  108. 4:24benefit from this value arises if
  109. 4:27everyone uses this copy, because it's
  110. 4:29as if there wasn't any before. So, if
  111. 4:32no one wants to use it, then that
  112. 4:34single copy is the value. And if
  113. 4:36everyone wants to use it, then that
  114. 4:38copy is multiplied by 8 billion people.
  115. 4:39The
  116. 4:40only problem is that as soon as someone
  117. 4:42controls this one thing, they
  118. 4:43start to use it to their advantage.
  119. 4:46When I was thinking about the
  120. 4:48internet's tendency to create
  121. 4:51monopolies, it seemed to me that a key
  122. 4:53feature is the absolute externality of
  123. 4:56business, meaning there are absolutely
  124. 4:59no costs to operating in different
  125. 5:01markets. If you take, say, some
  126. 5:04consumer industries—like clothing,
  127. 5:07cucumbers, or anything like cars—it's
  128. 5:11very difficult for an entrepreneur to
  129. 5:14service the entire globe. Sure, you can
  130. 5:18be some kind of Indietex and sell
  131. 5:20clothes everywhere, but it's impossible
  132. 5:22to completely conquer every continent,
  133. 5:24every country. The internet itself
  134. 5:26offers the ability to deliver a service
  135. 5:29to any computer for a small fee, for
  136. 5:31servers and the cost of what a person
  137. 5:33pays to the provider. It's as if this
  138. 5:36is a natural monopoly. And here I'd
  139. 5:39like to clarify: when we talk about
  140. 5:42corporations on a natural, intuitive
  141. 5:44level, it seems that a monopoly is a
  142. 5:47bad thing. What are the fundamental
  143. 5:50economic claims against large monopoly
  144. 5:54corporations? Monopolies, in fact, on
  145. 5:57the one hand, in products like these,
  146. 6:00they, again, oh, they, of course, and
  147. 6:03they, they're kind of necessary,
  148. 6:06because, in fact, with those same vomer
  149. 6:09AI models, if 1,000 companies spend 10
  150. 6:14billion on training identical models,
  151. 6:16that's an unnecessary expense, which—
  152. 6:19we'll all pay for it anyway. And it
  153. 6:23seems unnecessary, but the problem with
  154. 6:25monopolies, or even a small number of
  155. 6:28companies, is that, unfortunately,
  156. 6:31since their product marginal cost is
  157. 6:34zero—well, or very close to zero,
  158. 6:36copies of the next zero—these
  159. 6:39companies essentially have no
  160. 6:42incentive to set a low price. They can
  161. 6:48always set that price as high as
  162. 6:51possible. There's an example right now,
  163. 6:54and even an interesting trial that
  164. 6:56recently took place in Italy. And the
  165. 6:59successful lawsuit, when consumers sued
  166. 7:02Netflix, because Netflix, regardless of
  167. 7:04the fact that over the last 10 years,
  168. 7:07the number of people using it has only
  169. 7:10increased. That is, it seems like there
  170. 7:12are more copies, meaning that the
  171. 7:15per-user costs should be lower, but
  172. 7:17instead, every three months or so, they
  173. 7:20raise the price.
  174. 7:21Well, every six months, a year, what do
  175. 7:23you mean, they constantly raise the
  176. 7:24price, constantly raising it. And
  177. 7:26what's more, they don't even ask you,
  178. 7:27they just tell you that the price is
  179. 7:29now higher, and higher, and higher
  180. 7:30again. They justify this with inflation
  181. 7:33, although they are growing much faster
  182. 7:35than inflation. Well, that is, simply
  183. 7:37because they can, because it's as if
  184. 7:39you can no longer leave this platform,
  185. 7:41so they can. They need to be cheaper
  186. 7:43for everyone because more people are
  187. 7:45using it. Or the same thing happened
  188. 7:47with Facebook when that service was
  189. 7:49just launching; there were no ads.
  190. 7:51Moreover, the founder
  191. 7:54believed that if there were ads, people
  192. 7:56wouldn't come. That's it. And that's
  193. 7:59why there were no ads. Then, when they
  194. 8:01became popular, they added one ad, and
  195. 8:03everyone was like, "Well, of course,
  196. 8:05they're great, they created this
  197. 8:07platform for us, they must be making
  198. 8:08money." Then we added a second one, and
  199. 8:11now
  200. 8:12today we're talking about every fifth
  201. 8:14post being an ad, so they couldn't stop
  202. 8:16, right? Well, actually, Google
  203. 8:19followed the same path. They also had
  204. 8:22an initial attitude toward advertising,
  205. 8:23like, "We have a minimalist site, we
  206. 8:25don't cover everything with banners, we
  207. 8:27have minimal advertising." And that's
  208. 8:29also a very interesting question. The
  209. 8:31business model itself seems fairly fair
  210. 8:34at first. We give you a free product to
  211. 8:36the whole world, but we make money from
  212. 8:38advertising, showing you ads. It's
  213. 8:40generally logical, but at what point
  214. 8:43does it become necessary?
  215. 8:45Look, think about it this way. In fact,
  216. 8:47the product, search, and advertising
  217. 8:50are one and the same, because the
  218. 8:52advertising product is the product of
  219. 8:55search, from consumers to producers,
  220. 8:58from producers to consumers. That is,
  221. 9:00they are searching for each other. And
  222. 9:01now, from the perspective of raising
  223. 9:04this price in general, there's one
  224. 9:06fairly obvious and understandable
  225. 9:08reason, and it's natural, and there's
  226. 9:11no escaping it. A corporation exists to
  227. 9:14make a profit. Make a profit. For whom?
  228. 9:17To make a profit for shareholders. And
  229. 9:19the shareholders of all those
  230. 9:21corporations are retirees.
  231. 9:22And
  232. 9:23pension funds. Pension funds. People
  233. 9:26give them their money so they can have
  234. 9:29more money in their old age. Everyone
  235. 9:31invests in corporations, owns 50 to 90%
  236. 9:34of them, sits on the board of directors
  237. 9:37, appoints forces, sets goals for the
  238. 9:40very purpose for which the retirees
  239. 9:43gave them the money. Half of the people
  240. 9:45in the United States have put money
  241. 9:47aside for their retirement. And they've
  242. 9:50asked that this money be invested
  243. 9:52effectively to grow capital. There's no
  244. 9:55flaw in logic here. You just have to
  245. 9:58see the whole picture. We ourselves ask
  246. 10:00corporations to increase this price
  247. 10:02because we ourselves are their owners.
  248. 10:05But in whose interests is this really
  249. 10:06working? It works, let's say, in the
  250. 10:08interests of half the country's
  251. 10:10population. The other half, the young
  252. 10:12people who use these social networks.
  253. 10:14They're not the beneficiaries,
  254. 10:15because it seems like, well, what
  255. 10:17difference does it make? You're right,
  256. 10:18they gave you a free product, and the
  257. 10:21advertising, well, is hanging around
  258. 10:23somewhere. But the more expensive this
  259. 10:26search between the consumer and the
  260. 10:28buyer, the more expensive the
  261. 10:29manufacturer's product is for you, the
  262. 10:31more expensive the products are for you
  263. 10:33.
  264. 10:33That is, if this search were cheaper,
  265. 10:36then ultimately the market would
  266. 10:37optimize itself so that the clothes you
  267. 10:40buy, the cars you buy, and so on,
  268. 10:41everything would be cheaper. If
  269. 10:43marketing were cheaper, but ultimately,
  270. 10:46marketing isn't cheaper. It's just that
  271. 10:49all this, uh, marginality is being
  272. 10:51taken over by Google. And now Google,
  273. 10:54ah, $ 400 billion in revenue, and it's
  274. 10:57already surpassed, it's fifth, I think,
  275. 11:01or sixth largest company in the world
  276. 11:04by revenue. Previously, at least this
  277. 11:06company used these resources to create
  278. 11:09more and more of these free products
  279. 11:11for all of us. Google Maps, Google Maps
  280. 11:14, Gmail, and so on. But you can't. If
  281. 11:16you look at the last time something
  282. 11:18like this happened, it was a very, very
  283. 11:20long time ago. Meanwhile, if you
  284. 11:22examine the balance sheets of these
  285. 11:26companies, you'll see that they already
  286. 11:29hold 200 billion rubles in cash and 100
  287. 11:31billion in real estate. But what for?
  288. 11:34Even if you imagine these wonderful
  289. 11:36companies simply accumulating this
  290. 11:39resource, investing in scientists and
  291. 11:41developing humanity, unfortunately,
  292. 11:44that's no longer the case. The stories
  293. 11:53of Forbes's heroes demonstrate how
  294. 11:55international modern business has
  295. 11:57become. When projects, partners, and
  296. 11:59opportunities are located in different
  297. 12:01countries, additional flexibility and
  298. 12:03the ability to operate in other
  299. 12:05jurisdictions are becoming increasingly
  300. 12:07in demand. The rules of the game are
  301. 12:10changing, sanctions and banking
  302. 12:11regulations are tightening. Therefore,
  303. 12:14second citizenship or residency are
  304. 12:16increasingly viewed today not as a way
  305. 12:18to move somewhere permanently, but as
  306. 12:20an additional tool for long-term
  307. 12:22planning. While Caribbean passports,
  308. 12:24Cyprus, and Malta's programs were once
  309. 12:26all the rage, the market has changed
  310. 12:29significantly. For citizens of Russia
  311. 12:31and Belarus, the window of opportunity
  312. 12:33has narrowed significantly, but options
  313. 12:35remain. For example, Santa Cruz
  314. 12:37citizenship can, in principle, be
  315. 12:39obtained with an investment of at least
  316. 12:41$ 90,000, and Vnuka citizenship with an
  317. 12:43investment of at least $ 130,000. A
  318. 12:45less exotic option is Turkey. Here,
  319. 12:48citizenship is available through the
  320. 12:49purchase of real estate starting at $
  321. 12:51400,000. Golden visa programs also
  322. 12:54exist, but they are not as numerous.
  323. 12:56For example, Cyprus offers citizenship
  324. 12:57with a real estate purchase starting at
  325. 12:59€ 300,000, while Greece requires
  326. 13:01purchasing an apartment, house, or flat
  327. 13:03starting at € 250,000. Participation
  328. 13:06in the program is only possible with
  329. 13:08dual citizenship, but, as you've
  330. 13:09already understood, this is not
  331. 13:11impossible. Nevertheless, Greece
  332. 13:13remains one of the most flexible
  333. 13:14options in Europe. With a residence
  334. 13:17permit, you'll be able to travel
  335. 13:18visa-free throughout the Schengen area.
  336. 13:21The key is that none of the listed
  337. 13:23programs require permanent residence in
  338. 13:25the country. You won't have to
  339. 13:26radically change your lifestyle. These
  340. 13:28are simply additional opportunities for
  341. 13:30running a business and generally
  342. 13:31planning for the future. It sounds
  343. 13:33simple, but investment immigration
  344. 13:35isn't just about choosing a country.
  345. 13:38The legal structure of the transaction,
  346. 13:40verification of the origin of funds,
  347. 13:42and compliance procedures are all
  348. 13:44important. Such tasks should be
  349. 13:46entrusted to a company with serious
  350. 13:49legal expertise and real-world cases.
  351. 13:52Asns has been supporting clients
  352. 13:54worldwide for over 35 years. Learn more
  353. 13:57about the available second citizenship
  354. 13:59and residence permit programs by
  355. 14:00following the link in the description.
