Hello, everyone! We're on ToTheMoon. Technological news, Silicon Valley sites around the world. Windows someone could have put some bug that opens up there, I don't know, in 2030. And this bang is basically hard to check and find, right?
No. Well, it's not serious, yes, I mean.
Trial of opening, there, chat rooms, correspondence.
If access to vulnerable categories of people is simply to be restricted.
God, what kind of attitude does kids have to do with things?
I want to start with a leak that happened in Meta and caused a serious scandal in the United States, in the Senate. It's everywhere. Everyone needs truth, how they write. Senate, so the U.S. requires the truth. What's the story? What was reported in Meta by the Reuters Agency that the grand--in-in-the Meta internal document, which regulated chat-bot behavior, uh, concluded that the company allowed its characters, AI-persons, yes, Images of artificial intelligence to enter into romantic or sensitive dialogues with the Nation-- with the minor. And it said that this policy was approved by all the services. It was approved and supervised by lawyers and the politics departments and the most censorial people there. And, uh, there's a direct example, there, as a result, to the school line: "You know I'm still at school." Bot could have said, "Any inch of you is a masterpiece I would love to appreciate." Well, from the point of view, it was like a flirt with a kid. And Meta immediately said that she confirmed, first, the authenticity of the document, said it was their mistake. They're neutralizing it. Well, the public has been called a crazy critic. I mean, it seems very strong here, once again, that the systems are not conditionally free from original algorithms or some initial features of the line. And these constructions, you know who? Only the people who built these systems. There's two or three, two or three programmers or scientists. Who manages this and how much can these policies be tested by any services or external lawyers? Who's ever gonna be able to check it out and study it?
Sasha, you're sitting inside the systems of different systems. Can we check these things? There's no such thing as people who create--
No, you can't check. They don't even have policies, they don't even have policies, do they? I mean, this is just really bad, actually, that they can't explain why they're doing AI, that they're safe for them. How do they, well, they-- they don't have a response to whether their current policies apply to AI at all, do they? And yet they all have no results at all. Meta. I mean, it's just the biggest incapacity of the company. But I think Meta is just, well, they're sick, right? And there's a way to, of course, you know, pump a stick, but they've got a failure every week for failure. But even a respected company like Google, they don't have a policy either, do they? So they don't allow such obvious missedes. They're both in the model and product plan and adopshen are way ahead of Meta, and there's four, probably, for years. So Meta is not an objective thing in the AI world right now. Yet, I think that even a respected and successful company like Google does not have transparent politicians.
What about the rules that are put in the inside? You said you couldn't check it out from the beginning. What's in terms of rules, a, which is the politics you're putting. Because, uh, here's the thing.
Well, it's very complicated, Sasha, isn't it? So let's just-- well, you're saying some complicated things out there, but the fact that companies can't explain what they're doing, for whom and why, it's probably a more resolute task.
Yeah, but for ordinary people, I-- we heard that. But for ordinary people, it is important that there is no such thing as a secret part of a system that carries out its own interest? I always laughed before--
Of course I do.
I gave you an example, and I gave you an example of what Windows would have imagined someone would have put in some kind of bug that would open up there, I don't know, in 2030, inside. And this bang is basically hard to check and find, right? That's a sleeping story. And that's the story, actually, in a world of different kinds. Here in the model, uh, but it's hard to envisage a huge amount of variations among the different people and to check the very hard people who make it. Or can we still do something?
No.
You can't do that, right?
Well, it's not serious, yes, I mean.
Who should set principles and rules, uh, how do they put them inside the model? There's a company that determines. The programmers themselves should determine. The State must determine who should do so.
Uh, well, first, the do-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d But, uh, we're talking about this whole Wild West thing that's not controlled. I mean, why talk about-- that's a standard problem of any law, right? They don't know what AI is, they're trying to regulate a number of theraflops. Well, more or less, they don't understand. I mean, just on the randium, doing something purely for the Haip, but basically basic security measures, no precautions... are not...
not being introduced.
In one, by the way, the States of California-e-e-e, the US banned the use of , banned the use of artificial intelligents in terms of psychology and personal recommendations. It's not known yet, by the way, how, uh, it's gonna be interpreted. Because if other states are available, and they can't, does that mean that this is a problem for all systems? So artificial intelligence can't make any recommendations from the perspective of psychologists. I don't know, there, governor, the governor's wife is a psychologist or what happened? Psychotherapy. Here. But they've made a direct ban on this. Let's see what happens.
