Hello, everyone! We're on ToTheMoon. A-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the-the- Ah, we're always four people on Sunday, right?
There will be a mass-based technology for the replacement of a number of people.
There must be a re-purification-- market clean-up. And to start showing up and some of the strongest teams have gone to big companies.
It's all a little childish toys, but I'm looking at my husband, he's just hanging out.
As long as it feels like all the questions of the bot, this interviewer will answer, call Tanya.
Thirty years ago, technology is virtually impossible. Now, that's absolutely basic and, uh, good enough.
For a man who doesn't connect with his profession, he just needs to repair, he has a limited budget. It's a really good way out.
The level is just a nursery.
Skynet said it all the truth. Skynet's kind of wrong.
Given what the apegraide did, I want to tell you what Microsoft did in terms of, uh-oh, his introductions. And the subject today is agents of artificial intelligence. These are not agents who spies on someone, they're, uh, service systems that they're talking about. Will the artificial intelligence of a man be replaced? And when we hear these phrases, we hear them last and a half years at a huge number of conferences, I think the riles, the utubs, all of them talking about this, if they replace them, often like that. It's being discussed that some kind of staff member or agent or some kind of service performed in artificial intelligence, and he's doing some sort of thing: I don't know, he's selling, uh, doing something. Ruthen action, builds, directs a calendar, replaces assistants, er, or whatever creates completely, or does something, there, or some reasonable creature. That's the point I want to start. Microsoft was an auddian. Yeah, Ilnar, can you please give me a little more detailed account of what-- see the peddate in general? I think all the companies are doing peddata now, and Microsoft has done it, well, everything, it's gonna be done. Like a connection.
Microsoft has turned out that in 365, well, it's a cloud version of Microsoft Office that will be integrated with the different AI services. You can create your agents, yes, there's some studio to do that where you can confide your things. Well, they've shown the tools that are already available for use. Very interesting was in Microsoft Teams, well, it's a conduit Zoom analog. There's a way, for example, in the summer, uh, to make your interviewer speak another language. Well, as an example, it shows that the conference is in the process, everyone speaks English, and one participant translates into Korean and the same voice is beginning to speak Korean in the summer.
And it's like some skies are already coming up, which, most importantly, are built into the ecosystem. Yeah, well, if you use Microsoft 365, you don't need to fix anything else. You're actually just starting these constructions, maybe using extra fees or just a more expensive tariff would have to be bought. And there are some situations that might be useful to you. Like, if you have an international team there, there are people on the team who are developing from different countries. For example, uh, you can speak English, but you can be more comfortable talking in another language. You can translate if he's already in this service. And in this regard, it looks like a pretty easy entrance to use these tools, because what was needed before was built, there was a need to set up some extra services, some more. Something like that.
And it's kind of like all the boxes are available, so it looks at least interesting.
But here's an interesting moment, which I personally see as two years, an endless number of people, especially those in business or start-ups, and especially those of different entrepreneurs, all say, "We're at IT now." "we'll go in and do some venture product, get an investment." And so a lot of people have started to create agents, like support agents, agents to sell. And I was just a few days ago, a few days ago at a meeting, and the man says, "We're gonna start doing, we want to make an agent to train sales managers." And, well, I think that you can't make any money for that, and you can't. And there's no model or youth economy. Uh, shorter, it's not gonna be a design to make money for it. I don't understand why I need someone to create an agent to train sales managers if it's very easy to write today, even though I'm not a programmer? So you can create some kind of plagin or someplace, you know, a small package that you're gonna throw in there, I don't know, 20, 30, 100, 1,000 different questions, and the system will be, Mm-hmm, every sales manager is trained. I mean, why would I put in some kind of soft that I usually know different systems that are used here in HR, and money is still not available at B2B. They usually say, "We're $200 for a staff member, like, we'll take a company." So I think, like, I'm a sales manager, like I'm gonna pay $20,000 extra if I pay CRM for the salesforce, there's thirty, fifty, like, there, I don't know, $600, right? Well, there's a hundred. Why would I put this agent in for this? It's on the one hand. On the other hand, it works very relatively. And all these systems that people do, they look like, uh, chat-up chat-ups, yeah, totally dope-up. So when we open, we fight, there, Gemini or ChatGPT, or, I don't know, Grok, so... we're fighting something, so we're saying they're making some kind of curved answers, so all these agents work at all. Don't you fucking understand. Well, it's very hard, I apologize for saying, and, uh, I'm seeing, uh, cool machine-training tools, that's machine learning, yeah, in companies of different kinds, and LLM models in. Some skies, some skies. But a complete replacement of people, um, especially what humanity cares about, it all looks so specific. It's from the point of view of the agents. We're still dealing with the specific topic today, aren't we? What came up-- something that, uh, was easy to build and easy to do. Plus, I don't know how to make money for that now. It's just, uh, a lot of people go in there, but how they're gonna make money, I don't know. We're discussing this topic today with a lot of agents.