  356. 14:08So, for retirees, everything is
  357. 14:09wonderful. In reality, the current
  358. 14:11world, with the exception of the fact
  359. 14:13that it's
  360. 14:13divided not between rich and poor, but
  361. 14:16between young and old. The economy is
  362. 14:18being torn apart. Wealth and poverty
  363. 14:20actually lie in the divide between the
  364. 14:23young and old. And it's not so much
  365. 14:25about age. When you and I are old,
  366. 14:28we'll be poorer. So, in any crisis, say
  367. 14:30, in the American one, the government
  368. 14:33primarily seeks to protect the savings
  369. 14:36of retirees. Simply because they are
  370. 14:38the voting majority and the economic
  371. 14:40majority. And on the one hand, the
  372. 14:43internet and the technological
  373. 14:45breakthrough have created such an
  374. 14:47opportunity for young people—the
  375. 14:50industry is primarily filled with young
  376. 14:52people, but the end result still went
  377. 14:55to pension funds, because all the
  378. 14:57evening funds are mostly money from
  379. 15:00pension funds or endowments of
  380. 15:02universities. This, among other things,
  381. 15:04is why there's such a huge interest in
  382. 15:06cryptocurrencies, because they seem to
  383. 15:08be outside the scope of that regulation
  384. 15:10, outside the scope of that interest,
  385. 15:12and so on. Now it's not like that
  386. 15:15anymore, now it's like all pension
  387. 15:17funds, BlackRock is buying Bitcoin,
  388. 15:19everything is kind of mixed up with
  389. 15:22each other in the market, but young
  390. 15:24people had the right idea that maybe we
  391. 15:27can win from this now.
  392. 15:28Blockchain technology. It has a
  393. 15:30completely non-commercial idea behind
  394. 15:33it—ah, removing barriers, simplifying
  395. 15:35transactions between people, increasing
  396. 15:38security and transparency. From the
  397. 15:41outside, it looks like, firstly, a huge
  398. 15:43speculative market, and secondly, well,
  399. 15:46not just speculative, it's not entirely
  400. 15:48clear why it exists.
  401. 15:50If you think about it, if it's
  402. 15:52speculation, speculation, the entire
  403. 15:53market is in some sense speculation,
  404. 15:55it's
  405. 15:55important to compare apples to apples.
  406. 15:59For example, let's take
  407. 16:00cryptocurrencies and try to figure out
  408. 16:03whether they're excessively expensive
  409. 16:05or not. Speculation means they must be
  410. 16:07excessively expensive. But you take any
  411. 16:10currency in the world, take the Israeli
  412. 16:13shekel, the Australian dollar, the
  413. 16:16British pound, take the country's money
  414. 16:19supply and divide it by the number of
  415. 16:21users, namely, by the number of people
  416. 16:24in the country, and you get almost the
  417. 16:27same figure. What do you mean per
  418. 16:30person in dollar terms? If you compare
  419. 16:32countries with a high standard of
  420. 16:35living to one currency,
  421. 16:37you take all the currencies, convert
  422. 16:38them to dollars, and divide by the
  423. 16:39population.
  424. 16:40You divide it by the population of this
  425. 16:42country, which is almost the same
  426. 16:45amount per person, and then you take
  427. 16:47Bitcoin, you take Ereum, you take their
  428. 16:50, uh, active users,
  429. 16:52yes, for example, the monthly number of
  430. 16:54wallets, and you get almost the same
  431. 16:56figure, the same thing
  432. 16:57there. What does
  433. 16:58this mean now? It means it's the same
  434. 17:00currency mechanism. It's just a
  435. 17:02different currency. So if you call this
  436. 17:05speculation, then you call the dollar,
  437. 17:07the pound sterling, and so on
  438. 17:08speculation. Or you just call it all
  439. 17:10currency. And that's how much currency
  440. 17:12is worth. It's
  441. 17:12just that this currency is controlled
  442. 17:14by the federalized United States, and
  443. 17:16this currency is controlled by
  444. 17:17protocols. It's the same thing. If AI
  445. 17:19bubbles, we look at the companies that
  446. 17:22are currently the most, uh, getting, uh
  447. 17:25, a higher valuation because they're
  448. 17:28related to AI, you look at their profit
  449. 17:32multiple,
  450. 17:33and then you look at Walmart. It's a
  451. 17:36big store, a chain of stores.
  452. 17:39And it has the same multiple. Look at
  453. 17:41Coca-Cola, for example; it has the same
  454. 17:43multiple. The entire market
  455. 17:45is high in terms of multiples. But
  456. 17:48Google's multiple is no higher than
  457. 17:50Walmart's.
  458. 17:51And then there's the demand for a large
  459. 17:54amount of capital accumulated by
  460. 17:56pension funds and others. Pension fund
  461. 17:58accumulation in the US is roughly equal
  462. 18:00to all other savings. That is, all
  463. 18:02these ultra-rich people, all put
  464. 18:04together. All these savings are equal
  465. 18:06to pension savings.
  466. 18:08Pension accumulation is larger.
  467. 18:09Larger. Well, I'm ready to say that a
  468. 18:10comparable thing is that pension
  469. 18:11accumulation is larger than all the
  470. 18:13other accumulations. Few, few people
  471. 18:14understand that
  472. 18:15this mythical 1%who own everything
  473. 18:18can't own everything, because retirees
  474. 18:20have more money than everyone else.
  475. 18:23Therefore, this is a false perception
  476. 18:25of reality. When we look at the reason
  477. 18:28why stocks are expensive or, say,
  478. 18:30currencies are expensive, we need to
  479. 18:33look at the economic reason. This
  480. 18:35bubble occurs because so much capital
  481. 18:38needs to be transferred into the future
  482. 18:41with a return. What reason could one
  483. 18:44stranger have to give another stranger
  484. 18:49money, a
  485. 18:50percentage that is, to get more? And
  486. 18:53all modern thinking about economics and
  487. 18:55the market is based on the idea that
  488. 18:58it's almost a
  489. 18:59cliché that everything is built on the
  490. 19:01rate of return, right?
  491. 19:02This is the basic understanding of
  492. 19:06human interaction in economics. But it
  493. 19:09turns out there's another reason,
  494. 19:13namely, your labor, which is perishable
  495. 19:16,
  496. 19:16like milk.
  497. 19:17You worked today, didn't
  498. 19:20earn enough,
  499. 19:21didn't earn enough, and at the same
  500. 19:23time, you have an old age in which you
  501. 19:26need to somehow live. You won't be able
  502. 19:29to work there. Therefore, you, a young
  503. 19:32person, need to somehow
  504. 19:34freeze your labor,
  505. 19:36transfer it into the future. You give
  506. 19:39part of it to someone so that they earn
  507. 19:41more, so that you can return more to
  508. 19:43you in the future. This is what this
  509. 19:46effective transfer is. Accordingly, it
  510. 19:48turns out there is another reason to
  511. 19:50give money to another person.
  512. 19:52You simply need to do it. You can't
  513. 19:54avoid
  514. 19:55getting more right now, or even
  515. 19:57transferring it into the future, even
  516. 20:00if it's less. Because if you simply put
  517. 20:03money in an account or bury it in a
  518. 20:06suitcase, it will lose value. And if
  519. 20:09you want to get at least the same
  520. 20:11amount back, you have to find someone
  521. 20:13who will use it effectively and return
  522. 20:14the same amount in the future. What's
  523. 20:16happening with cryptocurrencies, for
  524. 20:19example? You have this idea that while
  525. 20:23there may be few people today who want
  526. 20:27independence from the SDO banks, there
  527. 20:31will be many of them, and that every
  528. 20:34crisis, every upheaval, the
  529. 20:37new COVID-19, and the printing of money
  530. 20:39will lead to greater and greater
  531. 20:40rejection of fiat currency and a shift
  532. 20:44to something that can't be reissued.
  533. 20:47This is the hypothesis that the creator
  534. 20:50of this currency, Satoshi, had. He
  535. 20:52created this protocol, and it turns out
  536. 20:55that it's working so far, and more and
  537. 20:57more people are choosing to use this
  538. 21:00tool to transfer their capital into the
  539. 21:02future without, well, printing more
  540. 21:05money. So, this was the main hypothesis
  541. 21:07that was initially laid down: that
  542. 21:09because it's finite, it doesn't devalue
  543. 21:12. And this was the idea that there
  544. 21:14would be more of these people. If there
  545. 21:17are more of them, if people who consume
  546. 21:20it buy more of it, then its price rises
  547. 21:22, since its supply is limited. It's the
  548. 21:24same, you know, with all currencies. So
  549. 21:26, the more users there are, the more
  550. 21:29monetary value is cut.
  551. 21:32For you, as a Bitcoin consumer who
  552. 21:34consumes it as a currency, it doesn't
  553. 21:36matter whether it costs $ 60 or $
  554. 21:38120,000, $ 160,000 or $ 120,000. It
  555. 21:40doesn't matter if you bought $ 100
  556. 21:42worth of it and paid $ 100 worth of it
  557. 21:44today. And if the price hasn't moved
  558. 21:47much during this period, then you got
  559. 21:49what you wanted—transaction currency.
  560. 21:51Out you go.
  561. 21:52But if someone expects more people,
  562. 21:54that's the idea. It's not a question of
  563. 21:58some person like Satoshi or someone
  564. 22:00doing something to make Bitcoin develop
  565. 22:03. No, you just have an idea. More
  566. 22:05people will want it,
  567. 22:07consume it. But more people will think
  568. 22:10the Central Bank is engaging in
  569. 22:13so-called tifzing and has printed 9
  570. 22:16trillion. That's a bit much and will
  571. 22:19end badly. Why did Bitcoin work?
  572. 22:21Because Bitcoin was based precisely on
  573. 22:24a big idea, an ideological one, not a
  574. 22:28belief that more people would want
  575. 22:31something,
  576. 22:32desire something, want to be outside
  577. 22:35the regulated state mechanism. But the
  578. 22:39state began to misuse this mechanism,
  579. 22:42using it for its own political ends.
  580. 22:46The same thing is
  581. 22:46happening now with artificial
  582. 22:49intelligence. That is, if you look at
  583. 22:52artificial intelligence today, on the
  584. 22:54one hand, something remarkable is
  585. 22:55happening. We all now have GPT or
  586. 22:58Claude chat, or we all use these
  587. 23:00artificial intelligences. They're
  588. 23:03incredibly smart, they answer many of
  589. 23:06our questions, they
  590. 23:08speak all languages,
  591. 23:09but the longer it goes, the more people
  592. 23:13will notice, uh, several problems with
  593. 23:16this. One problem already exists in the
  594. 23:18mainstream. What if these little clocks
  595. 23:22start replacing us all at work? And
  596. 23:25what are we supposed to do then?
  597. 23:27Robots will come along with them, and
  598. 23:29it's like, oh,
  599. 23:29and this idea becomes many times more
  600. 23:33complicated if you imagine it. If you
  601. 23:36have this chat, you're kind of
  602. 23:38superhuman, you can do more than you
  603. 23:40could a year ago, but that's only if
  604. 23:43you have access. But what if at some
  605. 23:45point your
  606. 23:45access is taken away, and you have to
  607. 23:48compete with everyone else who has it?
  608. 23:51It's the same thing
  609. 23:52that happens with social media, so to
  610. 23:55speak, because you're effectively cut
  611. 23:57off from the economy and you might not
  612. 23:59even exist if you're not on this
  613. 24:01platform. It's exactly the same logic.
  614. 24:03And you can be disconnected, blocked,
  615. 24:06and so on for various reasons. So, we
  616. 24:08see how
  617. 24:09people get blocked on Twitter. For a
  618. 24:11while, before Elon got scared and
  619. 24:13bought something to avoid being blocked
  620. 24:15there.