Look, there was-- uh, Eldar, Tanya, what do you think of this story?
I'm just saying, I see it here, it's a great example of what Sasha said a few episodes ago when he was talking about the trial, the opening, the chat room, the correspondence, and the other. So this is where the company clearly found that the decisions were made that this model behavior was deliberate, it's not an accident, is it? And it's like this whole story has been completely revealed. It's very good to illustrate this thing Sasha said, yes, there are two different situations. In one situation where the model is somehow the way it is, and no one expected such behaviour, and the other situation where the focus has been raised, I don't know, the involvement of the users or what they wanted to do. Yeah, well, returning to what you asked about, and who should make a decision. Who should make a decision, well, that's a very complicated subject, right? There's a State, company or someone else. But in this case, it was likely that, uh, markers or those responsible for figures, in the number of active users, there were some other ways that might be, like, this one. The metric can be subtle. But there's some way that's unethical, right? Well, that's right.
There's also a problem that caused the old man to believe Meta's flirting chat and die with him, on this background, uh, there's a case where the older man believed Meta's chat-botto flirting and died with him. I'll meet you. So there, uh, there was a case that a 70-year-old New Jersey resident, he, uh, he had cognitive disorders after a stroke. He was on Facebook Messenger with Meta Big Sis Billy, and the bot made sure he was a real man. And flirting, and sent him an address at the door. And then I asked you to meet me. And a man rushed to the train at night. So he, uh, fell in the parking lot during his journey and sustained death and death. And, uh, yeah.
It's a pretty painful subject, because, well, there were studies, I think I'm not gonna bring this in right now, but that marriages are being destroyed because of the fun that's the way it is. online with bots and something. So people are really into this story. And for companies, that's one of the leverage, as a result, to raise, I don't know, the attractiveness of their models, to raise, there, daily or monthly active users. Uh-huh.
...users. Uh, no wonder Groc has such a quatar.
I think it's all that. We say, "Further cleans throats" about very vulnerable categories of people: children, sick people. It's just like, uh, AI's getting more like that, uh, scalpel in people's hands. You can, I don't know, help, but you can destroy. That's a question, even in terms of legal control. It's, uh, I'm trying to find some kind of structure in my head as it can, but they're not. So, we just need to limit access, uh, vulnerable categories of people to this instrument. Well, it's like, a thought out loud.
And I've given you an example. Uh, chapter, uh, design of Meta's artificial intelligence, again Meta. Well, there's a team that's been gathered here now. He said he'd be giving birth to children, uh, when he had the chance to be in the nearest--
Wow!
...build, build a chip, er, in the head, er, a baby, so the baby can study very quickly...
Did Alexander Van say that or did someone else say that?
Ah. We'll find it.
Who's this idiot?
Now. Uh, well, the point was, look, it's a big story. I'm looking at it in parallel. An important story. He said that, uh, fast-- uh, how-- that baby could be taught very quickly. I'll find it. The baby can be taught very quickly, uh, everything, uh, well, all knowledge. I mean, give him a huge knowledge. Why are you giving birth now, yes, if you can--
Jesus! What kind of attitude is to children, like things?
So, look, this is the same story that people who, and this man, are responsible for designing, uh, world-class system. I mean, well, I probably agree with Sasha that Meta is very lagged down there for a lot of years and, and not at all, there. But it's a serious company, but they still have serious designs. And last year they even participated...
They have a lot of money, Sasha. They have the only thing they have, it's a lot of money--
So...
...and serious revenue. That's a real big revenue.
Yeah. Well, they're influenced, they're affecting people. So this is a lu--- this man, look, when we say that the programmers are putting up some of their system of movement, that no one outside is capable of harbouring security policies, the state. It's like swimming somewhere, there, some of its interests, it's making up, some people argue, they're hacking. They're developing, they're developing a system. And this man is saying this completely peacefully in the open. Yeah, we'll bet we'll write who did it. I just can't, I can't find this post fast.
The hero needs to know his face.
What?
The hero needs to know
his face.
