Hi everyone! You're watching ToTheMoon. Tech insights, news about artificial intelligence, Silicon Valley and the whole world. So, today we've got a kind of new format for you. And we'll go through the main events of the week in the artificial intelligence market. A lot, and fast.
OpenAI raised eleven billion at a valuation of one point two five trillion dollars. And according to Reuters, the company closed the first part of a new funding round, and it plans to expand it to thirty billion. When I personally saw the eleven billion, I thought: what kind of number is that, right? What does that number even mean for them anymore? Totally uninteresting. Actually, it's an important topic, yes, because it's one point two five trillion even before the company has gone public. A kind of interim effect.
And, well, maybe there isn't enough money inside, to be honest. And who thought otherwise? And it'd be great if you left your comments on the different news items here, here, because we're going to go through a lot of them together. And I recall the huge number of people and commentators saying that OpenAI will soon collapse and die. So, Anthropic disclosed huge revenue and huge commitments.
And their quarterly revenue reached seventeen billion. By the way, Ilnar and I were just remembering that they wanted to report much more revenue for the quarter than seventeen billion. And their commitments as of today, that's the most interesting number. Think about it! Anthropic has infrastructure commitments of eight hundred ninety-five billion dollars. Imagine, eight hundred ninety-five billion dollars. Although there are commitments for next year, and those are only seventy billion, just that.
What's seventy billion dollars, right? They show a big loss, but I want to tell you that it's hard to look at the loss right now, because, here, for example, their loss of a hundred and four billion. There was an accounting adjustment to the financing. It's really very hard to make sense of this reporting. But once they go public, it'll be a bit easier to look at. But with numbers like these it's already really hard to figure anything out. And in general, hardly anyone will dare do anything to them anymore. That was the story this morning.
You can treat it as news item number three. The scandal around the most important person at NVIDIA, Jensen Huang, yes.
And the scandal is that he publicly admitted he cheated on his wife. And the woman the media link to his affair appeared in reports by various agencies about possible ties to Chinese intelligence and China's Ministry of State Security. Bam, bam, bam. And, well, obviously nobody is going to do anything to Jensen Huang, even if this woman was from intelligence. Even if. Of course, he's not to blame. A hundred percent. That's why nothing can happen to companies like these on the market right now. And there's an interesting story here, by the way.
Should problems even happen to companies like these at all? Imagine if, well, Anthropic is one thing, although a huge number of government systems and security systems around the world already run on Anthropic. If something happens to OpenAI, what happens to the hundreds of millions of people around the world who use them all the time? That's the question.
Tech stocks went through a heavy sell-off. The Nasdaq 100 lost more than six percent on Thursday, and on Friday it won back only part of the drop. Well, I want to tell you that my personal portfolio in artificial intelligence lost a lot. I liked a post by one trader in artificial intelligence, let's call him that. He wrote: “What times are these?” And he had a number written there, minus one to two billion dollars in his retirement account. He says: “My retirement account has shrunk.” And at that moment I thought: it's funny for ordinary people to look at how someone's account shrank by one or two billion dollars. It happens.
The tech market really got shaken up a lot. And I recommend everyone look at the following thing here. There were years, I remember, maybe they still exist now. And by the way, those of you who have information on this, write in, write in, it'd be interesting to see it. There was a year when layoffs started at Google, and everyone said: “That's it, so, everyone is getting laid off everywhere.” But if you look at the overall hiring chart, it'll look like this. That is, you always need to look at what the overall chart of the value of artificial intelligence companies has been.
Well, everything has a correction. Let's see what happens next. I think the market will grow much more. And I support Elon Musk's position that companies can be worth ten, well, he said a hundred trillion dollars. I even support a hundred trillion dollars by now, but tens of trillions I definitely support. Anthropic, according to the Financial Times, has reserved gas generators worth three to four billion dollars.
So what, what is this about? It's about generators with a total capacity of almost two gigawatts. That's one point seven gigawatts for data centers. And the agreement itself was concluded in September, but only now have they talked about it. So, what's telling about this story is that artificial intelligence developers have to provide themselves not only with chips, but also with their own energy infrastructure. And now, ahead of the change in power, power has passed to the Democrats, well, a significant share of power has passed to the Democrats in the US. And just now there were disputes over a data center in Texas.
And I liked that the local senator, or the local governor there, said on this topic, the new, the potentially new ones. They said that this way we will only support the development of data centers, but most importantly, don't take power away from ordinary people. In your country, do they take power away from ordinary people or give it to operators like these? It's different in different countries.
