Hello, everyone! We're on ToTheMoon. Technological news, Silicon Valley sites and the world. We're going out on Sunday. They've made a whole series of peddians. They're now producing a separate product called Atlas. So it's a browser.
The impression from Atlas, I think it was good to see Althman's face during the presentation. I've told him twenty times, please don't call me that.
One of the coolest things that OpenAI can do is start collecting normal information from people and asking them how to work with them. And make this extended configuration for memory. By the way, we got a video on Wednesday last week about the GPT cupboard. Write it, watch, watch, and propose topics. It's always very useful to hear different things from you. We add them to our content, and we put them on, we make special releases, right? So, today we have eighty-first, I think, a graduation that we're producing on Sunday. What do you want to talk about tonight? I think it's a very important moment. There was another O.
OpenAI pd. OpenAI, by the way, came out, I think I've started doing some peddales. The peddians always do as Higgs Field, always do their peddies. Only the OpenAI has a little more, of course, company. And as for the peddate, they released their browser. Last week, we were telling you, they did a whole series of peddates, including development. We once again talked about the development of our applications, the GPT on the inside of the plaque, the use of such a serious agency, agent regime. And before that, they were talking about internal built baskets to buy inside GPT. They're now producing a separate product called Atlas. So it's a browser. I want to make an initial start because today there's a point for immersion in the realm of artificial intelligence. And we want to talk more about it today. It's long to be a little bit of a skilled to understand, and what the essence of this browser is, it's either not necessary or why I have such attention to him. I, OpenAI, will revert a lot to bloggers in Instagram, and all bloggers, they're like, I don't know, the feeling that they've launched a missile on Mars. They just say, "Wow, what a big thing! How cool! Do you know? It's all haunts." I also like it when someone says, "Hang searching Google." How's the browser with the search, tied up, huh? So there's a lot of speeches. I've been asking a simple question and write down, please, inside, now. You see, I've asked the simplest question and I've been trying for a long time to know people around and, including GPT chats. What cabs and the uniqueness of the case, what are the boxes for use in Atlas? I mean, I can do that in Atlas, right? And write if you know the case.
Because the OpenAI shows in examples, they seem to me to show how to buy instacarta food for sushi and make sushi.
Cross.
Yeah, I think the programmers are just sleeping at work, and they buy sneakers or sushi. So, uh, that's very strange.
Now they can do it through an agent, yeah.
Yeah, and they can do it through an agent. I'm sitting here, so I'm asking myself a question. I'll give you my word for it. So, to use Atlas, I need to set up a separate annex. Who I should be the middle of people in the world, now at GPT's GPT's, like, eight hundred million users, so I'll download a separate browser app. Well, there are leaders in the world. There's a leader in the world, so Chrome, there's a leader in the world Safari. There was a strong leader there earlier, Explorer microsophthovsky. So there are leaders, there are different things that are unique like Opera, for example, right? And now, it's a unique one that doesn't default, which people who know something. And the world, billions of people use the most common browser. And I, users, means I use GPT chat professionally. If I use GPT chat professionally, why would I put Atlas? To open a tab by myself and in it, I said, "So, I don't want to cry, buy me something." Why? What's the logic? I can get him out of GPT's chat to do something to me, something and bail. Who doesn't know, by the way, yeah, again. There's a agent mode inside the GPT. He lets you run on a separate virtual car, open up different deposits in the browser and move between sites, including if you need to collateral. I have to bail here. And their new browser can't be a hood mode agent, you still need to turn on. And then what's his point? Why would I do that? And I can tell you that I'm just a businesswoman, I think it's a mistake. The only plus is to get away from the word. Well, when there's GPT chatting, get away from the word. But at the same time, you're not a browser, you're doing a modern new infrastructure that's more of an operating system. I mean, it's a big, huge system. Why don't you let Brauser open up inside the GPT. Why don't you get an internal looker inside the GPT, huh? They did Atlas on the basis of Chromium, by the way. It's an open source of product that's made of Google Chrome. And why don't you, uh, that inside is, there's different things that keep the pages going, there's JavaScript going to start, and so on. Why, uh, not do this building inside and expand the product you store like applications like a basket checkout. Aah, a mode agent who can run through a large number of different sites. What's the point? Please write what you think as audiences. We'll move on to, uh, except me, Elnar, Tanya's gonna have his views on the browser. But I think it's a mistake because they're likely to, uh, millions of swings. And the mass of people, uh, she used GPT chat, usually, and she uses it or will, well, use it in a simple mode. I have this induction today, I think it's very important to the market as a whole. Which is why, by the way, I've never been to Perplexity. Why would I go to the back platform, the aggregator is only probable if I could use super stuff inside. I like no system, by the way. Elnar, I've run a lot of tests. You've started, like, a no-- Atlas analysis. And I'm supposed to be saying, "You can use a agent in Atlas, GPT chat can't." I say, "What is it? I can." And at Atlas, you, uh, he can walk in-- and you can bail on your services. I say, "I'm in agent mode, too, "I can bury GPT chat." In the end, GPT himself tried to protect Atlas with all the possible powers, but he couldn't. So?