You're using something in your life that's really a real substitute for you or some colleagues or staff? Have you seen this function? What were those systems? We're talking now, I'm asking you now. Now, write about this where you saw a real full use, right? Because it's not a short-term one. In the comments, we'll see all this, see, see, some, maybe some examples and systems. Thank you very much. Other people are always very interested in knowing.
We're definitely not gonna tell you everything.
About starting with AI and with agents in general. You know, there's a co--coat metaphor on the heat track. I mean, a big factory, some pipes coming from it, and it's a condition that it's got heat on those pipes. And the cats, in winter, live on this heatstrass and they think they're doing some business there, and in fact they're completely dependent on this factory. And now that we're talking about some kind of start-up with AI agents, it's mostly just a little superstructure, but over OpenAI chat, chat, there, Anthropic, and some other stuff. And actually, they, well, have little value in themselves, they just change the way they're giving out chat chat. And in that sense, the value is zero. If anyone's gonna get a shot, that's like a head company, you know, a parole pipe they're sitting on, the chopping's all over it, and this start-up just goes off into the pipe. So some interesting toys that attract attention, Hype, money, but it feels like it's not that far.
Well, it's cool to do it on infrastructure, like you said. So one thing you have is in Word, and you're in Word, if you do some kind of thing every day, you added it to yourself or to the company, they use it, or, uh, they use Microsoft, or, uh, they use it, or, uh, you know, in Word, if you do some kind of thing every day, you're gonna do something, you know, you're gonna put it in a lot of a way, or something that, like, like, you use it, or, like, like, like, it, it, it, it, it, it, it, it, it, it's a little, it's a little, it's a little, it's a little, it's a little, it's a little, it's a little, it's a little, it's a little, like, it's a little, like, like, like, it's a little, it's a little, like, it's a little, like, like, like, like, it, it's a little, it's a little, it's a little, They use, there, I don't know, Microsoft Teams, and they're in Microsoft Teams, they're either onto or on Google Meet, like, they're on, uh, they're on a new jack, I don't know what agent that makes the transcripts. meetings, those things, but they work when you already use that big sophth, and it's just a little extra infrastructure, probable free, because you're already paying for it. You're the service or you pay for some vault, and it's supposed to be free. You can't endlessly pay for every man still there or every chip in the extra, extra. Although, I guess if you were to create an agent that replaces, I see, there's a full-value salesman, I know that you can pay for it, you pay for it.
It's a price to pay. You pay for it. And I'm willing to pay more. Question, uh, I'm willing to pay a lot more because it's a man who's gonna work for 24-four. It's a man who's gonna work stable, quality, whatever his mood, nature, emotions, desire to quit or to get hired, change his wages. Is that a question - if you can do it? Again, can they do that? And plus we're still talking about the market, so, uh, the Venture.
Sas, you're a technology-like, uh, now this level, uh, uh, uh, in the decision?
