Google's personal intelligence changes search, browser and Internet skills. The scientific race for AGI is becoming more formalized. Who controls the use of the Internet content in the AI search and under what conditions? We'll talk about it very much today. Anthropic, the largest multilingual qualitative study of artificial intelligence. Well, at least they call it big. And three million messages a day at ChatGPT on the payroll topics.
And of course, there are many other news. Hello, everybody! We're on ToTheMoon. Technological news and sites from the Silicon Valley and around the world. We go out every Sunday and special Wednesdays. So where do you want to start? I want to start at the beginning, with me-- everybody hears me saying that we're on a graduation? One hundred, huh? What...
One hundred and two.
Oh, one hundred and two. What, uh, I don't like ChatGPT OpenAI having an uncomfortable model choice interface and, consequently, a bunch of models, a bunch of versions, no idea how to choose them. And then again, they, by no means, changed, at least on the web, on the web. I'll remind them they have an application interface in their entirety, in their mobile application, it's probably theoretically different from Android. Probably different. They have a separate and web version of their own macOS, yes. And on the web version, they've changed the choice of models now. I think he's kind of a little more comfortable, because I kind of default, I can choose what model I'm working with and what model I'm working on, what kind of regime inside. Because there is, actually, a model number, and there's a way to interact. It's instant, like a quick answer, thinking and... And there's...
More cars.
There's another guy-- you know what's a car now, why didn't you call him? Because now the car, uh, na-- no, I'm talking about the mobile application, by the way. So, the car on the mobile application and the web disappeared. I don't have a web car. I just checked. I'm on my web car, I don't see it. And the third one is a parameter, it's a choice of the story. So it's five four, five three, five two, right? And there's a way to pick it up. So you can choose where you work five four, five three, five two. Even to say, for example, that I-- for me, the basic model is to think and think. For me, the base model is the maximum heavy, for example. See how it's gonna hold on and work, if it's gonna be inside. You've got it up, too, haven't you?
Yeah, there was a model selection configuration, and I looked at it with one eye and I closed it right away.
From the beginning, it's a Google story. And I'm just-- I think Google's incredibly going off. Yeah, again. What did he do this week? This week, in the framework, we've already told you, by the way, that they have a notion like personal, personal intelligence. So, in all their application, they've all turned this AI mode up to date, turned it even bigger, right? I mean, AI's a way to look, there's AI mode in different applications, well, inside it's clear that Gemini has an AI mode in Chrome. And the fact that they're starting it now, and they're starting, they've already made it even bigger. I mean, he's in the pictures, he's in, like, a little more cool in the mail. What are they talking about? So I'm gonna give you some examples now. So what does Google say now? Google says that, for example, a product selection based on your past purchases and tastes, right? Or, for example, the device maintenance, even if you don't remember the exact model, but you had it in the checks, so you had some check and the maintenance works completely. They understand you have it. Or, for example, they're describing the case. Well, it's a case, of course, specific, but still. They say that you can recommend food at the airport, taking into account your tastes, knowing what kind of gay you are, you-- from which gay you're gonna be leaving and what time you're going to leave. Or, for example, they say we're gonna set up under you, like some personal story about your journey instead of, like, a top 10, and on the basis of your, uh, your search history, at least, What you're looking for, huh? Or they say, "We can find you something interesting, like a hobby or some interesting action based on your interests." I mean, you don't have to put your endless context in. It's just-- it's a little cooler. It is clear that in Google data, people have a huge amount of data, even Apple users. Like a friend told me recently, and as a matter of fact, there's one of Max Grigoriev's symbs, he said, "Well, you're like Apple's a user, but still you have a mail in Google the main, and you're missing a--- You have Google Drive big, and you're still missing. You're holding it right there." I just asked him to put me ten terabites on Amazon servers right now, downloading some of the Google Drive data there.
And he's on this, and he told me that, uh, look, that means the unfettered Apple user. And they have all your data. I'm sure I like this subject very much, but he called me to put in an app for Google, by the way, saying it's much more, it's already getting a lot more cooler than Apple. I'm basically, uh, I'm probably not sure, yeah. Apple used to revolutionize the photo applications. I remember that moment, they were really swinging. Now, maybe Androids are really cooler. I don't know. I don't know. Write, by the way, who thinks and who doesn't even-- there's a guy in Apple, Android, who knows more about who the user is, and who, for example, is. Here, uh, but I really like these things they offer. I use Google W-Gogle search, first of all, another, yes, yet. I'm still a Google searcher, even though I have Safari, not Chrome. I'm supposed to be thrown out here and spit. Here, but that's how the world works. Oh, uh, I'm still using Google search. And, uh, retracting this story is really cool. Here's the question: what volume can it pull, yes, what's the context window? Because I remember that ChatGPT is the contextual window of my search requests all over ChatGPT. And yet, periodically, when he makes a recommendation based on my personal data, I want to hit him or throw him out, or I don't know what to do with him, because he can even call me a woman or not. I'll figure out where I live and where I'm staying. Or if I go to a country, say it's my favorite country and I have to go there. Well, I mean, not a pony, not knowing what it is. And even change the context in one dialogue window, I'm not saying, I'm not going to go-- well, that's, without understanding the original meaning, and then there's a new set of questions. Oh, but they have a pdddata directly serious and big. And we're gonna talk about Google this week having a lot of very anons. The pddette is big. Elnar, you're here, uh...
Yeah, with personal intelligence, remember, we talked about it some time ago, and then they rolled it in a beta version for the fee-paying users. Now, it seems that the experiment has been well, and they're rolling it out to all users in the United States and even those who use free plans. I mean, well, just the usual Gugla users.
So it's just that, uh, the test version was good, and the company decided to roll it up for more people. It has a clear advantage when the search is more personalized. But there are minus. As you say, the company is starting to own more and more information about you, starting to pass it through itself, which means more personalized advertising. And maybe people will buy those things. Well, on the one hand, it's good when you're offered to buy what you really need, and on the other hand, you can get what you might not need, right? Consumption is increasing
Yeah, and you're the one who's gonna be making these requests. So they know how to manipulate you.