The main thing is, now it's a structured attempt to give up. First, uh, the impression from Atlas, I think it was good to see Altman's face during the presentation when the staff were sitting there and showing different cabins that we were here, I don't know, One kind of task will be transferred or here we are, like, eight people on the road, and we're gonna be running the other. He had a so much missing face that even some mems had come with it. So I think he's not much, too, aaaa, excited about how they love to talk, that product, right? Aaah, if you speak purely technically, I will be a little excited for Sasha today, yes, instead of...
Yeah, yeah, yeah.
I'm excited, inspired, uh, excited. That's it. And if there's some technical element, uh, Perplexity with his Comet, it's been released two months ago, right? After that, the companies did it, too. So technically, they didn't show anything new. All the demonstrations show you can do is--
Yeah. And GPT chat, I think the last one of the big guys has released their browser with their own designs. The only benefit that, well, is not good, the only difference that is positive from other like browsers is that you have a connection here, first of all, going to your GPT scout, you know, because it's rude. If you use Perplexity, uh, browser, you'd like to, uh, not be necessarily, uh, acquainted.
You'll have to have, uh, Perplexity's account, preferably a fee. So you pay separately for ChatGPT, pay separately for Agent Mode in the browser from, uh, another company, and you're doing it right now, right? If you have a payroll on ChatGPT, you're right in sync with the browser. That's a plus. Another, uh, let me try to get to the analogy. I'm not taking this as a new browser from ChatGPT, but rather a new app ChatGPT that can be installed on a computer. And there's a browser's function inside. Because I put it on my own. And, yes, I used to have a ChatGPT in a separate deposit, and I just got off the line. I'm just gonna switch between the browsers. I got this browser, it's for GPT chat, and there's a basic, good Safari.
It's for the actions I'm doing.
What didn't they do? What they didn't do inside the GPT chat, support inside, how to live normal scenarios, cool scenarios, when, for example, they're referring, you opened, you scattered, you looked at something. And so on. What's-- you know, the question is not to get back to the browser. We're going to a new world when a normal browser, Sam Altman himself says, "I don't use Guglom." He says I reduced the number of applications to the browser and used or a browser window of GPT, or an annex. I don't like using GPT on a computer. I use the window of the brow-- chat GPT as a deposit, as you said. Why do you want a browser if you leave the browser?
Let's try to figure this out. So, look, assuming that some users go to GPT browser chat, that's what GPT Atlas calls it, yes, not just Atlas, but GPT Atlas chat, you'll have more than that, their stories, uh, chatting, there's gonna be more information. The history of site visits, some activity, a search that a man does. So you have as much information as you start to learn about the user, it's getting bigger than just working with the GPT chat. I mean, there are still some extra activity that is not through GPT chat but in separate browser deposits. And now they're getting them. Uh, what's the added value? If you think that man has been riding himself of this browser, he's starting to use it. You have a new deposit always opens you with GPT. So you have a point of contact with GPT chat, you know, getting really active, so, let's just say.
