YT YouTube2Text

YouTube2Text → Transcripts → Por qué CHINA regala su IA al mundo

Por qué CHINA regala su IA al mundo — Transcript

by Gustavo Entrala · 4,766 words · 740 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:01Thursday, July 16, 2026, a Beijing
0:03laboratory called Moonshot launches
0:06Chemistry 3, the largest open-source
0:08artificial intelligence model ever
0:10released. In various tests, Chemistry 3
0:15is on par with the best American ones,
0:17yes, the ones you and I use every day.
0:21The key difference between this new
0:23Chinese model and the American ones is
0:26that anyone can download Chemistry 3
0:28for free. Friday, July 17, 100 km away
0:33in Shanghai, Xi Jinping takes the stage
0:36at the World AI Conference and
0:37announces that China is going to bring
0:40artificial intelligence to developing
0:42countries. China will provide them with
0:46training, technology, and cooperation
0:49centers in Africa, Latin America, and
0:51the BRICS countries. Furthermore, a new
0:55global AI organization is being created
0:57, headquartered—get this—in
0:58Shanghai. 29 countries have already
1:01signed on. That’s two announcements
1:04in 24 hours, one technological, the
1:06other political. And it’s no
1:08coincidence; they are the two halves of
1:10the same project. China is no longer
1:14just manufacturing top-tier artificial
1:16intelligence; the important thing is
1:18that Xi Jinping sees AI as the playing
1:21field where it will be defined who
1:23rules the world in the 21st century. AI
1:27is the ammunition of this century's
1:29Cold War. He already has his plan to
1:33defeat the United States, the only
1:34counterbalance to his influence. While
1:38the United States charges a high price
1:41for its models to recoup its
1:43investments, Xi Jinping's plan consists
1:46of launching powerful models—not
1:48necessarily the best—but lowering
1:51their cost until AI becomes a commodity
1:54like oil or gas. The model seems
1:58familiar, doesn't it? The timing is
2:01perfect. In this video, you will see a
2:04piece of data that Silicon Valley would
2:06prefer we didn't see, a ranking where
2:08number one is no longer American, and
2:10by the end, you will understand why the
2:13most dangerous thing about the Chinese
2:15offer is precisely that it asks for
2:17very little or nothing in exchange for
2:19its technology. The last time a power
2:22exported the infrastructure of its era
2:24to the entire world, the machines
2:26arrived by ship from England and were
2:28called locomotives. The world adopted
2:32them without blinking and took half a
2:34century to understand why that decision
2:37had turned England into an unbeatable
2:39power. Hello, how are you? I am Gustavo
2:45Entrala. By day, I help large companies
2:47and startups design their future, and
2:49by night, I practice a hobby that I
2:51love. Teaching what I know about that
2:54future to those who always want to stay
2:56one step ahead. If you are interested
2:59in the future, subscribe and hit the
3:01bell so that YouTube notifies you when
3:03a new video comes out. Let's start with
3:09the setting, because everything about
3:11it was very calculated. The first
3:14symbol is that this is the first time
3:17Xi Jinping has attended this Chinese
3:19conference in person since it launched
3:22in 2018. Before, he would send
3:24congratulatory letters or send his
3:26prime minister. When the leader appears
3:30in person for the first time, the
3:32message is that this is no longer a
3:34tech fair, it is state policy at the
3:36highest level. Second, the start of his
3:40speech. Xi began by recalling that 70
3:44years ago, a group of young academics
3:46first proposed the concept of
3:48artificial intelligence at an academic
3:51conference at Dartmouth College. Notice
3:55the move. China's leader claims AI for
3:59humanity by citing the founding act of
4:02a meeting that only Americans attended.