Google announced seventeen billion in investments in a data center in Wyoming. So OpenAI's eleven billion is somehow not interesting at all, right? If Google, so, is building new sites, to be honest, I've already gotten lost in all these sites. There's just some insane number of sites. And the US, of course, I'm glad they've started moving very hard on this, because, of course, infrastructure capacity is essential. It's strange that they didn't think about this before and were saving lizards or pumas somewhere, and that's why they didn't build data centers, but in reality, without this, the modern systems of the future won't work.
And there's a certain downside to this, because a huge number of people will suffer because of it. Those who previously ran various projects against these, against these development sites.
Apple opened up the Siri AI integration to all developers. Let me remind you that their new version of Siri came out in September. Those of you who've used it, tell us. I personally, so, signed up for everything again on purpose, turned Siri back on, because I had everything turned off. Well, it's awful! Awful work, poor-quality work. Here's what I didn't like. At first I saw this cool feature of theirs, that it would supposedly work great with photos, notes, and so on. But when I asked it to count all my notes or count all my audio files that had been recorded on Apple's infrastructure, it couldn't do it.
That is, it could only do very small things. And I want to say: look at your projects, at your life, where you're doing very simple, small things. Apple is an example of a company that is terribly behind. And clearly developers should probably start building something for Siri AI in order to position themselves properly in partnerships with Apple. But for now it looks like a company that's seriously losing. Kira now has permanent employees for development, and the new bots in Kira Crew get their own memory, an ongoing dialogue and a specific area of responsibility.
They can launch several work tasks, monitor execution, and work on a schedule. To be honest, I don't know who Kira Crew and Kira are, and I'm not sure I want to know about them. But it's interesting, by the way, that people often come to me and say: “Tell people about our platform.” Or someone comes to advertise, they ask about various platforms. And most platforms, the vast majority of all these platforms, are platforms that aggregated someone somewhere or created their own agents.
And so you sit there and think: “What can anyone even say about you? And what platform is under the hood?” Some of you can write in and tell us what Kira is. We'll look into it together. Maybe someone will even try them out. And maybe it's some cool company and I just missed it. And next time I'll definitely talk about it.
Anthropic showed how agents can fix their own memory. The company described a system in which Claude doesn't just save notes, but during pauses checks them against the work history, fixes mistakes and restructures its knowledge. This matters for long projects. By the way, I want to say that we'll definitely do a separate episode on this topic. This is a new world of how systems work, of course. The system is constantly listening to you. You all know my statement from, like, four or however many months ago, that I'd be willing to pay thousands of dollars for some system that constantly listens to me, chimes in and adds various details.
And it doesn't necessarily listen to my speech or voice, it's sort of constantly in interaction. ChatGPT Dots is trying to do this in a somewhat notable way. But, you know, I'm not really sure they actually built this concept into Dots and that it isn't just an ordinary pet project inside the company. Right now I'm just watching how it looks. So with this whole thing, when you're interacting with some chat right now, pay attention to it. That is, can the chat change the data along the way?
What's cool right now is ChatGPT in the Pro version, at least in their latest tests, they answer as they go, answer, answer, answer, and then they can completely change their answer, because they find additional data, and then they can correct themselves and point out that it's different here. Their model works better and better. There's a certain downside, yes, we know they're supposedly merging ChatGPT with Work.
And I see how ChatGPT can't do the things it used to do for me in large-volume data analysis. It wants to analyze them the way Work analyzes things, and that's a slightly different concept, yes. You can see they're cutting resources. By the way, it's become clear, of course, that ChatGPT by default spent a lot of money on big tasks.
The Grok developers showed a shared artificial intelligence analyst for the team, for the team. And I saw this topic, by the way. It's good that it showed up here. They have Data Bot, and I like Grok, by the way. They love to release some trendy, cool stuff. Their audience is small, of course, but still. Data Bot knows the company's internal metrics, queries the data, breaks down the results of ad campaigns and builds reports. Its memory is shared among employees. But new knowledge can first be checked and go through review by separate responsible people.
And the company estimated the work done by this new bot, this sub-type, sub-type, yes, of bot, at roughly seventy hours of human labor. Yes, but that's, strictly speaking, their own estimate of what it was. Google increased the length of generated video and lowered the price. Gemini Omni 1.2 Flash creates clips up to fifteen seconds long.
