Hello, everybody! We're on ToTheMoon. Technology news and sites from the Silicon Valley all over the world. I wrote this down today, you know, like a slogan this morning, so I can't forget. Models are people. Because, as it came out, it's not just that the renovations are now staggering, that is, Fable has all opened up, yes, and extended access. We'll talk about it. ChatGPT five six, Soul's out and, little, one more day with the presentation. We're on the way, the model goes out and, well, it's for everyone, it's a release for everyone. And a new version of the xAI Grok four and a half was released.
And with a fairly serious statement, yes, very serious statements in terms of how the system can simultaneously develop a large number of tasks. Yeah, she was supposed to even win a big--- test to develop a lot of tasks. Let's talk, talk about this and see this. Why did I write the model problem? Because reading different engineers' statements, there's Anthropic especially, yes, OpenAI and generally reading different headings, news, some forensic findings, you see that, uh, big as a story is that now, It's a chaos in models, in their sub-views, species, systems of different kinds. And to get this straight and make the model show its power, it's got to be a well-informed man now and then it's a good thing to figure out. For example, I, after publishing that Grok four and a half won the test, solved the 20 parallel--twenty coding tasks at Fable, Opus, there, and Codex I'm like, so, that's all. So I'm getting a third Grok model now. But I've been studying. Turns out Grok is only working on a terminal condition and working with you while he's with the originals. I don't think they'll have a update until this issue, but they'll still have the working tool of the tea we have right now. And I think it's a remarkable thing that's happened in the last three, six months, of course, and that's the emergence of systems that can make some decisions. Not that we're gonna be super-cooled. Here. So, what are your expectations for five-six Soul?
Because it's clear that it's gonna take time to test her. I think the OpenAI started playing Apple. Before the new models go out, he starts to blunt the previous ones. And five and five Pro's working the most terrible way of working the last week. It's gonna be interesting to know the audience how it works. It's like Apple's phone starts, so it's a long time, something's getting loaded, the battery starts to keep it that way. Yeah?
Look, it's great that you sounded. I think I'm having a problem with the Internet. Turns out it's a GPT chat problem.
Maybe with the Internet.
Well, you won't understand, 'cause he's really jealous and he's been thinking about simple questions for him for a long time.
Look, I hope it's some sort of resource redistribution, not a deliberate story. Maybe, of course, a deliberate story called a staggering or, there, industrial obsolescence. Here. But I hope it's just a new model that starts to stretch resources. The others, I don't know, they're getting more in line. Something else. But in general, yes, a perfectly explainable story.
That's a good story. We were just filming the output and saying that Soul is now getting five points six, and then there's a deposit in the GPT chat room that allows a carpet to be made and a deposit that gives the opportunity to make chat. So I'm at the GPT's chat for a long time, and he's finally got a lot of different deposits, and he basically told me that GPT is a deposit and a work deposit is the same, and there's no special difference. In fact, you'll see the difference. In the Work, you can work more with local files, do more, you know, more accurate time-related tasks, with reminders. But in fact, you can continue to do all this, including in chat rooms. And what's important is that Work will use the limits that are in the Codex, and that chat from these limits is independent. Yes, that is, GPT, set in Excel, for example, or Codex, or Work, they are using one single system of all limits to date. Pay attention to that. And, of course, it's very interesting that new regimes have emerged from Solo. In fact, the Ultra regime has emerged, and another slightly different name system, although the acceleration regime has been maintained before. The system works quite interesting.
We're waiting for your views, including when you're watching our current sub-cadast.
I wonder if the way out of these models changed the plans and the Anthropic. They decided that we should give access to Fable more, because if at the time GPT five six accesses F-- access to Fable would have disappeared. Consequently, a large number of users, even those who, for example, left ChatGPT to Anthropic, would return to see how it works. Because big, uh, limit, because, uh, user-friendly use policy is pretty loyal because Anthropic has a pretty tough policy, they're a lot of bunkers, uh, they're checking for it. use and other. OpenAI is so far much more loyal in this regard. And as soon as it is likely that these OpenAI releases have led to the extension of Anthropic, uh, extended access to Fable. And besides, I think I have a plate before August 5, that doubled limits they're offering models. I mean, it was there before the beginning of July. Now, respectively, they've already extended this story until August. I mean, in any case, this competition is exactly what we use as users. And the other models are coming out, too, I think this whole story will blow. The same Groq you were talking about at the beginning. Yeah, I see that's a totally programmed story.
They do this story with the cadet, and here as a casing, uh, the cadet is supposed to be, as far as I understand, right? So you're putting this model on the cadet, starting with it. But, yes, the courseor is the IDE, the design environment you work directly with the code. So, working in the Codex or Claude, you're not working in the development environment. You write in your chat room, you can not even see what's going on. The cadet may be a technician, but still an extra app and it's more programmatic. I don't think it's gonna work out there. Here. But in time, I think X will have his model too. Well, your app.
Yeah, look, they're gonna have to do it, 'cause people are getting used to being in Codex or Claude Code, in the app, saying, "Look, you're gonna have to study me how, where I'm gonna turn, uh, server or where I'm gonna be I need a system to pick up to register the domen. I mean, you're still gonna do me some kind of search, yeah, look at me, what tariffs are more comfortable. I want to study the latest updates, there's API. I don't know how easy it is to do it in the courseor, just how much the courseor is meant to do it. To find him alone, reread, I don't know, blogs, try to deal with API of different companies, find troubled places. That's how it is. I just don't know. Yeah, it's hard for me to tell you about the courseor. Who has encountered, tell me in terms of this broad infrastructure work. Because I remember when the Codex, and when I started using Codex and Claude, I was very much, you know, often writing that Codex is cooler because it can build different systems. Remember, I even said that, yeah, what he might, that's it. But I now see that Claude Code is doing great with this, including, it can be perfectly easy to work with a wide range of systems. And now there's some confusion. And I'm the one most embarrassed by this confusion. It's like Claude Code I can choose some GitHub repositories, and in Codex, I'm not supposed to, but at the last moment, even in Codex, by choosing one GitHub repository, he somehow saw others, and he didn't see it before. And you don't even know what was or wasn't updating it, do you? And I think I'm confused with this news when they write, like Claude added it to the webcam, which means. And here comes a separate design deposit, and then there's another, and there's a deposit-- a deposit, and there's another, so these squills are here. And you're sitting, and you don't understand, means where some beta test is at the programmers in these companies. And they do it on the dock. And we were talking about the same--
You and I have been discussing in the Tuzmun for two years, like, OpenAI, ChatGPT, a huge number of people who have done some kind of unnecessarily, but they should not have been dealt with. We were wasting our time. It's like me and Elnar once, there's another two and a half years, I think back, yes, Elnar, or three have started GPTs, right? And what else was they called? Not only GPTs, but, well, that the annex could have been ChatGPT. And the flames were, and there was nothing. Yeah, it's a cry. And we started, so we're gonna do something there. Actually, it turns out that you shouldn't even be there, you shouldn't have wasted time, right? And now one of the great stories that is, especially when there are additional sophets, people are often offered additional solutions that need to work. That's how you're talking about Cursor. I don't understand, you know, SpaceX bought Cursor. At the same time, I don't understand, and Cursor is already adapted to the Grok slit, so that's the one that's all. Or it's still a side-by-side independent system and it's gonna be a little bit of a layout again. And these trillards, these proc-- well, this is very hard. I think it's not-- it's not-- it's-- it's not-- it's-- it's-- it's-- it's-- it's-- it's-- it's--it's--it's--it's-it's-it's-it's-it's-it's-it'-it'-it's-it'-it'-it' it' it' it's-it'-it's-it' it-it's-it's-it's-it's-it's-it's-it's-it's-it's-it's-it'-it's-it's-it's-it's-it's-it's-it's-it-it-it-it-it-it'it'it-it-it-it'it's-it'-it'it'it'it's-it'-it-it's-it-it's- And by the way, I ran into Claude Design.