  621. 24:15And recently, yes, he just assessed
  622. 24:18that the risk of being removed from the
  623. 24:20platform is worth $ 40 billion. Many
  624. 24:23people just think, why would Elon buy
  625. 24:25it? He's some billionaire, buying media
  626. 24:28and this and that. Well, in fact, if
  627. 24:31you think about it, for Elon, it was a
  628. 24:33mechanism for
  629. 24:34communicating with his entire audience,
  630. 24:37with all his customers, with all his
  631. 24:40stockholders. So the Wall Street short
  632. 24:45market couldn't overcome him. And then
  633. 24:49they tell him, "The president just got
  634. 24:51disconnected from there, which means
  635. 24:53you'll get disconnected too for your
  636. 24:55next bad word." And he's like, "Is this
  637. 24:57worth $ 40 billion?" Especially knowing
  638. 24:59that it's not really that I lost those
  639. 25:0240 billion. I paid for the business,
  640. 25:04and I still have it. I can sell it
  641. 25:05later if necessary. And actually, it
  642. 25:08could even be profitable. I assessed
  643. 25:10the risks and decided to buy. Now
  644. 25:11imagine what it means for all of
  645. 25:13humanity if, at some point, these ten
  646. 25:16corporations—the five American and
  647. 25:18five Chinese corporations—begin to
  648. 25:21realize it's more profitable for them
  649. 25:23to stop giving you access to AI
  650. 25:26and simply start selling it directly to
  651. 25:29factory owners, like, er, taxi
  652. 25:32companies. Recently, Atropic, the
  653. 25:35company that produces Claude, released
  654. 25:38—well, not released—announced that
  655. 25:41they'd created a new, even more
  656. 25:44powerful model, but immediately said
  657. 25:47that access would only be restricted to
  658. 25:50select companies.
  659. 25:51Yeah.
  660. 25:52Yeah, there's a message about security.
  661. 25:57But in reality, how many steps are
  662. 25:59there before it becomes a matter of
  663. 26:02economic security for the company and
  664. 26:05its clients? There's no limit, not a
  665. 26:08single law that sets a limit. And
  666. 26:09what's more, it's in their commercial
  667. 26:11interests. Recently, the founder of
  668. 26:14this company, in an interview with
  669. 26:18aa Darkash, a well-known podcaster from
  670. 26:23Silicon Valley. And then I asked you
  671. 26:26about their business models. Is their
  672. 26:29business model valid? Will it remain
  673. 26:31that they sell
  674. 26:33API access tokens to other app makers?
  675. 26:38And where do they pay for the so-called
  676. 26:41AI tokens? User usage
  677. 26:42is a measure of AI usage. He said, "I
  678. 26:44think this model will remain,
  679. 26:46blah-blah-blah-blah-blah." And then he
  680. 26:48said, "But on the other hand, not all
  681. 26:51tokens are equally valuable." For
  682. 26:54example, if someone can help you fix
  683. 26:57your MacBook, it might not be so
  684. 26:59valuable. What if
  685. 27:02someone can cure themselves of cancer?
  686. 27:05Are such tokens worth more?
  687. 27:07And would
  688. 27:08people be willing to pay more for them?
  689. 27:10It's
  690. 27:10terrible. Well, you shouldn't even say
  691. 27:12such a thing, but
  692. 27:13he even laughs in this interview. So,
  693. 27:17he's laughing hysterically about this,
  694. 27:18because he himself understands that
  695. 27:20somewhere there is a line of
  696. 27:21unacceptability, of indecency, but he's
  697. 27:23already thinking about it and
  698. 27:24understands that this is where it's
  699. 27:26heading. This is what he's most likely
  700. 27:28selling to investors. But, but, but
  701. 27:31what does this mean? If today AI is the
  702. 27:33same as the internet, or even the same
  703. 27:36as electricity, then it roughly means
  704. 27:38something like this. The electricity
  705. 27:40that powers your light bulb costs the
  706. 27:42same as the electricity that powers the
  707. 27:45life support machine when you have
  708. 27:47open-heart surgery. Open-heart surgery
  709. 27:50is much more important to you. You pay
  710. 27:521,000 times more for those kilowatts
  711. 27:53than for the kilowatts you consume when
  712. 27:55you simply light your house. It
  713. 27:57sounds absurd. No,
  714. 27:58it doesn't sound absurd. It sounds like
  715. 28:01a world we don't want to live in. But
  716. 28:05that means we're not talking about it
  717. 28:07enough, you know? That's precisely the
  718. 28:09difficulty. And then there's
  719. 28:11electricity. Electricity wasn't
  720. 28:13perceived by anyone as something, well,
  721. 28:14something that love could produce. But
  722. 28:16imagine if it were also considered a
  723. 28:19matter of national security, as it is
  724. 28:22now.
  725. 28:22And everyone would be banned from
  726. 28:24using this electricity. What would the
  727. 28:27world be like today if the United
  728. 28:29States of America said, "You know what,
  729. 28:31no? Electricity is just for us." Or,
  730. 28:33well, 90%of the electricity is for us,
  731. 28:35and we'll give you the other 10%, and
  732. 28:37then you can consume it as you please.
  733. 28:39This is also a
  734. 28:40stunning dystopia. I don't even know if
  735. 28:43anyone has ever done anything like this
  736. 28:45. All the power plants are in one
  737. 28:48country, and it sells it wherever it
  738. 28:50wants.
  739. 28:52The GPUs needed for AI computing were
  740. 28:54built in such a way that all this
  741. 28:57computing power, 90%of computing power
  742. 29:00is now in the United States. Next will
  743. 29:03be China, which will also be doing this
  744. 29:05computing. They're almost catching up.
  745. 29:08It's still quite a ways off, but
  746. 29:10they'll catch up. It's obvious they'll
  747. 29:12catch up, that China will give it all
  748. 29:13away, or something. No, it's the same
  749. 29:15thing. And this creates a false
  750. 29:17impression, because China made its
  751. 29:20models open source to catch up. That's
  752. 29:23a common strategy. That's a non-profit
  753. 29:24strategy. You make yourself open source
  754. 29:26to catch up. But what people don't
  755. 29:30understand is that as soon as they
  756. 29:31catch up, they stop being open source.
  757. 29:34You create a model, you make
  758. 29:34it open source. Developers all over the
  759. 29:36world use it. When they use it, they
  760. 29:39train it automatically. It becomes,
  761. 29:42well, not exactly. Not exactly. It's a
  762. 29:44small percentage there. That's because
  763. 29:46the training actually happens
  764. 29:47beforehand.
  765. 29:48But people build an ecosystem, you
  766. 29:50understand? This ecosystem is very
  767. 29:51important,
  768. 29:52they show you how to use it, yes, the
  769. 29:53methods,
  770. 29:54because all your agents are configured
  771. 29:55for this model. Since all the models
  772. 29:57are slightly different, you fine-tune
  773. 30:01them to fit it, and you build your
  774. 30:02product.
  775. 30:02Basically, you make it part of a global
  776. 30:05infrastructure, and then you simply
  777. 30:07shut it down. And
  778. 30:09once it becomes so powerful that it
  779. 30:11controls a robot that replaces humans,
  780. 30:13why would the Chinese sell it to anyone
  781. 30:16? They'll lose the market. The cost of
  782. 30:18labor will drop to the price of the
  783. 30:20robot, which will then drop to the cost
  784. 30:22of the materials and the artificial
  785. 30:23intelligence it contains, the
  786. 30:25computations it contains. And if that's
  787. 30:27the case, then there's no genius in
  788. 30:29Chinese manufacturing, because all the
  789. 30:31knowledge about how to produce anything
  790. 30:32, very cheaply and quickly, is
  791. 30:34available to everyone all over the
  792. 30:36world.
  793. 30:36Therefore, the Chinese will also shut
  794. 30:38down their models, and the Americans
  795. 30:39will shut down theirs. Right now, what
  796. 30:41we're discussing is understood by a few
  797. 30:44thousand people. When Fzhe wrote, there
  798. 30:46were a handful of people who believed
  799. 30:48in the same idea. And ultimately, it
  800. 30:50blossomed into a community of tens, if
  801. 30:52not hundreds of millions of people who
  802. 30:54use Bitcoin today. All of us sitting
  803. 30:56here listening to this podcast are
  804. 30:58gradually beginning to understand that
  805. 31:00something's wrong with AI, that
  806. 31:02somewhere down the line, we're all
  807. 31:04being screwed. But then, if you imagine
  808. 31:06that
  809. 31:07in 10 years there will be hundreds of
  810. 31:10millions of people who realize this,
  811. 31:13and that all these hundreds of millions
  812. 31:15will be looking for an alternative, one
  813. 31:18where their access won't be taken away,
  814. 31:20or where their
  815. 31:21jobs won't be taken away,
  816. 31:22and where there will be an
  817. 31:24ever-increasing number of them, and
  818. 31:27their productivity will increase due to
  819. 31:30the fact that they will constantly be
  820. 31:33using more and more new tools in their
  821. 31:36work, we saw at some point that, on the
  822. 31:38one hand, you're slightly opposing
  823. 31:41yourself, just like Satoshi also
  824. 31:43opposed himself to central banks—they
  825. 31:46seem incredibly gigantic, like, well,
  826. 31:49entities that it's impossible to even
  827. 31:53imagine that you could resist.
  828. 31:55Funny. Like, well, his point was that,
  829. 31:58well, billions of people won't agree
  830. 32:01with this. Since billions of people
  831. 32:03won't agree, since currency and values
  832. 32:05are determined by how many people
  833. 32:07participate in them,
  834. 32:08then it will ultimately work. We're at
  835. 32:11a similar point in AI right now. Large
  836. 32:15AI corporations seem completely
  837. 32:18inaccessible.
  838. 32:20As you can imagine, they invest tens of
  839. 32:23billions of dollars every year in
  840. 32:25building data centers and purchasing
  841. 32:28chips.
  842. 32:28Hundreds of billions already.
  843. 32:29Hundreds of billions. Even now,
  844. 32:31everyone has started looking at graphs
  845. 32:35of how the costs of AI computing
  846. 32:38infrastructure compare with the Apollo
  847. 32:42program, and they're actually
  848. 32:44comparable, even in terms of the
  849. 32:47percentage of GDP spent on these
  850. 32:51projects. This mega-project is underway
  851. 32:54, but this mega-project, as it's
  852. 32:57currently unfolding, certainly won't
  853. 33:01provide equal access to everyone. On
  854. 33:05the contrary, we're already seeing the
  855. 33:08first steps toward exclusive access for
  856. 33:11certain groups of companies or certain
  857. 33:14groups of people. This is where the
  858. 33:16important question lies, precisely in
  859. 33:18the numerical order. That is, there are
  860. 33:21billions of people who disagree, but
  861. 33:24then, let's go to Bitcoin. Well, it's
  862. 33:27not billions of people, it's a large
  863. 33:29number of people who, uh, are
  864. 33:31enthusiastically engaged in this. It's
  865. 33:33a matter of geopolitical confrontation.
  866. 33:36And by and large, it seems like for
  867. 33:39most people it doesn't really matter
  868. 33:42what platform they're working on.
  869. 33:45Moreover, as far as I understand, the
  870. 33:48point of this huge Chinese miracle, IT,
  871. 33:51is that the Chinese population—
  872. 33:53however many people live there—is
  873. 33:56generally quite accepting of the fact
  874. 33:58that we have states, some
  875. 34:00quasi-corporations that own production,
  876. 34:03that own our data, and that manage us
  877. 34:06using algorithms. They build smart
  878. 34:08cities, offer us super-cool platforms
  879. 34:11on which we can do everything. But it's
  880. 34:13super-convenient. And I have a feeling
  881. 34:15that the rest of the world wouldn't
  882. 34:19mind.