By the way, when I saw fifteen seconds, I thought: “What nonsense is this? What fifteen seconds?” Why did I think that? Because it's now 2026, November, and everyone has been loudly claiming that minute-long clips are being generated. What's fifteen seconds? And it's lower-quality video too, yes. Hello to Higgsfield here. You can generate cooler clips with them. Although, honestly, I never managed to easily create super cool clips. Well, the way it's shown in their ads, it didn't work out. By the way, they just showed a cool video.
I watched it this morning, they made an update together with ChatGPT Image, and they showed a video of how an old offline business, a café or a restaurant, gets made over. And there's a girl sitting there drawing, she hung up a new sign, launched an ad. It looks great. But I thought, what offline restaurant is going to do this? And will it actually work at all? That is, will new visuals work to sell something better. Fifteen years ago that model existed.
OpenAI, right, we've got this news item here, look, released, as if I'd foreseen it, released Image 2.6, and images are faster, and animation appeared. Of course, I'm not very interested in going through all these pictures and GIFs. We were just going through GIFs in the last episode on Sunday, and my sister said she sent her husband the first GIF in the world, the first GIF of her life, yes. And he immediately realized that it was strange. Probably artificial intelligence sending it. But still, personally, I'm waiting for both photos and video to become ordinary.
Well, only the English word comes to mind right now, commodity, yes. That is, they'll become, the way ChatGPT became a thing for questions and answers, pictures will become like that. Because, of course, for now there are certain nuances, there are certain questions about the quality of execution. Although at the same time, when we talk about creativity, my Anya was talking about creativity in pictures. Everyone likes their own thing, after all, yes. And the question here is how freely ChatGPT will be able to truly understand what you want to make.
And it's impossible to draw everything super cool. If everyone draws everything super cool, everyone's stuff will be the same. Right now, everyone often makes icons in interfaces that were once incredibly complex and unique, yes, or uses some color schemes that used to be cool. We'll see. Although, by the way, some things are unclear to me. You pay attention to car colors. And some car brands have very beautiful colors. For example, the Mercedes Gelandewagen has very unique colors, a unique yellow or a unique green, a unique light blue.
And you wonder, why don't the cheaper car companies paint their cars so beautifully? I don't know, Ford or Hyundai. Why don't they paint them like that? Or Jeep. Why don't they paint them that beautifully? Maybe it's because people don't need such cool colors. The Chinese model Moonshot, a model we know. Max is with you.
Last week Max, remember, Grigoriev went through it and explained why the Chinese model Moonshot is unique in a certain way. So, Moonshot published the files of the Kimi K31 Pro model. This isn't a model announcement, it's the release of files so you can run it yourself. Kimi works with text and images and supports a large volume of input. This gives companies and you an alternative to depending on someone else's cloud service. Clearly you're unlikely to run it on an ordinary laptop. The model is very large, yes, and Moonshot has a license for using it. But why not?
And for many this will be, well, a strong step forward. Although, again, how much compute you need to run it and get serious answers, I don't really understand that. That is, you need some kind of cosmic compute for that. Next news. Thinking Machines. So, Mira Murati.
Mira Murati showed up not only here in our news, but Mira Murati showed up, she gave an interview, and Sam Altman referred to her interview and said it was the best interview, so, of, of all time, from anyone. And the most important topic overall, he said. He even used crude language. Strange. At first I thought he probably wanted to say something bad about her, since he was using crude language. What was the point of the post? That Mira Murati said the system of the future is a system that constantly listens to you and constantly interacts with you.
And, well, Mira Murati said, of course, this is new news, although the news isn't new at all, that things will develop in that direction. So, Thinking Machines introduced Incling 2 Small. It's a new model. Well, this is generally an interesting topic, of course, because there are cool developers there, and they've still been making services like this for cool, well-connected companies in the artificial intelligence space. We'll see if they go out to a broader market. The new model is meant for programming, working with documents, professional tasks, and images.
And it's available through the Tinker service. They have a service like that. And as for the files for running it yourself, the company promises to release them after external safety reviews. And what's interesting here, of course, is the go-to-market model itself. That is, they give this kind of managed access, and then possibly open it up. We'll see if this model appears on the market. Although for me… So many models, it's unreal, yes. Hundreds of models. And of course they're all insanely behind the cool, serious infrastructures, which… I really don't understand people who think they won't take the lead.