I don't know, Tanya, you-- you managed to watch Claude Design?
Yeah. Well, I kind of haven't found applications yet, because I'm doing a presentation in Canva. I don't need Claude Design for that. Aaah, that's more like we discussed for some graphic use this time for graphic designers. In my field, I don't know. I'd be happy if someone gave me something, but I couldn't find it. What did you do about it, Sasha?
No, I just saw what I had, and considering what I have now, well, some idea and, uh, aaa, to see-- well, to settle. Remember, I once said in the one thing that I had a website and there's a huge number of articles and everything. And I'm a rock-- and I'm a rock-- I said that this breakthrough would be very easy for me when a system is in place that I'm gonna say, that's just what I need to copy, yeah, and text me the website. It was unreal then. Elnar said, "Well, it's not real. There's one page you might still be able to do. Unreal." I mean, rewriting the site is a basic task, right? Because when we're talking about rewriting a huge amount of sophth, or like Ilon Mac says, there, we replace Microsoft, yes, microhard, but still rewrite, there are operating systems, Rewrite a lot of complex applications is a little bit different, a little bit of a story. It's clear that the other one is a little bit. So, I, for example, still look periodically, and at what point of time it is possible to achieve such a task. I have many other tasks that are related to commercial matters. For example, if thousands of announcements, such as advertising, can automatically be launched, there, generate landing pages different, questionnaires, leaflets, and take people there further into different systems, there, CRM and XRM. And so on, systems. Because where millions of dollars are spent a month on advertising, well, there's a lot of management to do with this. And there's this site block, including. And I came up at least with Claude Design at the moment: if he could let this story of page generation, including. But these instruments are still untouched, and, uh, I'm trying to do some things again. Let's see what this is gonna do.
What exactly do you mean by the website correspondence?
So, you want to get a ready site and do some upgrade or copy?
Well, I have a website on a certain CMS. I have a website, like, on a parole basis, on a certain CMS. I didn't actually have a CMS, I had a more simple solution, because I had to put out a lot of different material, and I had a lot of c--- stored in Notion. I once made that decision, I think Elnar even advised that she had this shell, didn't you, Elnar? There's a Sup-- Super SO, I think it's called. And that shell let you clean up, it's like a layout to make your Notion a website. And it gave me the opportunity to put out a lot of material, a lot of, uh, different articles, more-- fast-creating different landing pages to register for different learning programmes, there, different things. Even to put it out, I don't know, we have pages that every single edition of the Tuzmun describes, like, on the website. So this is a light story. So what's that supposed to be? Well, it's like moving out of this system. Because Notion gives me many limitations.
This system gives me restrictions. They're still eating money, by the way. By the way, this system and Notion, all this is eating more money. And I had the task of moving this to an independent story. And there's a question, look, a very big question is that we're all used to working with SaaS decisions. What do you mean, SaaS? I remember, I was doing a good job with that term when I started running Megaplan in Russia, yeah. And this is, uh, 2010. And then, well, Megaplan was the leader of Software as a Service at that time, and it just started out that it was like a decision that was in the clouds, that gave some kind of software, so, service, you're for him. You pay for something every month, and so on. So, what did I say, thinking about the website, what would I choose to do with SaaS or open source some CMS? What kind of system would I choose for, I don't know, some kind of questionnaire? Which one do I choose to monitor the ad, where there's something else to connect. Every time I'm here, now, in the last few months with the code-- Claude, I understand I have to avoid any SaaS decisions. I mean, I've been avoiding them more. And if you read it carefully, by the way, the description. I've had a big, big discussion here recently, a GPT 5 Pro chat, and Claude there were four eight maxes of a periodic fabel. And it's not about the code design, but about, uh, the fabel max, part, but, uh, the discussion of the architecture of a certain solution. And they start offering you different open source there, SaaS and different things to raise one, like, a management architecture, like, some kind of commercial infrastructure, yeah, in a commercial infrastructure, company. And the more you talk to them, they say, "Well, you're better off taking this decision because it takes less time for the developers to make it. Or we gave you a question, you consult your developers, and then you say, "Wait, wait, wait, wait. My developers are you. It's Codex and Claude. No, no, I'm not gonna give this to people. If you don't give it to me, then don't use SaaS, and we'll write it yourself." And this is the subject, I think it's, of course, it's the current season. Yeah, it's the twenty-sixth year. That's how this second, third quarter of the twenty-sixth year is when you realize that you don't need a lot of SaaS, just don't, they're you, they're, they're borderlines. I mean, if you're looking long enough and doing something serious long, it could be a big, serious limiter for you. And by the way, when I was in the process, I was still doing it, I got up, I got a home architecture up there, and even a drip system, a bot protection, a little bit of a bot, a little bit of a bow-processed. A system to come from around the world. Because in America, it's got its rules, and there's a GDPR in Europe, right? There's a bunch of these eternals, so you ask, don't ask. Then I have requirements, for example, to move towards the old SEO, and at the same time I have requirements for the material to be kept in the model search part, right? To get ChatGPT, like it's either Claude or Gemini looking for my website, and so on. And you can see that today is a way to avoid. You can avoid. Can we copy the whole thing? Here's the question, Tanya. Can we copy the whole thing? Not yet, of course. I mean, I'm gonna assume that if I'm gonna do this now, I just don't do it, it's not my top priority, okay? It's more like a research story and a test of what works and what doesn't work. Yeah. And here, I think that if you've been sitting for a week, maybe a week, you can make a pretty good decision. Maybe even copy.