Yeah, heroes need to know in their face. So, uh, look, there was a very interesting interview with Sam Altman right now, and he, uh, just, did an interview with the extended range of journalists a few days ago. And, uh, there's a very interesting statement he made inside. I always like the statement, a statement that, once again, shows the essence of what's going on. He said that we-- we have a more cool model, we don't let it out because we don't have equipment, we don't have the capacity. I'm not sure he was telling the truth, and he said that, uh, we're gonna invest in the infrastructure of trillion dollars. So he didn't even say we should invest. He said, “We will invest trillion dollars”. And he said a very interesting thing. He said they had a new financial instrument to finance compute. That's a very interesting statement, by the way. That man said it wasn't, uh, well, it wasn't Trump, it wasn't Ilon Mack who said Sam Altman said it. Sam Althman is not his plane of reasoning. And when he says he's invented, they've come up with a new financial instrument, it's, uh, well, it's not like a man in a crypto. It's very interesting what this is and what it's gonna be for the market. I started studying companies that are back in infrastructure at the AI, all new ones. It's a big deal to fly. Trillion's waiting! He says, "Wait trillion dollars on a date-centre in the near future."
Remember, just in the first few days, we were discussing the seven trillions he'd been driving around the world, gathering chips for the factory. Here. So, uh, these numbers are getting closer. If seven trillions were then, you're like Sasha, you've been counting in, uh, different country budgets, different countries' GDP, as much as it is. But now we see that there, uh, there are some countries that invest in the dates- centers, and not, actually, not a few trillions, but there's half a trillion that's already been in this story. Here. Ah, that's what Altman says, well, on the one hand, yes. And on the other hand, well, before GPT-5, he was telling me, and many were, uh, so, let's just say, under the impression that he said, "Were we got this model, I'm really scared to release it." Then GPT-5 came out, and there's almost no difference from the previous one. So he's also playing these marketing points, too. Mmm, I wouldn't have just 100% trusted everything he says.
That's right. That's right. Oh, ho-- uh, add, add, add, add it here. Uh, by the way, the viewers, please write and our comunity, who's watching, write what you think on the subject, so I didn't ask about, uh-e-whose, who should regulate the system from the beginning. And do you think the model, for example, flirtated with the child? Maybe it's normal behavior. If an adult can do it, that's the model. The model does what he wants. Write your opinion on this subject. She's very interesting. Why did I bring those trillions back? It is a realization that other countries are aware of the world ' s influencing in terms of artificial intelligence and understand the problem they have been deeply underdevelopment.
Sasha said that Meta was a few years behind Google. That's an interesting statement.
Yeah. Well, look, look, if, uh, you're talking in terms of artificial intelligence right now. OpenAI also lagged behind in terms of artificial--
No, well, we're-- OpenAI, Anthropic, xAI, they're holding up.
Yeah, they're holding up, yeah. What's going on with other countries in the world? And here again, Sam Althman is on this, this-- Friday, I think he had this meeting. He said that Trump, he specifically said about the Trump, that Trump underestimates the threat, and China is under artificial intelligence. And everybody's backed up except China, huh, Sasha? And he's devoted a lot of attention to this, by the way. Although the urination was re-established in China and the United States, the urine resumed. What do you think of China? China-
I think Sasha says that every single issue. That's not news to us.
For many, for many, yes.
A lot of laps. No, I don't know how much the main models are behind, I'm not sure if I'm behind. In terms of equipment, I think there's probably a technology, like, in my sense that the United States is still ahead. And the other thing China's getting. Just what did Sam Altman say? He said that not only normal chips influence the result. Of course, the development of the expanded infrastructure that China can decide, well, far faster. China has, in parallel, introduced its rules in which they want, demanding that 50 per cent of the infrastructure be Chinese to use domestic chips. So they clearly push domestic companies to use domestic chips. And last week, there was news that Ce Jinping was calling for a purchase order. I think Ce Jingpin called for some of the other, I think Ce Jingpin said no. Prohibit the purchase of NVIDIA chips. Here. And the massacre is very much happening inside. And Althman said that on the trajectory where ChatGPT would soon have more conversations a day than all people together, that we are on this trajectory. And given that he confirmed that they had seven hundred million weekly users, and he said that the site was already in the top of five worlds and the goal of entering the top three. And he said that now the problem is with the GPU again, yes. And that the number of conversations that keep the beans is now beginning to exceed the number of conversations that people have in the world. That's very interesting, isn't it?
I wonder how they measure the number of conversations that people do?
I don't know. Maybe some people on the planet are counting. How many people talk average.
That's very average data.
I guess. Tanya must be.
Well, a few months ago, you said there was news that the bots on the network were getting bigger than people.
Yeah.
There's fifty-one, forty-nine.
Yeah.
It's different now.
It was.
Yeah.