That's right. I mean, if they know you're on some dates, you're doing, like, impulse purchases or after you got back from somewhere, out of business or somewhere else, you're doing some impulsive purchases, you're gonna get you some kind of... They'll just put them up, knowing you're right now, you've had a long flight, like. You're back, so you need to show some things you don't need. But the company knows you have a weakness on this side, and it starts to throw you some commercials. But knowing the whole context that she can reach. Oh, I see that information is very much, right? And there's not all the information online, but this profile of a person is quite possible to form a certain amount of data, keep it in some of the user's clusters and actually show it to him. I'm gonna go with a rogue commercial, some sort of thing that's happening. So, uh, on the one hand, okay, right? On the other hand, it's not very, but it's where we're going.
Yeah.
Whatever it is, we are going to this side. The amount of information that companies are processing about us is getting bigger and bigger. Processing speed is getting bigger and bigger.
So it's only to accept and hope that the pros will overweight the minus.
Interesting, by the way, this is the extra thing you said about the window. They say it's not just-- they want to solve a problem called, uh, packing this whole content as a basis for a request. It's not just having access to your data. And for example, there's a problem when there's a million tokens, by the way, they've got them.
A million tokens, by the way, they have. But you can't just take the dodge, collect all your con-context that she read, right? Like I told you, I have some space number of different files and send a request. So the system, well, don't stop. So they say they're going to the-- to the history when the model uses, they call it a tool for use dense retrieval and long context. And they say that for real-- well, for this real personalization, the system dynamically collects the necessary pieces of data in the model's work memory. And here's the story, and the basic thing, is that I was saying how she was, as a matter of fact, when you don't have a million tokens, there, and a hundred million tokens, how she understands these basic work pieces. I mean, this is the main part. Although, in fact, it's built, all the models are built, right? They're the ones who give you the answer, they're the ones that need to give you the most important answer. I'll make a rizoning. I guess. By the way, the idea is, if there was a way a man could pay the money, like, and if she could get more of your data, that's what I'd be doing, I'd be interested in it personally, wouldn't it? I mean, if she was a little more trained, spent more time, wasting a little more of that resonance, setting up, developing some architecture or some kind of data, repacking and I knew where to go. Although, of course, it'll take time. And here's the next story that's there, uh, We'll be right back here, by the way, to Google, right? And about how he is and where, and what's going on with Britain, it's like a very good example of a world. I think it's gonna be a story all over the place, and it's gonna be done, right? There's a concept of opt out Gu--Gugle EI, Google EI, right?
And before that, I'm just saying that about understanding, uh, that's the Osle-- there, that's the concept of research, yeah, man. So, Antropic says they've done the biggest, even the biggest human behavior study. And allegedly analysed-- not allegedly, but analysed. Oh, I'm not sure. I mean, people are being asked. They said there were four sec questions, right? Ah, a total of eighty thousand people. And Lena, I remember, yes, that we interviewed half a million people in about six months and questioned them a month. Well, we have 100,000, over 100,000 registrations a month, and there's about 60, seventy, eighty thousand forms we've been doing unique. Plus, we understand we've been asking for another additional traffic of current, which is what people who study, there are hundreds of thousands of people, and we've been questioning them in 50, there are five, six hundred questions that are enough in there. different cuts. And plus the people who studied, they asked about a whole range of questions, wishing to have real information about how people use the system. And here I have a question: people who have dozens of millions of users. Well, Antropic has a good one, they don't have seven hundred million users, but they still have hundreds of millions of users, not really tens. They could have asked dozens of millions of people in sufficient detail, and people would love to answer them because there are millions of people, egregious users of the Antropic, as well as the epic users of Gemini, As the purple users of the GPT and X AI readers, there are millions of users. Grok, huh? I'm not saying, I'm not saying, there's Chinese systems. And these systems could already provide a vast range of research. And there's a moment or they can't do it, well, they just don't have a research product, although they're here to be buried. I'm gonna tell you what they did for the study. In general, there are some-- some interesting figures, there are some interesting figures. Or they don't want to question people, because they're gonna have to question, for example, five million people, to show some kind of problem people are facing and show them out. Because it's always a question. I mean, there are companies we think, which means this company, why I won't ask for it. She doesn't ask questions because she doesn't want to know them. So they don't want to face the fact that they're gonna have to open this information. I mean, that information will come out. Just like the huge sa-- the number of sa-- portals that collect data. There's Booking, like, yeah, at hotels. They're not asking all the questions. I have questions. Booking, by the way. I can still choose my numbers back. Booking asks a different type of questions. And they try, try to dig, yes, before the answer, but they don't show the user, by the way. They're digging, but we're seeing a aggregate scale. So we see a common score and we see a clip, there's seven, five, ten parameters. And then all the other things we need to look for through the feedback. Well, now we have GPT chatting that can scan two thousand hotel feedbacks and still find different points there, although they do it in a very specific way, but they can still. But they don't show us the details. They're important. Airbnb, for example, does not actually show much detail. Why doesn't it show? Well, because if they start showing all the details, then their clientele will say, "Look, we don't want these things to be put out. We wouldn't." That's what it's always about-- that's the brink, right? Dancing, right? Amazon and I will see.
There's a retraction, a rating on the product, three or four times a response and--
No, why not?
No information yet.
You can read all the feedback in detail if you're right.
You should read them. But, Tan, look, in the modern world, the information is structured-
Well, you just have to push one star and watch. Do I know what I'm doing?
I do that.
I'm watching one star and five. And I have-
I do that, too.
It's a bit of a feeling. And the cool pictures are really looking.
I do that, too.
The people who did.
But when you go to the hotel, you-- you, like, go to the hotel, and there's a pretty standard story. Like silence. We know there's no sign in Booking, like a silence. But silence is a very debated story. How quiet in the room. And silence has many different parameters. First of all, which side is not quieter at the hotel, what floor is not a floor. There's another street. I'm in Cairo right now, and Cairo, and I got a kind of fashion number, but he's on Neal. Because he's going to Neal, I'm talking about it, ChatGPT warned me about it. All the numbers going to Neal, noise, boat music, noise streets and everything. It's often better to have a worse view, but you're not gonna have a noise. Yeah, that story, it's defined, it's known in many hotels, in many cities. Not much of a modern artificial intelligence is pulling it today, but we don't see that parameter, we don't see that parameter. Because if there's a parameter, you're gonna put a hotel on the table, and then everybody starts making that noise, automatically clicking, right? Because we're seeing him now in the feedback. But people, people forget. Not all people write a super-designed revocation when you get structured answers, normalized, right? That's the structure that you're getting right. Booking, by the way, is cool, enough normalized information. If they take this parameter out, you'll have it.