  883. 34:19It only appears that way on the surface
  884. 34:21. Well, if we dig into what's happening
  885. 34:24in China, underneath the public surface
  886. 34:29, things are actually much worse. We
  887. 34:30don't even need to argue with that,
  888. 34:33because you're right that from
  889. 34:35Bitcoin's perspective, ah, since there
  890. 34:38are many central banks, many countries,
  891. 34:40and each country has its own central
  892. 34:43bank, otherwise, ah, well, most people
  893. 34:45will say, "Well, okay, in Europe we
  894. 34:48live with the euro, we live in
  895. 34:50Australia with the dollar, and so on."
  896. 34:53There are many different alternatives.
  897. 34:54And so, Bitcoin has ultimately
  898. 34:57accumulated around 50 million active
  899. 34:59users. And now they are 50 million
  900. 35:03countries worth a trillion dollars.
  901. 35:06And not China, not the US in terms of
  902. 35:10population.
  903. 35:11Well, like England, France, Germany,
  904. 35:14with AI,
  905. 35:15things will be much worse, because
  906. 35:17you're right. If we're talking about
  907. 35:19the topic we're discussing before the
  908. 35:21United States, most people, most of our
  909. 35:23interlocutors, say, "Well, we're doing
  910. 35:26fine." So, they'll have a UBI, they'll
  911. 35:28collect all the value from around the
  912. 35:30world and distribute it to some segment
  913. 35:33of the local population.
  914. 35:35A person gets into a car, and it's a
  915. 35:38self-driving car. The car drives, and
  916. 35:40he doesn't get into an accident. He
  917. 35:42essentially doesn't care how it's
  918. 35:44powered, what algorithm it uses, or who
  919. 35:47owns the model it runs on.
  920. 35:49But, but, but if you're part of the
  921. 35:52group that benefits from this, you'll,
  922. 35:55of course, feel more or less okay. But
  923. 35:59if we even imagine these two groups,
  924. 36:01the
  925. 36:01United States and China, and imagine
  926. 36:03that China also gathers a lot of the
  927. 36:06world's wealth through developing its
  928. 36:08AI and also distributes a little to its
  929. 36:10population, especially so that the
  930. 36:12population doesn't rise up against them
  931. 36:14,
  932. 36:14then the rest of the world—the
  933. 36:16remaining 6 billion people—will feel
  934. 36:18—well,
  935. 36:19maybe not so comfortable.
  936. 36:20Well, they need to choose where they
  937. 36:24won't benefit from this, you understand
  938. 36:26? That is, they will be the ones
  939. 36:29collecting the resources, while those
  940. 36:32paying for them will be the ones.
  941. 36:33This feeling is growing in almost every
  942. 36:36country in the world, except for the
  943. 36:38United States and China. Over the past
  944. 36:41year, we've met with ministers from
  945. 36:43many countries with varying levels of
  946. 36:45GDP per capita. And while just a year
  947. 36:49and a half ago, and usually the
  948. 36:52ministries of these countries, said, "
  949. 36:55We're doing our own thing, we'll
  950. 36:59buy our own chip, we'll do everything
  951. 37:01ourselves,"
  952. 37:02now they've realized that no, they're
  953. 37:04completely behind, and it's impossible
  954. 37:07to compete with these two fathers on
  955. 37:09their own. It's impossible to even
  956. 37:12imagine anything similar to these two
  957. 37:14policies.
  958. 37:16And what we're seeing is that almost
  959. 37:21all of them are now looking for an
  960. 37:25alternative and are turning toward
  961. 37:28declaring something more non-commercial
  962. 37:32. They're determined to
  963. 37:35cooperate with one of these two—the
  964. 37:37Poles, the Chinese, or the Americans—
  965. 37:39simply because they have no alternative
  966. 37:41. And that's why they're looking for an
  967. 37:44alternative, inviting us and
  968. 37:45communicating with us. It's precisely
  969. 37:48because it's obvious to everyone that
  970. 37:51an alternative is necessary.
  971. 37:54This allows for the same thing we
  972. 37:56discussed with monopolies at the
  973. 37:58beginning: that when you have a
  974. 38:00monopoly, you can raise your price
  975. 38:03to whatever you're
  976. 38:04willing to pay. Ah, just recently, the
  977. 38:06company Trobico and its product Clot
  978. 38:09had a conflict with the OpenC community
  979. 38:12.
  980. 38:13Yeah. The
  981. 38:13OpenC community built—if anyone
  982. 38:16doesn't know—it's currently the most
  983. 38:18popular system for agents. It's open
  984. 38:21source, it's not. People got together
  985. 38:25and built a product that's better than
  986. 38:27what all the leading AI labs have built
  987. 38:30. And then the company Ant Troopic
  988. 38:33released a new update and banned its
  989. 38:37use.
  990. 38:37Well, yes, if you're supposedly an open
  991. 38:40system, but you're working on a closed,
  992. 38:42corporate model, then it tends to close
  993. 38:45pretty quickly.
  994. 38:46Well, in this regard, all other markets
  995. 38:48have regulations prohibiting such
  996. 38:50actions. Because that's precisely one
  997. 38:53of the reasons people don't like
  998. 38:55monopolies: one is that they raise
  999. 38:58prices, and the other is that they
  1000. 39:00start pushing their other products on
  1001. 39:03others, crowding out the competition,
  1002. 39:05and offering bundles. Microsoft was
  1003. 39:09ultimately sued in an antitrust case
  1004. 39:13because they said, "If you install
  1005. 39:16our system, our system, it must have
  1006. 39:18our browser, it must have all our
  1007. 39:22products, and you can't pre-install
  1008. 39:24anything else, otherwise we'll simply
  1009. 39:26forbid you from selling our model."
  1010. 39:28Yeah. And that's the logic that's
  1011. 39:29working with both [and][them]. Well,
  1012. 39:31now it turns out that if you want to
  1013. 39:33use agents, you can't use OpenClone.
  1014. 39:36Use our agent, our genetic system.
  1015. 39:39But if you want to use our model,
  1016. 39:41this will inevitably lead to the fact
  1017. 39:45that, since the goal of any company is
  1018. 39:49to make the greatest profit,
  1019. 39:51Yeah, then
  1020. 39:52all competition will ultimately be
  1021. 39:54pushed out.
  1022. 39:55That's it. And I'm simply using the
  1023. 39:56Microsoft example as a starting point.
  1024. 39:58It's clear that Open AI has everything
  1025. 40:00covered by patents, it's all licensed.
  1026. 40:03The law allows it to charge whatever it
  1027. 40:05wants for it, because it belongs to it.
  1028. 40:07Are there legal mechanisms, like in the
  1029. 40:10case of Microsoft, that prevent the
  1030. 40:13owners of these language models from
  1031. 40:16preventing use? It's a difficult answer
  1032. 40:18. But the state can, yes, the
  1033. 40:21prosecutor's office, the state, the
  1034. 40:24consumer protection agency, and so on,
  1035. 40:27find something in this that isn't in
  1036. 40:30the interests of consumers as a whole
  1037. 40:33and sue and deal with it, and so on.
  1038. 40:36All these models are trained on data
  1039. 40:39obtained from all public data from
  1040. 40:42around the world. For example, they use
  1041. 40:46all the books ever published, both past
  1042. 40:49ones, which are now considered a public
  1043. 40:52good, and current ones, which everyone
  1044. 40:56actually pays for. And there are
  1045. 40:58authors of these books. Oh, and it was
  1046. 41:00precisely for this part of the authors
  1047. 41:03'books that they paid a fine. The
  1048. 41:06United States sued them, and they paid
  1049. 41:08a $ 4 billion fine. But if you think
  1050. 41:11about it, it turns out they paid $ 4
  1051. 41:14billion for some billions of books,
  1052. 41:16some hundreds of millions or billions
  1053. 41:18of consumers of these books. That is,
  1054. 41:21they paid some meager cents per person,
  1055. 41:24per book. When one of the creators of
  1056. 41:28one of the major models complained
  1057. 41:30about how Chinese model manufacturers
  1058. 41:33were stealing their intellectual
  1059. 41:36property by training their models on
  1060. 41:39their models. And someone wrote a funny
  1061. 41:42post about this, and I'll analyze the
  1062. 41:45statistics: apparently, Harry Potter
  1063. 41:48was written by this model because it
  1064. 41:51can replicate it with 95%accuracy.
  1065. 41:54And why can it replicate it? Because it
  1066. 41:56used it for training. It's clear that a
  1067. 41:57social contract is needed. It's clear
  1068. 41:59that we all need artificial
  1069. 42:01intelligence. It's clear that
  1070. 42:03artificial intelligence greatly
  1071. 42:05advances us. It's clear that it was
  1072. 42:07truly only possible thanks to the
  1073. 42:09contributions of
  1074. 42:10all these
  1075. 42:11very large numbers of people.
  1076. 42:13If you ask us, we're looking far ahead.
  1077. 42:20This question: is it possible to
  1078. 42:22achieve absolute abundance through AI,
  1079. 42:25and if so, and, in a sense, Open itself
  1080. 42:28—this group, they also gathered there
  1081. 42:30, with the same idea: there is an
  1082. 42:33opportunity to create HI. The means
  1083. 42:36would be good to reach this point: to
  1084. 42:39create shared prosperity for all of
  1085. 42:41humanity, this state of complete
  1086. 42:44abundance, in which robots, not humans,
  1087. 42:47toil, and everyone gets exactly as much
  1088. 42:50as they need and are possible with
  1089. 42:52current resources, so to speak. There's
  1090. 42:55this strange, incredible utopian future
  1091. 42:57, the possibility of achieving it. We
  1092. 43:00can all invest in it together: not wait
  1093. 43:03for corporations to invest in it, for
  1094. 43:05the government to invest in it and use
  1095. 43:08it to their advantage, but all together
  1096. 43:11, invest in a common structure, allow
  1097. 43:13all data to be used on the maximum
  1098. 43:16amount of data, train the best model,
  1099. 43:18make it publicly available to everyone,
  1100. 43:21build a public infrastructure. So, how
  1101. 43:24do you train a model to create a public
  1102. 43:25open source top-tier, but how do you
  1103. 43:27create the infrastructure and even the
  1104. 43:29physical servers in a data center?
  1105. 43:30That's a different question. This layer
  1106. 43:32, in a sense, solved Bitcoin, because
  1107. 43:35Bitcoin, in its technical
  1108. 43:38representation, in its technical
  1109. 43:41incarnation, is a huge distributed data
  1110. 43:44center. The same network can be thought
  1111. 43:47of as hardware that performs
  1112. 43:48calculations and makes them publicly
  1113. 43:54available, distributed, and unblocked.
  1114. 43:59Yes, you have to pay for this, because
  1115. 44:00someone installed this thing. And this
  1116. 44:03creates a market, within which everyone
  1117. 44:06who spent the effort and money to
  1118. 44:08install these servers around the world
  1119. 44:11starts making money. But the protocol
  1120. 44:14works in such a way that the cost of
  1121. 44:16the operation gets cheaper and cheaper.
  1122. 44:19Do
  1123. 44:19you have any kind of relationship with
  1124. 44:21your Chinese comrades in the context of
  1125. 44:23the race, or are they completely closed
  1126. 44:26off?