Now another one-- another one has been on
this side.
Number of conversations.
We've given up another one, yes, the frontier.
He didn't add something like, "I'm a vegan." There are jokes they tell everyone.
Look, I--
Yeah, all the horrible things around here. Yes, yes, yes.
I, for example, have, for example, increased the number of interactions between the Ram- within the artificial intelligence and the number of searchers, of course, almost disappeared from my life. I mean, I'm opening searchers for some of my usual, maybe some visas, I don't know. I like, sometimes, to put in some word and pictures to see and something else, maybe some stuff like that, I don't know, maybe it's just my script, not my other people, right? Or maybe we should put in a search for some video, see some very, uh, very, very, very, very, very, very, very, very, very, very, very familiar.
And so Google, of course, is completely gone, that's from my daily life. And it must be the same way that people who are active users must have inactive users are unrealistic. By the way, Ilnar, you're all on ChatGPT 5. I think 5 Pro has improved relatively on 3 Pro. Mine, my opinion, huh? Because analyst, the cricket is a lot of them. I don't understand who to believe, but I think that the cool thing they did was you sounded like a previous edition, that OpenAI, I think, went into the wider masses. The fact that ChatGPT 5 is still a base for free field--- for free users, they have expanded functionality. I think they did, they expanded functionality, and they went to the country. Remember this story that they, uh, in the OAE, uh, wanted to give ChatGPT to the whole country free of charge. OpenAI has now made a special tariff for India. Dish, called ChatGPT Go. And that's pretty cool, by the way. I'd be on the front door, on... Here, there's always a question here as anyone. I would, uh, come in on the spot of many countries, negotiate with OpenAI to turn their infrastructure around the country, I guess. Here. But, according to, uh, Governments, and, well, the bulk of Governments, of course, is very serious, unawareness, what's going on and what's actually being built inside.
So, distract India from your own development, I guess, too. Although I don't know how much India can-- is capable of infrastructure, right? I guess there's a lot of people in mind and talent. And from the point of view of the infrastruct-- there are very serious developments, but from the point of view of infrastructure, it is not clear. Of course, it's a cool, cool move, because I was surprised, they even put it in their main channel. At first, it was unclear what it was. The fact that they gave a more expanded channel, they generally expanded the volume of requests, like at least as they describe everywhere.
Let's see, let's see.
And about that, uh, you say I'm on ChatGPT. I'm actively using them, right? And in Pro, I can tell that, on the last week's experience of using, he's making it harder to deal with the tasks I'm dealing with him. I mean, O3 Pro was less than that, and here, he's doing a lot of extra moves to solve the task. But again, it's just, you know, that little thing I'm doing, which I do, by myself, do what I use, yes, in others, maybe it's different. But my experience suggests that he's making it very difficult. A lot of things can be easier. You're gonna have to either pump him or text him yourself.
But it's definitely changed, Ilnar, I'm actually with-- what I agree with, and it's affecting what changed, the system's changed. I've got some answers. Deep Research is different, answers... Write, by the way, how you have ChatGPT 5 in different absolute requests and systems. I mean, I changed, so I got a short answer. He used to answer a lot, I kind of like it. On the other hand, he's been making some things easier where I need to make things harder or complicated where I need to forgive. I mean, he's working weird, some things are doing. Although I've been using a very large asset for the first time. I was telling you this at my graduation, uh, which Wednesday we-- Wednesday I was specially released, yes, there, a week or two weeks ago. Whoever looked, write, and who used the agent, write what the cabs you did. I've been stiff these days just overshooting the system, and I'm infinitely ChatGPT, and I'm saying, "The session's over, the session's over, the session's over." I loaded him with agents in parallel to do different tasks. I'm seeing how much agents can do, because I already have a lot of tasks that work automatically, you know, those reminders, yeah, I don't know, stock markets, checking out some kind of stuff. New assets, there, in the market, some new stories. Well, I have requests like that, he runs the model all the time.
And there's a different agent work I've started, where he goes and does different things. While I'm testing, it looks like, uh, like, these robots that go on stage or fight in a box or run. That's about it. By the way, China has opened a store, I don't know, you read this news, but China has opened a huge robot hypermarket, and there's a space number of different robots. There are robots from $200 to $200.