You'll have nine at the hotel, for example, and quiet will be alone.
You know, people write feedback when something really bad happened, and when it was great, like, even leaning out those stars.
Look, Tanya, hotels still have, like Booking's got enough, Booking is known to be a strong feedback system. And if they have a hotel in the main, in the main cities, yes, there are some cities that are problematic. If the hotel is worth nine points zero, and it has 100 plus feedback, it's a real rating. And people are still being put. People do ratings and ratings for workers. Booking story. There's a system where you're right, right? I got a comrade yesterday telling me that, for example, in India, there are a few places, where the Google Maps rating at the restaurant doesn't say anything because they learned to do it, like they learned to do it. And that's the question. For example, Yandex, the Yandex feedback in Moscow was quite a jobless in the restaurant.
Booking, too. You know, Sasha, I'm, like, my acquaintances, ours, ours, ours, they're in Hawaii for New Year, they're out of the roof because Booking gave, booked. But.
It happens.
And I had that same thing in Rome at night. I mean, I'm just getting this system.
There's always a story. Look, there's always a story. And here ChatGPT makes mistakes, by the way, because, like when I picked the hotel, I have a lot of hotels on my trip, and when you have two noises, two noises, and he's on the 2,000-square, They're focusing on your attention. And then I have at some point, "Look, how many reviews have been?" He says, "Two." And you're like, "I don't care about that information." Because when you have two noises from thousands of reviews, you have that information. But look, we're not talking about this, we're not in the middle of a job, not a job of recall, a system of everything. Now all that's about is if you need to get this information out, you need to question her. The question of returning to the anthropic study is that these systems do not collect many data. What's that? Because they, uh, they're not mo-- no, well, they can't, they don't want to do these research, or because they're not interested in collecting and extracting these data. I'd certainly put a big moment here because they can't lie now. Well, it's not companies that can falsely report data, they're gonna be called to Congress, they're gonna ask them to get that report, there, and everybody. In the U.S. Congress, they'll just be broken into scapula. I saw a congressman in the US taking the Starbucks' glass of the most recent drink, the most popular. Well, everyone knows, yeah, probably riles-- racels and mems.
It's coffee before, and now.
So coffee is coffee. Now, the fashionable subject, and you have, you need to say 16 words to order something in Starbucks. And he's showing a glass in Starbucks. Well, he, he, he, he, he, he has a famous, well-known, clip-man, I don't remember his name, but I know him visually. He's got hands like that, hands shake, everything. He's probably standing with that glass and reading inside the ingredients, so this caramelize--caramelized latte, there, with vanilla taste. There's that--that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that-that- And he, and he says it's, like, eighty-five percent of the day sugar. And, well, he read it and said it on that subject. So, believe me, this research will blow them up. But what's interesting about this, on this subject? There's a story in this subject where people have a concept of light-light, shadow, right? Well, it's a positive-different series. And for example, people in the group showed that, uh-- fifty percent of the time savings vs 18 percent of the false productivity. That's interesting stuff, isn't it? When some-- someone says, like, that I'm in there saving time. And there are 18 percent of the people who, in this, are in this-- fifty percent versus eighteen, yeah, that in this couple, say that you're actually saving time, but you have There's a false productivity.
It's a personal assessment, yes, Sasha, so you're the people who answered the questions?
Well, people, people filled up, yeah, people filled up--
Well, that means there's a higher figure.
A pretty serious, serious report. But look, look here, look what they're talking about, and I read a pretty good study, you know, I've been studying it, yeah, there's a lot of data. They still say they were able to weigh these points well and they're good at collecting and formalizing. I'm pretty good at studying the big polls, right? And where the findings work, where they don't. Because there are some answers to a question that is actually falsely credible, right? And they're not making any decisions. I remember that. Remember, Elnar, we had this question: your family's computer is family-friendly or personal? And people, for example, if they had their own computer, their personal family, the likelihood of learning and the likelihood of online learning, has increased significantly. Well, I've been up straight, there, more than twice. And it was a certain amount of confidence. I mean, if people answered that, they've talked about it quite well. Yet a large number of people still have a common computer, one computer, just like a computer. I-- because, however, online learning, online learning, it required a lot, at least, programming and technical stuff, it certainly requires a computer. Or there's a couple of things like that: emotional support for 16 percent against emotional dependence for 12 percent.
A very interesting couple, I think.
Really very interesting all the numbers.
Yeah, very interesting. So on one, on the one hand, on the one hand, you have-- that's why I'm saying the research is generally good in terms of his quality of work, right? We'll talk. Antropic also swing-- AGI. That's really cool. Oh, not Antropic. And that's a very cool subject. Very, yes? They seem to have done an incredible subject. But here they are, too, young, in terms of conclusions. But again, I'm eighty, eighty percent, oh, eighty thousand, eighty thousand people are embarrassing. Not enough, by the way, they've been studying 100 people, now I'm telling you, 150 countries and seventy languages. And when I see this kind of dispersal, I always laugh with him, right? This reminds me of these studies, statistics, when they were posted there in Russia or Belarus, and they are posted everywhere in the States. We questioned people, so these people were talking about, like, the Trumpa wars. Well, I'll say something independently for the Russian language. And there, and then I'm in the horse-- I'm leaving right away, and there's a thousand, seven hundred and fifty people written. I'm like, "Well, guys, well, whatever..." Because if you're eighty thousand, you're gonna split it up for a hundred fifty-nine and split-- and you understand, yes, the number. So it's less than a thousand people in the country, that's nothing. So in the modern world, you could have been a little... a little bit of a bunch of people who sit in the system and still use it, and that system is still being used, right? And this research is very specific. Here's AGI tonight Google what they described, they've done more correctly. I'll tell you a couple more. Or Nara will supplement it here. Improvements of the outcome of twenty-two per cent vs reliability of thirty-seven per cent. And by the way, this is the only couple that's got less power than a plus. So you have the system that improves your decision and your reliability. I mean, people still say, with the reliability, something is unreliable, unreliable. So, economic growth of 28% vs is supplant, uploading of 18 percent of the labour market. There's another couple. One of the main ones is a contradiction. It's basically a contradiction.