4:07It's his way of saying, it isn't just
4:09yours, it was born from science and
4:11science has no flag. The speech was
4:15titled "Joining Hands to Build a Fair
4:18and Equitable System of Global AI
4:20Governance." The tone, therefore, was
4:23humanist. He spoke of a people-centered
4:26AI that seeks to avoid, I quote, "new
4:29historical injustices." And behind the
4:32rhetoric, Xi promised three things. Two
4:36seem harmless, and the third is what
4:38should keep Washington up at night. The
4:41first: 5,000 AI training scholarships
4:44for developing countries. The second:
4:47access for 30 countries to the Chinese
4:50AI-powered weather alert system. All
4:54this sounds like soft diplomacy,
4:56friendly diplomacy. And what was the
4:58third? AI cooperation centers with
5:02blocs grouping Southeast Asia, the Arab
5:05world, Africa, Latin America and the
5:08Caribbean, and Russia and some former
5:10Soviet republics. Think about it for a
5:14moment: these are the geographic areas
5:16where the majority of the world's
5:17population lives. China is not,
5:19therefore, just launching a scholarship
5:21program. What it is doing is planting
5:24its flag in a new pole of global
5:27influence. Even more important than
5:30these gifts was a signature. A day
5:33earlier, 29 countries, including Brazil
5:36, Indonesia, South Africa, Russia, and
5:39Pakistan, created with China the World
5:42AI Cooperation Organization (WAIC O), a
5:46sort of UN for AI based in Shanghai. Xi
5:49said this signing marks a milestone in
5:51the history of global AI development.
5:55This means that while Washington
5:57prohibits selling chips to China and
5:58limits access to its most powerful
6:00models whenever the White House deems
6:02it convenient, Beijing is founding
6:03institutions. Keep this asymmetry in
6:07mind because it is the key to the
6:08entire video. To explain the historical
6:15significance of what is happening,
6:17let's travel back a century and a half,
6:19because what happened then is about to
6:21repeat itself step by step, with your
6:23country and mine as the targets. Great
6:27Britain, which was the leading
6:29industrial power at the time, extended
6:31its influence across the planet by
6:33building railways in the second half of
6:35the 19th century in India, Argentina,
6:37South Africa, and half of Europe.
6:40Locomotives left the factories of
6:42Manchester and Glasgow to transport
6:45people and goods. The important thing
6:48wasn't the machines, but the British
6:50standards. Once a country bought its
6:54tracks, it sank into British debt,
6:56British technology, and British
6:58influence. Look at the mechanism,
7:02because it's a masterpiece. First, the
7:05standard. The width of English tracks
7:07became the mold for half the world. And
7:11once the tracks were laid to that gauge
7:13, every new locomotive, every wagon,
7:14and every spare part had to fit them.
7:17And who manufactured them? Exactly.
7:20Second, the debt. Many of those
7:23railways were paid for with loans from
7:26London, so countries spent decades
7:27paying for that infrastructure which
7:29tied them down completely. And third,
7:33the layout. The lines didn't go where
7:36the country needed them; they went from
7:38the port to the mine, from the
7:40plantation to the ship. The train
7:42belonged to everyone, but the map
7:44belonged to London. Nobody forced
7:47anyone at gunpoint to buy locomotives.
7:49They were the best technology available
7:52and often the cheapest. Each country
7:55that adopted them made a rational
7:57decision, and all those rational
7:59decisions added up to half a century of
8:01dependency. China is trying something
8:05very similar today with artificial
8:07intelligence. Their open models—I'll
8:09explain what those are in a moment—
8:11are essentially like free locomotives.
8:14The cloud, the chips, the data centers,
8:16and the standards are the tracks. And
8:19whoever builds the tracks decides for
8:21decades where the economy flows.
8:24However, there is something that almost
8:26everyone misunderstands, even people in
8:28the industry. The word "open" doesn't
8:30mean what you think; there's a trick to
8:32it. And in that misunderstanding lies
8:34the entire Chinese strategy. Let's take
8:38a quick pause in the video. As I told
8:40you a few weeks ago, I have been using
8:42Lumo for a while, which is the AI
8:45assistant from Proton, the secure
8:47company that markets Proton Mail. Lumo
8:50has gone from being one of the many
8:53tools I test to becoming the tool I use
8:55for the most delicate part of my work.