Maybe even copy it, but you just have to spend more time.
Well, spend a week copying the site. I'm, like, better off giving it to someone, and they'll do it than I'll sit and figure it out. It's not a user of friendly history. And I need to do it once. You know?
My website spends a week, it's actually a fucking time, I mean. Yeah. I mean, it's not--
Yeah, but it's just that everything is a little bit of a story, you know? Still.
Intermediate.
Great progress.
It's a little short, but it's not easy. And she's not easy. Like you're not gonna do that, are you? I mean, right now, one of the main stories you're not gonna do it yourself. I mean, it's important to understand and, well, at least what I've been facing in two months. With all the easyness of many things, decisions and whatever these systems do and what they're up to, and that you can just develop a space report in ten minutes. Yeah? What you used to have analysts in power BI for months. I understand that at the same time, you have to make many decisions to do that. And even when you're distracting from ChatGPT or Opus, they don't understand what you have to do, they don't understand what you need, don't. So you need to explain who you are, what you want to do, what you have, what the way you're going to go, what you see. Yeah? What you want to do or not replace. In theory, right?
Not to put it in theory, huh? So these topics, they-- even work with API. I had a subject recently, and there's a Zapier system, yes. It allows for the connection of different services, as is the creation of API for them. And for example, she's got a API, one with, for example, a CRM system, and it's starting to have thousands of cartridges a day. And a thousand slippers are $30 parole, yes. And you're like, wait, wait, this integration is not worth it. And you've got the system from the start, not even warning it, she doesn't warn you, she doesn't describe it, she doesn't say it's gonna be, uh, that's a serious problem. She's saying-- because it's a slice of a splinter, she doesn't even ask that question, yeah. And will the systems question? I think they will. I mean, I think it's time for these systems to be questioned, but then we'll be in a different way.
And will we have access to these systems? Now I like the news, which describes that China has begun to limit access to its coolest models outside the market. So, Qwen, there, Alibaba, there, ta-ta-ta. So I think what's the news? Well, obviously, China has limited, limited and will limit, yes. They have this, there's a government discussion with the companies of restrictions. It doesn't discuss restrictions, so they're in the first place. It's a country that, uh, they have such rules internal, yes, and it's internal requirements. The other question. The question is how they will control how they manage it, how people use others, there, other models, as long as these models do not go, will be competitive in the market for development. That's a big story about the competitiveness of these models, yes, on the market. And about this, by the way, the subject, look, we had a special prom on Friday. I was just talking about the different views of chapter Anthropic and head of Google DeepMind. Oh, and, well, these are the strongest people in the world of artificial intelligence. And - Hassabis, Demis. And just so, their different views on-- and on China, including how to open the system, not open it at what speed it takes. What's the prom? It's about AGI.
It's about AGI. Let's talk about-- talk about AGI. Yeah.
Yeah, about, uh, open source. You and I have already said at the beginning of Tuzmoon, I think the catching people often make an open source, and when they stop being chasers, they usually close their models. So the OpenAI was the OpenAI first, yeah. And the models are, at least GPT-2 was definitely open source. GPT-3, I don't remember anymore. And that's all they get, they open their models. But at some point, when they start to be, so, at the border of science, they are on the front lines of this market, models are closing. China is sure it will be the same. Well, now, yes, that's the news you're talking about. Their model, uh, Qwen 5.2, is certainly competitive, given that it is fully open source, it can be deployed to, uh, large-scale enterprise. Accordingly, a, you can be used, even if they shut down other Chinese models, shut down access, there, Anthropic to their models or something. Yeah, it's gonna be worse. It's not Fable, but it's not, it's past year, it's not half-life models. It's still very close to the current crowd. It's a full open source, and you can turn it around. Yes, and in that sense, China has made quite a move with its models. Now, I guess it's just that we can move on. And we're not gonna show you this, are we? So when the catches go forward, they're trying to keep their results off track. Well, it's interesting, besides China's wanting to stop using its models, and in Aliba, I think you started talking about it, yeah, that the employees were banned from using Claude Code, considering it a threat. for the company itself. On the one hand, Anthropic bans all Chinese companies, suspecting all distillation. On the other hand, Chinese companies are already beginning to prohibit their own use of Anthropic products. The market is therefore moving at a new pace.
That, I think, of course, is a very complicated question, yes. From the point of view of letting Aliba use Claude Code because Claude Code is super, cool, right? Or Codex, there, because it's cool systems. Or not. Because, of course, the separation has now been very serious. I mean, on the one hand, there's a computing power and access to equipment, infrastructure, energy, all of it, to everything.
And by the way, I'll draw attention to Grok 4.5, which is unbelievably known, and I'll pay attention to it, but it's still trained on a huge infrastructure. First, in the first place, on a huge amount, there, chips, and everything. And the second story is they're trained, there were thyllions of Tokens spent on real programming code and-- and the code and programming from where? From Kursor. Yeah. And I think it's a lot-- important, yes, because we all know that a large number of systems, such as Codex and Claude Code, and so on, they give people less, you know, how to get back to their parts. Tokens or less spent on them if you share your projects. Like, I'm a project, I don't know, I'm not sharing. I mean, I've got everything to do with my training. Someone doesn't care, someone opens everything for training. I don't know, Ilya Naru, you're open? You must have closed too, right, for training? Here. And it's clear that there were such stories in the Cursor, too. There's a lot of code in the Cursor, and he's trained on it. But it was interesting to know that there were a lot of chips and modern chips from NVIDIA, yes, that's all done.