Well, how many search engines in the world do you know? Well, I started using a search engine back in, like, ninety-five, ninety-six. I don't remember in which year all these Apollo, Search, AltaVista and so on started appearing. I don't… They were searching for some codes. I don't even remember what anymore, yes. Rambler was once around in Russia, it appeared. How many search engines are left in the world? Do we really think there will be millions of models? These models, they're going to have to die.
So, the Spaces feature lets you describe the tool you need in words: order tracking, a trip planner, a school schedule and so on. Muse builds that kind of section and can update it with data from connected apps. I don't consider Meta a cool maker of artificial intelligence models right now, but Meta owns Facebook and Instagram, and that's a huge, serious, big resource. And good models appearing inside may, may improve this system. Although I believe in the restructuring of messenger and social network systems within the next five years or so, yes. Messengers for sure.
With things appearing like System Dots, a whole new era of new messengers may come, yes. If it comes, of course, it will be a strong blow to infrastructure companies like these. Well, for now they have a lot of traffic and a lot of money, they can test and invest. That's why they do releases like Genie.
Let's move on to robots and various industrial equipment, and defense. So, shares of the maker of the Digit robot started trading on Nasdaq and, by merging with a listed company, the related financing brought in about half a billion before the costs of this deal. And well, robots are growing. Recently Max Grigoriev told me to buy a robot dog that runs around the property. I found a fairly inexpensive one for four thousand dollars. And at first, of course, I thought he'd set me up with some little dog for three or four hundred dollars.
I needed to work through a couple of scenarios here, but somehow I didn't want to spend four thousand dollars on this dog. But I still think it ran pretty well, for that money, on various slopes and so on. But I'd still like something more than just a dog. A dog, what can it do? It mostly watches the property and chases away, well, pumas, deer. Who else can it, what else can it do? Turkeys, coyotes. An opossum came to my place a couple of days ago. That's the one I'd like to chase away at night. Ha, I have to think about whether it's worth paying four thousand dollars. Although my work and my flowers are already worth much more.
And also, by the way, the kids sometimes want to sleep outside in a tent. And a robot like that would be great. My daughter slept in a tent with a friend the day before yesterday. You worry a little, you think, what if some animal comes up to them. So, Zipline got one and a half billion and agreed to expand drone delivery for Walmart.
Let me remind you, drones are a very widely discussed topic when it comes to delivery. It's been launched and shut down, in different companies, in different countries it's been launched and shut down. So, Zipline overall has now reached twenty-one billion, a valuation of twenty-one billion dollars. And the new agreement with Walmart provides for launching another three hundred delivery points. And unlike, by the way, a demo of the drone of the future, what's being discussed here is scaling an existing service, manufacturing the aircraft, everyday ordinary delivery, yes.
By the way, where I live, I live in Los Altos Hills, in Silicon Valley. There are hills here and the houses aren't as densely packed as, well, in the very center of the valley. And it seems to me that delivery here would be fairly convenient. Although, of course, it's unclear how, how convenient it would be, because how does a delivery driver go around now? Three, five, ten delivery cars, for example, come to my place a day. And that same driver isn't delivering only to me, he's also delivering to all the neighbors and so on, yes.
And usually you check where he is, and they tell you he still has fifteen or twenty stops. So with drones, of course, the question is how efficient it will be. But we'll see. Plus there's something flying in the air, something constantly buzzing. If you launched all these drones here now, launched all delivery by drones or flying cars, you'd go crazy, you'd really go crazy. You can't even hear the airplanes, there's an airplane somewhere, there are two airports nearby here, yes, San Francisco and San Jose, big ones.
And you can't hear the airplane flying, you can just see it flying over the bay in San Francisco. And what will all these deliveries be like? I don't think this will get super wide adoption right now. We still don't live in a world, you know, like when you watch some movie and everything there is flying and constantly buzzing.
China blocked the sale of ABB Robotics' Chinese business to SoftBank. The ban applies to the Chinese robotics part of this business, and it complicates the global deal. This is, of course, a good topic about how industrial robotics is becoming a strategic asset, and we can all remember a number of deals that were blocked involving China. And the most notable one was the deal where a team built an agent and sold it to Meta through a Singapore company. And the Chinese blocked it, the Chinese government blocked the developers, well, the ones who sold all this, from leaving the country. Very funny.