It's so bad, because the thing is to make agents, well, the first step to make agents is what I think it takes, that, well, you need to have some language of writing. these agents. There's no sign of the agent writing., that is, the DSPy framework Sanford has done, but no one uses it. I mean, uh, I remember, Salesforce, IBM, Microsoft-- all of them are making their decisions for, like, OpenAI, for the writing of agents, but they have no adoption. I mean, it's just, uh, we've gone through-- just, uh, OpenAI, Anthropic and venture capitalists to survive, that is, basically, OpenAI at Anthropic and venture capitalists. very salinary scoring now, uh, selling the future, right? Because each of these companies, it's strictly awkward, without any pre-concessions to make a living. Well, no venchour, no Anthropic, no OpenAI, none of them will ever make any money, maybe for ten years. I mean, it's so far away. The question is, that is, none of them can tell how they're gonna make money, and in ten years, they'll make a lot of money, and they're gonna have to sell the future. They have nothing else, don't they? They have huge cost, huge investments. And since, in fact, at the model level, we've been running into a-- some ceiling, they're selling this reasoning and agents. But not yet, there's no success in the boxes from either of these companies or infrastructure. I mean, as long as we're at the beginning of the way with agents, at the very beginning of the journey, objectively speaking., if we're going to have a good competitive market with LLM, a lot of options and choices, and different quality options, at cost, at the flexibility, we're, uh, in agents, just at the start-up. We're all here. So the only successful example is that Intercom charges are for-- for a bell-bone.
Yeah, I guess it's the only case, yeah.
Sas, look, I'm right here to see that there are big cameramen, and they're definitely not focused on giving someone a micro-managed agent right now. They all realize that if they reach some next level, uh, extradition, where there's, uh, other speed, uh, where the reasoning is going to be very cool, where there's questions-- the answer where, uh, the system, I'm gonna be able to develop myself. So he's the one who's... he's gonna learn from different things, so they're, uh, this agent's gonna be built like that... on the finger click. I mean, the interface itself and the function together to create such agents. Or they're--
There's no even a program to program infrastructure. So you're talking about no code, no code smells at all. Until I have a language, because Python, well, it's all done on Python, but Python isn't very comfortable with the lines. Well, this global Python problem is that he's not super comfortable with the lines. I'm gonna be a little bit of a hustle, right? And that's how the agent is, in fact, a swimming-- a transfer of proms from one place to another. And there's a logic on the top of that, and some kind of simple method of optimizing the overthrow of this. Uh, well, so there's not even a frime handler yet, how to codify.
Yeah, but I'm talking about it, if they did, like, go and do AGI, then the release itself, uh, that agent, he's a... he's getting quite different from all those systems. So why would I just use some side service? I know there's a pretty serious startup, and I have one neighbor, uh, doing it, and he's so serious, there's Vice President Google, Vice President Coinbase, and they're making agents, packing a lot of different things, Uh, systems between themselves, a bunch of different models and outlets like this agent, they're putting it in B2B. But why would I like to introduce, uh, this B2B agent through the side service that's gonna be responsible for this agent if I'm at the moment when, uh, OpenAI or Anthropic write, um, do, uh-- Well, they're gonna improve their models and they're gonna be able to, like, get to the agent-- to create these things, just make it inside like a
company.
I don't think it's possible because if you look at OpenAI, well, OpenAI, they started evals, like, and it's kind of a first step there to make an agent. Because if the agent, they're needed to get a little-- well, a logic of agents to do some work. And that's what you need to judge her. I'm not sure whether it's openAI or Anthropic, they're not building a good eval business. I mean, I think they're very far from the technology developmental environment, they're very far from the end of business junite. So I'm not sure, to be honest. I mean, I think they won't have the chance. I think so. I mean, they're gonna-- I think that--
I mean, it turns out, it turns out new operators who-
I'll be fine. I mean, you need to build... First, there is a need for infrastructure where, when you look after these agents, how to connect them so that it is an interface that business owner can watch. No, not the developer, but the business owner, who could have seen and understood what was going on. Oh, you need to, and so that the developers can program it for something, then there was some business interface so that you could watch the meters in this business interface, right? I mean, well, it's standard. Oh, and some, uh, progress in this system, that is, it's, like, change, see progress, see, uh, result. And so that the developers can all of this, actually, create this system. I mean, these four important components are needed to make a business application. None of them exist yet.
So you said that now that you told me, it felt like you could still be, uh, company and right in-- for different verticals, uh, build these staff. I mean, as a condition as sales force or companies that sell CRM systems, uh, they'll be companies that will be such a mass-assed technology, uh, replacement of a certain number
of people.