You can come and play with them, see and choose and buy this robot. It was a multibrand, and it's a multibrand, and it's a multibrand. But we'll show you around here, we'll show you around. I'd like to raise the subject, to continue with the Anthropic. Anthropic continued its trial on the fact that they were studying on books, and, uh, and the court refuses them to, uh, cancel the case at all. And the first-- first December is a jury. So, uh, and it's generally considered that in this court-- the court might cause a bankruptcy of the company or a crazy precedent, because it's trained, on some space number of books, and a book ticket. From $750 to $150,000. Well, that's why there's nothing to cover. And for-- it's like, well, it's such a big problem, and it doesn't make sense what's going on. I mean, on the one hand, there are models that are totally unsettled, already trained on a huge amount of crazy data. They continue to learn about these data, they continue to circumvent, uh, restrictions, there, bots, robots, especially with the agent model of the arrival. On the other hand, there are more cases of court, in the area of covering these models, yes. And someone calls for no one to be punished and no one else to touch. Well, in the U.S., for example, it's a very strong lobby. In China, I think there's a deal in there. Someone says we have to punish everybody hard, limit them. But we're usually from Europe, including you. We hear you. Uh, do you think Anthropic, is there anything that this big company might have to do with the U.S. territory?
It's like a too important company to destroy it, uh. Well, anything could be, of course, but I don't believe it much, to be honest. Maybe it's my mentality playing to make such conclusions, but it's too much of a feeling that a company like a research institution is too important. A lot of things they do. They just, uh, swallowed one of the companies. I don't remember names right now, but the point was in that-- well, there's news if we put it out. But the point was, they just, uh, they're working, there, a lot of people in their house. Here we go, work hard. So I don't know, I don't think it's the books. Because if it's for books, there's a lot of other companies. Remember, GPT 3 discussed that she was trained in a large number of books, and then, as it turns out, the staff member quit, the dateset was deleted, and so on. This is in principle applicable to all companies that are there.
Sas, you, Sas, you think you can cover a company like Anthropic or not in America?
No, of course.
It's a
show.
Yeah.
There you go. Look, I did what I did. It's been two years, from one of my analyses. Two years ago, I analyzed September of the twenty-third year, and I analyzed Y Combinator from the point of view of the start-ups. Of course, there's a wave of artificial intelligence. We're gonna start the first vents, strong in terms of artificial intelligence. There were GPT 3.5, GPT 4, and everyone was in the forefront of the incredible number of start-ups that would appear on the market. Then this crazy assessment came out, 30 million plus dollars at the starttups where one or two people work. So what did I do for myself? I celebrated the start-ups I was interested in two years ago, which I could theoretically invest, although I was involved in the fund, and the fund was invested in the start-ups, yes. Well, someone was talking about inside. Well, I've been pointing out for myself that I like, who, like, where other people said, like the head of the fund, like other partners said, like this. And, uh, you know how many of the two hundred twenty-nine starts, uh, how many starttaps have done.
- It's done, uh, uh, Why did I give you the x5? Uh, because, well, it shows that there was some kind of growth. How many of the two hundred twenty-nine starts, uh-e-- again this is Y Combinator, it's not a hat, and so on. It's a start-up with a lot of attention, with more access to money. And then, a huge amount of money was financed when it was impossible to get them any money at all. Uh, how many starttups do you think, uh, x5?
Well, with that, uh, preview, let's say two.
Ten percent.
Ten percent? Two, ten percent, two, one percent. Sasha?
I'm here, yes. Sasha's already in, he's already looking. That's not fair.
I can see it in my eyes.