Thirty-three percent of the people study and the fact that 17 percent are cognitive atrophy. Speaking of which, I think cognitive atrophy is much higher than training.
Yeah, Sasha, 100 percent.
I think I and I would personally say ten-ninesome, right? Because I think that in the country-- if the less developed the less access to education than less education is less well developed than the less money invested in teachers and access to teachers, the more it will be. Cognitive atrophy, right? Someone with me might argue and say, "No, in developing countries, it's gonna be tough." But there's a story that says modern artificial intelligence systems help to train in small-scale countries with no access to learning systems, with poor learning systems. Teachers. And, in fact, human-- people learn inside these systems, but they get used to using them forever. And yet I have another straight model. Yeah, well, my daughter's studying, well, I have three of my four kids in school, but the oldest daughter is such a more spectacular example. She's in the tenth-- E, tenth grade. I'm not allowed to go there. So she doesn't even fit the GPT chat and she doesn't open it and try not to open it under any conditions. They can open it, they can open GPT chat, but they have direct restrictions that they can use, what they can do. They're taught not to use them. Well, because you-- uh, I'm, like, now, so if I was going to study somewhere, I'd have, uh, I think there was a big problem. Uh, I don't know--
You know, it's good when you're already wearing GPT on some base. We're still growing up.
Yeah, yeah.
In classic education, we've already had some structures in our head, which means people don't even have the capacity to form, and they're the ones that think.
Yeah. And one thing you do, we're in the offline, I'm putting on GPT, I don't know, I'm learning piano, and I'm using GPTs to find out, I don't know, composers, music--
Yeah, yeah.
Theory of music and so on.
Additional sources.
Or I work in the garden, sat, sew the plant, and I use it. Or I paint...
But he doesn't play for you, he doesn't make up for you. I mean...
Yeah, yeah, yeah.
There's no
cow.
I'm drawing, yes. Here, here's the type, and here's the type. Or, I know, there were more than 10,000 kids. I never wanted to go to school, although I had a daughter there on-- but it's more like I just put it on the dock and I don't really think it's a useful education when you go, and you. You start, for example, drawing, painting or drawing with the help, a, prom, yes, proms, that's, uh, another subject. And I'm, like, very drawing, and I'm gonna find out about paint, about a whole bunch of theories of everything. I use GPT chat, he's really helping me. But, but at the base, if, for example, I work on accent and work, use GPT chat, well, like, well, again, you're probably gonna be, like, over the accent, you're gonna say, yeah. Just over my tongue or I'll be teaching math there, I guess. That's a big question now in the contemporary subject. And or I'll be preparing for the rights test and teaching GPT chat. That's a big question. Here, uh, Anthropic, Ilnar, what do you think about this research?
The study is great. Aah, it still causes me some distraction when I read it. You know, from that series, people gave that answer, right? And here, you have to keep what you said under the star, just eighty thousand people. You don't know what people are, but you always stop yourself, that, well, some people have given such answers. I don't, I don't always agree with them. That's how you say it's there, it's atrophy cognitive and other stuff. I have a different feeling, too. But as a whole, there are a large number of people, relatively large numbers of people give such answers. And I'm curious to see it anyway. I'd like to drop something else about training. First, on the other hand, yeah, look, when I got the license when we were at school, there, on the car and stuff, it always felt like we should learn to ride the mechanic first, but on the mechanic. I learned, then you're getting on the machine. And it was supposed to be that when you drive a mechanic, you have a car understanding, other things that are much deeper, you start to understand. But in this case, if you take the last one, I don't know, five, ten, fifteen years, do people who start driving around with a machine have any problems? Well, obviously not, right? Because no one needs mechanics now. And start with mechanics, well, only those who are there, I don't know, are cars running, riding or something. I mean, just a specialized part. And it might be like this training with AI and without AI, it could be something like that. After that, I don't know, there's ten, fifteen years of age, we'll look at Aboriginal people who, like they're starting to learn from mechanics, so then we'll get a better understanding of the car. But on the other hand, the same Anthropic had a great study. The reference will be made, too, about the one, and the...
- to suggest how, uh, the use of AI affects the john. Aah, they've had a very fun experiment. I mean, well, we're starting to have a technician. They've collected some of the starters' software, aaa, they've got a mission to work with some library, yeah, well, so, with some technology, we'll talk like that. Aaah, they've made sure they're not in the process and they're giving them a mission, that using this technology, we need to do this. They were smashed into groups. One group of cartridges, uh, gave nothing. You can use the old Guglom, read the records, get it done, try to solve this task, there, on your own, on the old-fashioned, like it was, I don't know, there, five years ago. Another group was accordingly given the use, uh, of AI as a consultant. I mean, he could have been questioned if something was unconscious, something else. And the group that said you'd use it, you know, if you like. You can have him write everything for you, as a matter of fact. But everyone was told that there's gonna be a little test at the end of this technology. So, besides having done the job, they were also tested through the interviewer how much they, well, at least learned something. I mean, you could only answer the questions of the interviewer on your own. And the question is, what's in the head?
We're leaving.
...people? Well, it's like, yes, the is a very obvious conclusion: those who, uh, didn't do it on their own, and, accordingly, just delegated it, uh, Claude code, yes, because the anthropoics were, they used, Nothing in the head is practically left.
Yeah.
