Hello, everyone! We're on ToTheMoon. Technological news, Silicon Valley sites around the world. We're going out on Sunday.
Well, he's a genius, he's a genius, right? I mean, he's a genius.
Can assessments ever grow?
This plane has a lot of drones, and it's never been, like--
Yeah, there's no social conditioning where you're either hired or hired on the basis of what LLM will answer.
And AI will determine national security in the potential of 10 years.
I want to start with the market for investment in the technological world and move on.
There are a number of topics that are clearly interested in discussing in terms of restrictions and legislation and the different models that are now available. And if we do, we'll touch the robots, because NVIDIA has this accent on the robot presentation. And in general, they will proclaim that the twenty-fifth year will be the year of artificial intelligence in such physical facilities.
So, about the investment. Here's Anthropic. Anthropic now attracted three and a half billion dollars in early March, and it has an estimate of sixty-one and a half billion. And in fact, the evaluation was made out of the half of the OpenAI series. Although, in my opinion, OpenAI is probably worth a billion, oh, trillion. But the market is so active in terms of money. And while the US stock market has been under way in the past few weeks, it's been experiencing some kind of crisis, clearly visible. What do you think about these big amounts that are still being scattered like investing in artificial intelligence? Even in Europe, even in Europe, there are increased estimates.
In my opinion, it's still interesting in the sense that, well, Anthropic, I think Google's still attracted a lot of money. Before that year ago, we were discussing how much money they had from Amazon. And it's like competing companies investing in these assets from different sides, right? And it looks like a diversification. The company may have a department, and they may also be in the outside company, and they're carrying a lot of money into the start-ups. And here, it looks like, from my perspective, that's how the diversification goes. You're just trying and investing, and you're trying to invest in some different companies. Because there was information, remember, we were also discussing that Manus is under the hood, maybe Claude works from Anthropic, right? And he can be a steamer, too, on the popular side, on the amount of money, on capitalization. Perhaps somehow, companies start investing in different assets without focusing on one particular company.
Well, there's one of the fund managers, by the way, leading, and, uh, Lightspeed Foundation. It's a fund that once invested in both Snapchat and Figma and Mistral and Epic Games in different projects. And here, well, Amazon was four billion too. Sasha, what do you think about these
assessments?
Well, here, that's Anthropic. Well, I mean, it's like we should first determine if we believe that a closed-source model will have a significant value, right? So, if a closed-source model is at least 10 per cent of the market, right? That's the question. I mean, well, basically, we know that in the history of the Internet, well, whatever it takes, well, something's not always like that, yeah, it was just that, like, very often, open-source solutions win, right? So we knew that at the browser level, Microsoft tried to lock-in, there, Mark Andrissen tried to figure out a lock-in to do something, but it didn't work out. So I'm saying that, according to the fact that open-source models will have a big field of application, Anthropic is worth sixty billion. That's it. And according to what, well, if you push the picture now, that closed-source models don't win open-source, objectively, and if you push away from the fact that closed-source models have their problems there with them. privacy, yes, potential problems, unresolved copirate episodes, and the fact that the Internet, as much as it is, is all a fundamental solution, they're all open-source, then, well, all these companies will be You're just worth zero, yeah. Well, no matter what the zero is, they won't cost zero in any case, because these models will need to be survived. And, well, there, Anthropic, the OpenAI has a huge expert on how to grieve the models. So, well, as much as zero is never gonna be, it won't cost more than ten billion dollars. I mean, of course, like infrastructure start-ups that can just be grieving, servous that can be fluent, writing infrastructure, they're worth a few billion dollars in any way. I mean, that value is definitely in all these companies, yes, because they do things that can do not so many people in the world, but in America, it's almost nobody. I mean, of course, their value will be huge for cloud providers anyway to improve their infrastructure for LLM, LLM Fynthening, etc.
But, but, but, yes, there is a very high probability that the value of a closed-source model will be very low.
How much more can investment be made without real global implementation and real evidence of implementation? For example, Satya Nadell, head of Microsoft, was now speaking, and he said that he was, in general, saying that the success of AI models should be measured by a concrete contribution to the global economy. For example, productivity growth should increase, for example, rather than the number of abstract-obstratic tasks performed by model models, or winning tests. And he's straight--
He's a genius, he's a genius.