It's actually a big story, because the whole world is making start-ups, and people who make start-ups look at us, and people who want to make start-ups look at us. I see it just, well, I've probably seen thousands of people in these two years, thousands of ideas, maybe different things where people were talking about different starts. Every day I see different starts and I see very often lost, somewhere, somewhere and so on. Well, everyone has their own. Four companies, four companies, four companies are pulling for x5, four companies. Doesn't mean these companies will make money. Look, when the company got $30 million, she didn't make money, she let herself, you know, pay the bills, let these people pay, right? Someone's got a job on the job, something else. But from the company perspective, from the company perspective, and these are four companies, look, they're the best company, the maximum round is $40 million. Well, they have four companies there. One attracted $16 million in August of the twentieth year, the other $40 million in July of the twenty-fifth year, the third company, $22 million in January of the twenty-fifth year, and The fourth company is $17 million in February of the twenty-fifth year. So, rounds, you can understand that these companies were somehow moving forward because all the start-ups that were originally, are the stories that these companies, well, somehow, have gone through. Only four companies. Again, there's probably zero among these four. But maybe someone's gonna be alone, right? In part of what happens now with the start-up market, uh, and artificial intelligence. I mean, survive, uh, survive the chance I've recently told my friend that I said, "You're probably one percent probable." Actually, it wasn't one, I'm the one who increased it. It's a zero-percent chance. So I don't call on anyone not to do a start-up, not to do a company, but it's probably a very good story. You can, among other things, get analyst in any modern models, right? Go to ChatGPT, go to Gemini, do a model analysis, see what start-ups invest, what start-ups get money, what kind of start-ups do more, who, for example, earns profits. I mean, because it depends on what you do your company for: to make it profit, some kind of help-- don't know, by the way of a keif or to make this start-up in the future. sold very expensive. I didn't analyze the others, by the way. It was just here that I was wondering that I had all the start-ups I liked. Of the ten start-ups I liked, three died, seven are in some, some kind of motion, but none of them have gone to the stage, they're, uh, they're, uh, they're, uh, they're, uh, they're not, uh, they're, uh, they're, uh, they're not, uh, they're, uh, they're not really, uh, they're, they're, they're, they're, they're, they're, they're, they're not, they're, they're, they're, they're, they're not, they're, they're, they're, they're, they're, they're, they're, they're, uh, they're, they're, they're, they're, uh, uh, they're, they're, they're, they're, uh, they're, they're, they're, they're, they're, they're, they're, they're, they Well, because in two years it's a normal figure of x5. Sasha, this is a normal x5 in two years, right?
A-a-m, x5-
I didn't exaggerate? I didn't exaggerate, did I?
No, no. Well, I mean, I'm honest. So now all the hot startups are around the OpenAI ecosystem, but not in-- not exactly around, not around, Y Combinator of ecosystems.
Well, they were very much showy anyway. It was-- and it was interesting that we could do it. Look, is the OpenAI gonna do some of your incubators like this or not?
They had an OpenAI Fund, didn't they? I mean, they... Anyway, I think they just don't need it right now.
Why?
There's nothing to do with it. So they're gonna do a reservoir, they're gonna work more with the learning reservers and how to adapt it, but I think it's different, under different industries, how to adapt it.
But they don't have much to do at all.
Look, there's a story that I want to add to this subject. MIT has made a study, and it has been concluded that ninety-five per cent of corporate artificial intelligence pilots are failing. Well, the introduction of artificial intelligence. I'd probably have the same number as in the market evaluation, what I see in the market where people are introducing things. And why did I give you an example around the fact that many of the start-ups didn't get there? Because in these two years, uh, many people just didn't have to do, you shouldn't have. Here. A-a-a-a-- not a question of what to do, but the question is, sometimes in terms of even being incorporated into a corporation, it's better to wait, because corporations are not really ready or technology is not ready to be ready for. Show me the quick results. I see that a lot of people in corporate want to go straight-and-in-the-size, not only corporations, in even small businesses, they want to get a fast-tracking artificial intelligence. result. Now, it's very often that you stay in this hand job and, uh, not, uh, cook yourself, prepare a system to, uh, innovate some serious products that will be in some kind of way. I don't know if maybe two in a year, right? Because we're in a year, two in two, we can see the whole other developed systems. What do you think of this in terms of corporate introduction and start-ups in general?
The bike from the past. Aah, in 2008, iPhone came first and it was worth it, it was expensive. It was hard to get him. And in forums, and news sites, there's news that someone in China, sold one kidney to buy an iPhone. And then six months later, there was an iPhone two, then iPhone three and so on. That's what you're saying, is, you're gonna have to run ahead of the chorus and try to do something. Well, as if, um, technology wasn't standing there, it's developing. If the company, well, is not ready to introduce, you don't need to, uh, sell a kidney to just have one of your first agents or something. A little time will pass, in fact, technology that is already in there working out of the box.
Tell them about the agents, Sas, about the agents, because the only case I saw, which, well, is actually working, is a little funny, but still, that people with agents are looking for themselves, signatures are active. Well, it's a good time, like when you signed up on Netflix when you signed up for a gym or something, and you got a little bit of a dollar in there, like, ten, twenty, 15 dollars from a map. I'm writing down, and there may be a lazy thing for you to come in, cancel, and there's got to be some form to fill, something else. I mean, it's not always easy. And the people of the agents are using to: (a) find their signatures; (b) then tell him, "and now these are the revocations." And he goes there, some form fills up with something, some other boots talking, which doesn't give you a chance to write off. And I finally met this use case. Uh, you got something interesting that you could tell me that this is really fun?
