Hello, everyone! We're on ToTheMoon Channel once a week news, the Silicon Valley sites and the world. Technological news, sites, technology. Hey, guys!
And for me, it's amazing that, like a fanding, it's not enough to say that. In Human AI, I think we've invested in the estimate of three hundred million. Just what they do there is pathetic AI. But it's like everything is still at the point level, not at the level of what the final product looks like. That surprises me very much.
What the hell? I'm sorry, where did they even go? Why does this system start working through visual interface?
This memorandum sets out restrictive measures in terms of artificial intelligence, including the Pentagon. There, intelligence services prohibit the use of artificial intelligence for the launching of nuclear weapons. There...
That the integration of LLM with absolutely no programmes is getting closer.
Well, it's got a budget, well, $100, 1,000, 10,000, 100 grand, and you see what the agent can do when he's got a blunt maximization of profits. And there, there were actually positive results.
So many episodes have passed, where we have always been four. But today is the first edition of the three of us. We'll be the fourth final issue again, and Tanya will be back in. Here. We're usually four people here, and there's Tanya.
Look, I want to start with this quick-and-talk thing. I think she's so interesting, worthy of attention, especially now, given that elections in the United States in a week and just a week later, at least this film will be left for a few days when it comes out. I'll get a video, you'll be out for a few days. Byden signed a memorandum on national security, setting restrictive measures for artificial intelligence. That's a very interesting story, I think, especially that's always important. Why does the president do that at the end of his term? Well, I mean, if you do at the end of your term, especially given that there's no real clue about who's gonna win, and that strong candy--show-show-like candidates against each other, well, how much they are. - You're not. Yeah. There were candidates when they were so funny and they were standing around, and they were very counterbalanced. And this memorandum, if it is to be brief, that it establishes restrictive measures in terms of artificial intelligence for the state, there, including the Pentagon, there, intelligence services. The instrument prohibits the use of artificial intelligence to launch nuclear weapons, there, for all surveillance, religious attacks, different things, yes. And plus, it's important not to interfere. Protect the technology sector and AI from foreign espionage. I mean, there's a way back. What's the state level to protect? I mean, on the one hand, it's like not to interfere, on the other hand, to protect. There's a lot of memo points there. What do you think of a use? Why did Baden decide to let it go and get into it? Although Baden didn't even talk about the past this year, I'm not talking about the past, I haven't spoken about artificial intelligence. Well, some micro things are there, just micro standard things.
Yeah. I mean, it looks like a political story on my side. Maybe that's what they wanted to do for a while, but since he's not in the election now, maybe they think this is a measure of unpopularity. Yeah? If they wanted to do it and think it's a non-popular measure, it's logical to do it now, not later. Because then, well, it's gonna come to Camal, yes, if they believe they're gonna win, and it's gonna be hard for her to start with it, maybe. Maybe I don't know, but it's like it's just something to do with him wanting to put something in there now, for now. Or, or I don't know, or maybe he thinks they won't win. And in general, I have some sort of conspiracy theories, I don't know. But it's kind of a yes, it's totally un logical. I mean, but, unfortunately, that's what's happening in American realities, it's quite possible to expect that this happened, yes.
Yeah. It even looks like it's not Haris' project, like it's not connected to it. I mean, it's like it's on the side, it looks like it's, and it's also a matter of question. Not much, I think there are those items that are listed, that is a big, big story that would be really cool with the new president, yes, and tell.
There is, by the way, a very interesting point about the involvement of foreign professionals working in the field of AI and it is indicated that foreign expertise in the area of artificial intelligence is being attracted. Engineers and not to work for the competitors. There is a paragraph that is directly separate, yes, at the State level. That's interesting. And, by the way, with regard to private companies, there is a very clear indication that private companies are obliged to provide access to the State to analyse, do, you know, state control. On the part of artificial intelligence. I don't believe that artificial intelligence will be used in, uh, defense industry. And, uh, especially with what they say about intelligence services and espionage. I think they're using it all. Whoever uses data, there, machine learning and above all, will be stiff. I don't know. They'll be getting away from China on the move in these things. Noo--
Sasha, they're the first.