Says that, for example, conversations. Yeah, but he says that talking about, like, AGI, that AGI would replace a man, that's a bachelorette, very early, like, he's not gonna be here right now. Well, unlike the head of OpenAI Sam Altman, who says that AGI will be or there, Elon Musk says that in the coming year we will definitely replace, get some kind of model of this level. That's how much it takes? I understand, for example, that if there's something ultra relevant and estimates as high as possible, they're coming forward. Can assessments ever grow without this story that Satya Nadell's talking about?
Yeah, maybe. Yeah, maybe. I mean, we're already very close to the autonomous agents who do something. We don't know what, but the code writes or makes a business analysis that can cost a thousand dollars a month and still do. I mean, let's just say, the huge potential for the introduction of such agents, just like a-a-l virtual staff, right? Just, you know, what I mean, right? I mean, whatever. I mean, there are staff who check for you that you have no data that's been compromised.
Yeah, yeah, yeah. I mean, from a kind of infrastructure, but he's paying attention. So Nadell's just saying that it's micro things, and they don't show the real value of the company. Well, we do not see, for example, that some company using artificial intelligence has begun to increase its profits or that EBITDA has been increased or sales are increased. I mean, there's no express expression like this right now.
Yeah, he's right about that, yeah. And in that regard, he's ahead of the way he's yelling. So he's doing a big haj like he doesn't focus more on models, focusing on business. So it's putting him on the OpenAI automatically, right? Well, that's why it's very good for him, and it kind of reduces him as a function of OpenAI, at least in the public field. That's very good.
It's very, very, very important.
There you go. No, make sense. I'm just curious from the perspective of so different discussions, we're going to move on to the debate on the law field. I want to place a direct emphasis on the difference between U.S., China and Europe ' s legislation in the area of artificial intelligence today. But there's still a completely different polar view, right? That's how Satya Nadell is or is Richard Saton. He's got a triumph, a triumph of this professor in the field of artificial intelligence. Turing's award. He got the Turing Prize. It's analog to the Nobel Prize at IT. And he and the American Andrew Barto, he's Canadian, and he says he has one, you know, saying that he's a probability that by two thousand thirty years, I'm gonna notice, look, I mean, I mean, I mean, he's got a saying, you know, that he's got a chance that by two thousand, like, 30 years, The man who received the Turing Prize means in the AI area, and who is concerned--
Oh, and who's concerned, well, with a backup pionic, yeah, in artificial intelligence. And, uh, from-- training with backup, we can explain if people don't know. This training, one of the types of car training in which an agent learns to make decisions by working with the environment, and he receives rewards for good behaviour and punishment for failure or absence Awards. Well, as an example, I have a good comparison, in the example of a dog. I mean, I have a dog now, and I have to drain it. She's got good taste for good actions, so you're teaching her, right?
What are you giving AI as a reward? I'm just curious.
Some bonuses, I guess, some bonuses, probably inside, you know, maybe some balls. Here. And I'm saying Nadell said that by the 20,000th year, artificial intelligence will be available with a mind at the man level. Probability of 25%, probability of 50% by 2 thousand to 40 years. I mean, uh, completely remote de-de-activity. And he, uh, is an absolute scepticism that, uh, well, as the current direction of the development of large language models. He, for example, said that it would not be possible to obtain AGI through current methods. Well, you're gonna have to find new ways, right? He says, "Let's get on the new methods." So, again, returning to the topic of the assessments, does it mean that these current companies will find new methods or new companies will come? New companies will be coming. Because the scores are big, yeah, huge money is now investing. Well, there's a lot of transactions. I mean, I'm actually looking, first of all, a lot of transactions, fifty, 100, two hundred million dollars. Here. And even IPO is planned for someone. Question is... Databricks are scheduled to have IPO. Well, at least there's rumors that they'll do the second half in the twenty-fifth year. But the question of the assessments is very interesting to me. Oh, they're gonna just grow up, they're, uh, I-- I don't think they can fall.
Who, who's looking at the other side?