No questions at all. From the point of view of the presence of a browser inside his memory, old deposits, right, right, well, work from the point of view of your memory, okay. Why did they not add the deposits in their GPT chat?
A-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-
Well, because the technology is different, right? Because the technology is different. Different app. I should have rewritten everything. So they would have said they'd have made a full peddith, uh, GPT tea apps. He's got an internal browser now, and we've got an entire ecosystem. Then I know, oh, shit! I mean, I'm opening up now, not-- I don't want to open it. I'm used to using Safari, and I have synchronized and everything. I don't need this Atlas to open the site and make a summary of it. I'm, uh, so weird canes. I need to get some translations on the page sometimes, so let me make it possible. Develop an app for Chroma, develop an annex for Safari, and let me do it. And then gradually in your big app on Mac, for example, develop a huge infrastructure there and de-de-decrate WeChat, you know, where there's all the outsourcing, and the stores, and the browsers, and the apps, and the maps.
HR systems.
Well, HR system, yeah. These, they're lawyers now, yes, Elnar was the cool thing that was, like, throwing out that they hired 100 lawyers there.
Financials.
Oh, financiers, yes, financiers from different cool offices, there, like JP Morgan, and so on, to build a system within itself, uh, sys-- well, to adapt the system to help, there, in finance, in the Legal affairs, etc. So do great.
It's a great tool.
You're gonna have to develop your huge operating system. As I set up the macOS operating system, there's a default weather. Okay, I like her being default or there's notes, I think there's some notes, right? Or there's a mail. Do the system like that. Your cabins look, um, very strange. Well, do it inside. And if you're gonna do the post office, you're gonna be, like, a postal client now? And it's like, and you're talking about GPT, which is why I've always been there two years ago saying that it's, uh, the future operating system, because too much is done with different tools, a large number of cabs. use and in this way that can be expanded within one system.
So finish her up, then finish her memory inside the GPT. I'm freezing out this week. The thing is a rat. I'm asking for chat, he's like, they've expanded their memory. Who doesn't know, again. So the GPT chate has memory. In the plus and the extension of your memory now, the vast amount of your old chat is retained. And when you're in touch with the system, she's using your old chat as a story. Like, if I'm traveling somewhere now, it knows I'm on my way. In fact, it even sees my return ticket when I go back to the valley, right? I think. And, uh, and I say, "Can you give me a summary of my chat?" It says, "No." I have no access to them. So wait, do you have access or do you have no access? I couldn't get access, by the way. I finally came in. Who doesn't know, we can come in. The chip is really fun. This is the new case I've done these days. You can go down to the construction, down there, click on your profile, pick up the data controls, and there's an export of all your data, and you can unload all your chat. I've had six hundred and forty megabytes, six hundred and forty megabytes. You're getting a little bit of a story. I'll see. I've run a full analysis of my history, in terms of what I have inside the data, right? I mean, there's a thing inside that archives and what volume I've been doing, and what topics I asked for, so on.
But I can't turn inside, within one of those. That's what they did in the browser, in Atlas, that he kept history, including your browser visits. Well, like, okay, but I'm ninety-nine, ninety-five of the hundred tasks in the GPT chat room. Why would I go back to the browser to migrate? Tanya, will you set it up? Tanya, will you set yourself a browser?