9:00As you know, I work as a business
9:02consultant, where I handle very
9:04sensitive information, unpublished
9:07strategies, client data, and ideas not
9:10yet on the market. And before, dumping
9:13all that information into another tool
9:15that saves it—and sometimes trains
9:17its model with that data—clearly held
9:19me back, but that doesn't happen with
9:21Lumo. And for those who didn't see the
9:24first video where I mentioned this tool
9:26, Lumo has the power of a ChatGPT or a
9:28Gemini to actually get work done. It
9:32has project-based chats; you can upload
9:34your documents and images, but with one
9:37difference that changes everything: the
9:39conversations we have with Lumo are
9:41ours. The tool has zero-access
9:44encryption, meaning not even Proton can
9:47read them. The tool has no logs, no ads
9:50, and they never use it to train an
9:53artificial intelligence. It is open
9:56source, so you can verify everything
9:59the tool does, and it is located in
10:01Europe, out of reach of Big Tech. I
10:05have taken my clients 'research to
10:07another level without ever betraying
10:10their trust, and with Lumo, I have all
10:12the power of AI without you, or me, or
10:15my clients being the product. Try it
10:19for free at the link I've placed in the
10:22description, which is
10:23www.proton.me/lumo/gustavoentrala. And
10:29let's get back to the video. An
10:35artificial intelligence like ChatGPT,
10:37Claude, or Gemini is, to put it simply,
10:39a frozen artificial brain. It is
10:42manufactured in two phases. The first
10:44is training. For months, supercomputers
10:48have it read astronomical amounts of
10:50text, books, and web pages. Code, and
10:55the system fine-tunes billions of
10:57internal connections until it learns to
11:00write, reason, or program. This phase
11:04costs hundreds of millions of euros.
11:07Second phase: usage. Once trained, that
11:10brain fits into a file. It is a giant
11:13file, but it is nothing more than a
11:15file, like a video or a game. Something
11:18that can be copied and sent. To be able
11:21to answer our questions and tasks, what
11:24is known as inference, that file has to
11:26be hosted in a very large and very
11:29expensive data center. And that is one
11:31of the keys to why China is giving it
11:34away. And we will see that later. The
11:38big American companies—OpenAI,
11:40Anthropic, or Google—usually don't
11:42give you the model; they rent you
11:43access. You write your question, the
11:46question travels to their data centers,
11:48and it sends the answer back to you.
11:51When you use ChatGPT, the brain never
11:53leaves their servers. To understand it,
11:57imagine a restaurant. You choose the
11:59dish, they cook it in their kitchen,
12:00and they serve it to you at a price
12:02they decide. You never enter the
12:05kitchen, nor do you get the recipe. An
12:07open-source one works differently.
12:10It’s as if the chef handed you the
12:12pot with the ingredients already
12:14measured out. You can cook the dish in
12:17your own kitchen, adapting it to your
12:19tastes. Change an ingredient, serve it
12:22to whoever you want, and even set up
12:24your own restaurant. The kitchen, the
12:28stove, and the electricity bill are all
12:30on you. But pay attention to the nuance
12:33, because that is where most of the
12:34debate lies. They give you the stew,
12:36not the recipe. You don't know exactly
12:39what text that brain was trained on, or
12:41how, or with what criteria it was
12:43taught to behave. And even if you had
12:46the full recipe, you don't have the
12:48kitchen. Repeating the training would
12:51require supercomputers and budgets that
12:53are only within reach of a handful of
12:56actors on the planet. You can have the
12:59copy, but you cannot manufacture the
13:01original. And why is China giving away
13:04its best brains while the United States
13:06keeps theirs under lock and key? It is
13:09not a matter of altruism; it is a
13:11long-term strategy, and it is quite
13:13brilliant. First, if you arrive late to
13:16a market, giving the product away is
13:18the fastest way to gain customers. You
13:21cannot charge more than OpenAI, but you
13:23can choose to give your creation away
13:25for free. Two, the United States
13:28prohibits China from buying the best
13:30Nvidia chips, but that ban has led
13:33China to do what it does best. Imitate
13:37the Americans by optimizing AI so it
13:39can be built without Nvidia chips.