Shield agreed to buy the radar maker EchoDyne for five billion dollars. That is, the company develops various autonomous systems for detecting and countering drones. This is, of course, an incredibly big, huge market, which will be, a crazy market, yes. And together with the agreement, financing of two and a half billion dollars was announced. And, of course, now the topic there isn't just about, building a drone, yes. Now the topic is also that new tools are needed for detection, control and protection.
And this deal, which they plan to close in early twenty twenty-seven, that's exactly what it's needed for. In general, we need to do a separate episode about drones. I, of course, understand how far ahead the military industry, which will use the technologies of the leading companies, will get ahead of everyone else. When this artificial intelligence system constantly interacts with drones, a mass-scale artificial intelligence system, with drones, with ships, with planes, with missiles, with tanks, with bases, with people, with various infrastructure systems, with messages, with television, it will interact with everything as one all the time and fully control everything.
Of course, that's a different era, and it will take trillions of dollars of infrastructure money to run such a system. That is, it's not just about deploying some model, for this you'll need to allocate a permanent resource that will spend, I don't know, a billion dollars a minute, for example, a minute on various detection or various analytics. This is serious. By the way, I didn't say a billion dollars a minute for nothing, because it'll be huge money, some cosmic money will be spent on this.
To understand it, you need to move to some other numbers. OpenAI showed the design of a physical device using artificial intelligence. And that's cool news.
Two employees using GPT-6 Astra designed a test device in less than, than a week, working on it as a side project. Well, of course, OpenAI has its pet projects. Although OpenAI, by the way, presented this topic as if they had created entire separate modules for building such systems, and a manufacturer made a prototype in less than a month. Let me remind you, by the way, that a lot of people have already made various devices. It's a very important topic, very cool. I recommend that geeks and fanatics of new technologies create something, some devices, and order their production.
Well, if you live somewhere where you can order anything at all, I recommend doing it. I personally want to make a few, microphone devices of that kind and put a few sensors inside. We'll see how it goes. Because every time you want to make something of your own, you end up using some existing infrastructure or board. Yes, and here you need to make something entirely your own. We'll see. Google showed detection of signs of high blood pressure from ordinary watch data.
This research system uses an optical heart-rate sensor and a motion sensor, accumulates data over several days and can detect blood pressure. This is, by the way, a very interesting story overall, that we're moving into an era when we'll be able to diagnose things using completely different systems. And all the diagnostics that existed before were based on old experience and on certain expensive equipment. Although in fact many things will be doable very easily, yes. And I'm, of course, deeply convinced that, whatever else, they'll learn to measure blood pressure with some super basic, the most basic things.
And the question, by the way, that stands right now in terms of various measurements along with this, I think, is creating serious batteries, or some temporary devices, yes, where you stick a little patch here, for half a year, and it, or, say, once a month, for example, yes. Because sticking one on every day is inconvenient. It might cost pennies, but you'll forget. And this patch collects some insane amount of different data. The question, of course, is when this will happen and how much time it will take, how quickly it'll get approved.
And this is also a very serious restructuring of the market, and the restructuring of pharma, I think, is much harder than the restructuring of the ordinary tech market. But artificial intelligence will push everyone very hard toward it here. There's a startup, so.
Physical Superintelligence reported a proof of the Aaronson–Ambainis conjecture, formulated in two thousand eight, in '08. A research system with a non-public Anthropic model was used. The proof passed machine verification. The result concerns the conditions under which quantum computers can significantly outperform ordinary ones. Let me remind you that if something serious works out in quantum computing, then everyone's done for right away, instantly, yes. Well, if they make it public.
And, of course, it's unclear how this movement will move in parallel. That is, what if artificial intelligence appears on such, in such computers, on such compute, that is, what will happen? Current artificial intelligence models, current artificial intelligence systems, what will happen? But I think this is just kind of a discussion. We once talked a lot about quantum computers, back with Sasha Sagun too, and you can see that for now it's just a kind of, a hypothetical story.
Axiom translated a huge mathematical reference work into machine-checkable form. The company reported formalizing roughly seven and a half thousand pages of the Stacks Project in Lean, a system that checks the logical steps of proofs. The result is a library of roughly sixteen million lines of code, created with the help of many artificial intelligence agents. The interesting point here is that artificial intelligence, it turns out, not only proposes new mathematical ideas, but helps build the foundation for rigorously checking them.