I mean, no, I'm not gonna fix it because I'm, first of all, a Safari fan, but I had to set up Google Chrome because my programmes, they have extensions like that, they're all clippers, Only Google Chrome. And it's a terrible stress for me. I need all these passwords to carry, remember, reconstruct, and so on. That's one thing to turn to another. I have a very good one, as you have ChatGPT. I switched this once, asked what I needed, copied out the other window, and it's working really well. About memory as much as on the one hand, well, it's probably the only thing that's gonna be the only thing that's gonna make me think, but it's not enough motivation for me to be the third one. Brauser set. But the idea is good. And then about the memory of the GPT chat, I don't know, I'm in there, like, I really don't like it when he calls me, it's some kind of story in Russia that all the professions now have to be added to the end. So he calls me a designer. It's just driving me crazy. I've told him twenty times, please don't call me anywhere, never under any condition. He's like, yes, yes, good. And the next one calls me a designer like you again. And la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la la I mean, it doesn't work at all in my case, in principle, I even have this function to interfere because I have 100 fifty projects and there are very similar things, there, I ask questions. Well, there's, like, every house, basically, a kitchen or a bathroom or something. I'm asking some detail about something, he's remembering it. Then I'm asking about another project, he's giving me information on the project. And that's the cushion. And I'm super uncomfortable. I don't know how this can be regulated, cleaned up, and so on. I can't tell him, don't use your memory, can I?
He's remembering it again, and then he's reappelling to something that has nothing to do with my particular wa-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b--
Yeah, Gemini's doing it, of course, in people's. So that's the thing he was. People are gonna sit like that, "What's the occasion?" So, Google-- people have a huge infrastructure, and you have a Google Drive, you have Google Carts, you have Google Post, Google--- people Google Meet, And there, well, a lot of things. And browser, first of all, super-crunch, still made. A huge number of extensions for the developers.
Yeah. I'm Safari, too.
Because I'm a loyal man in principle.
Yeah, me too.
I'm used to something, and it's all over.
Please write, users, very nicely comedian, our people watching us, write what problems you've encountered in your memory and where you've seen this incorrect work, like Tanya now. I told you about the designer. Because of course it's--
It's Tanya.
No, I--
I'll keep saying.
That's a tough word.
I think I think that, uh, one of the coolest things that OpenAI can do, it's gonna start collecting normal information from people and questioning them how to interact with them and work, and Make such an expanded configuration for memory. And memory to be managed. And I think models now allow a great deal of memorial management and a lot of different things, constructs, unique features to make you understand, uh, what it's like. - Yes, sir. Because what you're saying is really a problem. I do a lot of research, and he can smash me like a single window on another window.
Mm-hmm.
Or today, a man told me, I was on Wednesday, I was on the market for shares. And the man says, "Have you asked or not in one chat?" Well, in the comments, one of our viewers asked. And he says, "Maybe if you asked in another chat, he would say that this political situation has affected the stock market. And in this chat, he told you she didn't influence it." So he could have said a different opinion, and, like, if you want to, you know, agree with you, wanting to say something, be better with you, right? That's how he tried to sell Atlas to me. Well, he was sure of Atlas, though, obviously, it's just some kind of... even the news market. We're still doing an analysis, and we're running a few more runs. I'm very good at running, Elnar makes a lot of runs, making Chinese news blogs, a lot of things. He was not even important with deep research atlas. Although, from the standpoint of OpenAI, it's a serious story. Remember when the graduations were all about the browser when they started, Google said when Perplexity said it was like, "Oh, cool thing!" So it's gonna be called. I'm asking everyone again, what's the case?
Tell me what the box is, why would I put a browser? Give me one, please, Case. Well, except for the case in the series, make the summary news to me, right? Because it's a pointless case to date. I can start at GPT chat room if I know there's some news in the world today, I don't know, that's the news. Louvre, I was just passing him, and I robbed the Louvre.
Well, yes, Sasha's near Louvre.
Yeah.
It's a wonderful coincidence.
I was going through, yeah, I was walking this minute, and, uh, they're ogre-- there, guys, amazing, I'm telling everybody, nobody believes me. There's no cameras outside the Louvre building. And, uh, why would I have to give the summary news? I mean, if I use GPT chat, I'm writing GPT chat right away, tell me what happened to Louvre. The system will go through analysis, analyze a large number of different sources, including French, English, any, and give me some kind of result. And if I'm not a GPT reader, what the hell am I gonna put Atlas? Why? How do I know?