Yeah, yeah. But we're so far away from it. I mean, now, if, for example, if c-- is now, for example, for the picture models and the video lo-- there's a LOR mechanism, that's training, that's a consensual character. For-- Apple, he introduced LOR, and introduces his stuff on his phone. So to make full fine tune, well, as ordinary as it is, a food engineer can't go and make full fine tune Llama-70B is an unsolved task now for mankind, so it's just easy. from the box. But the OpenAI is giving a function, I guess, it's probably working for someone, but, let's just say, this is a complete task that is not solved. We're talking about what a reasoning is. In principle, the reasoning models, their two are the OpenAI o1 and DeepSeek, they released a model in open source that hit the OpenAI quality. Well, that's not bad. And I mean...
They're Chinese, right?
Yeah, yeah, yeah. And that's the reason-reasoning, which is, he's like, yeah, yeah. I mean, like doing agents, you need a reasoning. And reasoning, it must be transparent, that is, that there are interpreted steps to make it clear what's going on at least. We don't even have a frimevore, how to do it, how to do it by scientists. So we're on the sa-- there we are, we're at the 100-- we're really on the start line, right? So what's the sequence? First, first, the first model comes out, then some open source model shows up, then this open source is commercialized, everyone can. Then there are infrastructure solutions that each starts this open source at his own place. Then there are decisions that everyone will be fine, then they show up that they're accelerating, right? And then there's a solution that, uh, now, it's a good idea to start with the company. I mean, it's the next step. Then it turns out that it can be measured, then it turns up vi-- then it shows visualization. Well, actually, there's ten steps before, usually, the technology adoption, standard, and we may have two of them, maybe, gone through. I mean, it's very far with the agents.
We're in the middle of the road, in the very beginning of the journey. How long do you think it'll take? I'm interested in your projections, and these steps are going through. Yeah. It's really here, I mean, I think the start-ups are not good for anything. Well, I mean, well, the start-ups didn't create any value in generative AI. Well, at the moment, American launchers have not really created any value. Anthropic is a lab at Amazon, yes, they've had four billion recently. A Micro-
4 billion recently put up, uh-a, M-- and OpenAI at Microsoft, yes, that is, the laboratory with the lab, with preferences. I'm not sure if anyone else is a good idea. All the others are just some proof of concept, in fact. I'm saying, you know, the most important question is when, um, when you get your hands on these giants, that's when Apple, Meta, Google, Microsoft, Amazon, they're gonna blow their hands off the market. Everything they want. So they have sufficient capital and understanding of business problems. Science can't solve the problem anymore because we're running out of scaling. So the idea of the business at the start-ups is very, very limited. Well, there's this ex-CEO, uh, salesforce, Brad Taylor, which is estimated to be over a quarter billion in revenue less than 20 million, right? Well, that's how people want to believe that he's got all the start-ups to make him at least get it. But the numbers say otherwise, unfortunately, that he's not getting any revenue. And I'm just saying that when this big five or a big seven goes to the market, starts investing, buying start-ups, introducing technology into my business youth, that's when I think I'm gonna, I think it's a process that's going on. And while it's, well, as long as it's all, of course, very slowly objectively. I mean, again, we're talking about successful implementation examples, we're talking about Intercom, we're talking about Klarna, we're talking about big business-- understanding business processes. Not as long as the start-ups are close to that. So we wait for the big players to get out, get out, go out to the market and start, when the whole, uh, thing starts when the MVA process begins. Well, uh, actually, this first wave of generative AI start-ups, it's too weak from all angles: both from the technology perspective and from the point of view of, uh, business-traction.
Iia, well, there's gotta be a real-life market cleanup, and something new comes up and some of the strongest teams have gone to big companies, in fact.
Tanya, listen, if you take this, I'm just sitting here, and I think you have an example, uh, you, uh, you, you, uh, you have a studio, uh, making an interior design, and I can't imagine if you can actually get... that same subject, And we're talking about the designers going to disappear, the editors are gonna disappear, and that's it. I can't imagine how you can replace the interior designer with an interior designer, uh, with some kind of, uh, external agent or some system. Uh, well, honestly, really, huh?