  1127. 44:26Look, there are a lot of problems with
  1128. 44:27China. Well, we know little about China
  1129. 44:29. We try to go there, try to understand
  1130. 44:32what's going on there. But the problem
  1131. 44:34with China is that there's a huge
  1132. 44:35firewall there. For a long time,
  1133. 44:38blockchains as such were banned. So,
  1134. 44:40for us, China is kind of out of the
  1135. 44:41question. In our current protocol,
  1136. 44:43which we've just created. Many Chinese
  1137. 44:46were there at the beginning of Bitcoin.
  1138. 44:49Many Chinese are still involved in
  1139. 44:51Bitcoin, but outside of China. Many of
  1140. 44:52them are now participating in the race.
  1141. 44:55In one way or another,
  1142. 44:56similar effects are occurring among the
  1143. 44:58Russian-speaking population. Everyone
  1144. 45:00who usually leaves their countries
  1145. 45:03because they see obstacles, and they
  1146. 45:06become even more involved in building
  1147. 45:10systems that would liberate people,
  1148. 45:13make them free. That's why there are a
  1149. 45:16lot of Russian-speaking and Chinese
  1150. 45:18people in blockchain. Among Bitcoin
  1151. 45:22miners, there are many, uh, very strong
  1152. 45:26entrepreneurs from China who created
  1153. 45:29chips for, uh, crypto networks.
  1154. 45:33And here's the thing: right now, for
  1155. 45:36example, the main center, the point of
  1156. 45:39control, of artificial intelligence
  1157. 45:43development is precisely in the
  1158. 45:45computing infrastructure. And almost
  1159. 45:49all of this infrastructure is
  1160. 45:51controlled by a single company, NVIDIA.
  1161. 45:55It also played a very significant role
  1162. 45:59in Bitcoin in 2013. Until three years
  1163. 46:03ago, say, around 2013,
  1164. 46:06NVIDIA GPUs dominated Bitcoin mining,
  1165. 46:11whether it was from AMD or AMD. And
  1166. 46:15then, ah,
  1167. 46:17specialized chips arrived in 2013-2014.
  1168. 46:20And to this day, NVIDIA is no longer
  1169. 46:23there.
  1170. 46:24And at first, Eastern Europeans also
  1171. 46:27played a significant role, and then
  1172. 46:31Chinese entrepreneurs became
  1173. 46:33increasingly important. And now we
  1174. 46:37think that
  1175. 46:38more or less the same AI can be
  1176. 46:39achieved. But these are chips,
  1177. 46:42accordingly, created outside the
  1178. 46:45corporate structure specifically for
  1179. 46:47Bitcoin, not NVIDIA.
  1180. 46:49Some of these people had never created
  1181. 46:51a chip before. It was simply a very
  1182. 46:53strong engineering thought on the part
  1183. 46:56of some whiz kids who saw an
  1184. 46:58opportunity—let's call them
  1185. 47:00enthusiasts, let's call them that—who
  1186. 47:03solved a super-complex engineering
  1187. 47:06problem
  1188. 47:07better than the company that makes
  1189. 47:10money solving it. And they solved it so
  1190. 47:13well that over the 16 years of
  1191. 47:15Bitcoin's existence, the computational
  1192. 47:18efficiency of this protocol and chips
  1193. 47:21has increased 300,000-fold. That
  1194. 47:24is, these calculations are performed so
  1195. 47:29incomparably, insanely efficiently. And
  1196. 47:32all this is mainly thanks to the
  1197. 47:34mechanics of this open market, in which
  1198. 47:37anyone can participate. Oh, and it's
  1199. 47:40direct money right away. Participation
  1200. 47:41equals money. You earn money right away
  1201. 47:44. And this system eventually grows to a
  1202. 47:47size where it begins to form
  1203. 47:49infrastructure, where tens of millions
  1204. 47:52of people are already using it as a
  1205. 47:55currency. And this is not going away.
  1206. 47:58Here, perhaps, in our understanding,
  1207. 48:00the same thing is true. AI isn't going
  1208. 48:02anywhere. We'll only use it more.
  1209. 48:05Much more.
  1210. 48:06So much so that we can barely imagine
  1211. 48:07it. It's just that every light bulb
  1212. 48:09will eventually contain AI, just like
  1213. 48:11electricity. Simply because the result
  1214. 48:15is fundamentally more productive. Yes,
  1215. 48:20this is truly a new stage of
  1216. 48:22dramatically increasing the
  1217. 48:24productivity of each person. If we look
  1218. 48:26at it through the prism of corporations
  1219. 48:28taking away our freedom, we'll lose
  1220. 48:30everything, then our productivity will
  1221. 48:32be zero, it's as if the market doesn't
  1222. 48:33need us at all. But if we look at it
  1223. 48:36from the other side, if it's all public
  1224. 48:38—yes, roads, electricity, the
  1225. 48:40internet—everyone can have access to
  1226. 48:42it, then the productivity of each of us
  1227. 48:45will increase, say, fourfold.
  1228. 48:47And then we'll interact in a new market
  1229. 48:49, in which together we can produce four
  1230. 48:51times as many services, four times as
  1231. 48:53much of everything. We'll simply become
  1232. 48:56four times richer, just as we once did
  1233. 48:59with industrialization. Now we're
  1234. 49:01seeing the next transition from working
  1235. 49:04in offices and companies to working
  1236. 49:07through AI systems. Somewhere around
  1237. 49:102015 or 2016, I, ah, had just started
  1238. 49:13studying blockchain, and I happened to
  1239. 49:16attend a lecture by some philosophy
  1240. 49:18professor who put forward this theory
  1241. 49:21that with the advent of blockchain as a
  1242. 49:24technology, as a philosophy, the need
  1243. 49:27for fiat currencies, which define a
  1244. 49:29person's economic value, would
  1245. 49:32disappear. Well, the easiest way to
  1246. 49:34understand how much you're worth is
  1247. 49:35through money. And the downside of this
  1248. 49:38process is that inequalities arise:
  1249. 49:40when you have a blockchain, a smart
  1250. 49:43contract mechanism, decentralized
  1251. 49:46organizations, then you can refuse
  1252. 49:48money in this decentralized
  1253. 49:51organization that spans the entire
  1254. 49:53world. Your value is determined not by
  1255. 49:56money, not by some boss, but simply by
  1256. 49:59your usefulness, which is already
  1257. 50:01accounted for in this system. It was
  1258. 50:05such a futuristic idea, but what you're
  1259. 50:09saying seems a bit similar.
  1260. 50:11When Bitcoin appeared,
  1261. 50:14it actually introduced a mutable ledger
  1262. 50:16—an irreplaceable ledger. This is
  1263. 50:19immutable money, immutable transactions
  1264. 50:21, uncheatable transactions. Everyone
  1265. 50:23knows who has how much, and no one can
  1266. 50:25cheat. Then Sirium went a little
  1267. 50:27further. They said, "We can make smart
  1268. 50:30contracts." This is the unchangeable
  1269. 50:34execution of a program's script.
  1270. 50:38Yeah.
  1271. 50:38Ah, an
  1272. 50:39unchangeable obligation. There is an
  1273. 50:40unchangeable obligation. The contract
  1274. 50:42will be fulfilled. It doesn't matter
  1275. 50:44how important it is that the conditions
  1276. 50:45are met. It will definitely be
  1277. 50:46fulfilled. There is no middleman. Now,
  1278. 50:48if the same thing happens in the chain
  1279. 50:51from AI, then that's where we get that
  1280. 50:54very crypto-utopia into this
  1281. 50:56vetri-utopia, within which we can build
  1282. 50:59an incorruptible official. We say that
  1283. 51:02we need to remove the middleman, but we
  1284. 51:04can install an algorithmic,
  1285. 51:05incorruptible middleman. And this
  1286. 51:08middleman will constitute almost zero
  1287. 51:11cost. Well, almost zero, because he has
  1288. 51:13a cost of securing it. We need to
  1289. 51:15install hardware, spend electricity,
  1290. 51:17buy chips. If all of this has a price,
  1291. 51:19then we need to do everything possible
  1292. 51:22to drive it down. That is, build a
  1293. 51:24market in which the price of every
  1294. 51:26token, every job performed by this AI,
  1295. 51:29inevitably tends to zero. Under these
  1296. 51:31conditions, we can truly achieve what
  1297. 51:34the crypto-utopians are talking about.
  1298. 51:37And in this regard, with AI, there's an
  1299. 51:40interesting point. Danya often talks
  1300. 51:43about this future world, where everyone
  1301. 51:46has their own robot that works for us,
  1302. 51:48that goes to our factory, and so on.
  1303. 51:51That this world is such an abundance.
  1304. 51:53And people constantly ask him in simple
  1305. 51:56terms: "Well, why doesn't there just be
  1306. 51:58a factory? Why would a factory need
  1307. 52:00your robots? A factory just...
  1308. 52:02Why are they robot masters when it can
  1309. 52:04buy its own?"
  1310. 52:05Yes, that's the most obvious question.
  1311. 52:07But this obvious question seems to be
  1312. 52:10looking at the consumer, right? It
  1313. 52:13looks at the consumer as passive.
  1314. 52:16As if there's no difference. And to
  1315. 52:19understand this world, there's a simple
  1316. 52:23exercise. Let's imagine that there are
  1317. 52:25only settings. And Dania has come up
  1318. 52:29with one that can do everything so that
  1319. 52:32neither your work nor mine is needed
  1320. 52:35anymore.
  1321. 52:37And let's say you have everything you
  1322. 52:39have left: the land, your home. You
  1323. 52:40live in it. I can come and say: "Give
  1324. 52:42away your houses, because everything
  1325. 52:43that can happen to me, everything that
  1326. 52:45is needed, only I can give you. You
  1327. 52:46have nothing else except your house.
  1328. 52:48That's it. But it is very important
  1329. 52:50that if he does not need our labor, yes
  1330. 52:53, our labor will not add anything to
  1331. 52:55his life, then
  1332. 52:56nothing bothers us.
  1333. 53:00Nothing bothers us. The world has not
  1334. 53:01become worse for us, because you and I,
  1335. 53:03even if we do not have this great" yay
  1336. 53:05", you and I can exchange this labor.
  1337. 53:06We do not need it. The factory that
  1338. 53:09takes us all away. But at that moment,
  1339. 53:11when we think that we do not need it,
  1340. 53:14then this appears, what is described in
  1341. 53:16that example, yes, that m Dao is a kind
  1342. 53:19of closed system in which we begin to
  1343. 53:21exchange when we do not need universal
  1344. 53:23money, in which
  1345. 53:24we have really all disappeared, right?
  1346. 53:27But we can create money between us,
  1347. 53:30which denotes an exchange between us.
  1348. 53:32And in this regard, the consumer is
  1349. 53:34also a creator, You understand? Because
  1350. 53:37it's only because you want to use my
  1351. 53:38labor, and I want to consume your labor
  1352. 53:40, that's the only reason this labor is
  1353. 53:42valuable at all. And just imagine that
  1354. 53:44in reality, if you fast-forward 30
  1355. 53:48years—it's not like that, not that
  1356. 53:50much, but you still have to
  1357. 53:51fast-forward as far as you can. Where
  1358. 53:52exactly this super-super-artificial
  1359. 53:54intelligence is, it's not exactly a
  1360. 53:56question of whether it will be in 2
  1361. 53:58years, 5 years, 10 years, or 30 years,
  1362. 54:00we can guess, but let's assume we've
  1363. 54:02definitely moved ourselves far ahead.