13:42Three, every company, every university,
13:45and every government that builds on a
13:47Chinese model will get used to its
13:49technology, improve it, and train its
13:51engineers on it. And four, every
13:55customer who switches to a free Chinese
13:57AI is a customer who stops paying San
13:59Francisco. And there is a fifth reason,
14:03the most important of all, which
14:05connects with Xi Jinping's speech. An
14:08open one spreads on its own, without
14:10treaties, without embassies, without
14:12conditions. This is crucial, and I want
14:15to make it crystal clear to you. Any
14:18country and any company in the world
14:20can download Kimi or DeepSeek. A
14:23Chinese alternative, today. There is no
14:26need to be an ally of China, no need to
14:29sign anything, belong to any bloc, or
14:31vote with Beijing at the UN. A startup
14:35in Texas, a German bank, a Chilean
14:37ministry—all of them can download
14:39Kimi, and that is precisely why it is
14:41such an effective tool for influence,
14:43although this whole plan has one
14:45prerequisite: that what you are giving
14:47away must be top-tier. Is it? Let's see
14:52. Until recently, the answer from the
14:58West was that Chinese AI was two steps
15:00behind. But that answer has just
15:03expired. When? Thursday, July 16, 2026.
15:09That day, Moonshot, one of China's most
15:11important labs, launched, as we
15:12mentioned at the beginning, Kimi K3.
15:16They present it as the largest
15:18artificial brain ever released
15:20open-source, and it can be downloaded
15:22since July 27 with legal permission to
15:25use it with almost no restrictions. Is
15:28Kimi 3 on par with the best? Well, look
15:32, there is an independent ranking that
15:34tests all the AIs in the world with the
15:35same benchmarks. It’s called
15:37Artificial Analysis. In that ranking,
15:41Kimi 3 scores 57 points. Do you know
15:44how many AIs on the planet surpass it?
15:46Only two, the best from Anthropic and
15:49the best from OpenAI. Fable from
15:53Anthropic reaches a 60, and ChatGPT 5.6
15:56reaches a 59. That’s all. The third
16:00most powerful in the world is Chinese.
16:03It’s free and can be downloaded. And
16:07in some tasks, it doesn't just get
16:08close, it wins. There is a blind
16:10competition. If this is the ranking
16:13that nobody in Silicon Valley wanted us
16:15to know about, there is a competition
16:17where thousands of programmers vote
16:19that it designs web pages better
16:21without knowing which is which. Number
16:26one, Kimi K3, ahead of Fable, ahead of
16:29ChatGPT 5.6. It also beat the American
16:34models at searching for information on
16:36the internet, working on spreadsheets,
16:38and on long programming projects. And
16:41what is the price? Because while you
16:44can install it, you have to house it in
16:46data centers, and the company Kimi
16:48charges you to use theirs. Well, using
16:51it over the internet costs the same as
16:53mid-range American models. But let's be
16:57rigorous because this channel doesn't
16:59sell smoke. Moonshot itself admits that
17:02K3 is still behind the two best
17:04American AIs in general performance.
17:08Independent analyses point out that it
17:10makes things up more frequently than
17:12its rivals, what the sector calls
17:14hallucinations. And I made a video
17:17about that which I’ll put here so you
17:19can watch it, which is a problem that
17:20remains unsolved, and American AIs keep
17:22winning in the most difficult tasks
17:24where a mistake is very costly.
17:27Furthermore, Kimi has an engine that
17:28consumes more fuel than the others.