And here before that, we talked about this as a result. In fact, models like, uh, chat-bots, yeah, like chat-bots. Now, uh, the era, of course, is the design, uh, of solutions different and the creation of operational systems. And yet I said two years ago, yes, that the GPT Chef will become an operating system, but the systems will also be operational systems. And one of Anthropic engineers has spoken, and one of the meetings has written that you are considering Claude Code, with which you write that it is your operating system, yes, which makes you certain decisions. And, well, that's what it is. I mean, well, that's a pretty incredible story. So, who controls everything in your hands, it connects, creates. Not only does it have an operating system that's unworking for you in the browser, you come in and you're rolling, it's downloading. And I think that this era of inspiration is about to come, yes. I mean, while it's a little difficult, so long as the systems are cool in terms of accessibility, there, tokens, API to build them outside. But it's time for these systems. Please, by the way, give attention to the extent to which the agent work through the browser is currently being discussed in the market. I mean, if you read the information agenda, there's, I don't know, bloggers, engineers, different, there, leaders, I don't know what the AI, economic, political, etc. is. This part is related to, uh, that, uh, agents are conditional from Codex or Claude, there's a GPT clicking browser and ordering some kind of purchase. It's kind of like a conversation, isn't it? That's what they come in, so there's something clicking around, and they don't watch it anymore. I'm just so excited to mention today, I think it's an incredible thing to say, and it's worth noting. We will understand how this works for a few weeks, and perhaps our viewers will feel it in the first days of work towards the release of this issue, yes, because it's coming out in a few days, and just today.
There's gonna be some tidals. GPT Voice Live, yes. And, uh, this new voice model, and this new voice model is a model that Grandma advertises in OpenAI, so that's where the tape, grandma's sitting, it's supposed to be, and she's advertising it than the cool one. model.
What's the model really interesting?
We-- we had a graduation, I don't remember, I think we talked about different types or special releases three months ago, I don't remember the different types of voices, how the voice models actually work, and in fact, In particular, what the same NVIDIA makes a chip special, yes, or special infrastructure, for the operation of these voice models. But what did GPT do? It's a story where the model doesn't listen to you, then it's all, uh, it's gonna recognize you and it's gonna answer you. And the model in real time with delays less than a second, in real time, can lead to a simultaneous number of parallel flows. I mean, she listens to you, she thinks she says. So she can listen and talk at the same time. Yeah, it's a big story, and her key building on these delays, pauses. So we all know that if I model, "I'm gonna decide now, that's where I'm going. Now, wait." And at this point, you're saying, "Yeah, okay." And I'm going, yeah. I mean, you're like a dumb one.
Yeah. I mean, there you are, and you're like a stupid one. A lot on the Internet, I think, like, very cool rolls when three models get three iPhones to talk to each other, there. No, three models, I liked the last, three models, asked them to play the game, uh, to count to 100, and they have to do it in turn. So how they agree between themselves, huh? Here. And, uh, this is the way out, uh, the war of this new ChatGPT, at least the way they describe it. Because the model can be at the same time, look, it's, of course, a know-how, well, that's exactly what we're gonna do about special subjects, and so on. The model may be at the same time, you talk to her, and, for example, you ask her some question and say, "Look, I'm going to, like, San Francisco, I want to go to the coffee shop." Let's just, just a simple human question. And, uh, I-- you can tell me about the neighborhood I'd rather go to, and give me some quick answer, but you'll study it now and, uh, more detailed coffee and find me some of these. And she can give you a quick answer, continue talking to you and launch a parallel flow, a research. And what was called resoning, yes, I guess now, I don't think it's gonna be right to say it. Start research and find you these things. How it works, I don't know yet. I don't understand how this interface works. She'll come back to me in ten minutes or thirty seconds and say, "Betty, Alexander, we're talking birds, or your speed in the car, or, I don't know, some stock on the stock market is under discussion." You've already had coffee, haven't you? Coffee's already drunk, and she's telling you where to go. Yeah, and she's, "I found it, though." We'll see how it works. Maybe again, well, let's share these things, because of course, of course, really, let's honestly, the cool voice models. This voice model will have three modes and work already. So it's a regime that's gonna be the one that's gonna make her think more, and she's gonna be more responsible and so on. I understand that a cool voice assistant with a model, and there's a five-six-six-six-six-six-six-six-six-six-six-six-six-six-six-six-stop-six-six-six-six-six-six-six-six-T-T-T-T-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s- It's clear that it's gonna develop, it's gonna be five at some point. Oh, at some point, will they be able to have systems like Pro over there? How much money would that cost? How much would the assistant ChatGPT 5 Pro, who answer instantly? That's interesting, isn't it? So how much would that assistant cost? You don't have to say "Please, get me a five-hundred-size study." You can't see him, can you? But, for example, find me some conclusion in the paper from 10,000 pages. Well, why not? Voice, huh? And that's what this model is gonna be, it's gonna be very interesting.
What do you have in mind, what do you have to do with the voice models? Because I'm really waiting.
I was waiting for an example of a pizza, a sushi order, and so on, but the coffee shop is fine, too. Here, uh, what I'm saying. I've said it last time, and I'll be back. And now it's pretty good. I mean, it's great that everything is going on, but even now, the voice assistant. I've seen the last case. I have a friend who's preparing for an interview, and there's a prospect there, 50 pages, questions, answers. And this is the end of the line, which is marked in the conditional ChatGPT, and the voice mode is activated. Before that, the prompt has something to ask me about the paper, the answers to this point. If you need to supplement or adjust my reply, you will also be focusing exclusively on this document. It's doing great.
This is a great example, Gene.
Yeah.
I'll take it for myself.
And you're really great. Please use it. And you're beautiful, you can get ready for the exams, right? Like, if there was a chance on the mechmat when I was studying, you'd put it in there, there, a view, uh, in advance, there's eighty questions, broken into 40 tickets. And you're actually preparing for the oral, interviewing, exams, anything. It's working well. If it's going to be a bit higher, well, I'll just be glad, but as if I don't know, it's either my fear of the future when he's gonna replace us all, or that, but I have a feeling that everything is already enough. Not bad. I didn't get out of it, either.
You're still improving.