  1364. 54:05Yeah. And in this distant future, there
  1365. 54:08are two very different scenarios. One,
  1366. 54:11in which all this is owned by
  1367. 54:13corporations and everyone else lives,
  1368. 54:15maybe, only on some kind of benefit and
  1369. 54:17, in general, has lost the meaning of
  1370. 54:19life and is very dependent on it. Or
  1371. 54:23the other, in which we have a future
  1372. 54:25future, in which we have all become so
  1373. 54:28much stronger, precisely because
  1374. 54:30intelligence is so democratized, and
  1375. 54:33people create very different things,
  1376. 54:36and consume each other, because it's
  1377. 54:38interesting consume each other. So this
  1378. 54:41labor, as I say, is four times. Why am
  1379. 54:43I saying this figure of four times?
  1380. 54:45Firstly, statistics from the last 90
  1381. 54:48years show that every 30 years, human
  1382. 54:51productivity has quadrupled. Now we are
  1383. 54:54on the verge of a similar breakthrough,
  1384. 54:56which will be achieved in the next 30
  1385. 54:58years. Why? Because if every person
  1386. 55:01simply has a robot working in their
  1387. 55:03interests, both intelligence and a
  1388. 55:06robot. And why will there be a robot?
  1389. 55:07Because it will cost 10-20,000 dollars.
  1390. 55:09We are already seeing this. The market
  1391. 55:11is already moving towards this. Ying
  1392. 55:12promises that it will cost 20. The
  1393. 55:14Chinese are churning it out. 500
  1394. 55:16companies now, I'm not kidding, 500
  1395. 55:18plus companies are now engaged in the
  1396. 55:21creation of robots. Robots will be
  1397. 55:23accessible. The question would be, why?
  1398. 55:26And the question would be,
  1399. 55:27understanding how to use all this. And
  1400. 55:29every person can have a robot. If every
  1401. 55:32person has their own robot, which is
  1402. 55:34exclusively in your interests If it
  1403. 55:36works, your personal robot, then 8
  1404. 55:38hours of your work, you sleep, 8 hours
  1405. 55:40of rest, 8 hours of work and another 3
  1406. 55:42times 8,
  1407. 55:43yes,
  1408. 55:43your robot, who just works with you,
  1409. 55:46maybe you take another 8 from him and
  1410. 55:48say:" Let's have fun together, my 8,
  1411. 55:51your 8, let's have fun together. "And
  1412. 55:53you only work 16 hours, not all 24. But
  1413. 55:55actually, just a second, you will have
  1414. 55:58fun with other people, and the robots
  1415. 56:00will teach you productivity together
  1416. 56:02with you. Build four times more houses,
  1417. 56:05clean the ocean four times more, plant
  1418. 56:07four times more lawns, replant trees.
  1419. 56:10And there is a very big gap between
  1420. 56:13these two versions of reality. And you
  1421. 56:15are right, yes,
  1422. 56:16that we don't care
  1423. 56:18how much money we earn, but we still
  1424. 56:19want to be in the second world. Well,
  1425. 56:21that is, we definitely need to make an
  1426. 56:23effort not to end up in the first, but
  1427. 56:25to end up in the second. And we don't
  1428. 56:27have the option of
  1429. 56:29not ending up in any of them, you know,
  1430. 56:31just stopping so that I didn't develop
  1431. 56:33any further, and so on. This is not
  1432. 56:34true. This is not an option. There is
  1433. 56:36no Switch killer who will kill this,
  1434. 56:38like, and it won't happen, yes. Someone
  1435. 56:41will periodically throw Molotov
  1436. 56:43cocktails there, threatening the people
  1437. 56:46who create
  1438. 56:48these, and they, they, they will be
  1439. 56:51created.
  1440. 56:52Fact. We will all use the GPT chat. In
  1441. 56:54doing so, we are paying for the
  1442. 56:56development of the GPT chat and the
  1443. 56:57development of these future
  1444. 56:59corporations. Simply by doing nothing
  1445. 57:01or simply using these tools, we are
  1446. 57:04paying for the construction of their
  1447. 57:06infrastructure, when we could actually
  1448. 57:09pay for the construction of public
  1449. 57:11infrastructure. And the mere presence
  1450. 57:14of public infrastructure leads to a
  1451. 57:17completely different market.
  1452. 57:19Now hundreds of billions are being
  1453. 57:21invested in this infrastructure. The
  1454. 57:23first thing to think about is who will
  1455. 57:25have to pay for these 100 billion. That
  1456. 57:27in fact, every investment is bad news
  1457. 57:28for us. Every time the capitalization
  1458. 57:30of these companies grows more and more,
  1459. 57:32it's bad news for us all, because we
  1460. 57:34will all have to pay for it.
  1461. 57:35Well, yeah, you We'll have to pay so
  1462. 57:37that the people who invested in this
  1463. 57:39can profit.
  1464. 57:40While they're competing with each other
  1465. 57:42, and there are five of them here, five
  1466. 57:44there, we all essentially have to pay
  1467. 57:4610 times more. Well, because we could
  1468. 57:48have built this public infrastructure
  1469. 57:51once, but now they'll all build
  1470. 57:53capacity with Drum, competing to steal
  1471. 57:55each other's clients. By the way,
  1472. 57:58there's an understanding of the use of
  1473. 58:00computing power for AI, that is, what
  1474. 58:03it's mainly used for now. All this
  1475. 58:06chatting, all this chatting, is a
  1476. 58:08pittance for the current consumer.
  1477. 58:10Got it?
  1478. 58:11Ah, it's already a pittance. It all
  1479. 58:13started with that. It allowed us to get
  1480. 58:14everyone involved. Ah, but when you
  1481. 58:17send 100 words to a chat, you get 100
  1482. 58:22words back.
  1483. 58:23Yeah. Well, not 100, 1,000. Now that
  1484. 58:25they're still thinking, yes, there's a
  1485. 58:27process called rendering, otherwise
  1486. 58:29there are already 1,000 words,
  1487. 58:31including what, you know, he's spinning
  1488. 58:33around and thinking. There, what you
  1489. 58:35think is right.
  1490. 58:36Right or wrong, but what did the user
  1491. 58:38mean? Now I'll think about it some more
  1492. 58:40.
  1493. 58:40When people use this professionally,
  1494. 58:43for example, for programming, then
  1495. 58:46I send 100 words and I get at least
  1496. 58:50100,000 words back for every word I
  1497. 58:53write. Agents. That is, when you launch
  1498. 58:58it and tell this system:" Let's have
  1499. 59:01you report information to me on this "
  1500. 59:04on this, about—when you ask it to do
  1501. 59:08certain things for you: work with
  1502. 59:10tables in Excel, create websites, etc.
  1503. 59:14—for every 100 words you write,
  1504. 59:16millions of words are sent back.
  1505. 59:19Millions." And, accordingly, it's like
  1506. 59:21a huge risk that everything, at some
  1507. 59:23point, will all close, because these
  1508. 59:25models, they are corporate, because
  1509. 59:27they are designed to make money. This
  1510. 59:30is how the race itself is structured in
  1511. 59:32this regard, this decentralized story
  1512. 59:35with, and as we mentioned, exactly what
  1513. 59:38happened, and this is what happened
  1514. 59:41just in the last few months,
  1515. 59:43and that the emergence of OpenCore, and
  1516. 59:46it gave a powerful tool to a very large
  1517. 59:48number of people to start using
  1518. 59:51artificial intelligence to automate
  1519. 59:53their work, create new businesses, as
  1520. 59:56if to do some kind of...These are more
  1521. 59:58serious things than we had all done in
  1522. 1:00:01chats before. But at the same time,
  1523. 1:00:03people began to notice that it was a
  1524. 1:00:06bit expensive. That is, they use it and
  1525. 1:00:08they already have hundreds of dollars a
  1526. 1:00:10week, they began to see that their
  1527. 1:00:12credit cards are being withdrawn.
  1528. 1:00:14Many were surprised that for a couple
  1529. 1:00:16of nights, a couple of hundred dollars
  1530. 1:00:17so far. That's it. And from this, a
  1531. 1:00:20noticeable movement arose to connect
  1532. 1:00:24open models to OpenCore and look for
  1533. 1:00:28some other alternatives.
  1534. 1:00:31And, that is, the same In our network,
  1535. 1:00:34it's clear that the main usage is
  1536. 1:00:37agents right now, precisely because
  1537. 1:00:40agents, unlike people, don't have any
  1538. 1:00:44loyalty to the Open AI or Antropic
  1539. 1:00:47brands. Often, when you tell them
  1540. 1:00:50directly: "Find yourself a
  1541. 1:00:53cheaper model, here's a credit card, go
  1542. 1:00:56find it, pay," they'll go and find it
  1543. 1:00:59cheaper.
  1544. 1:00:59That's it. And in this regard, the
  1545. 1:01:02advantage of blockchain protocols is
  1546. 1:01:07that this growth is ensured by the
  1547. 1:01:11issued coin. Fewer and fewer of them
  1548. 1:01:14are being issued. And therefore, when
  1549. 1:01:16you ask agents: "Find me
  1550. 1:01:18the cheapest, the
  1551. 1:01:19cheapest route," then for them, uh,
  1552. 1:01:21they'll search on the internet, they'll
  1553. 1:01:23compare prices, and they'll quickly
  1554. 1:01:25come to the conclusion that it's
  1555. 1:01:27necessary, of course, to use decimated
  1556. 1:01:29networks, because that's the best thing
  1557. 1:01:31. And we usually, when we finally talk
  1558. 1:01:34about this topic in the testing of
  1559. 1:01:36Soviet scientists Intelligentsia, we're
  1560. 1:01:39not talking about our protocol and our
  1561. 1:01:41product specifically, but we're talking
  1562. 1:01:44about how de-clasticization is the way
  1563. 1:01:46to go.
  1564. 1:01:47Yeah. Yes. We've tried to implement in
  1565. 1:01:49our product
  1566. 1:01:50everything that we believe will lead to
  1567. 1:01:53the construction of this public
  1568. 1:01:55infrastructure. That is, this is how we
  1569. 1:01:57see the specific rules of this protocol
  1570. 1:01:59that should lead to the construction of
  1571. 1:02:01an infrastructure worth hundreds of
  1572. 1:02:03billions of dollars.
  1573. 1:02:04But we strongly encourage everyone to
  1574. 1:02:07double-check, look at different ones,
  1575. 1:02:09and encourage your agents to find the
  1576. 1:02:12most effective. And this is what's
  1577. 1:02:15already happening and will only happen
  1578. 1:02:18more. Every time people encounter
  1579. 1:02:20higher prices or every time they
  1580. 1:02:22encounter being disconnected from the
  1581. 1:02:24model, just like the OpenCell community
  1582. 1:02:26at some point discovered that they were
  1583. 1:02:28all disconnected from the OtropiC model
  1584. 1:02:30. The more such cases there are, the
  1585. 1:02:34more people will join in the
  1586. 1:02:37consumption and, in particular, the
  1587. 1:02:40protocol race.
  1588. 1:02:43And this is the same thing that
  1589. 1:02:44happened with Bitcoin. With Bitcoin
  1590. 1:02:45What was happening? You can see that
  1591. 1:02:47every time there was some kind of
  1592. 1:02:50economic crisis, yep,
  1593. 1:02:51people would join in because, well,
  1594. 1:02:53it's clear, yeah, in calm times you
  1595. 1:02:54don't really think about it, well, what
  1596. 1:02:56are the numbers, a
  1597. 1:02:56dollar, a dollar, they pay salaries, I
  1598. 1:02:59buy coffee,
  1599. 1:02:59and then a crisis happens, you realize
  1600. 1:03:01that no, there's some big problem with
  1601. 1:03:03modern finance, something needs to be
  1602. 1:03:05changed and alternatives needed to be
  1603. 1:03:07found.