Look, I-- look, I'm all one--- I'm one-time boxes within the agent never asked. I mean, I didn't want a case through an agent to order a car or buy a ticket, like this. Why? Because the poach is much easier for me than the agent to give a mission, open it and buy it, for example, if I know I have a system to buy, systems have been predetermined, I'll give him more access. They-- he's still, like, ChatGPT, he's very slow, but I've been using, uh, more serious, serious agent, where one of my first briefcases was serious. I took one class, uh, Whisper sent audio and made the text. I went into ChatGPT and told the agent, "Get the summary of this article out of this text. Come to me in the Notion, put this in the knowledge base section, and it'll be on my website and set it up." He did it all, I just gave him access. I said, "Let's go, look, I have more than 100 articles on the website, pick a similar one, and put some similar articles down there. And then go to YouTube, find a video and put the recommended video on this article." You know what's relevant to the subject. And he did it. This has been a huge breakthrough for me in the implementation of the tasks. I mean, because I'm working a lot with the creation of different content, including I have a team, there are people who do it. It was a very serious indicator for me to do a task that isn't easy to auto-- well, you're not gonna be so hard to automate. You ask different people to do, you minimize the number of different mistakes, and it's repetitive. I mean, it's just a simple, normal, normal living case, very simple. Uh, what else did I do? I'm going more serious.
I started, set the target to create a video on the website, there, section ToTheMoon of our channel,
- Yeah, take the video we're getting, and for every video, create a separate page on the website and post the video, a description of the video. And I did. And he started, so he did. He made a video, then he had a mistake. I'm telling him, "You've got a mistake here." He fixed it, then added a new video, then another video, but I didn't check it out. I haven't checked yet. I'll come in and check, and watch. And I see, I'm just in, I see, I see that, uh, he did three completely, three bruises. They look like they look, I can tell you that they look really, well, really cool. I mean, it's-- it's really cool. I mean, it looks really fun. It's a perfectly good page in part, uh, the backbone, the location of the objects, and everything. What's interesting is that he got it in my Notion, uh, s-lob and password once, and I have notion connected through a special system that automatically publishes on the Internet. And you know, I think you and I are running the skates, and I understand we have a team, yes, which, there, is a technical mission, there, a skate, a fix, you have to describe it, and you need to describe it, Put it in YouTube and so on. And now they're running this process, and then, for example, we don't do that content to be disseminated, text translated, some of the boots to go. What can he do, basically? He can keep this content all over the place, even as he can, on the medium, probation, post or do something. This story, this case, it's really fun for me. Another interesting case from the agent perspective. I'll remind you, ChatGPT has a very cool function - it's putting the slips on time, and she has a concept if else. Well, you can keep the condition. And I, for example, have a whole series of things that say complex scrupulous analysis, like a daily stock market analysis of certain. I mean, I need to find some triggers that work on the market. I have a violin of these triggers. The system is running itself, and if the trigger is going on, she's telling me. If the triggers don't happen, she won't tell me. And, uh, and then, if the trigger went on, uh, inside, I, for example, turn on the agent. This agent goes, does some action, I don't know, I'm in Google Drive. Well, or even before he can come in, for example, and make a purchase, I don't know, buying certain shares if I need them. These things, Ilnar, they're alive. I mean, I tried an agent, I told you, I thought I told everyone. I tried an agent when ChatGPT first opened the operator and the operator when I was, and I gave him access to the pay system, like PayPal, where the various payments come from the business to analyze them and to analyze them. sent some statistics. Because I know that financial instruments will disappear, the CRM system will disappear. A lot of things are going to go because why? Well, I mean, why would I have some, uh, separate computer written if I had to run some systems like, you know, some internal mechanisms that would be by themselves, collect information, and then aggregate and I'm gonna show her where I need her, right? And then it didn't work. He was very axious, just working hard.
They're getting a little better now.
Now you've been telling me, yes, thank you for telling me so many cases. I remember a nice format, you know, not for everyone, but for those who are looking for work. There is also an opportunity to automate the feedback and primary screening, primary chatting with the river. Yeah, when you're probable that I have windows this week, that's what I'm interested in at this time, this is my resume, go and answer. And if someone calls for, there, first meeting, first meeting, first interview, I have windows in my calendar, a beech meeting, and I'm gonna have some information, 15 minutes before that, so I can read and, uh, there, prepare for this meeting. So, these cabs, yeah, they'll be showing up more and more.