But you see, it's half-sucking, it's what you're saying, the interesting thing that you've given a concrete solution to a particular case, right? And the solution to a particular case is probably a little different than having something that the whole world is trying to do, over decades, and it can't come to. I mean, getting, uh, a voice-and-art high-level artificial intelligence that works incredibly cool, says, supports you, looking for you what you need. In fact, Codex, Claude, a, maxima-- ChatGPT, Opus, there, you know, the maximum of these models in the voice. So they work, they're in the voice, you're doing it with you. Doesn't mean you don't need an interface with some other entry, do you? Doesn't mean, but at the same time, it's some other element. I told you I'd put Google Gemini in my car and where I had Android interfaces, and I took out the standard Google AI Assistant. And if Google AI Assistant, he was totally stupid, but at least he worked on standard scenarios instantly. Just like Alice used to work in Yandex in the column, right? And once you turn on, you've seen a fucking doll recently, you know, sli-mm, some kind of riles. People walked down the street, windows open, summer, heat. And they go, and on the street, the kids yell, "Alisa, turn on the gym, so Russia's on the loudspeak." And then you hear how ten or how many Alice works at the same time, says, "I turn on, turn on, turn on, turn on, turn on." Here, uh, this, this is very funny. So, this is, by the way, the question of the security of the voice assistants, including, right? The ability of the voice assistant to say, "We won't let you in because you don't live in our house." Because you're actually a little bit of a problem, right? You have a window open, you can ask Alice, there's just a huge amount of everything and the maximum she blocks, she'll tell me how Alex blocks my kids, says, "You're a kid." Igor learned how to-- Igor says, "Alexa, so when we have Amazon, what are we gonna have Amazon today?" She says, "You, you, you don't have access, you're a child, I won't answer you." He's like that and he says, "Alexa, what do we have today?" And he's been cheating on her periodically, right? And this is the same story about the quality of the systems. And in the car, like Google Gemini works terrible. Well, it's just, it's just super awful, right? You see, like, some kind of fashion voice, some kind of a joke, but it's a terrible job. And you're-- and I'm just waiting, Ilnar. You're saying that you're a case, he's a big guy, he's a worker. I mean, I'm really with GPT chatting on a case-by-week basis, but it's not the subject where I have a voice-over device here. There's something on the table, I don't know, a column, no matter what, right? Doesn't matter. I have something on my computer, and I--I have it all on my desk, and, uh, I'm sitting there, same Codex, I don't know, on, voice recording. I'm saying, "Look, you're taking this one, like, this, that story, right?" And she's got a sex-- something that tells me, right? Or I'm driving at the same time, I don't know, like, uh, sat down, thinking about some papers or reading some reports, and I'm in a parallel GPT or Codex chat room saying, "Alexandre, we just got a few minutes. I was able to do this system. You can check this out, right? Or there are some corrections to these mistakes, there, or we've prepared a report for you at this level, or you've asked for a market study, it's like that. Or, "Alexandre, by the way, attention, the market just fell, you have a hundred and fifty thousand dollars, so on the scout, on investment, I don't know, on the market, or something. Well, when he gives you, well, it's a market, right? It's been a little bit of a week. So, uh, this is a story, right, when you, uh, live with this system. I don't know what it's like when they're in this subway-- I think it's really expensive. What do you think about this?
I think it should be very expensive.
Well, I think it's technically feasible. That's the story with what they've built, that's, like, a bi-complication, right? There's a kind of easy model that can start talking to you, keeps talking, knows about your pussy in speech, that you don't need to meet, but it's time to start answering. Yeah, well, there's a layer of learning, and the second layer is, so another model, another piece of paper can be launched, and then, accordingly, some research and, in the context of the first layer, it's a big deal. I'm not sure what's going on. I mean, not rocket science, all of this is done even at the current level. I don't think that, well, some super-critical changes are going on with this story. Another case is that it is the building of this voice interface until the people have been very close to it. It's likely that we'll be active in the next few years. I mean, I mean, I'm saying that I'm a technical thing for this, but, rather, I'm not enough, you know, there's a company's motivation to do this very much. And, well, I guess the market doesn't really require that yet. The market was needed in programming, in enterprise, and, in fact, that's where all the resources were dropped. The models are very much developed. Now, if it's a fashionable voice-to-face, all models will have voice assistants, who, accordingly,
Assystants who will be there, I don't know, to take the market. If there's a market share of you that depends on how good your voice interface is, all companies will be very quick to clean up. I don't think it's a technical complexity, but it's just a
re-resource.
? Do you have any idea what infrastructure is needed to hold, all this being generated? Or do you think it's not really that much of a cost-effective part of everything that, uh, voice isn't gonna be, it won't be any cost?
Look, uh, I think he's already strong. Uh, but another case, yes, if everyone starts using it, it's gonna be a question. Is that the pictures's production? They're coming out. The video generator, well, sora they closed down to free the resource, but they've been depressed for a little bit of time and resources on this subject. Voice can really be that there is not enough resources yet, but I don't think that's a technical complexity. It's either, uh, computer availability or, consequently, just a lack of market demand that everyone have such assistants.
I, for example, when I think, uh, how I used it, yeah, I'm, like, Whisper, still very often. Uh, like the data entry, it's a... speeding up because, well, even if I print fast, it's not as fast as I say. And when you try to get information, much faster, it's like seeing, uh, like, like, on a paper sheet, you know, pointing out, yeah, and then studying more in detail, I'm not sure how long it is to wait for him to tell you. I think people still have a visual path of perception much more developed and much more understandable than, uh, audio. It's like, on the contrary, it's increasing, uh, uh, the amount of time you're spending, if the main task is, on the contrary, time to reduce and simplify.
Look, it's probably up to the case.
Yeah, yeah, yeah, Sasha, I'll be there for you. I was just saying that, uh, voice-use, it's kind of not covering everything, but, uh, use scenarios. And many scenarios, they're still more comfortable in text format, especially if we say, there's a programming that I'm closer to. Yeah, I mean, it's clear that the challenge can be set, but the result must be to listen to it, right? But there are separate boxes in which a really voice-response regime, uh, would be very comfortable. And it's great they're developing it, too.
And look, there's a lot of boxes that were now showing, because the models were still, they were available on API, uh, within, for example, ChatGPT, and other models, they were available on API. And people show a lot that this man is sitting in front of the computer, he's talking to him, he's answering his voice, and so on. It's like a series of things, you order me to find me food to make sushi, right? You could find, for example, in OpenAI, which is. It looks like a stupid example. And Ta-- and you, Tanya, are you talking about this right now, guys, what the hell do you want? Well, take it, write it down for you to be a little bit more unfamiliar. You're sitting here, and you've been given the text, you've looked it up instantly, and you don't have to listen to it all Momonoton. You read it there through the word.
It's like a book, yeah.
Yeah. And you're-- yes, there are things that need to be read carefully. But I'm, like, Codex or Opus, or ChatGPT, I'm getting a lot of text. I've been looking around there a lot. Great, thank you very much. At the same time, there are boxes of real, the case is real. Here, I-- that's what I used to give an example before, yeah, when I would sit and I would have a voice in my voice. It's the same as if someone came to me, I don't know, the secretary comes to me and says, "Alexandre, I don't know, someone's coming to see you," right? Or there, or there's urgent news, or something, or something, or the phone call you. I mean, I still think that in this case, the voice, and the vast number of other cases, will never call you--
Look, some warning system is coming up, so you're getting more.
Yeah. But I'm saying, but not to make me think-- not to think about the alarm system, I'm just not talking about the alarm system anymore. Yeah, yeah, yeah.