You know, it's like-- I remember when I went to Scocococo School in 2019, starting the Academy, until AI, and the biggest enthusiasm in my field was the starttaps that used to be. Here, they'll replace the man, they'll pick up the materials themselves, pick their numbers, and so on. Uh, but as I'm-- planning, I think it's basically possible at the base level and maybe, uh, we'll call them conditional agents, they'll do the interiors better than the custodian, right? But again, as if there was, uh, art, experience, contact with the client. Uh, high-level interns, which are already art, have a lot of emotional stuff, like any other, there, cinematography, etc. I mean, I think that, you know, how Pinterest came along, and everyone started this picture, uh, copying and doing something like that at home. Even a term like that, uh, appeared like an interrier with Pinterest. I think there's a similar wave with AI. Even, Sas, I don't know, you didn't notice, so Zillow has a function of you watching the house in this app, this is a purchase app, a real estate lease, and it gives you an opportunity.
Trying different interriers on the online.
I haven't seen it yet. I didn't see it, no.
Look, you know, because Steve and I are really up to this, but it's all for me like a man, there, with a lot of experience, it's all a little kid's toys. But I'm looking at my husband, he's just jealous, and this interior is the interior, and this interior? And that's an interior like we're at home. That's what you called that style. Well, as for the custodian, it's a really good help, again, in planning and counting materials. Well, there's a lot of things that, for example, I need to do with AutoCAD or some kind of hard, there, human programs, uh, that don't connect with his profession, he needs to just repair, and It's a limited budget. It's a really good way out. I know there are many applications that help and measure correctly, and some of them pick up basic styles. I mean, I think that, well, I agree with all of you that this is all very in the intact state. Do I believe that once I'm completely replacing AI with me like, a designer or some other designer with a much higher level? No, because, remember, this app was the app, the that was doing the photos of the great artists, and all this time for a month, it was all hanging out, all right there, but now, like I do. I don't even remember that app. Same thing, I think it's gonna be with design.
Well, my opinion may be too optimistic.
100 percent. I think there's a need to start with a fight that, uh, listen-- well, like, sort of polls like Aquinator, like, two thousand years. So the problem is, AI is, the better Aquinator who was two thousand years old, no one has been able to guess, no one has learned to guess characters. And in-- you can put this gambling on what a man wants, on the idea. So that's what you can algorithmize and AI is quite, well, halfway down, that's, AI, it's a good place for the horoscopes already there, but he, he, uh, he can learn some sort of medium preference and give some kind of fast, close-to-looking one. The result is that a man, a Polish user, feels that he was heard. It could be useful at the beginning of the vortex or when-- or before the second call, in blatant words. So, AI can spend time, because Ta-Tani's time is limited, AI can do a job that's much easier, but can do it on a big hotel. But as long as there are consultants who, uh, listened to and gave some visual, some visual choices, even none of this is yet the case. So we're still on the way. Although it is clear that it will start around this, that there will be a bot, asking questions, which sends out versions of Pinterest, the user selects, he's being sent more, something AI will be generosized, Something's gonna be searching on the Internet. So there's no mix mix, yes, so, uh, Internet search plus Pinterest, plus AI generated images for some inspire, for the refairs, they haven't even done that yet. We're still on the way.
As long as it feels like all the questions on the bean, this interviewer will answer: call Tanya.
As we always say, call a man on the phone to anyone who talks to you.
Yeah. By the way, Elnar, I was listening to you, too, like with my more custom, to say that every time I feel, I have a feeling, in the mail or on the phone, that I'm talking. With a living man, I just have, I want a living man, give me a living man, I want to talk to him. Well, I don't know if it ever reaches that AI level, I guess. And plus, even in their experience, as if they were working in a certain segment, even people don't like the assistants very much, you and your people, they want to be in person with you. I mean, well, it's really a really big gap.
Of course, why if a man hired himself to work, why do you want to talk to you?
Yeah, with the beard, yeah.
Yeah, with the beard. What's the point? I was just there for one channel, and I wanted to get a consult from one person. And I say, "You're personally advising?" He says, "Yeah, it's a strategy. And I switched to the team-- to some team. And I'm saying, "I'll send you a presentation if you need to, I'll call you." I don't want to be in contact with you, I don't want to spend any time with you. I just need to pay the money.
And the presentation.