They didn't know anything about this technology. Aaah, even though they were able to do the job. Uh, those who do everything on their own, spent more time, spent more effort. Aah, not everything, I don't think I remember the details right now, but not everything, in my opinion, even if I was able to solve the problem. But the polling man was much better, because it's the whole story of the whole universe. Also, the people who used Claude as a consultant were very well shown. Who did not ask for ready answers, and who, as a matter of course, was dealing with the details, trying to figure out further, either then his code was throwing it on the test or something. Well, I mean, more, uh, domestic investment in this story. And there's like, on the one hand, the obvious results, right? On the other hand, it's just a story with atrophy cognitive abilities. If instead of being able to deal with it, doing it alone, we're starting to do something completely, then yes, there's atrophy of cognitive abilities in this direction. I don't want to do it on my own. You're thinking, "Oh, now, I'm gonna throw up." Although it's a very similar task, you're starting to throw it away again. And how hard it is to get back from this story is enough.
And here's a perfectly clear effect.
I think the more these and agents will improve, the more they come up, the worse it gets. Now that these instruments are not perfect, but when they become perfect, then everything is gone.
Yeah, the question is, well, this story is that one thing is, when we use artificial intelligence rationally and when you have a base, there's an understanding. That's how Tanya said you were putting it on base. Just the question is, people understand the base. And it's another thing when people do, you know, how a man doesn't understand that basis, right? And here or he needs to be confined to school, right? It's just like the people who were always writing or, uh, walking, and the ones who didn't write, walk, whatever, there, limit it, not restrict it. He's just like that inside. Yeah? A man like a man. Or he's got a family that says he's got, or something else. There's a different question about these rules. I was just reading, that was a very large article, which said, as well as a very strong presence of artificial intelligence, an increase in education as access to education in a vast number of ways. countries. And the other thing is, Elnar, what you've been telling from the point of writing now, when it's like a task, and actually, there's a huge amount of detail inside. And then what happens to it? This building will be standing in 20 years or not, it'll fall because it didn't solve the question inside. This, this story, yeah, well, the building is just a clear example. Yeah? It's clear there's a lot of other rules, but it's a good example. Yeah? Or, as for medicine, did some of this delicate nuance and a delicate component really have been put in there and taken into account? Did he really take it? Did he really understand? Yeah? Did that really limit itself? Not much of a medical condition, you realize that since you have this system that can still be hallucinated, that is, it's not 100%, the current artificial intelligence for these tasks is not generally appropriate. The base doesn't fit. He, he complements, he gives extra opportunities, he reinforces a man, but he doesn't fit as an end-state solution. Because in the final decision, it might not happen, like the video was just a little bit out. I don't--we'll find him and show him. At the California restaurant, uh, the robot was there doing something. This robot got mad and started dancing forever, just to fight. And then his staff tried to turn off and quiet, so... Well, people took it all off, didn't they? He was there, so he was just driving his hands like that for no reason, right?
And, and I did a lot of different things. And here, and there's a moment, yes, what this car can do. And it's good there's no harm to anyone. But it's a dangerous story, isn't it? When you can even move your hands like that, it's all over, it's already accidents, it's, there, a broken, broken thing. And people can get hurt clearly and have a big problem. We still have two, two incredible topics today, which need to be heard and addressed. And we'll be back on HR. History is first, so... In Britain, we started discussing the topic to make a concept like opt out of AI Google. What's the main design? Well, they say Google's dominating the search. Well, obviously, there's more than nine percent in Britain. What's the subject? They believe that, aaaa, the site has the right to abandon the Google AI Index, while remaining in search. It's a very interesting subject. Yeah, how do you know what AI is and how different summarys are issued, and how it all happens? And the subject itself, aaaa, it's kind of unattractive, and-- but it's really not really important. That is, it is affected by the new context of what is going on. There are extreme topics. In New York, for example, one, uh, one of the politicians introduced an act to prohibit the general provision of responses to the model in certain subjects, such as medicine or psychology. But, well, there's a subject in medicine and psychology, there's a science and engineering thing. That's why it's very, you know, like that. How interesting is that? Answering what does that mean? Well, or legal answers. It means you're not getting answers, you've got 500 diskleimers writing, restrictions, yes, about where you can't use it. You don't get the system to answer. So you're asking, for example, some medical question, and you're saying, "I'm not giving you that question at all." But I think it's just tough, hard-core and thrash, right? It's like going against the world at all. Because obviously, it's obvious that will obviously be models in other countries that will answer these questions, they're already answering them. And it's just a race loss. Never in a lifetime, it's not gonna be agreed, at the government level, but it's a line. There's a real story, right? This is the subject that the British Government has opt-out, and they are asking for, by the way, four sets of measures that are different. And about how and how it is, how it is, how it is, how it is, how the content is generally accepted in terms of extradition, aa, then how the screens are hit differently and how the data are transmitted inside, Yeah? And there's one of the things they were, first of all, in the request that they were default, well, they don't even discuss that when Gugla refuses, this story should automatically be transferred to another system. It's a fun law, by the way, and all systems should support the data transfer between themselves. I mean, conditionally, if you worked in Google search, you have a search history and when you move on to another Google browser, you have to transfer your whole story to another system. And on-- that's a really, really cool subject, isn't it? I mean, if you've been working in the same system, you've got the ability to change easily between these systems. Here.
Illar, do you have anything to add to this British story?
I wonder if we'll watch. So the deletion of the AI index site is probably linked to the fact that you--AI Google could not use the data from this site at least. Here. But as much as Google would probably be the result of ranking above those he knows more about. Here. I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-I-
Yeah, but they just said you had to be... not. You must take it all properly and that's not gonna happen. Yeah? I mean, here. Well, there's a subject that Europe is very much asking to reveal to them the endless algorithms. Yeah, there's a massager out there, searching, everything else. Well, there's an endless struggle, yes, this is the American system with Europe. Like the indicator of some kind of requests, where they're a bunch of details. They say, "Look, you just want to get into our protocols and, uh, hurt people, so you can watch them more." And these people say, "You're watching." So, shorter, that's what I was saying, "You don't want to take down this story, floating with a doll? So finally, I'm not gonna have to make it go anywhere." Here. And I don't know, some kind of boxing, and I don't know any more about a bunch of people in the world. Actually, it's just not--
He's asking about cookies. They're not asking. I don't know, there's a tie like that, maybe some, uh, maybe some browsers are, yeah, there's a tie like that.