17:30This is very important. Independent
17:33studies that have started using it say
17:35that to perform the same task, Kimi
17:37consumes triple the tokens of the most
17:40advanced American models. But the
17:43general framework has been set by a
17:45prestigious analysis center in
17:47Washington, the CSIS. and it says that
17:52the gap between Chinese and American AI
17:54has narrowed to weeks, around 12 weeks,
17:574 months. Before we were talking about
18:01a gap of 6 to 9 months. And how has
18:04China reached this point without having
18:06the best chips? I talked about it in
18:09this video about DeepSeek. First, with
18:12ingenuity to do more with less. Second,
18:15with a global community that improves
18:17their open models with their
18:19contributions, and third, with a
18:21controversial technique called
18:23distillation, which consists of asking
18:25the rival millions of questions and
18:27training your own with its answers, a
18:29kind of reverse engineering. It's like
18:33the student who learns by copying the
18:34notes of the top of the class. OpenAI
18:38and Anthropic have accused Chinese labs
18:40, Moonshot included, of doing it on an
18:42industrial scale. And the White House
18:46accused Moonshot a few days ago of
18:48using Claude, an AI model from
18:50Anthropic, by asking it millions of
18:52questions. Whether this description is
18:56exact or not, what seems clear is that
18:58Beijing is learning, among other things
18:59, from the American AIs themselves. And
19:02this hides a final irony. Washington
19:06banned the chips to slow down the
19:07Chinese advance, but it didn't slow it
19:09down, it redirected it. Now the Chinese
19:13engines are good and they are free, and
19:15what nobody tells you is what happens
19:17after you install them. That comes next
19:20. And here we reach the underlying
19:25question. If I can download a brain
19:28tailored to my taste and install it on
19:30my computers without asking anyone for
19:32permission, where is the catch? What
19:35exactly does China gain? Because nobody
19:38gives away hundreds of millions in
19:39technology for the love of art. The
19:42answer has five layers and I'm going to
19:44tell you them from the most obvious to
19:46the most disturbing. The fifth one, by
19:49the way, has had a proper name in
19:51Silicon Valley for 20 years, and when
19:53we get to it you'll understand why
19:54Google knows it better than anyone.
19:59Layer one, the treadmill. The one you
20:02download today is a still photo of a
20:04field that moves every 6 months. K3
20:08will be old when K4 comes out, and K4
20:11will be old when K5 comes out. You have
20:14today's copy. Moonshot has the
20:16intelligence factory. If your bank,
20:19your administration, or your company
20:21builds on this family of models, every
20:23update brings them back to the same
20:25provider's door. It is the logic we are
20:29so experienced with in mobile phones.
20:31The phone is yours, but it doesn't give
20:33you the right to use the next model
20:35without paying again. Layer 2. The
20:38ecosystem. An AI doesn't live in
20:41isolation; it lives surrounded by tools
20:43, manuals, forums, engineers who know
20:45how to adjust it, and so on. When half
20:48the world builds on Chinese AI, the
20:50best tutorials are for Chinese AI. The
20:53best tools are designed for Chinese AI,
20:56and the engineers your university
20:58trains learn with that Chinese AI.
21:01It’s what made Windows or Android
21:03unbeatable in their day. Nobody forced
21:05you to use them; they simply managed to
21:08become the reference platform. That is
21:11why the 5,000 scholarships announced by
21:13Xi Jinping aren't about educational
21:15quality, but sowing that human
21:16ecosystem. Layer three, factory
21:20settings. Everything arrives educated
21:23from home, trained to respond in a
21:25certain way, with certain limits, with
21:27a certain worldview. And that vision is
21:31embedded in millions of invisible
21:33decisions that were made during its
21:35creation. You can re-educate it later
21:38—that’s what open source is for—
21:39but that readjustment only touches the
21:41surface, it doesn't rewrite the deep
21:43education. It’s what’s known as
21:46post-training; you can make
21:47modifications to how the model behaves,
21:50but let’s say the essential core of
21:52the model remains the same. And let’s
21:55be realistic, the vast majority of
21:56people will never touch those settings.
21:59They will use the Chinese one exactly
22:01as it came from the factory. And when
22:03the tool with which millions of people
22:06write, summarize documents, study, and
22:08decide, comes configured from the
22:10factory in Beijing or San Francisco.
22:12Watch out, because this applies to both
22:14models. Whoever sets the factory
22:16configuration shapes the global
22:18conversation a little bit. Layer four,
22:21the standards. Remember the track gauge
22:24? Once half the world laid their rails
22:27with the English measurement, every new
22:29train had to be manufactured to be
22:31compatible with England. It’s the
22:34same with AI. Whoever dominates
22:36adoption ends up writing the technical
22:39rules: how AIs are evaluated, how they
22:41are identified, what formats become the
22:44common language of the industry. That
22:46is the track gauge of the 21st century.