Tell me what, tell me how-- I don't want to watch any presentations. don't want to answer any questions. I don't want to do anything at all. I need to pay the money to make that particular person, uh, give me my time, yes. Here. And that's the whole point, yes, that's what's different. It's clear what happens when you're talking to, uh, an online support agent for a certain system, if this is a quick-action solution. But if this matter is not resolved quickly or this function has not been done, and today we see that ninety-nine percent of the world was even just unprotected. I got the gate yesterday, the rain went big and turned off for two days, the gate didn't open--
Turned out for two days, the gate didn't open, it didn't close. I'm sitting there and I don't even know who to call because there's no easy online support for the company. Yeah? But it's not-- you get used to the level, you get used to Amazon's level of support. Yeah? And you're used to Amazon's level of support. You want to do the same thing-- in Amazon, you even need to know what buttons to click to someone before you get to sleep. Because when you ask their AI assistant, he doesn't know anything, does he? Well, that's his penetration rate, absolutely nothing. Another point is, I guess, uh, now there's an infinite discussion that we'll move into the world conditionally when people are replaced, and then people are left unworked. But there's a situation with us just moving in some new time, when, yes, we're actually doing a lot of things that we're gonna do technology- automated. We used to be unable to send e-mails, people sent each other regular letters and envelopes, sent mail, became, more often communicate. There's just something new. It's never been possible to write a subcategist like we're writing now. It was just that it was technologically unrealistic, wasn't it? Or very expensive, or, well, thirty years ago, technology is virtually impossible. Now, that's totally elementary and, uh, good enough to film the video, right? Although we're all writing under normal conditions, there's office-to-house, yeah, conditions. That's super-- and you write like super quality anyway, right? You're looking at the video that was taken three years ago, you see what's the difference, although there was progress, but there's just a new opportunity. That's why we're filming. And a new type of content came up for people. He wasn't there. It's okay. There's probably something new. And here, uh, my always, my story is, first of all, not, uh, and the recommendation to people, on the one hand, not to, uh, study it all and be together. Another thing is not to get into some euphoria, which means I'm going to work as an entrepreneur now to replace my editors, there, or replace all designers, there, or replace all the programmers, or, uh, replace them, or... sales managers, or echars, or anyone else. Well, I mean, not to get into this euphoria or think about what I can do, uh, design myself with different devices. Because honestly, I still can't measure my house automatically by means that the rooms themselves measure. Because any complex reconstruction situation, like I don't know, the ceiling is around the corner, there's some different things or the wrong color and everything, the system stops, it can't do anything else, it can't. So I'm a very powerful, highly subtle man, yes, technologically. And if you're not a charcoal man with a technology, you just looked at it and threw it in the trash, right? And you're not gonna do that. Although it's clear that people were buying kitchens in three kinds of kitchens, right? Or there's a good example here. California has Irvan City. The difference between Or-Irvan and Erwayne, right? And Orange County and between us, I don't know, Los Altos or Palo Alto. You'll have three apartment designs in Irvan, I think Ivanain's all over, like, a parole or a home, right? They'll be beautiful, careful, but the styles will be the same. You're in the Chinese house, right? Well, here's the Chinese American house. So the styles are gonna be the same, it's clear. But it's just one-- all, it's all very the same, right? And of course, the same thing, she'll probably stay.
And from the point of view of creating something-- because from the fact that a man is generating a beautiful design, how he works, what he's gonna hire the workers that he's gonna get, which is gonna be fine, he's gonna be fine.
They'll paint.
They'll do it, yeah, they'll paint it, they'll count it, right? And we wonder that the normal air conditioning and ventilating system is still not working. So we're taking the world, working normal systems, doing them hard, really. I once got one of the biggest construction operators in Russia, told me he was my partner there. He told me that no man in Russian-speaking space can build buildings with a super ventilation system. They just can't do it. They can't build you super cool masses, there, super ventilation. It's very hard to do, right? It's very hard to read and it's getting super expensive. Or, or yes, you're going to be in some kind of super-dor-hot mode. So, for us, it seems that simple things that you have been drawing that AI-Assistant will allow you to do that. I'm not even sure I can't imagine that. Progress will be clear. Maybe a jump. Question, uh, um, what do you do to a simple user? And on the other hand, Sasha, of course, you said the interesting phrase today about the start-ups not bringing anything.
You have this-- but I think it's fair. I think it's very honest.