And I hope that law will soon come, yes, where I won't, I won't have to let cookies do it because I have a window open, a window opens up, I click, spend time, and, well, it's just, It's a terrible thing. And Ilona Mask is here to support. That's just, uh, uh, uh, An important topic of the, uh, that we will touch with you today and, by the way, before we move on to this topic on AGI, and it needs to be addressed unequivocally today. I was just saying it in our ancestors. I'll tell you what, uh, OpenAI put out an interesting subject. They said that every day, and this is an official, official, full notice, every day, almost three million reports a day with wage episodes. And, uh, what--what's interesting about this topic now, I'll tell you. She looks like she's, uh, obviously asking about wages, about the bonuses, about their formation, about the details. But, but OpenAI says that they are now in fact getting the opportunity and they are developing their index, and they're putting in a workbench that is a new sign for the labour market and a useful tool. workers. And in fact, they say that ChatGPT is five points four, they compared it to the data of two thousand twenty-fourth year. Oh, uh, there's a system that shows all the average wages and everything, and the system gives very correct, correct numbers. This is a very interesting subject, by the way, for the world as a whole, in terms of vacancies, summaries, search, job search and the dismissal of people and the employment of people. I remember, Ildar, there was a theme over a year ago that the OpenAI would launch its portal for recruitment. They never let him in. This topic was, like, in Russia, the HeadHunter or the American LinkedIn, is indeed and so on. Yeah. But the point is, the OpenAI is obviously moving, but this baccalaureate is very cool. I mean, if artificial intelligence systems can be able to figure this job-related topic cool, the resume, they can do it, they can accurately call wages, some kind of comparison. And, uh, this is an incredible movement towards combining vacancies, a summary between themselves, on the one hand. On the other hand, of course, creates even greater chaos, right? Because, on the one hand, it's kind of a plus when you can consider a million AI résumés to find yourself people. On the other hand, it's a little early and it actually takes away the perseverance and the living. I just see how many of these systems are being used in a huge number of companies and how inaccurate they work, right? I even know about the OpenAI in terms of their responses, how they have portals, how their system responds. She's responding very well, and their HR is not really automated.
You know, when there's a system that serves people, the idea is they, they'd be the first thing they'd do for themselves to do. They don't have that. I see you want to supplement it here.
I want to, and on the other hand, I'm a little worried that I'm gonna get a little out of context right now. But let's try, since I still have such thoughts coming around, associations. There was a "Doctor Who" show, I think it's still coming. But I was watching him when the university was finished, there's David Tennant filming. So, BBC, fantastic science show, so scientific, so, let's say. And there was a series of pretty busy places where he was dating a creature that was reiterating his words first. Well, I mean, I'm saying something, and then I'm there in a second, all the words are repeating. At some point, it actually comes to what we start talking to this creature at the same time. And then this creature starts to say what I'm saying before I do. Yeah, well, there's a pretty busy series, and, in general, the feeling is pretty interesting from watching this series, I had it back then. So what does that remind me of? People are asking about wages in some industry. ChatGPT is looking for this information, aggregating, and he's giving, yeah, he's following the market. At some point, he will be with the market to be in and show up with the exact relevant data. At some point, the market will start to follow what ChatGPT shows. I mean, if he says that plumber for this job, you have to ask for so much money, well, people will start asking for so much money, and at some point we can be in a situation where ChatGPT leads us and us. Which indicates how much money you need to ask. Well, it's just associations that I have occasionally started about Doctor Who. I was gonna say a few more episodes ago, but it was all over. I think there's not enough viewers who know this show, especially with this series. And if you're ever familiar, write in the comments. It'll be interesting if anyone remembers this.
This, this choro-- this is actually a good complement, just like complementing what we've been saying. And the subject of the labour market in some of the episodes will probably be considered, perhaps some special issue, even in our output, when someone in this subject is going to move very much. But again, this is the story that ChatGPT has now put on people's queries. Well, first of all, it shows that three million messages a day, actually, it's not that big. It's 700 million users on the weekly, right? At the same time, it's a decent number. I mean, they have, I think that the daily users are less than four hundred million, probably, yeah, well, less than twice, they would have shown that number. And yet, it's a number, it's already, it's pretty good. Plus, I think a lot of people don't understand what they can compare what they can do, how they can make calculations, like this. And what ChatGPT insists that people have the opportunity to count, count, make some kind of assessment of their hours. Because there are still jobs and countries where it is very difficult to verify the quality, how well the wages were paid, how it was counted, what details were, whether or not they all took into account.
I mean, now that's the way people can get a more detailed look. Write by the way who uses it, who makes such requests at all, who does it? So we'll go to another subject. Which means Google, Google DeepMind. It's a very super-brain, a very important subject, because everyone's always called AGI, so it's super-intelligence. They say he's a story, they say it's a thing. It's a little abstract, right? We know there were a lot of people who said he'd be there by the end of the twenty-fifth year. Then they said they'd be there in the twenty-sixth year. Someone says he'll be in ten years. Someone says there, man, he'll be there in a year. And we all know that all people, these people are the big ones, they say what they want, means whenever. This is the example of the big Mark Zuckerberg, which means the company has changed its name to Meta. Uh, to get on Facebook to make Metaverse, and now Metaverse is completely closed, right? This is, by the way, set out as the biggest, most huge phase of the world in technology. And the largest amount of money spent. It's worth 80 billion. Well, some money, huh? Although Zuckerberg last said that if he had to spend $200 billion in artificial intelligence on a task that would eventually be a loser, it's okay, right? Uh, here, but still. I don't see any problem with Metaverse not working. It's a question of market movements and different things. But still, the sound of AGI is about the same, in the same zooteric knot. So, what's their research? Interesting, huh? And they, by the way, started a hackato at the same time to help the community build the necessary assessments on their own, yeah. Kaggle, Kaggle hackato's been launched. So their main point is that there's a lot of talk about AGI, but little empirical tools. It is therefore difficult to understand honestly how close systems are to human common intellectuality, right? The idea is simple: not to ask binaryly all the time, is that AGI or not. And that's what they did, it's not a new definition, there, AGI in one line. That's the whole point, isn't it? And to decorate the common intellectuality on a set of cognitive abilities, to test the system for each of them. Speaking of which, the subject is interesting, because when I'm modeled and texted these baccoons, it looks like, well, really, you're already-- I just, I'm just, I'm already stopped watching them. Grok puts it out there, says, "We're the coolest in something," right? Then you see that in the general evaluation they're very much behind, there, I don't know, from the model, from the last version, there, Gemini or ChatGPT. And these baccoons, they're very specific, right? And here, which means, uh, to unleash, like, a set of cognitive abilities and to test the system for each of them and compare the results to the distribution of human results. This is the interesting thing they say about a chick-human. They said, "I'm gonna need to compare people's IQ." I'm always impressed, by the way, yes, compared to people about IQ. I think it's, uh, a tough bachelorette. Oh, come on. Or, for example, they said, "Don't compare people to a region, there, use and compare people to something, somewhere, super-trained or untrained. And they found some medium degree of secondary education. It's about people who graduated from school completely, there's 11, right? There, medium-sized, that is, they started looking for a more or less statistically average person on Earth. And that mid-statistical man is compared with, uh, re-moving IQ, right? Because the task of AGI is not an IQ. There's one of the moments of the very cool ones, by the way. They said, "Well, if we give AGI a calculator, then we compare the man with the calculator." That's really cool. I mean, this is what I think is a great subject, right? Because it's what you say, it's a professor. Wait. If you let the professor give ChatGPT to a companion and professor, then the professor, together with ChatGPT, is like a hotter ChatGPT professor or not a hotter ChatGPT professor, right? Ia, uh, this is very, well, that's a good, good mark they've been doing inside. So, uh, what they're doing next. They're, uh, and, by the way, they're writing straight that the Freumworth is focusing on what the system can do, not how it does.