  1604. 1:03:08Each such event was also a big
  1605. 1:03:11development for us for the protocol we
  1606. 1:03:15created. Take Open Claw and all this
  1607. 1:03:18open source movement, the creation of
  1608. 1:03:22agents. And at the end of last year,
  1609. 1:03:24everyone thought that agents were
  1610. 1:03:27something completely under control, and
  1611. 1:03:29not these large existing corporations.
  1612. 1:03:32And now it's not just OpenClaw, there's
  1613. 1:03:34Gstack, there's Hermes, there's that.
  1614. 1:03:37That is, in fact, it's already
  1615. 1:03:39unstoppable; there will be a huge
  1616. 1:03:41number of free open source projects,
  1617. 1:03:44because what is it? Essentially, it's
  1618. 1:03:47just someone sitting down (not even one
  1619. 1:03:49person, but a group of people) and
  1620. 1:03:51putting their own vision and knowledge
  1621. 1:03:53into in the form of text, simply
  1622. 1:03:55formulated processes, protocols, and
  1623. 1:03:58operations in text—the projects are
  1624. 1:04:00written, and from them a certain
  1625. 1:04:02personality is created, a certain
  1626. 1:04:04behavioral structure, work processes,
  1627. 1:04:06which are then passed through AI models
  1628. 1:04:09to execute your tasks, which you
  1629. 1:04:11already say in the line: "Do it for me,
  1630. 1:04:13please." So, it, having all these
  1631. 1:04:15processes, runs there in the model,
  1632. 1:04:18calculates, returns the results and
  1633. 1:04:20continues to spin inside its set
  1634. 1:04:22processes. And then you really have two
  1635. 1:04:24paths. One is to try to make the most
  1636. 1:04:27of this, or we all share this knowledge
  1637. 1:04:30and we all only make each other
  1638. 1:04:33stronger.
  1639. 1:04:33And it turns out that there are enough
  1640. 1:04:36people, including those who have
  1641. 1:04:38already earned money, but know how, who
  1642. 1:04:40sit down, write and post it for free
  1643. 1:04:42for everyone. Here is precisely the
  1644. 1:04:44clarification. That is, the agent to
  1645. 1:04:46whom you say: "Find me a cheaper model,
  1646. 1:04:48" has the opportunity to choose. That
  1647. 1:04:51is, as if the world is not reduced to
  1648. 1:04:52these three or these three models.
  1649. 1:04:57It turned out that all these companies,
  1650. 1:04:59they can spend Well, billions of
  1651. 1:05:01dollars already spent on model training
  1652. 1:05:03.
  1653. 1:05:04Yeah. But as soon as they make it
  1654. 1:05:07available to anyone else, through
  1655. 1:05:09communication with this model, to
  1656. 1:05:12create a model that is practically as
  1657. 1:05:15powerful as the copied product, yeah,
  1658. 1:05:18where did we start,
  1659. 1:05:20right? Well, even, that is, you can't
  1660. 1:05:22even protect it, you know, it turned
  1661. 1:05:24out that this could have happened. We
  1662. 1:05:26predicted the emergence of DeepSky more
  1663. 1:05:27than a year in advance, everyone at
  1664. 1:05:28home was saying, "It's impossible not
  1665. 1:05:29to catch up with everything open, not
  1666. 1:05:30to catch up with Anthropic, not to
  1667. 1:05:31catch up with Google, forget about it."
  1668. 1:05:33And speaking of which, why? How easy is
  1669. 1:05:35it to catch up? Just imagine that
  1670. 1:05:37everything they've done is based on
  1671. 1:05:39publicly published scientific articles,
  1672. 1:05:42publicly available data. Then you need
  1673. 1:05:45a lot of calculations, and then, once
  1674. 1:05:47you've done these calculations, you can
  1675. 1:05:49test your own against their model,
  1676. 1:05:50while your model communicates with
  1677. 1:05:52their model to train itself. That is,
  1678. 1:05:54for now, this bottleneck has turned out
  1679. 1:05:57to be precisely in computing power. And
  1680. 1:05:59this is where reality control exists
  1681. 1:06:02now.
  1682. 1:06:02And Where will it remain if we don't
  1683. 1:06:04build a public infrastructure? So you
  1684. 1:06:06understand, Deepsk became possible
  1685. 1:06:09thanks to the fact that the company
  1686. 1:06:12that was trading on the exchange bought
  1687. 1:06:16GPUs for trading on the exchange. And
  1688. 1:06:19they put them there for a while. And so
  1689. 1:06:21they decided, well, let's just leave
  1690. 1:06:22them idle and calculate. And
  1691. 1:06:24why don't we calculate our model, like
  1692. 1:06:25those guys, put the ones posted on it.
  1693. 1:06:27We're interested, we're all scientists,
  1694. 1:06:29we all study mathematics here. Let's
  1695. 1:06:30take all this and calculate it on our
  1696. 1:06:32GPUs.
  1697. 1:06:32They had a lot of GPUs. They had about
  1698. 1:06:361,500, uh, the
  1699. 1:06:39most powerful GPUs at the time.
  1700. 1:06:43And that's how they managed to do it.
  1701. 1:06:45And what we want to show people is that
  1702. 1:06:49the power is in them, yes. In our
  1703. 1:06:52community now, it's not in chips, not
  1704. 1:06:54in Chinese, but in people.
  1705. 1:06:56People. There are 15,000 people in our
  1706. 1:06:59community. And this community is
  1707. 1:07:02already They assembled 3,000 of these
  1708. 1:07:04same
  1709. 1:07:05chips and connected more of the
  1710. 1:07:06existing ones to the common network.
  1711. 1:07:08That is, they didn't have to buy them,
  1712. 1:07:10they didn't have to build anything new.
  1713. 1:07:12Most people even simply borrowed these
  1714. 1:07:14chips from data centers and cloud
  1715. 1:07:16providers.
  1716. 1:07:17That's it. But then, the next step is
  1717. 1:07:20that yes, there will be 5,000, 10,000,
  1718. 1:07:2320,000 of them. But the protocol itself
  1719. 1:07:27, because it's open, has no barriers.
  1720. 1:07:31And what's happening now? That
  1721. 1:07:34engineers around the world are just as
  1722. 1:07:37skilled as they were with Bitcoin—
  1723. 1:07:40yeah,
  1724. 1:07:41who just made chips for Bitcoin or
  1725. 1:07:44Ereum. They're like, "But this is the
  1726. 1:07:47future." And there will be hundreds of
  1727. 1:07:50times more computing than Bitcoin ever
  1728. 1:07:53had. "And they started making chips to
  1729. 1:07:56connect to our protocol, so that people
  1730. 1:07:59no longer have to buy a modern chip
  1731. 1:08:02because it's expensive. That is, to put
  1732. 1:08:05this in perspective, if you try to
  1733. 1:08:09build the Bitcoin network today with
  1734. 1:08:12all its computing power, but on Nvidia
  1735. 1:08:15chips, you'll need to spend 5
  1736. 1:08:18quadrillions of dollars.
  1737. 1:08:22This is not the way we need to build
  1738. 1:08:25infrastructure for AI. But what this
  1739. 1:08:29crypto community, which everyone
  1740. 1:08:30perceives, has shown, you know, they
  1741. 1:08:32haven't done anything. No, they've done
  1742. 1:08:34a lot. These are incredible engineering
  1743. 1:08:37breakthroughs in chip construction. And
  1744. 1:08:40now similar breakthroughs, yes, this
  1745. 1:08:43equipment needs to be redesigned,
  1746. 1:08:45because the equipment used for Bitcoin
  1747. 1:08:48cannot be used for slahiTektu, but the
  1748. 1:08:51principles they came up with, how they
  1749. 1:08:54layer memory on chips and so on, they
  1750. 1:08:57are now using everything to create ASX.
  1751. 1:09:00Already this year, already this year,
  1752. 1:09:03new chips will be released, not from
  1753. 1:09:05Nvidia, not from Google, and not from
  1754. 1:09:07Amazon, but from completely independent
  1755. 1:09:10companies. Nvidia even bought one of
  1756. 1:09:12these GRKyu last year, this year they
  1757. 1:09:14even integrated their new platform. And
  1758. 1:09:17there are many new manufacturers like
  1759. 1:09:19this, and these chips will be
  1760. 1:09:22competitively much better, cheaper,
  1761. 1:09:24like this, more productive for their
  1762. 1:09:27cost. The same people who produced
  1763. 1:09:29chips for Bitcoin in the past have been
  1764. 1:09:32working for many years to apply their
  1765. 1:09:35knowledge to the construction of AI
  1766. 1:09:37chips.
  1767. 1:09:38And now they are releasing them this
  1768. 1:09:40year, this year, next year. Edchet in
  1769. 1:09:42the US, NTR in Canada, Talas in Canada,
  1770. 1:09:46accelerators somewhere in Europe, I
  1771. 1:09:49don't remember where, some Chinese
  1772. 1:09:51companies, Taiwanese companies, Korean
  1773. 1:09:55companies are producing new chips. They
  1774. 1:09:58all, uh, when they started, might not
  1775. 1:10:01even have thought about decentralized
  1776. 1:10:03networks. They might have thought:"
  1777. 1:10:05We'll just sell this to those same
  1778. 1:10:07Facebooks, Googles, and others, "
  1779. 1:10:08whoever needs it." But now it's
  1780. 1:10:10becoming clear that all these Facebook
  1781. 1:10:13corners, they've either already bought
  1782. 1:10:15whoever they wanted, or are even
  1783. 1:10:16producing new chips on their own in
  1784. 1:10:18this club of theirs, a closed club of
  1785. 1:10:20boy-bunnies, where even the five of
  1786. 1:10:22them have taken up the entire SMC,
  1787. 1:10:24taken up all the memory, and so on. So
  1788. 1:10:27they sit there and it's all their club.
  1789. 1:10:29And these small manufacturers are
  1790. 1:10:32suddenly realizing that they can use a
  1791. 1:10:35decentralized network, connect
  1792. 1:10:37non-their chips and earn coins. Because
  1793. 1:10:40now it's the same as it was with
  1794. 1:10:43Bitcoin. And what is the coin imitated
  1795. 1:10:46in Bitcoin here? A token.
  1796. 1:10:48Well, it's the same in our example of
  1797. 1:10:52our protocol,
  1798. 1:10:53the same thing. This protocol has its
  1799. 1:10:55own coin, which has a limited supply,
  1800. 1:10:58and only those who mine receive it.
  1801. 1:11:01This and this coin, it represents a
  1802. 1:11:04certain opportunity to either get
  1803. 1:11:06computation now, or wait 10 years and
  1804. 1:11:09get computation when this network is
  1805. 1:11:12much larger. That's exactly what
  1806. 1:11:14I was asking. That is, it's not In
  1807. 1:11:16itself, in a vacuum, it's tied
  1808. 1:11:18specifically to access to
  1809. 1:11:20this currency, which is then used to
  1810. 1:11:23pay for its consumption. Hey, you
  1811. 1:11:25either use it now, when there are 3,000
  1812. 1:11:28GPUs in the network, or you use it in 5
  1813. 1:11:31years, when there are millions of GPUs.
  1814. 1:11:35And it's clear that, accordingly, you
  1815. 1:11:36can get significantly more computing
  1816. 1:11:38power later and then
  1817. 1:11:39sell these coins to someone, and they
  1818. 1:11:41will use the computing power.