Yeah, they're the first ones there, they're the first ones with the first seats. I mean, this, uh, interesting thing is, I think it's a good story, I think it's a good thing to watch. It's the first, well, that level, at least, the politician, yeah. Before that, we said that Governor California, we had the releases straight, we were talking about how they are, what they're affecting, how California, as a front country, on the one hand, supports very limited-- The viewpoint protects private companies from design and that the AI develops. That's a topic that deserves attention. And, well, we'll keep an eye on her. We'll keep an eye on her, watch what's going on.
About the military and intelligence services at the AI, and so on. It's obvious it's not an artificial intelligence, it's a machine learning. That's the example you're gonna be bringing. But when I was doing computer vision, it's very popular, and now it's a very popular model, there's YOLO. This model is for the detail and classification of objects in images, and in particular, it is used on video, yes, there, for, uh, again, for the objects. To understand if there's a camera object or video or not. And, in fact, there's a training of objects, tracking of objects. So, the maker of this fraimvor, I regret to remember his name, but he was involved in this break-up project until the third version of YOLO, and then he left and he left for moral and ethical reasons that The military started using this fraimvor in drones. Then it was just a surveillance in such systems. And he, by moral and ethical standards, has declined to develop this project further. The project is developing just now, there, by other researchers.
I'm just saying that someone, and the military intelligence agencies and all the other guys are getting the first thing to do, just what comes up. What? Anthropics have produced a new model. GPT released o1, and it was interesting what the Anthropic could answer. The anthropics responded, but the answer was very unusual. Yeah, well, we, uh, had news that Claude 3.5 New had come out and, uh, other than some, uh, not very visible change, there was an opportunity, uh, to start a system so that it could access a local computer, yeah. So she's a conditional mouse to move in front of you. I used the software instead of you, I went to the browser, uh, started some search, and I was looking for the extra. Well, basically, you did the same thing you could do. Only, in fact, your computer is being operated by LLM. What do you think? How far is this going to the LLM to go? Where do you think the interface should be?
You mean in the com-- yes, in the commentaries, that the guys wrote it, yes, who's watching. Look who's watching, ours, and, uh, viewers, please write in the comments on this subject, too. And that story. What do you think? You still asking?
Yeah, yeah, yeah. And you're thinking, of course.
In my opinion, it's the right move. In my opinion, this is the right move, because, in a rude way, with a GPT reader... The main instinct was that the search should change and the space of personal assistants should change. Many start-ups have raised a lot of fading about personal assistants. No one's got any tractions. I mean, I don't know that. So, there's a difference--
So, you know, there's a different direction, right? The first-- the first line is RPA. That's a rude word of people doing a lot of such a repetitive job, yes, for example. 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- And it would seem that the model could look a lot, there, I don't know, we're talking about tens of thousands of hours and learn how to reproduce the mouse. And the second is a concuit line. It's just how the manager's work is done in any American company. I got one of those information, transferred it to another location, made some list of the lobbys, made those lobbys. These lobbys need to be updated every day, you know, to update the context that happened, see-- see, uh, well, as many pedits have come, you need to get your own freeboards off these apdits. So, and this is, like, information management, what actually everyone's doing is all medium, whole middle management, roughly, in all companies, at least America's exactly the same. Yia, yes, well, as long as it's been premised, it should be like a personal assistant who's a sha-- who's kind of structuring it on daddy's hands, yeah, like it used to be, but that's what it was. This structure, proposes some action items and helps to prioritize and call for these projects. Because there's still a call-up, especially in corporations, that's a very complicated process. Everyone has to be under each other somehow, with an endless number of meetings. It's in principle work, that is, in principle, working with tables very simple, without, without any difficulty with the dock-- that's the text. Ah, I'm almost out of a visa, right? I mean, there's no work with-- almost pictures, so, unless there's a schedule. Uh, no video work, no, maybe audio work, but only at the meeting level, summary and calendar work, right? I mean, it would seem like models are really good with that. I mean, the calendar is, he's got metadata there, and it doesn't have to be like, uh, getting a picture of the calendar, yeah, to take it. There's still metadata on it that describes it. I think that, as for the past two years, this progress has been kind of cold. I mean, that's what I've described, he's been like two years, but nobody's made real progress. I mean, uh, no fun demo. And now Claude, like, has said it. They-- until I saw a working flow, but at least people are already trying it. That's a lot of progress. And that they're... ripping up, that they did, uh, let go, like they did, that they'd really push it, and that Google. And there's gonna be a serious competition. Ah, it's still very unrealistic, um, who wins and what the company's interface looks like is gonna win. But at least it's from a dead point, and we're both-- and, like the corporations, they're not changing. Corporates are the best responders to competition from other big companies. And I think Anthropic is the one that's the one that let this industry down, right? Because as long as the ventures, they moved very slowly, well, just slowly. No big deck in two years. There was nothing to discuss. It's really cool that Anthropic moved the de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-de-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do It's already giving users a try.