I think both will be developed. So new companies are gonna be coming. The same guys that went off with OpenAI opened up every single one of their own start-ups. The same Sutskever, respectively, with SSI, I think he has a company called Safe Superintelligence. Then, as appropriate, Mira Murati with his company. That is, it is precisely the search for other architectures, other ways to deal with such challenges, yes, and closer to intelligence. When you say that, uh, Satya Nadell said that the current models were not artificial intelligence or anything. One Jan Lekun is definitely blushing his hands and says, "I told you last year that you call a linear algebra an artificial intelligence." Here. So, uh, new companies will be definitely coming, but it is unclear whether they have reached AGI, whether they can do something in principle. Current companies are unlikely to disappear immediately, too, because what's now probably gonna be able to find a use that's gonna cost a lot of money. At least that's about Manus, we're talking two episodes, yeah. And like Sasha says, it doesn't look like it's a bench, because there's, uh, some statements from those who already used it. And at least at this level, what is now with LLM can already be applied. Or OpenAI is opening the API for, uh, voice generation for transcribation. And these models will find some value on the market, yes, so they won't disappear at all. But this is how we get to something fundamentally new, what you can call an AI, it's not clear yet. And apparently the main question is where all investors look, yes, that's where to put the bet on who can get to it.
I don't quite understand Nadellah's rhythics, because from the point of view of the current, Sasha's example is, there are agents around here. I see, for example, an incredible break in the use of models everywhere, in general, in the world. It's certainly not an artificial intelligence, it's not what we see in the movies, it's not what we see in the Black mirror, it's not what we see in Levion and so on, yeah. But the point is, there's something that I see she's in the movie. But, uh, that models, I think, have been effective and productive for hundreds, about thousands, about tens of thousands, depending on the spheres. It's a fact.
The question, uh, and I think that the question is not that it is expressed in the growth of the instantaneous, there, that some particular business has arrived or that of the country ' s GDP growth. The question is how people start to use it all the time. Those who need to use. For example, there is a direct increase in investment in defence industry, in defence industry, in the February-March industry. There's an Israeli deal, a big deal. There's a American launcher Shield-
One of the biggest venchour stories in Vis, that's, like, 30 billion.
No, she's not defensive. I was talking about another example. There's an Israeli company Spark Beyond. She did, did an AIA., and she's taken over by Rafael to use the data in the intelligence. That Vis deal, which, well, that's not defense, right? Here. Not defense.
Information security, if I remember correctly.
Maybe, too. Well, yeah. Although they may be used again, they're used, Sasha, in defence?
Well, Vis is no, it's more cybersecurity.
Yeah, I was talking about Spark Beyond. Yeah, I was talking about another company that was consumed. And there's a $200,400,000 investment in Shield AI, and that's a lot of market attention, too, in terms of the defense, they're making their artificial platform. Intellect, which allows drones to survey buildings, evade TSB, coordinate their work in battle without direct operator control, and so on. By the way, Trump's released new Fs now, what's the 40th, right? What generation? Seven. Yeah, forty-six, F forty-six. And he even paid attention, said, "Well, this plane has a lot of drones, and it's never been in a lifetime. Like, you know, you know, you know, you know, like, uh, logically, obviously, I just, I'm gonna keep thinking that the models are now in, uh, different casees, they're gonna do what they used to do. It was fundamentally impossible. Well, I don't know, I'm, uh, very, very simple, I guess the cabins are doing from the point of view of the scale of countries. But I think the boxes of size of countries, the size of intelligence, there, the size of the defense, they're phenomenal. The problem is, these cabs aren't being used very well--- would be great if they said that they would use it in the educational system, right? On a scale, while health care is also, it means it's starting to use it, watching, but there's a lot of regulation.
So I guess that's what I'm asking about Sati Nadella. But I'll go to the next bush. That's a very interesting fact. We didn't pay attention to him. We talked about JD Vance being a Vice President of the United States, making a slogan in Europe about removing barriers to US leadership in the area of artificial intelligence. That's where McCrone said we were doing this, like analog, analog of the American Stargate, which said that we'd invest $100 billion--500 billion. It's worth $19 billion, there from the United Arab Emirates, there from the major United States funds, Canada, etc. And I'm honestly speaking, I'm hit by this whole amount in Europe. Given that Europe, well, what they're gonna spend $100 billion, they're not very visible in the artificial intelligence market right now and they're in a terrible state of affairs. Now, there are Europeans who are putting their stakes on very strict rules, yes, the prohibition of any risky different applications. Prohibited, now it's a total ban on social ratings. I mean, now they're very demanding that all high-ris- high-risk artificial intelligence systems be certified. There's a US that took all the rules that Baden introduced. And there's not much of a lot of services that don't even know what to do now because they've all been rolled back in terms of artificial intelligence everywhere. Although there's a number, a number of nuances. Here's Meta again, political restrictions, there, made elections and so on. And there is China, and there is China, which, where they come, well, like strong models now. Everyone knows that China has a strict rule that any model in artificial intelligence must be registered at the State level, censored and must reveal its own. Data. And so Sam Altman said, and he said, "Let all friendly countries go on with that political slogan, Sasha, just in the past issue, saying that this is an inadequate statement, you know, expressing your opinion." He said, "Let's all friendly countries, so we'll block the Chinese models." There are three of these different vectors now.