That is, as a kind of part of it, right? I really like projects like this. I do a lot of this in my businesses too, in my projects, in terms of analyzing and structuring various data, translating it. And, of course, we in the world will still face the aggregation and analysis of a huge amount of information that, that doesn't exist anywhere. Here's what happened to me: the other day my sister sent me some data, a person wrote memoirs, and in these memoirs my dad, well, he remembered my father and, and remembered the time, the seventies probably, or something, and, or the eighties.
And the interesting aspect is that later, when I started, sort of tried to go deeper into this data through artificial intelligence, not into the data of his memoirs, but into, into the system, first of all, where my father worked, or, like, the companies, well, it was still the USSR back then, yes, this data basically doesn't exist. Because obviously a huge amount of archival data in Belarus hasn't been transferred into artificial intelligence. When all this gets transferred, very interesting facts may surface.
The question is whether it will be transferred, and who will allow this data to be transferred. On the one hand, it's a very big plus for the state, because there's no chance, right? You have to enter all the data into the system in order to draw conclusions. Last week I shot what I think is a really great video about finance. And in that video I said that one of the financial rules was what? To first enter absolutely all the data in there. On the other hand, we understand that if they enter all the data, someone risks some information surfacing, well, and that could surface at the level of a country.
What artificial intelligence can do now is, of course, incredible. Today I read a not-so-nice post.
A person wrote there that artificial intelligence took away his dream and his life, that he had been a programmer, a very strong developer, he put years of his life into it, like ten years into learning how to write code, object-oriented programming, algorithms and everything else. Now you don't need any of that. And he says he won't use artificial intelligence at all, because it takes life away from people. It's really, it's a fact. It's a fact that it will take away a lot of things. Things that you once enjoyed doing and could still make money from.
There's this idea that you should do what you love. And a huge number of people suffer because of it. I urge you all to live in an objective, sane reality, yes. You should do what life gives you to do right now. And you should love life, all of it, regardless of what happens.
OpenAI proposed testing artificial intelligence not with exams but with everyday tasks. They've released a new test, Life Well. It evaluates help with everyday matters, choosing a kindergarten, organizing a move, shopping, family finances, dealing with government services. And, around four thousand people used this thing there. It's not a new assistant, but an attempt to measure how much artificial intelligence actually takes the load off life, rather than just answering questions well.
To be honest, studies like this make me doubt the quality of the people, the researchers at OpenAI. OpenAI definitely has more than a billion monthly users, yes, guaranteed. Hundreds of millions of power users. You can run serious tests at that scale, real tests, real evaluation test systems. You know why they still don't do them? Well, in some cases people don't understand how to do it. People are simple when it comes to real-life cases, and in other cases they don't do it because they don't want to show the truth, that with four thousand people there's one truth, and with forty million or four hundred million people the truth can become completely different. So, a bit about security.
OpenAI released GPT-6 Cyber. I think I literally shot a couple of Shorts on this topic. And this system is a new generational level of security systems, available only by subscription. A scary, scary system in terms of what it can do. It found a huge number of problems in Linux, in Firefox, in the browser. And, of course, I want to say that everyone's done for, guys. A huge number of countries in the world don't update their software. Most of the software in the world doesn't get updated, the stuff installed on Windows in lots of places, wherever it's installed on Linux, old software that was written long ago.
And how many websites don't get updated. No access, the websites keep running, and there's no access to them. Of course, that's a big problem. Right now one of the problems is, imagine how many sites are hanging around the world, old sites made once, that nobody uses. For old businesses, some people's companies, someone lost access somewhere, something else. So they keep running, especially on free hosting. Well, of course, if they're on free hosting now. Or somewhere they might even be paid for and maintained, but nothing can be changed there anymore. Nobody is going to go into that code.
And in the published report, this, by the way, I think is a really important topic, really important news has come up here. The published report describes how the model added unnecessary conditions, abandoned working solutions and reported nonexistent obstacles. Who's run into this? I've run into this many, many times in ChatGPT. On comparable tasks it performed worse when they were presented as security research. Look at that, how funny, right? That is, if you present a task to OpenAI as security research, it starts working worse. Essentially, it's a kind of sabotage by artificial intelligence. An independent audit.
So, artificial intelligence is already noticeably speeding up the development of the next artificial intelligence. Meta estimated that Anthropic is improving model capabilities roughly twice as fast as before. And a significant part of the speed-up is due to using artificial intelligence itself. Let me remind you that Boris, one of the lead developers of Claude Code, wrote that, of course, the development of Claude Code is held back by the poor-quality code of Claude Code itself, and that it needs to be rewritten.