Well, is that really gonna start using and using? What will be the development of this functionality now?
I don't know. I haven't seen any good examples of use yet, unfortunately. I mean, it's just that I can only speculate while
I'm at it.
Here I am, you know, when I first saw, the first reaction was, "What the hell is that? - I'm sorry, where did they even go? Why, is this system starting to operate through visual interface? " That's if, for example, you want to find the weather in San Francisco on some date, for example, right? How will a man do that? A man will come in on parole, open a browser, uh, Google will be hitting, there, "San Francisco's dog, that number," and looking for some results in the extradition. The program won't do that. API will knock in, uh, service that provides weather information and will get it from there. This will be much faster, more accurate and, accordingly, kind of a right approach. What did Anthropic do? Yeah, Anthropic showed that Claude now has a way of not knocking on the API system, but he can do the same as the user, open the browser, and, in fact, googling, um, some information he could have been able to. I'd like to get it, wouldn't I? And it looks like the system's going around. Then, uh, uh, you start thinking, uh, what kind of profiling is that? What's the use of the system getting in touch with each other, they start to duplicate the user. The only thing I could think of is that systems are now accessible, LLM systems are getting access to any software, even if API has not opened them. As a matter of fact, Microsoft Office, yes, it is unlikely that Claude will open direct access to work with this information accordingly. Or some kind of service there, huh? Competitive companies are likely to open up access only to their services, their LLM, their systems. And here it is that if this works well, uh, interaction with the system through the user interface, through the graphic interface, then with any program that a person can interact with, and LLM is also. starting to get the opportunity to work together. And that's what I'm saying. In this sense, LLM integration with absolutely no programmes is a step closer. Yeah, well, it turns out that whatever, the manufacturer has given it that opportunity or not. If a person can use this system, then LLM can also be trained to use the system, using this approach, which Anthropic has shown.
And in that sense, it looks interesting.
Not the end. Oh, and as you say, it seems strange from the start, like Sasha said, that, in fact, no one knows what future interface W-- actually in these systems and what it looks like. But there's a clear need to have this integration missing from API or someone gives access, right? Because a large number of systems, uh, no, will have no access under any circumstances and no API. And that's probably some kind of decision. I just don't... I'm very much a little bit of a repetition, and, well, I just don't know how it's possible now, at least, to put on something, as it must be, that it should be. Be for a system that will all be fully understood. It's probably a search, right? It might be a search, maybe some additional data on the use of experience. Maybe this scenario that we've been opened up is not over. So it's some story that's gonna let models and, uh, in the AGI, right? To provide some more data on actual user experience. Because, basically, uh, now these are the current chat windows or some sort of, uh, programmed interface, instead of interaction with LLM-ca, it's very limited. I mean, it's a part of something all the time. So the system doesn't know why you're getting those data. I mean, you want to know this weather because you need it for, uh, to go somewhere, or you want to know the weather, because you're putting it in some kind of report. I mean, she doesn't understand, she doesn't know these data. And I guess if you give this access to a broad format, it's getting pretty good, interesting, yeah.