22:50And now connect this with Wiko, a
22:51government organization based in
22:53Shanghai, dedicated precisely to
22:55coordinating standards among its
22:57members. The full play: open models lay
23:01the tracks and the organization decides
23:04the width. First adoption, and then the
23:08rules of use. And layer five is the
23:10most useful. Every company in the world
23:13that solves its problem with a free
23:15Chinese tool is a customer that
23:16American companies won't bill. Less
23:19revenue means less money to pay for the
23:21next generation of models, which costs
23:23a fortune. You don't need to catch up
23:26to the leader if you can bleed their
23:27business dry. And this strategy has had
23:30a name in Silicon Valley for decades.
23:33Make free what your rival sells. Google
23:37used it with Android. They gave away
23:39the operating system so that the mobile
23:41business wouldn't be exclusively
23:43Apple's. China is using it right now
23:45against the entire American artificial
23:48intelligence industry. Let's summarize
23:51the five layers of China and Xi
23:53Jinping's strategy. The pace of
23:56versions, the ecosystem, the factory
23:58values, the standards, and the
24:00financial bleeding of the American
24:03rival. And what is the weak point of
24:06China's strategy? Because it has one.
24:09Look at this news. The success of Kimi
24:12K3 has been such that Moonshot had to
24:14turn off the tap to new users. Why?
24:18Because it doesn't have the data
24:20centers needed to service them. The
24:24United States prevents them from buying
24:26AMD or Nvidia chips, and Huawei's AI
24:29chips aren't at the same level yet, but
24:31not for long. American AIs have
24:35invested and continue to invest
24:37billions in their data centers to
24:38provide service. That's why the
24:42competitive advantage of American labs
24:45will be their massive investments in
24:47data centers for a long time. There
24:51isn't enough chip supply for everyone,
24:54and Americans get priority in the
24:56American cloud. In other words, it's
24:59hard for any company to buy Nvidia and
25:02AMD chips, and the Americans have
25:04priority. And it's also difficult to
25:07set up those large data centers that
25:09can service all our questions and
25:11agents. And in that, obviously, the
25:13Americans have a much more extensive
25:15network. Chips go to those who can pay
25:18for them, logically. And here is the
25:20trick I was telling you about at the
25:22beginning of this video. It's one thing
25:24for China to give away its models, but
25:26it's quite another for the world to use
25:29them massively. It will still take a
25:31few years. But that is precisely Xi
25:34Jinping's long-term plan: to form
25:36alliances with third countries so they
25:38are the ones investing in installing
25:41those data centers with Chinese
25:43technology. Machiavelli was an amateur.
25:50And for the rest of the world, do an
25:52exercise with me. Imagine you are the
25:54technology minister of a mid-sized
25:56country. This week you have three
25:58offers on the table, in addition to
26:00China's. Let's see what each of those
26:03offers sells and what the fine print
26:05entails. Offer one, The Club of
26:08Democracies, an alliance linked to the
26:10OECD that brings together countries
26:12committed to AI that is respectful of
26:15human rights and transparent. What is
26:18its strength? Legitimacy and the seal
26:20of quality. Its fine print is that it's
26:23a select club. You get in by sharing
26:25values, and that has left out a good
26:27part of the developing countries. You,
26:30as the minister of that mid-sized
26:31country, were never called. Offer two,
26:34the European way. Europe no longer
26:38competes by giving away technology,
26:40because it doesn't have it; it competes
26:43by writing rules, its major AI Act and
26:45an international human rights treaty
26:48backed by over 50 countries, hoping its
26:50huge market forces the world to adopt
26:52these rules, as it did with data
26:55privacy. Its strength, so to speak, is
26:59legal certainty and its own market. And
27:01the fine print is that it offers you
27:03rules for a technology it doesn't
27:05produce itself, and that, obviously,
27:07limits Europe's ability to influence
27:09the world. Offer three, the American
27:13offer. The United States has decided
27:16not to play at global governance. It
27:18offers country-to-country deals with
27:20the full package. It sells you the
27:22chips, gives you access to the cloud,
27:24and sells you the most advanced
27:26American models under the promise that
27:29you will have your own AI within its
27:31ecosystem. At this year's summit in
27:35India, they even presented a sort of
27:36technological volunteer corps, an AI
27:38Peace Corps. Notice the symmetry. The
27:42two giants offer you the same thing:
27:44capacity in exchange for an ecosystem,
27:47with one difference. The American offer
27:49asks you to be a partner; the Chinese
27:51one only asks for a download.