So they try not to tie the evaluation to a specific architecture.
assessment of the specific architecture of the model, which, too, I think, is incredibly important in modern reality. And they've identified ten cognitive, uh, blocks, right? They call them faculties. And they write that it's a base eight building blocks and two building blocks that show how a few abilities work together, right? Well, I'm just gonna say a little bit to start this. I think that in different ways, uh, we'll be raising this subject. And, uh, because this is a very interesting subject, yes, and she's gonna move, and we'll see what other companies are gonna do. But still. So, first, perception is extraction, sensory processing inside. They're telling me what's in there. In the I-- visual perception, there's brightness, contrast, color, depth, movement, format. Then ID's on the objects, huh? There's a lot of things going on, counting, understanding statistical dynamic scenes, segmenting words, how much the text is understood correctly. Audio-view. It's the same way, that's the perception. It's loud, it's thin, it's a sound-speaking, sound-speaking, speech segmentation, rhythm, speech understanding, talking-speaking recognition, audio-sage understanding. Then, uh, reading the text as a separate fashion for artificial intelligence. The different symbols, the smashing, the understanding of the language and the code. And there are other mods, like, there, pressed, aaa, odor, which is, what temperature, multisensorship, etc. But they're not yet on the first version, there, taxonomy. Why? Because, well, this is definitely a very specific topic. I guess the robots are already there, right? But the equipment needs something else. The second theme is generosity, it's the production of exits, that is, text, speech, action. And there's a natural language in it, a code, a speech with clarity, grammarity, choice of words, emotional expression, right? And, for example, they-- they even separate, and, uh, thinking, internal thoughts and thoughts that help, uh, make decisions, right? But they stress that it's more like the execution of the chosen action, not the choice of a better action. Go ahead. The third theme is attention. This includes, oh, volume, and, there they include, uh, a choice of attention, and, uh, a constant focus and, well, different other parameters in terms of attention. And they say that, not just looking at something, but being able to keep a focus, ignore interference, switch and celebrate important changes in the environment. That's incredible, I think, is it cool, right? And we talked about audio in a few episodes ago. For example, what happens to talk about what modern models are now trying to answer quickly in the conversation. And, uh, this is a very interesting story that really shows AGI's not AGI. Uh, uh, noticed that the more-- I just called three points. When you call the paragraphs, you realize that it's not AGI now, right? Well, that's not AGI. And some phrase says that there's a GI already. You're like, "Whoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo-hoo- They include, a, a large number of different parameters for training. And by the way, I like that when you're reading about it, it's even hard for the systems to read, and, uh, it's very much of a language that's even hard to translate in Russian. They say that for real adaptive behaviour, the system is not enough to learn only during, but, in training. And, uh, or inside the context. I'd like to learn and w-- and keep new knowledge in time. That's a really cool story, right? I mean, not just learning about the model, not just your data being kept on you, not just you in the context of go-- in context, there, there's a window that's old. Well, that's what you're doing here. What's changing in space now? Somewhere else, if something changed, right? That at this point of time you should be told. When you changed right now, you're told to stand there. You're asking, you know, safely out there? They say, "Wait a minute, we're all data, so we've found everything. Or, there, based on what's going on somewhere, it's not working." The fifth is memory, the ability to keep information in time. And they separate it, there, semantic, world, occasional, a, uh, forward-looking. They separate even one such-- separate notions as, uh, in Russian forfeiture, right? It's a forgetful. And, uh, there's a lot of things that she's looking at inside. And I wonder, as they have this oblivion, it is considered important to be able to remove obsolete, erroneous or irrelevant. I mean, when the system is actually when it's supposed to forget some things, right? And the awesome subject, huh? When you're a million-- these million contexts you're giving out the window, and she says, "Here, this is not the right thing to do, yes, that's not the point. I mean, it's no longer relevant." Or when you're in one context window with the system, you're in another conversation, and she's kind of, uh... she's very clear that it's a need to forget. There's a reasoning next. Big subject of reasoning, but reasoning is less understandable, yes, what is it? They have a metacognignium that's so separate. Well, I'm saying, "Methcognition," it's obviously not translated. It's a system knowledge of cognitive processes and how to track and manage them, right? And there is a knowledge of their limitations, their learning processes, their metadata, their meta-methanes in their entirety, and for example their mistakes, their learning strategy and everything. I mean, understanding the system, like me, this system itself, works, yeah, I'm working. And that's what we're going to do with the system like a weapon, right? You know, here, that's the point, I think this is the layer, it's the metacognicia, I think that's why even the Russian word is not. Well, metacognition, I think it's a modern term. He shows that this is AGI or something higher, right? Something, that's something higher. That's the-- that's the power, they're, of course, unreal, right? Because there is, uh, aah-like judgments of learning. To make the decision on learning and what training, right? There's a separate point of performance next to that, it's a top priority, a ability to behave, right? And they allow for any planning, targeting different conflict situations to be addressed, and so on. The next one is a big, separate problem, solving problems, and it's a composite. That, by the way, I've been severing from the beginning that there are some constituent capabilities, and she needs to understand and present the challenge, to get real knowledge, to break the challenge