I have a wife who uses GPT chatty, but if she asks, "You put Atlas?" She says, "What's the matter?" She won't even know about him.
Well, they always praise that they have eight hundred plus million active users. I think there's a number of them. Especially considering they're expanding, uh, cheap tariff plans for Asia. Yeah, well, we got a lot of Indonesia, India, and I think they're expanding for 16 countries.
They're very strong, they're Elnar, they're fifty, uh, I've read some news for the last time, and I haven't even opened it up. They've turned on some fifty-eight other countries. I think there was something, well, that's it. They're like this tariff to the super-all cheap countries.
Uh, so I think that with a high probability, there's gonna be a lot of swings at Atlas. It might not be in the US or the Silicon Valley, but it's likely that there's a lot of things around the world. And here I have a hypothesis, which is likely to be a frequency of contacts among people who will deliver Atlas, it will be more than among people who use GPT chat in a separate deposit.
That's right!
Just because you're in your browser, you're always touching him.
Yeah, that's a fact.
Any situation, opening a new deposit, you see GPT chat again and again.
And involvement, it's likely to grow up substantially in these people.
It's a fact. I mean, what you're saying now is that it's basically one of the places of people being involved, even though it might seem strange. They'll say, "Don't worry, we'll add browsing in ChatGPT, too. In the Annex of the GPT. Still, maybe some people are being appended. I mean, if we see, they don't talk about the app at all. Remember when they showed the app? Remember, first time they wanted to make an annex that works as my own, with my entire infrastructure that I have on the computer, and then it doesn't work with anything. I mean, it's a regular application, a normal deposit. What's she more fun? That the deposit in the browser can hang. The page stops getting cold if there's too much to do, it stops being loaded. And here's the app, by the way, I never have it gluing. I mean, even if I have some deposits I've been using, there for four months, and there's a lot of data, a lot of data, thousands of pages. So this, this app doesn't stop when you open it. But there are other glues in the annex. I have glucs in my own place when you switch, different models, inside, in the projects, it doesn't turn them over. Well, there's a problem, but maybe they'll reintroduce this app, maybe they've made Atlas, saw the app is not working well. Maybe a group of people who make an annex no longer works in OpenAI. Judging by the way the Valley developers jump. Meta, it means they're downing six hundred AI developers. I have a question again. Another question. What's going on with you? You spend billions of dollars on hiring, then you want to hire someone, then you fire someone. What decisions are made on a continually spontaneous basis, right? Well, maybe that's why, by the way, that whoever wrote, like Mac's app, he's no longer working there. And when
The origins are lost. Let's re-enact the browser.
I don't know. Well, according to the applications that were being developed, in two weeks, you can't see more of the news in this area. I mean, you don't see the peddets all the time. I don't have enough in OpenAI. Well, we're in Russian. Maybe we can't scream at anyone. I didn't fill it up, by the way. I've been sent some interviews from the OpenAI. I forgot about them. They're doing these ten days, and I think they sent me, like, a day, ten days ago, to turn on, tell them about chat rooms. I really don't understand. They sent me a request, told me to tell you about the chat. Do I have to make my own chat? I have data there, different, special.
Well, that's what makes me embarrassed. Once again, it's important that everyone understand what they're doing. That's the subject we're talking about about finances, and there was more about lawyers, too, I think it says, Elnar, right? The lawyers were there, too. I saw a lot more about lawyers, yeah. The headline is that ChatGPT wants to fully create a super-cooled legal platform. They employ strong market experts, including those involved in the development of these different verticals. And the educational verticals were also. By the way, they've released the peddium in education. They have a huge team in the study and Learn, and nothing else. Although, in the idea, there must, well, be constantly updated with a lot of things, some kind of chips are being created, some infrastructure should be, its designs should be in the process of learning. Well, so you can build some of your...