And something new, something different. By the way, there's something new about Apple. They're adding micro different chips there. Uh, the fact that they're anoncily works partially, but at least they started, like, working in-- I started working. I don't know how anyone who has it. There's too much updates, yes, different kinds. Started working on a story that you can write when I, when I-- because I'm a user of the Apple Postal, Postal Interface, I can edit the letter inside, in fact, in this interface, right? It would seem like a perfectly elementary function. And how many sophths it was. But one thing is, I need to open up some extra program. Another case is the beginning of working inside the usual application. I'm not sure how much I'm saying, at the time I saw it, had to check my English language, there, at the professional level. Oh, I sent it to GPT chat. Then I returned, I was putting it in Apple, and when I was in, I saw that function. I mean, I realized that, uh, cool, I wouldn't have sent it next time.
I mean, this is how it works, all these new scenarios, these skies, and they're, uh, going back to some kind of interface, which we're not gonna have to figure out.
Yeah. Well, I'm very human-like, I really want it to be, like, so that the input can be voiced. First, it's faster, and second, well, it's obvious that, well, as much as I have, and there, I mean, people sitting behind the computer, especially the laptop, that's people, there's a mess that's going on, and, yeah, problems. I really hope that
Yeah, it's a problem. I very much hope that this will lead to such new models of interaction that are more useful to health. Well, I think it's just that chat-chat line, it's all really getting really up in one place, actually.
Yeah, yeah, yeah, yeah, yeah, yeah. Of course you need a voice. Sas, Sas, obviously, needs a voice, obviously. I need to.
OpenAI Canvas, there, I don't know, Anthropic Artifacts is great, but I don't know, it seems like Perplexity, Spaces is all great, of course, but it doesn't look like major breakthrough. I wonder, but with a voice, there's a misunderstanding about the fact that I can't really comprehend anything that's happening with a sound, with a voice. I don't have a picture in my head yet. I'll tell you. I've analyzed, well, we're dealing with a copper-space, right? And I understand I'm following all the competitions. And my original assumption was that ElevenLabs was like superpool, superpool.
Because, well, everyone like you have to do, well, you have to write a voice down, yeah, there's a ch-human. I'd say some voice over put something in other languages. It would seem like superpopular use of case, yes, ElevenLabs, it seems like everyone needs it. Oh, well, Runway ML, he's got all the Twitter, too. I, uh, as it feels like he's just everywhere, I have a tape all over me, like everything, all videos, it's like it's like it's all about Runway. Okay. Well, I understand that algorithm is teaching me, right? But I'm just getting a feeling like this. But it always needs to be with numbers, yes, real numbers, like real figures. And I found out that Suno has more usage than Runway, all video editors, Udio and ElevenLabs together. And it was a real revelation for me. Oh, and I found out? Which turns out in public places, uh, you have to pay royalty for tapping the track, right? So we're Spotify just switching on it like that. I remember, I--we had a birthday party, too. I said, "Let's just get Spotify on, "Right track, normal sound--" there, 'cause there's already some playlists like, "Aaaa, there, for the birthday and so on, yeah, music." I'm just saying, let's turn it on. No, it's not natural to do that. There's a special program, and there's some kind of economy going there and there. And, yes, I understand that Suno is popular because it's just a nice music to turn on, and, and that solves a specific economic problem, especially in countries, where Economically, there's no way to turn on copyright music. What am I saying? I'm saying that in YC, there's some sort of unfamiliar situation for me. I mean, on the one hand, the recognition models have become very good. Even my strong accent is already well recognized by models. ElevenLabs is a good voice-generating, castomizing. Emotionality is already good enough. OpenAI has good emotional in their app. API is not very good, but it's a good app. And I'm really very strong, and I'm surprised why there's no adoption. And for me, it's amazing that, like a fanding, it's not enough to say that. In Human AI, I think we've invested in the estimate of three hundred million. Just what they do there is pathetic AI. But it's like everything is still at the point level, not at the level of what the final product looks like. That's what makes me so surprised. I mean, there's a ogre-- that's all the start-ups in YC, who make the voice of agents, they got, like, a customer support, voice agent, that's all. This whole thing is getting really good, but it's like B2B, yeah, that way. There's, like, in-voice consultants, voice customer support is growing. There are some points in-- on the technical side, like this Human compasetic, and no products of the Consymers at all. That surprises me. So the Suno and ElevenLabs are already on a cool level. This is not a foundational company, is it? I mean, they've reached it, there, with a rash that's worth not more than $22 million. I mean, it's not-- we don't talk about hundreds of millions of dollars for training.