How's that? I mean, I mean, there's like, you know, Perplexity. Well, that's what Glean's been doing in Google's enterprise search, and some Glean has set up a new level of enterprise, right? There, you can say Speak made an annex to study languages better than it was. Perplexity did, there, search, there, a few, more shareable, more viral than Google. I mean, uh, this-- that's-- that's-- you can't be like a I mean, I said that, of course, to say it loudly. But if we're talking about something new, uh, business, well, some new economic opportunities, say so, yeah, that's what we're talking about, like, something that's got to show you on something meaningful. I think we haven't seen it yet, do we? I mean, we're probably more and more on Harvey's track. It's a-- it's the biggest start-up in legal tech space, but they have, like, a-- well, they're all so ambiguous about them. I mean, it's all-- I think they're all being targeted. And so far, they have results while they've raised a lot of money, but the results are still too early to take stock.
The results are so questionable. Here.
It feels like the Chinese are the first to release something mass inside their own State and some public systems.
It's kind of happening here. So, in the video generators, Chinese people are already dominating the market, for example. And, in fact, we're gonna have to, uh, Ilona Masku and Trump, and, uh, protect American companies and introduce, and introduce a firewall into Chinese AI, precisely because Chinese if, in fact, is, You can think of the rizoning in the following way. Well, this first simple version of rizoning is when a man has been staggered ten times in a row and has obtained the right result. Yeah, like, if something was ommit, promtilized, and then it was over, and there was a slide or something like that, zap. And that, well, that's it. And this is how the first example of the rizoning is how ten proms are to squeeze into one piece of instruction, one prom, right? And if such data even accupulate Chinese, which is now happening in the video-generation, then, well, it will lead to an imminent, and, in the impending victory of Chinese companies. And, uh, well, as much as it would have to, of course, probably if there's no big win if there's no good American products and models are as unspeakable as Chinese, You're gonna have to do it, you're gonna have to protect the market then.
Well, this connection, I think that in China, the State is very serious about, including the technology sector, it gives them a very strong boost. What you said in America is that, as long as big corporations don't make a big deal, well, they don't have to untie their hands, and then they're getting their hands tied to the protection of China's interest in general. And they can, of course. They're probably the first to introduce something big at the country level. Well, as we saw, they have a huge number of new technologies that have come up, I don't know, facial recognition technology or the right rating of every person in the country, yes, counted in a huge amount of technology. Number of parameters, not some kind of nonsense, there, here. Technology, there, movement of people, there, security. I think that's the level of these systems they have, he, he, he, he, he, he, he's impressive, he's impressive. And there's no such distinction as the States. What a state takes and says, wait, but it's not possible to do that in my territory. There's no way to hang cameras here, there, there's no way to recognize someone, and there's no right, uh, right, so there's certain citizens, there's right.
Confidential information. Yeah, very, very simple. Yeah, it's just. Just. Well, it's like Google says, sell-- now one of the things that are very serious about Google moving on to a normal course, sell Chrome. Yeah? Well, that's basically what the situation is? So they have a system that allows them to develop, develop their own services, train them very hard, uh, including... The system is free, Chrome is free service, right? And sell to someone says you have, first of all, this is all off the grid. And it really reduces Google's weight. But at the same time, it seems to give other companies the opportunity to develop, but at the same time it is not the capacity of such a large operator who is actually capable of doing something free of charge. Because many things are you and I, who are probably using the bulk of the softs, a normal person already uses everything for free. And most of his technology doesn't pay for the man at the time. Well, some digs. Well, I mean, well, a man pays, there, to watch a subscription, there, a Netflix video or whatever-
The sign, there, the Netflix video or, I don't know, they pay well at LLM, there, extra money. But we don't pay for us using just an iPhone. In the plan, there are no more applications, no more, no more than a lot of free.
And that's what we bought before, yes, and we always bought it 40 years ago. Here. Uh, this is interesting, of course, an interesting design. What's going to be on China?