  1819. 1:11:42You just said 3,000, yes,
  1820. 1:11:44every time we mention a number in an
  1821. 1:11:45interview, by the time the interview
  1822. 1:11:47comes out, it's already higher. Now, in
  1823. 1:11:49fact, it's already 3,700, and most
  1824. 1:11:51likely, by the time people hear this,
  1825. 1:11:53it will be 5,000 or more. And what's
  1826. 1:11:55the safety net? This decentralized
  1827. 1:11:57network, it must always remain
  1828. 1:11:58decentralized. Roughly speaking, there
  1829. 1:12:00shouldn't be an imbalance in favor of
  1830. 1:12:02someone who has more of these coins,
  1831. 1:12:04who can build more of these capacities.
  1832. 1:12:06Coins have nothing to do with it. More
  1833. 1:12:07Coins, please, this is for the market.
  1834. 1:12:10If you want to buy up all the coins,
  1835. 1:12:11buy up all the coins. The importance of
  1836. 1:12:13this network is that it is not defined
  1837. 1:12:15by the coins, but by who provides the
  1838. 1:12:17capacity. That is, to have a voice and
  1839. 1:12:20influence in this network's vote on
  1840. 1:12:22where this network is going, you need
  1841. 1:12:25to add hardware.
  1842. 1:12:26When we launched this network, from day
  1843. 1:12:29one, we made it completely independent
  1844. 1:12:32of us.
  1845. 1:12:33Even regardless of what we say about it
  1846. 1:12:35and so on, we have no control over it.
  1847. 1:12:37We were often discouraged; they said, "
  1848. 1:12:38No, well, you have to have some kind of
  1849. 1:12:39control."
  1850. 1:12:40All other cryptocurrencies are like
  1851. 1:12:41this, back and forth. You're like, if
  1852. 1:12:43we're creating a physical
  1853. 1:12:45infrastructure,
  1854. 1:12:46and people should believe that no one
  1855. 1:12:49will ever use it,
  1856. 1:12:51complete control. And that's exactly
  1857. 1:12:53what needs to be given. Then it's all
  1858. 1:12:55determined by the protocol. And then
  1859. 1:12:57it's all determined, accordingly, by
  1860. 1:13:01these many independent actors that
  1861. 1:13:04define this network. Our thesis is that
  1862. 1:13:08even a non-profit organization, even a
  1863. 1:13:10charitable organization, wouldn't be a
  1864. 1:13:13good form, because in a charitable
  1865. 1:13:15organization there's still a board of
  1866. 1:13:17directors,
  1867. 1:13:18and then the board of directors usually
  1868. 1:13:20elects itself. And that means it's
  1869. 1:13:22still a certain power. And the only way
  1870. 1:13:24is, in Bitcoin, there's no charitable
  1871. 1:13:26organization, there's no organization,
  1872. 1:13:28and it works perfectly well. And
  1873. 1:13:29everything is determined by the
  1874. 1:13:31protocol. And this protocol has one
  1875. 1:13:33goal: for the infrastructure to grow
  1876. 1:13:37and for this infrastructure to be free,
  1877. 1:13:40accessible, and for everyone who
  1878. 1:13:43wants to buy up all this capacity,
  1879. 1:13:45let's assume how this is regulated in
  1880. 1:13:47Bitcoin, by the way. Not at all, not at
  1881. 1:13:49all. It's regulated. It's regulated by
  1882. 1:13:52That as soon as it grows...
  1883. 1:13:55Yeah. The
  1884. 1:13:55number of participants increases, so
  1885. 1:13:57the price of the coin increases. As
  1886. 1:13:58soon as the price of the coin increases
  1887. 1:13:59, the others look and say, "Oh, it's
  1888. 1:14:01growing. Why am I such a fool that I'm
  1889. 1:14:02not making money?""And they join too.
  1890. 1:14:05And what's more, at the same time,
  1891. 1:14:07whenever you want it, whenever you want
  1892. 1:14:10it...This is a theory, but it has been
  1893. 1:14:12empirically proven by Bitcoin. Oh, and
  1894. 1:14:14we had the same situation, especially
  1895. 1:14:16at the beginning of the network. At the
  1896. 1:14:18beginning, there were very few. Eight
  1897. 1:14:20participants, whom we found independent
  1898. 1:14:22of each other, experienced people with
  1899. 1:14:25a reputation. We found them and said:"
  1900. 1:14:27Here you add the first GP, you launch
  1901. 1:14:29the network. "They launched the network
  1902. 1:14:31, and it seemed like there were only
  1903. 1:14:326480 GPUs. There very quickly became 80
  1904. 1:14:34in a couple of days. And critics
  1905. 1:14:36were saying that
  1906. 1:14:37someone could add 100 tomorrow and
  1907. 1:14:38that's it, and they would have half,
  1908. 1:14:40and everything would be gone. Oh, and
  1909. 1:14:42our thesis was that the moment someone
  1910. 1:14:45wants to add 100, someone else is
  1911. 1:14:49already thinking about it too. And
  1912. 1:14:50that's how it happens. It doesn't
  1913. 1:14:52happen in a vacuum. We go out and give
  1914. 1:14:55interviews so that as many people as
  1915. 1:14:57possible know, so that the idea Whether
  1916. 1:14:59this network was ready for 10,000 new
  1917. 1:15:01GPUs or not didn't come to one person,
  1918. 1:15:04but to many at once.
  1919. 1:15:06It works in a surprising way. So, from
  1920. 1:15:08those sixty GPUs, the network grew to
  1921. 1:15:103,000, and there wasn't a single moment
  1922. 1:15:13where anyone gained significant power.
  1923. 1:15:15Conca, meanwhile, isn't really a
  1924. 1:15:17company; it's, so to speak, a protocol.
  1925. 1:15:19It's an OpenS protocol, hosted on
  1926. 1:15:21GitHub, and people are already using it
  1927. 1:15:23and contributing code to it without us.
  1928. 1:15:25Overall, there are no restrictions on
  1929. 1:15:27anyone's participation. Right now, you
  1930. 1:15:30can go to Nibius's website, that Neo
  1931. 1:15:34Cloud in Volzhe, rent a couple of GPUs
  1932. 1:15:38for a couple hundred dollars a day, and
  1933. 1:15:42go mine the race. Anyone can do it, and
  1934. 1:15:45you
  1935. 1:15:45don't have to negotiate with anyone. We
  1936. 1:15:47don't have to give anyone permission.
  1937. 1:15:49It's happening.
  1938. 1:15:50You go to the website, open the
  1939. 1:15:51documentation, read how to install,
  1940. 1:15:53then, if you don't understand it, you
  1941. 1:15:54go to the community, there are 15,000
  1942. 1:15:56people there. Everyone has long
  1943. 1:15:58understood everything, explained it,
  1944. 1:16:00agreed, made the instructions simpler
  1945. 1:16:02than ours. Ours is too technical, too
  1946. 1:16:04scientific. Someone made it simpler,
  1947. 1:16:06someone recorded YouTube videos. All
  1948. 1:16:08this exists. What's
  1949. 1:16:09my logic? Let's say I'm a venture fund
  1950. 1:16:12or a pension fund, and I invested in
  1951. 1:16:15Open and look at the race and say:"
  1952. 1:16:19Okay, the logic is that if this product
  1953. 1:16:22takes off and starts working, then
  1954. 1:16:25these billions invested in data centers
  1955. 1:16:29will turn into zero. And all my money
  1956. 1:16:31will turn...Well, not into zero, they
  1957. 1:16:32won't turn into zero. They can all join
  1958. 1:16:34the race,
  1959. 1:16:35ah,
  1960. 1:16:36and also update the race. Well, your
  1961. 1:16:39idea is this: if this desire for AI to
  1962. 1:16:43be free is not among 15,000 people, but
  1963. 1:16:47among billions of people in the future,
  1964. 1:16:51then this will be a much more valuable
  1965. 1:16:55network. And then, as usual, they all
  1966. 1:17:01come here and want money. But you can't
  1967. 1:17:04invest money here. You can participate
  1968. 1:17:06in mining or buy coins from other
  1969. 1:17:09miners.
  1970. 1:17:09So, to continue my example, I'm looking
  1971. 1:17:12at it, I understand that all these data
  1972. 1:17:15centers won't go broke, but I don't
  1973. 1:17:18believe they'll join the race. I
  1974. 1:17:21understand that, basically, I'm losing
  1975. 1:17:24this profit. So, I need to somehow...
  1976. 1:17:26But I can't give you money. Buy.
  1977. 1:17:32Well, you either go to these miners,
  1978. 1:17:34you go to these miners who have already
  1979. 1:17:36mined, buy out their coins,
  1980. 1:17:37or rent a GPU and start mining too. It
  1981. 1:17:41was very inconvenient for them, you
  1982. 1:17:42know, because the eternal capitalists
  1983. 1:17:43are used to the fact that they just
  1984. 1:17:45need to deposit money. And it's kind of
  1985. 1:17:46funny, really funny, you understand.
  1986. 1:17:48When they came in the first or second
  1987. 1:17:50month,
  1988. 1:17:50they were like, "We want to deposit, I
  1989. 1:17:52don't know, a million dollars." We say,
  1990. 1:17:54"There's nowhere to put it. If you want
  1991. 1:17:55, you can go buy it on the market. But
  1992. 1:17:57wait a second, there aren't enough
  1993. 1:17:59coins on the market yet, not enough
  1994. 1:18:01coins have been mined for you to buy
  1995. 1:18:03them with your million dollars.
  1996. 1:18:05Therefore, your only chance to invest
  1997. 1:18:07your million in these coins is to go
  1998. 1:18:09and mine these coins yourself. Okay, I
  1999. 1:18:11get it. It was such a, uh, process, too
  2000. 1:18:14, because it's a strange structure, yes
  2001. 1:18:17, it's designed precisely to create
  2002. 1:18:20some, uh, public domain, but in some
  2003. 1:18:23kind of framework, in which there is
  2004. 1:18:25still some, uh, commercial motivation
  2005. 1:18:28for people. Those same venture
  2006. 1:18:31capitalists who were looking for a way
  2007. 1:18:33to somehow get involved in this project
  2008. 1:18:35. We also usually respond very sharply
  2009. 1:18:38that, first and foremost, what's
  2010. 1:18:39interesting is the participation of
  2011. 1:18:41people who believe in this future, who
  2012. 1:18:44want this future to happen. Here,
  2013. 1:18:46taking into account other restrictions,
  2014. 1:18:48here, you can't just come and bring
  2015. 1:18:50money, you need to do something. This
  2016. 1:18:52was not the case, and at first, it was
  2017. 1:18:53especially It's a complex natural
  2018. 1:18:55product of some sort. Well, well, well,
  2019. 1:18:56it was
  2020. 1:18:57A who put in the effort to make it what
  2021. 1:19:00it is now. Overall, again, it was a
  2022. 1:19:03difficult process, two years in our
  2023. 1:19:05lives and so on, but we've already
  2024. 1:19:07earned enough money to not worry about
  2025. 1:19:10the time we
  2026. 1:19:13spent on it. But the future of this
  2027. 1:19:17protocol depends solely on how many
  2028. 1:19:20people believe that this alternative
  2029. 1:19:24needs to exist for this infrastructure
  2030. 1:19:29to be truly meaningful.

About this transcript

This page contains the full transcript of Братья Либерманы. Почему ИИ сделает одних сверхбогатыми, а других – нищими? by Forbes, generated from the public captions YouTube serves with the video. The transcript has 12,346 words across 2,030 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.