That's why I'm very surprised.
Look, well, now even in terms of voice, when we talk about daily use, it's still good to recognize and translate voices, for example, into text there or from text to voice. But this is the real speech, however, alive, and it's very problematic. And I mean, no matter how much I test the interface of this voice-over, which even released the last OpenAI, but it's still... Anyway, it doesn't add up. They can't resolve this issue. So they can't let this real communication go, well, as real time, alive, with a vast amount of detail. What's so cool about any accent is a fact. Not enough when you say English-language words to him, he's very cool with translations, he's putting them in. Or when you, you know, he's doing anything to you, yeah, in real mode. He might be able to speak English, he'll write you all the Russians that-- well. But in real mode, that's a question. I mean, production must be audio, yeah, production, and what happens to this is I have no idea what the industry is. Still, what you call money in the coffee shop is that you're gonna pay for the public event, so I think it's just for very high-developed countries, right? Because it's clear that people in ordinary economies think, what is that? How is that possible?
They're using now, yes. But nah-- but look, I know, I do. If you have regulation and there is a big regulation, you'll be able to introduce history when you have a system that generates, Suno will generate music like you. Not much of a time you're just gonna know what it is and what it is, who did it, and when, and when you did it, huh? It's a little hard for me to say, like a seventeen Son Seven listener in the car, yeah, only radio seventieth and only in the car. And that's what my music ends in a hearing. But people who consume massly, of course, and supplement what does in social media, right? These trends are different music in social media, of course, that's some other world. And I think he's different from the perspective of the interface of a common human user.
Well, that's what you started, yeah, that's when you said I didn't want to interact with that line there, and that's enough, right? But look at the cars, the cars.
I grew up if I could take more.
Look, look at cars, guys, in cars, this is voice recognition, it's just something, I don't know, ninety, but it's just very uncorreasy. Well, Volkswagen now says he's gonna put the Open Air Identification on himself, but he's gonna put it in there, right? But I understand that now a lot of people who will be building and what to do if it's gonna work okay. But I'm not talking in the usual life anymore, yes, when it's a conditionality. We were just sitting with my friend here at my house, on the balk-- on the terrace, and we were talking about what the phone was for? If I had some microphones and some columns, I'd just be able to get some information. I'd have heard that. He says, "This is Alexa." He came back to me right away. He's been doing a lot of voice-making. He says, "This is Alexa right away." I say, "Well, only at this moment, Alexa is, it's a thrash. Well, as they say, I'll keep saying that.
So, maybe it's gonna be a breakthrough. The one who learns to do it. We just started with the subject, and here's Anthropic's new interface. I think this interface is for a new kind of a deta. Otherwise, I don't understand some things, why are they giving them up? That's how OpenAI Canvas released this Canvas.
No, that's all, that's not what it is.
Just, well, how would there be a working script, but just why? Well, let's get a full search out. Okay, here's a full-time search, spend time on it, do the interface cool, let it out. It's gonna at least increase your user volume, I don't know, give you another kind of action, but some of these scenarios are some kind of, uh, I think, micro-kills. Not much of a scenario will not work, that's how it's handled. So, no cool service has ever been turned. Until it's working inside. I mean, it makes sense that Google Mail, there, Apple, will all have their processors, letter processing, yes, and recommendations, and that are recommendations. Obviously. Well, that's it.
Well, it's obvious they'll do it themselves.
Yeah. And I wanted a little bit, because we closed this topic today, and we touched it. An important subject. So Apple Intelligence came out. I don't know how to describe it.
That's very bad.
That's very bad. That's very bad. And here I owe you, honestly, a shit.