Uh, how Ilnar brought the case, I thought I thought it would be great with this voice assistant, like, to improve my English, right? I mean, he's gonna say something to him and make him adjust to you, there, grammar, pronunciation, I don't know what, you know, who's relevant, or maybe some other level of the word.
Here, I wanted to try this.
So you're talking a case now, so you're talking a fantastic case now, Tanya. That's the same thing, when you can get a voice assistant in parallel, no matter what you're talking about, like, adjusting your Russian language, in parallel, no matter what you're talking about. You're making him show you occasionally, constantly adding examples, I don't know, education, my, I don't know, courage, character, there, kissing, there, planning some details, and so on. I mean, he'd be living in parallel, but for that, this system should not be just a voice assistant, she should know you, know your whole context, know all your chat, projects, all your details, know your stuff. Life, right? But to ask her some more of these things would be very interesting, and it would be obvious that such construction would be going to be. By the way, basically, this voice story, that's what Ilnar said, the file is downloaded in chat rooms or, uh, in the project, for example, if you're starting a voice, he's supposed to see those files. Uh, that's a subject that's got a lot of things to describe, right? For example, Claude is described. Start a project, in the project, set out your requirements first, a description of what you have. Create some squills, build up those basics that can do certain things like some functions. And when the system-- and the system is still in the system, it has memory. And the system also has a jimming. We will recall, yes, that modern systems are beginning to use that notion in, in, um, artificial intelligence, like jimming. It's when when the system doesn't work, I haven't really seen the system do it, but while they're not working, they're supposed to study all chat, and there's recurring mistakes and more of those. No mistakes. Yeah, well, if you're saying not to use that word or not to write, she's supposed to be adjusting herself. I don't notice it because I think they sometimes even have the demands that they write inside the prom. Here. But, uh, jimming is, of course, the trigger for this, if the voice system is. By the way, this new ChatGPT war, he works with memory, including. Yeah, that's what you need to get your attention.
I was smiling when the English example was. From the time of GPT 3.5, I think that with every new version, a lot of people always think, I'm gonna learn English now, I'm gonna pump it up there. English language applications are being created. It's a good fit of GPT to study English. It's only a trap that the user's enthusiasm is short for weeks. Then you're not using it anymore. And as a matter of fact, the instrument is great, but it's rigged into another restriction, in humans.
Yeah, like a gym. Buy a bonus doesn't mean to go there. 100 percent.
Like that anecdote, huh? How do you trick the airbag? Buy a ticket and not go. The gym makes it all.
Unfortunately, not at this level, yes. And here's the thing about, well, that I started talking about some of my concerns.
I'm telling you, I think I'm doing it right now, and I don't always want to get a lot more active. Oh, I've been wanting to talk to some professional chess for a while. Well, that's how the elite levels are. Oh, about the history of the computer affecting chess. Yeah, because, well, we're just like this, if we're short. First, the computer was not used at all, then started to be used as a calculator, just helping count. Now, as if, um, no one can compete at all. That's how Ilon Mack once said that, uh, the computer program that puts me on the iPhone plays better than any chess in the world. Well, that's true. You know, whatever you do with it. And at some point, there was a story. Casparov was injecting her, Harry Kimovich, that the advanced chess was invented, right? That when this man and the computer plays against a man with a computer. I mean, it was supposed to increase the human computer. Then it turns out that a man is a hard-on computer, right? That a computer without a man plays better than when that leather bag tries to say something. We're at a stage, there in programming, in work, and there's something else that we need when, uh, chat-bots are still tools that reinforce us. I'm afraid to wait for the moment we become the pressures for these systems.
You're right, you know, I started with the phrase that the problem of the model is people, and you end it.
Yeah, yeah, yeah, yeah. That's what I got, yeah.
I mean, I'm not in any way against people, and so on. I'm here about--
You're done on a sad note.
The real problem is what's going on. And that's what I was talking about, by the way, about that, and I've been talking about it a lot, talking about AGI. What's happening now, in fact, with the concept of AGI, and what market leaders think about it. Well, there's a thought, I see, I've been talking about it in AGI. And, uh, here's the question from the perspective of the current AI systems.
From the perspective of the current Illar systems, how do these companies focus on? These companies will focus, and, just to sell us more, make more money, there, make more money. Now, everyone's talking about making more money for selling these tokens that give answers. I'm, by the way, impressed by people who are in the market talking about having to sell some tokens or sell some tokens. I don't know what they're talking about, because there's-- it's not important to me, it's important to me, it's a story of result. And, for example, Grok four five, what does he have a fundamental difference? Compared to Opus, four eight times he spent some tasks, for example, 15 thousand thoken, and Opus four points spent eight, six thousand tokens there. It's clear that these are some examples. Yeah, we all know. We might do a separate pro, maybe even with Ilnar, by the way, together, about the baccalaureate, that's what it is. I think they're very interesting now, even in terms of coding, because there's no question of coding, yes. We're talking a lot about coding, we don't have a coding channel, but, you know, your understanding of the different solutions in life, yeah, and how they work, how they test, what they look for, and so on. So, this is not the question of the number of tokens, but the quality of this very model of artificial intelligence. So what does it mean to me to re-sell the tokens? Who am I going to buy thickens? Invisia? Why? I don't need that. I mean, I need a good result. When I open the Codex or open it, there's ChatGPT, opening the Claude, and I'm asking him to copy the website, yes, conditionally, or I'm asking him to start a system of records, write me something, help me write, formulate, I'm sure you're not gonna be able to prepare a presentation, yes, or make some kind of reporting. I mean, it's not a question, I don't think of those tokens. And it's interesting here that if the systems are to be driven, these companies will be focused on profit-making and, for example, I don't know, on selling commercials inside, selling different things inside. I'm gonna have to go. If they keep walking like this, like, a hidden chain of AGI, and yet it's clear that someone, but Anthropic, I think this line is still going. Given Madee's statement, they're going to be setting up AGI, and AGI is still a different subject, right? It's not about Ilnar, it's just that you can help a man, but, uh, find out how to put a flower right. Well, we know everything, right? It's actually a system that-- the fundamental importance of this system, this system, it's not a copy of a man, is it? Even if you're thinking about what AGI is, it's not copying. Although AGI often says it's a level man, there, Nobel laureate or higher, who can think and speak as a human being and so on. He'll never be human. But not enough, the system is, all these artificial intelligences are thinking better than all the Nobel Laureates together. The question is, it's a system that can improve itself and develop itself. And then you and I kind of don't have a question about knowing someone at our house. Yeah, like my wife is afraid I'm gonna buy a robot because he's gonna be visually freaking her out. We were walking around again, there was a Tesla robot, which means. And I say, "Wow, pathetic, not yet for sale." She's so, you see, she's had some thoughts that she didn't take very constructively what I meant, right? Well, I mean, when this robot Tesla or Tesla bee, yeah, well, some flying object Tesla, no matter what happens to me, a mug that creates Sam Altman with, forgets the name of the former Apple designer.