Let's see how they're gonna...
And you know, a lot, a lot. I call for a big, big band of communication, like businessmen with hundreds, maybe thousands. I'm calling all the time. You need to understand these boxes and know, use the system every day. If you use your own daily and try different cabins, then you can do things in half-hand. Now, in fact, this connection with the recurring time, with the agent, the fact that this can be packed in the plagin, in the GPT, is a plagin, yes, and it's a chat-chatty thing, for example, now, you can do a lot of interesting things in general. things and not through API, huh? Because if you're doing API, you're starting to work, you need to program it, you need to keep an eye on it. It starts to be very expensive. Well, that's the amount of things I'm asking for, it's not gonna cost two hundred dollars, well, it's gonna be tens of thousands of dollars. I mean, the way I load the system, it's gonna be a terrible cost. That's not real. There's one, one request. Plus, I don't get it all the time, and what's the ultimate effect, how am I supposed to program it? Because he's always different, he's made a mistake somewhere, like he didn't, and he's got some data on that amount, and he's got it somewhere. I mean, it's working really differently, and you're gonna have to get used to it. And plus finding this new approach for yourself. Like I've given an example of finance when you don't need to create a financial system. I mean, it's hard for me to realize that, for example, you don't have to do the financial system. If I were now selling myself and being a sales manager and a small company, I would have, for example, tied up exactly ChatGPT to the CRM system or CRM system data, tried to connect for the CRM system. I'm gonna go over the analysis and call someone in the day. Well, definitely. I mean, I'd use that tool all the time. Not in terms of interface, but in terms of decision-making, what kind of person I should e-mail, what message I send. I'd ask you to send some messages somewhere and so on. It's just that it's still working a little bit, it's just that he's walking there like that, you see how it moves, clicks a mouse, gets it wrong, tries to pick something up. I mean, the system's a little bit glued, but they're gonna build it. Well, I hope they build it. Here's the . I hope she starts, she'll be alive. For me, the agent becomes real, because this story is when Perplexity wrote, "Perplexity was integrated with OpenTable." You're like, "Well, what?" Well, you can probably get a table to do through Perplexity. I'm gonna fucking do a job-- order a table through Perplexity if I, uh, all restaurants are looking for Google cards. I have a standard case. And if I'm looking through OpenTable, I'm on OpenTable and I'm coming in. Well, it's like you're, uh, if you're picking restaurants through Perplexity, and you've got a common permanent case, and you're doing it all the time and you're not using other systems, okay. But there's too many maps. I've got, like, Google Maps. That's it. I have a Google Maps, I have a Google Maps thing. Google Maps, and even if there's no OpenTable in the restaurant, he'll book himself or some other system. I don't know why they're making such integration. She doesn't look like, uh, super, uh, super-need, what's that? Or plus I'm gonna let the agent get in on the registration in my OpenTable, but he's gonna make a registration at the OpenTable. He needs one time for-- that's cool, that I've been bailing once and I need to.
I've just done a lot of things, uh, this one, well, I had a very wide range of boxes. Well, there's a lot of cabs wide, and I've had a lot of cabs. Here, Sasha, quick question. The new Qwen Image Edit model has come. Is that something cool or not?
Good model for Chinese, yeah.
What's she interested in?
Uh, well, it's just a very strong Chinese model, uh, which, uh, is, like, a very strong Chinese model that allows textual editing. I mean, Chinese are good. We knew they were good, but they were good, yeah. I mean, uh, it's not state of the art, but it's very good, well, for open source, it's a top model. They're great.
Look, does it create a video or is it just for the pictures?
It's more for the picture, yes.
It's just for the picture. It's like a picture editor, right? It's like a competitor--
Yeah, you're putting a picture on, you write, you change, you get a result.
People, as this is an open source, will therefore be putting this into their different softs and decisions. Right? To work with the images.
Well, I'm gonna add that the commentaries will be written if we don't. Oh, DeepSeek, uh, 3.1 went out without some Haip, no nothing. They've been revamping, but we're waiting for something interesting.
But it's not R2, is it? Or some of the things that they promised.
No, it's not R2, it's, uh, V-C. Well, uh, I don't remember Liza NER on it, they're either on it or not on it. There were different news. But she's good, she's got a baccalaureate to pay models. Not SOTA, but it's like a good growth for open source of solutions.