I have to be honest here, I'm here to be honest with you, I want to-- that's, that's, in my eyes, yeah, OpenAI fell. Anthropic, as it would be, Apple fell in zero. I mean, I don't know what they're doing. That's, uh, that's bad. I'm telling you, I wanted to praise Google. Just if we're talking about the application cla-- some kind of breakthrough. First, I was asked to write a reference email, and I worked there a long time ago with a dude. I can barely remember what I was. And I still can't write anything specific. Google answered me with this email that it had to be fixed there less than a minute. And, and ready, it's gone. I liked it. I mean, it's really good. And, of course, NotebookLM is a good use. It's, like, a pet project, probably someone in Google. But that's a good use, too. And frankly, well, as it is, while it feels like the foundational company, they're not trying to start the product, they're trying to start the prototypes. They're all working prototypes now, they're all like a start-up of the seed studio. Something's been set up, so light interface--
Yes.
See. It'll work. No workflow, nothing, purely interface with some kind of knob, modelled, improvised and ready. Yeah. So, it's like everyone has a $3 million fanding of this seed-strapp, uh, that needs to just, I don't know, get a thousand users. Aah, well, some kind of approach, yeah. I mean, unfortunately, don't understand that. Apple disappointed me very much. Maybe you have a-- maybe you have a more positive experience in using? I don't get it.
I mean...
No, I don't have a positive, positive experience. Not much I mean, look, I'm a Apple lover, and I'm often talking about it. Given the number of their equipment and everything. I still feel that Apple is taking into account their teams and everything they're gonna do and let out something serious. I hope they release their model, and they'll all end up saying that Apple is not a model, they'll have their own model. At the same time, Sasha, you're right. Like, they released this Apple Intelligence, and I have it every day or every two days, but at the time I didn't ask, I don't know what it is and I don't know what it wants to offer me. I'm just sitting there, uh, sat some text reading, and it's a stupid application, totally starting-up. That's when they're with the starters, which means they offered some idea, they come to you, they say, "Get it." You say, "I want to use it. I don't need it, it's just gonna make me feel more difficult." Here I am Apple here to ask, ask Siri, it looks very strange. There are some things they've improved. I have $80,000 on my iPhone. They've done all the fun things that you can do, I don't know, quick memories of videos there or something. But it's microscopes. Not much of a job, they're really working on a real curve. Well, they don't work great. It's not a system that really analyzed your content. Not much, I'm ready to open a lot of my content. And that's because I know it's gonna give me the opportunity, you know, to get better, yeah, faster to move. While I'm still in command line, alas, command line, it means ChatGPT, where the Jay, uh, I, by the way, became active in using o1 in some of the topics. And since it's not worth putting files in there, I'm copying this big text, giving it to him, and he's gonna recognize me. Well, I'm using some kind of plastic. Well, I know people use things differently. I mean, I don't have Apple, Sasha, so that's... And they're all there. I mean, if you look at everything, I think the last six months, the most huge number of applications, everyone in Google, Google X AI to Apple, everybody has a huge number of statements. But no-- you brought-- cool, by the way, this comparison, I didn't think about him, uh, they seemed to be acting like a start-up on a phase seed. So show me your idea. Yeah, that's how YC listens, Y Combinator, two hundred companies of these that they're telling in two minutes.
That's a very similar subject. Hmm. The Apple leader went to the demonstration, told me that was a thing. Where?
While you're talking, I was sitting, thinking, and how could a new interface be used as-- well, a new way of interaction between Anthropic that was given? As long as I've come up with this, it's possible, for example, to frillans or Upwork, uh, to give this cat a payload. And it only makes the question in the economy, doesn't it? I mean, money--
Yeah, yeah, yeah.
Which you get for the order you made, they cover burnt tokens or not. But if they cover, they make a money-making loop.
And som-- and you see what an agent can do when he's got a job to maximise his profits. And there, there were actually positive results. Here I wanted to do a little segue, aaaa, so a little law-- we're in large companies, basically, almost everything but Microsoft. Maybe today was the one that wanted to finish Microsoft. What I like about them in the approach that they're in the Copilot, I think they've announced in the last 12 hours that they're gonna use all the models. And Anthropic, and Google. That's really cool, of course, it's--
Seriously, huh?
It's killing everything now. It kills all the start-ups. All right, more start-ups that coding do don't exist.