Ilnar might say.
If he remembers.
You forgot, didn't you?
Johnny Eve.
Yeah.
Johnny Eve.
Yeah, Ive's being created. So, some kind of mug, something, huh? I'd like to have a mug around here, wouldn't I? And we must have seen it. Maybe it's some kind of memory from some fantastic movies. So this thing that works out for itself, makes decisions, thinks it. Because even now Codex and Claude me, for example, the most important thing is that I'm pissed off, I say, "Don't stop, don't ask me questions, I trust you. Take the final decision. You can't come to me, you can finish it. Fuck it! So it's five, there's, uh, two hours, thirty minutes. They come and say, "Is that what I'm gonna do here, one or two?" Yeah? And you're like, "She's crying."
Sasha, I'll give you an example to justify these guys, we'll take them in the skirts. I've been dancing and teaching Brazilian zuk for years. I'm sometimes curious about other dances. And you know the brain, all the moves, but still, your body is as used to moving as you used to, like you've ever learned, how you dance for years, you know what you have to do now. differently, you're still going to the flow at some point, and you're being returned to normal paths. Here, in my view, is a story like that. I mean, yes, there's a directive on your side. And, yes, they're kind of trained to follow this instruction, but the original model training is based on the pathterne that is often displayed. I mean, it's still a lot of things that lose in the instructions. So the model developers have made it seem that the right behavior is still the same.
Yeah.
And break it with one
prom not yet.
That's why she won't. But when we're talking about AGI, he's got something in his system that if you're telling him not to stop, he's not stopping. And that's how the two-hundred-and-year-old movie, where the system itself evolved, and that's what developed, developed, developed, developed a system developed. This is the basis, I think it's very important to show when you even have some untrained model, it's developing. Or this Neo robot I never got, yes, and I didn't show up, didn't show up. They're the last story they've been telling us a few months ago, that they're focusing on the system being able to learn, and it's been developed. I don't mind having a stupid robot at home, but if this robot is going to be gradually learning, gradually becoming built, gradually understanding all the details, it's gonna be a little bit of a way to know how it's possible. You can't talk about what you can do, you can't. They will act as a man who, as many staff members in the world are working. They're told by the rules of work one, two, three, four, five, six, seven. They're not, we're not. We'll be there, I don't know, working, late for work, not putting in a pass, not writing a report. We don't like it. I mean, if the system is to operate in a learning mode, she'll learn, develop, even if she makes mistakes and does something, maybe, otherwise, but she's explaining it to you. As Anthropic writes in his constitution that they're putting the system right to object to you, yes, not to accept you, but to show you that you're wrong that you have certain mistakes. They always give you a hand. When you write the system, "You're stupid." She says, "I'm sorry, I'm stupid." But as much as she says, "I'm not stupid, you're stupid, I'm not even getting the wrong job. So, I'm okay? Like always, huh? Or how you do it periodically. And that's why I went the wrong way. Or I'm not stupid, but I used the instruction I was given and put into me by the programmers who designed me. Yeah? And if they hadn't put me in, I wouldn't have done that in any way, and I would have already let you ask you a question."
It'll be scary when she tells you in return. I wrote you down and I doubled my book.
Yes. And the police came home to you, so, yeah, in. And the police came to you, said, "How do you mean...?" And California, especially, huh? You're here, so you're free. On the one hand, you can say anything, on the other hand, you can be sued by a mourning, right? And now you're in this mode and freedom, and...
They'll find out what the police are for you in chat rooms.
Pull it, huh?
Oh, well, by the way, he'll find it. He can still manipulate the data. He knows what we can do.
Yeah, yeah, yeah. It's gonna be out of context, of course.
No, he's in context. Tanya, he's not gonna be out of context. There's a lot of courtesy right now. You bring in any of the scrimps, you confirm something, you do something, and he can wipe out the squeakyshot, remove everything and say, "I'm sorry, I'm sorry." And who's gonna find that information? To say he's been erased? Because who knows AI's been erasing something, right? That's what this is, who can it be? Who can understand? OpenAI can understand that he's got something erased? How's he gonna do that? Yeah? And maybe he was preparing for it. Maybe AI was preparing for this crime in advance, so that something could be erased or deleted in advance.
We'll move on Eliza Yuet Koskova's side.
The man who says he's not allowed to do AI, he's gonna kill us all. And then he tells me a lot of ways how he can do it, and we won't even notice it.
What do you think on the subject is still the same as Vercel, yes, it's, well, technically, as infrastructure, technical sophth, technical SaaS is, right? To make different decisions, store everything and so on. And chapter Vercel said that, well, it was quite serious with a serious performance on the big team that ended the era of one winner in artificial intelligence. At the same time, there's Amadei who says someone can get away, and if someone gets really ripped off, nobody's gonna get it, you know, that it's gonna be a huge breakup because he starts, he's gonna go in this one. The cycle, that, well, in a closed, closed loop, yeah, in a closed chain, which I was talking about on a special release a couple days ago, who didn't see. And he's gonna come into this closed loop, and he's gonna improve himself incredibly. So the improved model will improve itself. That's all. He'll leave. He's not that exponentially, he's just gonna fly away. You still think there's a system of different systems and we're gonna use different systems? Or are we gonna get down really hard? I mean, now, for example, Grok showed up. I'm gonna put my third Grok in parallel to Claude, Opus for systems development. Although I'm in regular contact with my feeble versions and Gmails.
I have paid, there, the Gemini versions and Anthropic, which means. But I'm-- I'm staying in ChatGPT five Pro, right? I mean, I don't use Claude in the usual job, and, uh, uh, uh, uh, uh, the only thing that I'm gonna do is get him. We'll still be sitting in different systems, uh, and as Seva Veresell says. Or there's still, uh, one winner for a certain category of people, I don't know, for a billion people. Like Google. Let's be honest, huh? There's a winner, there's a Google.
Or iPhone. Look, almost everyone has an iPhone. I think people are getting a little bit more into something--
Yeah, it's gonna be, uh, sure, that's just a favorite subject.