for all the purposes and so on, right? And here, to solve the problem. And there's a separate subject called social knowledge. And includes all, there, agreement, partnership. It was noted separately that in some contexts, this is a partnership, and it may be harmful separately. And they're going on there, and they're doing it, and all the protocols are all over it. We're not gonna, uh, expand the prom. I've already taken a very big air of this story, but it looks like it's really cool. And this, I'd call it, uh, essay. Well, I'd call it a real, real, real, real, real research. It doesn't look like a steamer, but a questionnaire or a questionnaire or something that is, like, uh, that, uh, there's two thousand questions asked or looked, something weighed on the parameters. I understand that's why they're doing hackato now, some people are gonna do it. By the way, OpenAI started some parallel research, stuff, yeah, maybe it's in this, in this field. There are two young guys out there, crying, trying to close it right away. Here. And we won't even talk about them today. Here's Google DeepMind. Everyone, I think who's looking very much in this field, everyone who wants to know what AGI is a little more, to explore a little more, of course, we'll be talking about it first. I think I should go to the AGI name a little more in terms of not a Haip, but rather to pay attention to it. Because I think it's a basic study to take even from him, maybe some blocks and their example, it seemed to me that it would be great even in our-- Tuesday Moon. So much in Cairo, huh? In our, Tuesday Moon, what do we do? It's a when someone let something out, s-- look at it on those weighted things. And I hope there will be a lot of automatic stories that can be downloaded and asked, and how much AGI parameter is how much he's hitting on Google DeepMind statistics.
If this is done, it's incredible plus Google DeepMind and the whole AI market.
Everything that's measured is improving. Once this becomes a benchmark that can be acted upon, all models will be pounding and, indeed, improving in this direction. I just like Google out of this task on Kaggle. That means a lot of people are gonna start to decide this hackato. And besides, it means we'll see a lot of decisions. Yeah, not everyone will publish their decisions, but, in fact, the professionals will be able to come in, read, see, and how many of these tasks are being done, how many points they get and so on. Besides Gugla, the NVIDIA has a big competition, and that's a great story, too. It's, first of all, the promotion of technology. Well, if we're talking about NVIDIA. And second, again, the money for the competition. Enthusiasts can certainly try to get involved. From the OpenAI competition you're talking about, I think there's an open AI that's gonna hire those who are.
To hire, uh, those who decide best, um, those, uh, those, uh, competitive biddings. Well, they can say that nobody won, but somehow, it's quite interesting. I mean, if there's a specialist in the direction--
If you're cool, if you're cool, you can do it, they'll see. They said we'd see it later, yes, inside. Did you notice NVIDIA? I'm gonna say one comment on NVIDIA that it's a word, their shares, I guess, right? This is not a financial recommendation. There must be a banner like that. But generally, from the perspective of NVIDIA and NVIDIA. NVIDIA, of course, is not real cool. They're now released their operating system for AI infrastructure inside, and they're, uh, they're all in the world now, just until they have a huge joint project to produce in the twenty-eight cities of the United States. together with Uber cars. I mean, NVIDIA, and I think we'll start seeing them on the market, uh, physics, yeah, we'll start them on the Consymer market, what's called, we'll start seeing them. And, uh, the horizon is very close. They're still, yeah, they're kind of in a whole B2B sector like the next level. But according to their presence, they're, uh, contracted, they're all looking everywhere, right? They gave Carpathoma the first new computer they made with Dell. Cool! You didn't see, Ildar, the picture, did you?
Nice m-m-m-marketing
move.
Yeah, Jason Huang signed it personally. Yeah, yeah. Well, they're just very tight everywhere. Well, they're definitely comrades there, for a while-- well, working together. It says, based on what's written on the computer, they worked together and made a picture together. Well, there, Carpath is unrealistic, there, a man in the AI world and from a-- in terms of name, and, uh, in all companies. Uh, but the fact remains that NVIDIA, of course, uh, is directly involved. It's not a video card once, is it? It's not something to do with 3D. It's now a completely different world, and maybe we'll see them on the market, which is how Tesla is doing his AI chip and plans to release it by the end of the twenty-sixth year. Although the lock-in now has a huge contract with NVIDIA, there, for their new equipment. Ah, and that's how we see that Ilona Mask's company, they're just spreading everywhere. I think NVIDIA is a player of a similar, similar class. Yeah, maybe even they'll be in space. So I'd even guess where they are.
Yeah, here, uh, what's interesting about the videotape you said? Uh, this is the current GTC 2026 conference, which is now, well, ends by the time it comes out, and now it's the last day I think it's coming. Uh, it's one of the first conferences when the game video cards were not even available, and it's been years, well, until the last Haip, the main direction NVIDIA was marketed was ahead. market and so on. So they're getting bigger and bigger from this story, you know, like more money, in AI.
Of course, it's a lot of producers right now, remember, right? They're not gonna deliver the memory for the usual companion because they're more comfortable from there, all that matters is, there's a complicated AI, right? And I think it's an example of NVIDIA, probably very much what you're telling me now.
Yeah, and a pretty interesting story that they're open to new things. And, uh, we once said they bought a Groq, which, with Q at the end, which Cerebras and I always call, for twenty billion, and already on this exhibition, they presented their server counters on which, Uh, Groq ideas have been used. I mean, they're not only developing their technology, but those they seem interesting are buying in the market and adding their products quite quickly. This will allow, um, speeding up inference, you know, speeding up the NVIDIA. I mean, it's gonna be a little faster.
Thank you very much, everybody. I'll see you in a week.