And so we're now entering an era when all software will need to be completely rewritten. And the new software we build with artificial intelligence will need to be rewritten too, yes. It becomes outdated, and business cases become outdated. And what am I urging all of you to do? Check that case of yours that was trendy six months ago, it may already be completely out of fashion and unnecessary. So… Meta paused a controversial rule on collecting information from business websites.
The company postponed a change to its terms that, according to publishers and lawyers, gave it broad ability to collect the content of the websites of business account users. The rule won't take effect, take effect on November 20 as planned, yes. And in general, of course, this is a global problem now. Just recently there was very interesting news that a consortium of top, cool systems backed an initiative to, I think I shot a Short about it too, to set unified rules for website owners on the limits on various models collecting or not collecting your information.
For example, on all my sites I try to enable information collection by artificial intelligence. Why? Well, to show up in results in ChatGPT, or to show up in results in Claude Code. That is, on the contrary, to enable it, not to exclude it. Or on my YouTube channel, to enable artificial intelligence models to analyze my videos. It's important to me that artificial… Some people say: “It'll learn from me.” But if it doesn't learn from you, it'll learn from others and recommend other people's videos, not your videos, to someone, yes.
Well, that is, I, for example, am very interested in this. But Meta, as we know, pokes into information collection everywhere, unbelievably. You say the wrong thing near the microphone, and it's already offering you some product or some ad. So it's good that they've been restricted.
The European Commission proposed a new package of measures on artificial intelligence in education. Well, the European Commission, I think, is always proposing something. So, the package combines fighting the decline of basic skills, using artificial intelligence in learning, supporting teachers, and updating vocational education. And here, to be honest, there's a problem in the world: of course, governments and national economies, like a huge number of company leaderships and companies themselves, can't keep up with artificial intelligence.
It's going to be a disaster, because whatever they propose, they're always somewhere behind. Getting to grips with artificial intelligence the way they're doing it should have been done three years ago. Now you need to understand new things and see the future, what lies ahead, what systems, what problems there could really be, and what you'll do about it. For example, people think: let's not let artificial intelligence take over some business. But imagine it has already suddenly fully switched on, this system has appeared, and it has taken everything.
What will you do at that moment? American voters overwhelmingly backed restrictions on data centers.
As we know, there were, so, elections just now, and the restrictions were backed in forty-four of forty-five local votes. That's what we were talking about earlier. That is, the main complaints were that when a data center gets built, water gets more expensive. Also funny. Data centers are built with tax breaks, because laws were passed earlier for these breaks, yes. Jobs are supposedly created, but they didn't think it through, whether to give these people breaks. As one senator said: “Why are we giving tax breaks to the richest man in the world?” And here, of course, there's a choice, yes, how to keep human control while building up, infrastructure and capacity. How do you do that? By what method, in what way do you do that?
Is humanity needed here, or the opposite? Sam Altman: “The future must not be handed over to the control of artificial intelligence.” He said this in Abu Dhabi.
And, of course, this is a contrast with Anthropic and Dario Amodei. So, he stated that artificial intelligence should serve people, not the other way around. I have a question: whom does Sam Altman serve?
Musk and Altman argued over who should get RAM. So, Musk responded to Altman's reasoning about the efficiency of memory use in data centers: “People deserve their RAM too.” And Altman then clarified that he supports increasing production and people's ability to put as much memory in their computers as they need. Who's going to have that kind of money to put it in?
Amid investors' doubts, which are generally around right now, so, OpenAI's CFO announced 1.2 billion weekly users and a doubling of the enterprise business in a quarter. Her position: the company needs to keep investing to meet demand. And that's what we started with, look, and that's practically where we'll finish, yes. But I still want to share one last news item. The thirty-fifth for us today. But I added other news items here and there.
IMF Managing Director Kristalina Georgieva, so, stated that central banks will have to take into account a possible weakening of growth linked to artificial intelligence when making decisions on inflation. That is, her construction, as far as I understand it here, is that investment in artificial intelligence is already so large that its slowdown can't be seen as an internal problem of a few tech companies. Bam-bam-bam-bam! That's what we've been talking about. Support our channel. I'm looking forward to your likes and comments. How do you like this news format?
Today it's forty-five minutes plus, maybe there'll also be two-hour ones, and maybe ten-minute ones too. See you in our next episodes. Bye, everyone.