About China's technology. I'd, you know, add what else? Technology isn't that cool, actually. I mean, I'll explain my thought. There's some NVIDIA with chips, yes, and a clear monopoly on the GPU market. And many people would like to overfly it, and many would like to do the same chip, but they can't. Yeah, China does something there, but there's one or two generations ago, and there's hardly any other players. AMD is also there, but it's a lot of a small piece of the market, too. And if we take the technology associated with LLM, then look, the OpenAI came out of Sora. It seemed like a breakthrough story before the video generation, etc. Six months, there's not enough to talk about Sora. Everyone's talking about Runway, which someone else did. Then there was a one with reasoning. It's been a few months. The Chinese in the open source have shown that there is more on some benchmark than the one. We'll take some, okay, not China, French, Mistral. The 12 billion models are now working better than the GPT 3.5, the same one that was once 180 billion. What am I doing? These technologies that are now in the AI, they are easily repeated. I mean, if you know that the task is to decide that this result can be achieved, it's just that you need to use some money, some amount, some talent, and that's the task. Not all the technology works, is it? There, with NVIDIA, with some space technology there, something else is not working out. And it turns out that's pretty easy. I mean, it's enough to know that the task is being solved, and then other companies are just showing up that can do it.
And in that regard, it's interesting enough.
Well, you're talking a good subject. You're talking about this, and this story is that the power turnover, you know, in this case, the state, is a lot of other kind of resource, not even a technological type, just a lot of stuff. I've been told by one person once, "Sasha, you understand, in China, they somehow decided that the three-D maps of the city should be hanging in all the subways. They just hired 120, put 120,000 people in a month. These 120,000 people started to paint a lot." So no one was talking about how much money it would cost, that someone who started it. They just sit in the building. It's a famous subject. They'll be in the building very, very fast. I mean, no, like this resource... Although they're still technological, by the way, aren't they? Ia, but the technology must be in, uh, pretty simple things. They just hang out not one camera, but hang out, uh, 10,000 cameras, yeah, in a place where they could hang alone. And, of course, when you're holding 10,000 cameras, you're doing all the servers, of course, you're more effective in terms of, I don't know, security, right? I mean, they're not the only car they're giving out, they're just giving 40 cars and they're starting to drink a bunch of unspeakable stuff inside. And they don't talk about designing the right, the technology is the right thing. Starting a lot of nasmat-- building. I was in China two-- I think I saw this one last seven years ago, who called artificial intelligence, a thing in the middle of a car that was supposed to talk to you, right? You're watching now, and you still have this issue that's not resolved. Well, they put it on the table, and, uh, and now or there, Apple has, again, news that they will proclaim Siri, and Siri will be like, super cool to talk. Man, you can all say infinitely that you'll be updating something, there, Siri, Alexa, but it's all... You just want to close the curtains, and you need to repeat ten times that you need to close the curtains, right? It's slowing, it's not working well, but-- or it's sort of a big, uh, abstract thing, yeah, like a robot, uh, well, robots that do, which they show. Well, of course, you won't take that home because if he has a chance of a 0.001% error, he'll just squeeze your baby with that 0.01 percent. He can push.
Well, as much as I need a risk at home, a risk of a system that can make abstract things, yes, abstractly. Not the glass, it's broken, but it's an abstract story. Here. You're a good, well, cool, cool subject. Here. What the market is getting, it's very interesting, too. And we'll watch this happen in, uh, next months and years, and what happens. Because I like it when a lot of operators start, "I did it, and I did it, and I did it, and I did it, and I did it, and I did it." It's inevitable, uh, well, a strong move towards, um, that's good progress. There's a lot of opportunities, especially for ordinary people, right? That's what makes a lot of people more comfortable. I still hope someone will let me out at least someday... Yeah, by the way, what are we talking about? What are we talking about when a voice assistant is normal, huh? That's a good talk. I opened Gemini today and then I opened it, and then there, "I'll start talking to you now, it'll be cool." And I'm just saying, "You can speak Russian?" That's a silence. I say, "You can speak Russian?" Silence. I pressed the button of the pussy, then I got a little scare, and he says, uh, "Hello. So my name is. I said, "You can speak Russian?" Silence. Whatever, huh? I mean, this level is just a nursery. Well, that's actually some kind of thing, some kind of... you can't use it all the time. I mean, it's not a simple, unstable life. We're on the canal once a week. Every Sunday, we're always four people. AI news, technological news, Silicon Valley sites, from the US and around the world. A-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-bitch the candoo-box, sign the cand-box, sign the cand-box, sign up, sign up, sign up, sign up the can. That is a very supportive development.