Well, so you can build some of your learning process. It's not yet coming. Maybe just OpenAI does AGI, right? Like they planned. So they wanted to do AGI and said, "We're just AGI to do it. Or maybe, as they say in the Valley market, everyone doesn't care who does. The most important thing is investing in infrastructure, buying a lot of chips, lots of money, building a lot of new plants, and that's what happens, that's gonna be a success. As long as it's done, it's a success.
What do you think about this again? Oh, see, we may be leaving again at ChatGPT, not talking about Grok or telling Gemini right now. It's more of a story here, again, that's the main monopoly on the market in terms of volume. Gemini, by the way, counts four hundred million users. Talk about Gemini. Well, it's Google anyway. I think they're even less, probably, a little bit of time spent, well, it's a involvement, a time spent in the app, probably less likely Gemini, because there's still a lot of people out there. It's, uh, inside Google infrastructure. Here. Although Google's doing a cool system, they're not lagging. They've got a new pdddate out this week, video generators. Very cool. And they're obviously gonna catch up and be on the market, and then, along with Sora, they're gonna be there and so on. Very serious.
and, uh, the apegreid will do for the system. So I think that we'll have a separate issue on this subject soon, too.
Yeah. How do you, by the way, have ChatGPT Plus and Pro? I've been with Pro lately, and in the last few weeks, I've been... I feel like Pro's become, uh, uh, yesterday, I got right on a whole series of queries, he was right in the middle of a stupid, well, he was right wrong. It's not a question of hallucinations, it's a question already, I'd say, some logic or a mice-- like a thought, right? Because this system of reasoning, reasoning and Pro is not the usual model that's been generated there, is it? There's a big data analysis thing under the hood and the, uh, the information you're getting. Noo, um, the information was very strange and has become recurrent. You don't have one, Ilinar, do you?
Oh, it's hard to say, actually, clearly, because it might be two options. One option is that they're scrubbing, imputing power to Sora, and there's less power left, more, weak models, less reasonit or something. One story. Another story. Maybe you're just trying to find places where the system is weak. I mean, where she's strong, you ask a question, you get a good answer, you like him, and then you turn a little sidewards, you ask a question that seems like a similar to you, but the system doesn't take him out. It could be. So, to be clear, we need to build a baccalaureate, right? It's only a matter of the same kind to run with the same input, there, I don't know, once a month, once, there, two weeks or something. Here. But it feels that it's not doing all the tasks, of course it's present. And the last two weeks, I've been confronted with the fact that, uh, the system can't solve the episodes I want to discuss with her.
You have to do it alone, do it yourself, decide it.
Of course, there is a history that is very much developing in the world, which concerns various laws of information and writing down information. And go-- go back to the browser, the browser also keeps all your movements, keeps them very, very structured, I'd say, 'cause they still have models, I'm sure you're the one who's gonna let your data be super cool. Given that some countries are not allowed to, uh, open up some resources. In Russia, for example, since July there has been a new law, there cannot be some resources. So you're gonna get some kind of resource, you're gonna open it, you've already broken the law, so you're starting this link. Whatever you're asking. Of course, we're going to the era, uh, when our personal data are just in some form of huge stock. And our personal information is personal.
And Louvre doesn't have cameras.
I was wondering if I could do one little browser experiment with ChatGPT Atlas. I'm on the Lichess website periodically, it's a website with the opportunity to play chess, to do chess tasks. I'm sitting on the task and I think, and let me, in the agent's mode, open up the task I'm sitting there, doing it, and I'll tell him, let's do this. And I was wondering how he's actually visualizing what's going on on the screen, right? So he's there, impulsively, making a squeakyshot, this vironchot will then recognize. Or he's in the HTML code, like, getting the information from there. And interesting conclusions were that he was not using any scronyms. He's trying to analyze it for HTML, respectively. Uh, speaking of the board on the Lichess website, she's through HTML, as far as I'm concerned, she's not getting any steam. So you can't get the figure. There's more information on the progress made, it's still watching, the system, I mean, that's, and it's very hallucinating. She says where the pieces are. Sometimes it gets, sometimes it doesn't. There's a lot of pieces where they're not. Then, uh, when she's trying to make some move, it's a little bit of a field-- there, E-one, right? She starts pretending that she's okay, the field is about that size. E-one is a lower line, that's eight-block. So, somewhere in the middle, we need to try to get in there by a mouse. And that's where she's trying. And quite interesting experience. It was still a trivial task, although the situation of the figure was hallucinated. But it was so she made the right move. When I was a little bit, there, the harder the task was already opened, the system couldn't get away with anything. But it's interesting to watch her try to understand, and what's-- what's going on here, right? So if the chess is interested in the iatric agents, try to open some website you're on, uh, either playing chess or doing tasks and trying to set up a system, huh? And there's a process of reflection. It looks very exciting.