Unfortunately. Because how? You have a distribution, right? And you, well, as you might say, "Oh, I'm vertical, I'm there, there's all, there's a vertical glow under the task and all that." But how is that? Well, I mean, um, and that's where you can do, say that OpenAI is not focused on the codeing. What's the truth, because Claude wins them in the coding. But put the foundational models against all.
In a skirt that is obviously important to Claude and Gugla. That, that-- and they can connect other models. I'm sure if some start-up is gonna work out its model, they're gonna be on it anyway. I'm sure Copilot will. It's a very serious step that will cut off all the competitors. I don't know, it's, uh, that's, of course, that's not all the co-coding of the starttaps either. And frankly, now the main examples of the same Claude, these, this Claude, what he can do, not only the writing code, but he can work in the IDE, yes, that is, he can, you know, it's on me. Look. It's, uh, it's a serious step.
Sasha, you're saying they don't have a strategy either. They're not strategic either, because Microsoft said six months ago one year ago, another, now says the third. So that's what they're looking for, too, and they never know where they're
going.
They're looking, yes, they're looking. See, you see, it's kind of interesting what's right. I mean, heard there's a big problem, you know, with the OpenAI, right? I mean, there, I see, when you're running out of your leadership, the strategy is changing, and the distribution of people between projects. Sora didn't get a release, she didn't release, so on. Sora's no longer the best evaluation model, not the best. The Chinese model, which is Kling, is considered an evaluation objectively better than Sora. So, we're just a little bit out of the way, but the point is, you know, Microsoft plays it in his game when they have this distribution, and this is the viti-form structure, there's a connection to all of them. software ventor, integration, et cetera. I mean, now mean, there is no such approach that I have vertically, that we can be vertically and defeated. And I think those who say they're gonna do something on-prem are there, I don't know, VPC, that this is gonna be super-conscious. I think that Microsoft's game is gonna be very hard to win, too. I'm as good as I can...
In which direction, Sasha, can you get for the audience to make it clear more clearly in which directions can win?
I think Microsoft is a little bit of a winning place right now.
Yeah.
I think it's--
I don't know, I don't know, well, that's not all.
Well, it's in the codeing. And now I don't understand how to defeat them. So they have a huge distribution in the coding, they own GitHub. GitLab, another kind of alternative platform, like GenAI, does nothing at all. Now, how to defeat Microsoft in the coding, I don't know. So they took off the open-AI lock and used all the models, which, in my opinion, is a titanical change that, like, I don't know, which... I mean, it's clear that Open, that there will be OpenAI, Anthropic, Meta, Google, Chinese companies that train their models. And if Microsoft connects the best, competition is very high, of course.
Competition is very-- I mean, a plank, a stack to compete with Microsoft, it's getting really big.
Yeah. Although, by the way, we'll recall that a large number of start-ups, especially the start-ups that were funded, were in the coding. And how much coding is funded is just a thinner.
Yeah, I would. I mean, I'd say, "With a point of competing," I mean, I'd really like to go. I mean, there's a chance that it's just to cover the workflow, but workflow, with integration into such very complex systems. Well, for the cryotor who wants, I don't know, I don't know, the cryotor who wants to do his own, his own mediocronym, yes, for example. I mean, you need to put your content on, make your mobile application, website, sub-cadast. Here's a new one.
You can make a new media platform like that. There, yes, there must be agents who write the annex and the site, but it's really got to be-- I think it should go beyond the direct coding, and design should include some QA. Oh, I mean, it's, I think, it's getting a little harder now. I mean, I-I'm just gonna be clean-- if, like, a start-up is doing something really codeing, I think Microsoft is not gonna win.
That's fun, that's interesting. A very interesting ending of our-- today's issue. Please write in the comments. I think we've raised this subject today about the players. What do you think about the players? Where are you going? We're all reading, we'll watch it together and remember it. And, uh, who's gonna win? The most important thing is who wins. You'll have to write the comments you're thinking about what Baden did. This is important in terms of the memorandum. I'll see you in exactly a week, Sunday. We're always four people here. Today is the 30th, we have thirty episodes, a anniversary. And we're not Tanya, but she, uh, she's as likely as possible and she's all right. See you again! Technological news, Silicon Valley sites around the world. Guys, bye. Thank you very much, everybody.