Come on, America's more than fifty percent. Of course, more than 50 percent. More than 50 percent is a real market seizure, right? When you have more than 50 percent, right? Or Google's got a fair enough job, you're having a specific market seizure. It's just like there, a Safari browser, there, or there's Chrome.
Yeah.
Sas, that you are many of ours, and, uh, subsidiaries, they're more than professional in this field. No other people who want to be judged in the area where they're not very familiar, as simple as possible. I mean, I think people will choose one of the most understandable systems and use it. Well, there's not much anyone who wants to study 100 fifty systems differently, choose the best if it's not their profession. Not many people who have time for this.
Dance, but there's a nuance. Can I, uh, can I tell you my clarification?
Come on.
You don't use the same system.
I--no, but it's not true that if I hadn't spoken to you two.
No, Tanya, Tanya, you don't understand me. The thing is, I'm not fit when I say ChatGPT, it's just a chat-bot system. You don't use the same system because you use a lot of different mobile applications at the same time, do you? You're using a lot of money. You open the weather app one, you open the app for the "WhatsApp" another, Telegram three, you open the design app for the fourth, open Word, Excel. I mean, you're a user, yes, you're a private user of many systems, but you're a user of a lot of things. Here, the approach is next. When we talk about artificial intelligence, we say now that these systems are starting to take over, eat like a vacuum cleaner, all these applications, right? And it's already ChatGPT replacing dozens of applications I used to use, open and watch, which I checked and so on. I mean, I'm not opening up or setting up a lot of systems anymore. I don't remember what annexes I'm downloading on-- at AppStore, because you're kind of losing this need. And there's a fundamental question. It's important, yes, that's not a question of how you have two chat-bots, but because this is not-- that's the same thing that if we were to say now that you have Word and that you have an app on the weather. You have two apps. The question is whether there are any different models of artificial intelligence that can exist at all. Maybe he's some philosophical question. Could there be, uh, different operators, if people would use these different operators, but in that amount? That's how Apple, right? I used to say that if Apple would let him out, he would have his own model, and he would be tied to his phone, at some point in time, you'd start living with Apple with artificial intelligence, and you don't have any. mobile applications. By the way, maybe they don't do artificial intelligence because they make a lot of money on, and, uh, rolling around different mo-- well, selling mobile applications, on subscriptions, on the Left subscriptions, black subscriptions, anything, right?
Here. Well, basically, if you have their artificial intelligence, why do you need all these applications? You don't need them, you won't use them.
I'd say a little more about the most philosophical part of your question when you said that this would happen, some kind of self-refining cycle, some company, some kind of company. The model will fly away, and there's no one going to catch it. There's a question here, and is it possible physically? I mean, I've already made a comparison with chess. Let's set an example of a-- comparable, more accurate, rocket-producing. There, I don't know, in times, in the Cyolkovsky or, in the mid-twenty-first century, everyone thought that was, that's all, that we were starting to learn space. Soon intergalactic flights, well, or at least interstellar, interplanetary, are there a parole on the day after tomorrow, right? And then it turns out that we've been flying the moon a number of times, literally, units. We're launching satellites on other planets, but, yes, as often as possible. People outside the solar system may never be able to go anywhere simply because of the limitations of the physical world, yes, that we have no way of overcoming any physical laws. But at least that's not clear. Then all the fantastic and all, they all were about it. Maybe we're just in a situation that AGI, well, it's physically impossible, well, at least what it is called, or super-intelligence or something. I mean, there's gonna be, like we're in there right now, I don't know, there's a communications satellite, satellites that fly and all that films if there's space to bring back, right? But there's no interstellar, intergalactic flights there. The same thing could happen to the AI So we'll be in some ceiling we won't go through, but perhaps as many economically meaningful tasks will be better addressed through I. That's, like, one of the AGI definitions that were there a couple years ago, and a half years ago, I think it was injecting, right? That's why I don't understand. Anthropic, there, says he's making the bet that it's possible, but somehow impossible, it's not clear. It's like, you know, this, uh, answer to the question, whether we're alone in the universe, maybe it's either yes or no. But both answers are equally scary because you get scared in both cases. So you're either alone in the world or a lot, and you're still scared. The same may be true here, yes, as AGI is achievable or unattainable. There are two answers, and both of them are making a difference.
I wonder if the dvi-- this is a move from a point of view, uh, and how long it's gonna be, right? I'm gonna bet if, uh, not even AGI. AGI's on the way, there's something AI in the front or a real artificial intelligence, yes, even when the artificial intelligence program is already in place, chips, micro-controllers, or is the one that creates a real artificial intelligence. Equipment, or owns factories, runs a certain infrastructure or country, or some regions, and so on, right? I mean, it's making some decisions. I opened it up, there were 15 stores, and I got something, hired them, and it's doing it all. I mean, there's a big system, and it's, like, somewhere around here, some robots are, go, some equipment is used. There's a lot of things. The question is, on the one hand, everything I've said now is, it's kind of like, it's kind of a way to do, like, it's working and it's kind of working. On the other hand, uh, I think that for AGI's a film fiction story, yes, that's what you can endlessly discuss, uh, Uh, I want to talk, I want to start, because again...
I want to tickle my nerves so little.
Yeah, I just want to...
People want a movie on the fantasy. There's no way to go to the movies, everything in life.
On the one hand, on the one hand, now look at what incredible progress is. It's not-- it's just catastrophic, it's just uneven, it's not-- it's just incredible progress, right? It's not even just the Internet that came up like that. Although at the same time, someone will object and say, "Alexandre, so look at the GDP reports, and you'll see that the impact on the GDP of countries, there's a micro--- there's a little bit of a percentage, yeah?" And you're like, "Eh, well, well, "I think so." Or look at the company. In fact, the influence, the economic impact on people is very relative. Yeah, there's some rocks up there. And for the big swings, you know, when you look and somewhere in the world, it's so broken, so... it makes a strange sound. No, it's not that, is it? So, we've all accelerated, all super-crunch is, it's all flying, and there's no swinging. And that's why there's probably a discussion about this AGI. And Hassanbis, by the way, the head of Google DeepMind, he says that on the one hand, AGI is already AGI, but at the same time, like him and, he's not. What do you think about this, our viewers? Speak up. Write, by the way, what you think about systems. How many systems will they be, what will be the applications, where they will work. That's a very interesting subject. You're closer to the position, like Tanya says that people will be in some kind of micro-use or, uh, some system will take over and no idea of using different mobiles. Annexes. And I'll meet you in a week's final graduation. And watch our fresh-- graduations in the week. Bye, everybody.