And if you threw the squeakyshots in GPT chat, she'd probably have made the right moves, right? If you'd throw a scream in GPT chatting. Not really.
I'll try. But LLM in chess is disgusting. It's really bad. Here, for it to work well, they just need to give the instrument, the
chess, the moves and the sky so that it can just move from him to the translator, yeah.
But actually, Sam Altman and the team said, first of all, that GPT chat, uh, Atlas might be wrong by the agent. So when he offered me a long-formed box with recurrent visa-type blocks, accreditation, conference or registration of people, there, on the medical portal, or return, trade of goods, I'm I thought you were talking about what? If an agent can be wrong. I mean, basically, if you have an agent wrong, I'll never let him in any real action. It's like you say, chess. But... And plus then, uh, the logic of the game goes away. Why would I play this game, then?
Well, I'm the one who-- there's yesterday, uh-oh, this place in the valley, in the Cremenyum, called AGI House. A-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- This house, the house, the house, and, uh, there's a lot of different engineers, AI-Chiefs. There are some very famous people coming to do the show. And they put their cartoons at their house yesterday, that people made a project to-- play AI in poker. And I thought, what's the point? There are already a lot of moves, there, for poker games and understanding, next step. And that's right, well, what to do and how to do. Mmm, what's the AI doing, huh? And, uh, I play pro-- or man just plays like a game. I'm not a player, for example, but if someone plays, he plays for pleasure or he plays for use. But if you use an outside other car, like chess, yeah, if I play online and use a separate car, I'll win.
But I was wondering, technically, how the agent interacts with
what's happening on the screen.
Yes, yes! That's, that's really cool. And Tanya, by the way, asked a nice question about the picture. Carpathians released the post this week, that DeepSeek had opis- a description, yes, a new CR. What was the logic? What, well, this is technical information, of course, maybe it doesn't matter to everyone that it's cheaper to get up, uh, models, language models not texted, but text pictures, right? And once-- what makes it much easier to work with a picture, especially with a structured one, there with a known format than then, rather than the original text, and then,
To the text, uh, to deal with and translate that, like, would be more beneficial. Well, while as much as I've tried to-- some of this, some of the conclusions, it looks very, uh, technically. I'm gonna have to go to some kind of internal technical reason. I don't know if they're needed by ordinary users.
No, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, Noo's a really cool subject. DeepSeek showed a somewhat different approach to, uh, memory inside, yes, when visually they can, uh, rent pictures to keep memory, uh-e-e-e- contextual. It's much more efficient than keeping the text, yes, there., art and stuff. Well, it's technically very cool breakthroughs. Uh, from the perspective of the usual users, I think DeepSeek will soon release the model on it, locked, and the rest of us will also catch that know-how. Anyway, in a few months, I think it's gonna be a common phenomenon.
Yeah, but it's fun that DeepSeek came back and showed up with what hit, uh, that engineering market, right? Therefore, at least in terms of analysis. The Carpathian dia is a very serious person and has been directly dedicated to a separate, whole, distinctive description and has given that respect-- well, that is a respectable characteristic, right? Who reads, please write some interesting information or some kind of fun references that are available. Uh, don't forget to slide our canal, support it, and see you in exactly a week. And don't miss our special release. Bye, everyone!