27:54Furthermore, you have no guarantee of
27:55being able to access their best models
27:57forever—the American ones. We saw it
28:01recently with Fable, which we couldn't
28:02use for two weeks because Donald Trump
28:04decided it wasn't safe and that he
28:06preferred—that it wasn't safe for the
28:08national security of the United States.
28:11And that is why he banned it. And this,
28:13of course, makes any potential partner
28:15very nervous, whether they are
28:17countries or companies. I cannot make
28:19large investments in a technology if,
28:21at any given moment, you are going to
28:23take away my access to it. And who is
28:26winning customers? This is a fact that
28:28we in the West don't want to look at.
28:31Wiko is born with 29 founders, which
28:33include heavyweights from the
28:35developing world. It includes Brazil,
28:38it includes Indonesia, it includes
28:40South Africa. Exactly the countries the
28:42club of democracies never invited to
28:44the table, and analysts point out that
28:47this support will weigh heavily at the
28:49UN, where votes are counted by country.
28:53The game of AI rules will not be won,
28:55therefore, by whoever has the best
28:57model, but by whoever has the most
28:59seats occupied in these international
29:01forums. And today, the room that is
29:04growing the most is in Shanghai. And us
29:11, the Latin American countries and
29:13Spain—well, I did explicitly mention
29:15Latin America and the Caribbean,
29:17because Latin America is not a
29:18spectator in this game; it is on the
29:20board. Three possible endings for our
29:24countries. One, several AI poles
29:26competing, with each country choosing
29:29sides, the United States or China. Two.
29:33Two technological blocs facing off like
29:35in the Cold War, but with AI instead of
29:37missiles. And three, it doesn't matter
29:41who has the best model because the
29:42rules of the game are written at a
29:44headquarters in Shanghai, and whoever
29:46writes the rules wins without firing a
29:48shot. I'm not going to tell you which
29:52of these three options will happen, but
29:53I will leave you with the question that
29:55your government and your company will
29:56have to answer sooner than we think.
30:00When they offer you artificial
30:02intelligence for free, what exactly are
30:04you accepting? That is a very good
30:06question. And when they charge you for
30:08it, too. At the beginning of the video,
30:14I told you that the most dangerous
30:15thing about the Chinese offer is that
30:17they don't ask for anything in return.
30:19Now you know why. British railways
30:22worked, they moved the world, created
30:24wealth, connected continents, and for
30:27half a century they decided from London
30:29where the economies of dozens of
30:31countries would flow. Nobody called it
30:35dependency; they called it progress.
30:38The question today isn't whether
30:39Chinese locomotives are good, which
30:41they are, and they're free. The
30:43question is, who is maintaining the
30:45tracks? Locomotives are gifted; tracks
30:48are inherited. China's plans are very
30:52attractive to countries that have fewer
30:54resources and don't want to depend on
30:56the decisions of the White House, on
30:58the whims of the White House, to make
30:59their day-to-day work. How do you see
31:03it? If you were a leader, what decision
31:05would you make? Would you go down the
31:07Chinese path? Would you go down the
31:09American path? Would you try to
31:11generate your own artificial
31:12intelligence? It is not an easy
31:15decision at all, because the level of
31:17investment that artificial intelligence
31:20requires means that the options are
31:22necessarily few and that they have a
31:24lot of hidden compromises behind them.
31:28Well, see you in the next video, which
31:30we will try to make even better than
31:32this one. Please don't forget to
31:34subscribe. There are still many people
31:36who watch my videos and are not
31:38subscribed. Bye.

About this transcript

This page contains the full transcript of Por qué CHINA regala su IA al mundo by Gustavo Entrala, generated from the public captions YouTube serves with the video. The transcript has 4,766 words across 740 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.