What vector wins? You viewers, write, please, do you think what is the most appropriate approach? The approach now that brought in the new, new administration of Donald Tramp. The approach that continues in the EU, from the point of view of, well, in the European Union, is how they started their first artificial intelligence act in the world last year with a great number of limitations They're injecting. Or the way China is. So what are you looking for? Write your opinions and write as good as you, minus. What you see in this risk and risk. And we're gonna say our stories in parallel.
I'm probably gonna go a little bit, like, a little bit, and I'm gonna go back, yeah. I'll just say that if it's a little bit of a-- adding politics to our conversation, yes, of course, it's clear that Anthropic.
Of course, it's clear that Anthropic, OpenAI will be doing great in the next three years! In the sense that very much, well, Trump actually restricts the American market. Well, you know, yeah, what am I talking about? Some protectionism is good. And, well, of course, as a good open source of the developer, well, there's no one in America. No, it's not possible for a number of reasons. And, so, Google, like Microsoft and Google, they'll be doing 30 billion parameters, and even I think it's realistically up to 20 billion parameters. But still, models like DeepSeek won't come out. Good science-based models won't come out. And protectionism will certainly help very much. In the next three years, it's 100 percent. Just because it's true, well, because, well, for some people there, if there's a saying that AI will be national security in the potential of ten years, yeah, it's even a trillion. Well, it might be worth a few companies of $3 million that cement this national security. Well, there's 100 billion, and there's a lot of companies out there.
And it's like there's no difference, which, for example, China is very focused on national champions, right? They have this idea where they just cut a lot of money into their private sector for the development of artificial intelligence companies, so they can come in the world, too. I mean, it won't stop, it won't go ahead, it won't beat American companies.
Sasha, you mean?
We don't know.
Just knowing about China, they're making a bet on the open source, but they're making 100 percent of their internal models, putting them in their systems, putting them in all the way, right? Aliba, Baidu, Tencent, to the government. Everything is full of inner models. We just don't see them. We don't understand how the market works. It's a man who once says, "You see Aliba in Russia, some percentage of the real Alibaba, which we have." I have a Chinese Aliba offer, I remember opening up there twelve years ago, and he says, "I can borrow a million yuan with one clique to take a perfectly microcent, and I'll start working at once. "and then." And that was very remarkable. Or does he say, "Sites, what sites? Here in Tencent, turns out to be WeChat. First, we need to set up microsites there, do it." So we don't even feel this infrastructure as huge as ever, we just don't feel it, we don't see it. And I, and I understand how ten years ago it was a forward-looking move. I don't understand what's going on in these systems now because we're not in them, we don't use them, we don't consume them. I've been living in America for two years, I've only been alive two years, able to feel certain, certain aspects. When you're on the side, you're watching a little on the side, and we're not even in China. Well, let's go, they say, "We've let you out a model. Look, she's kind of cool, DeepSeek." But in parallel, there's something that doesn't believe they have closed systems. I mean, these countries, these, uh, these people, these people, they're very interested in such super national services. It's not gonna get cold, Sasha! OpenAI, Anthropic won't slam. It's not enough that they'll be able to get to some kind of cloudy space assets and numbers. That's obvious! Something has to be-- something has to happen, I don't know, they have to, or really start, the proceeds to generate. Anthropic, by the way, with a 60-milliard estimate, is making a billion dollars in models like that. But at the same time, the OpenAI states that they have four hundred million users and several million, and I think they have a few million corporate users.
That's interesting, by the way, too, yes, that the OpenAI has already had a few million corporate users. Tanya, Ildar, what do you think in terms of what model is more appropriate in terms of development, that is, in the context of artificial intelligence, from the point of view of law? I'm talking about the law now, because I've been looking at Europe again, and I've been throwing them out for a while, and now I'm looking at them, like, getting a lot of money back there, maybe they're gonna do something.
I think that this is a ambiguous approach, you know, because I, for example, have a personal opinion that is not a very attractive political structure for China. I mean, like a so compare... Protectionism, it's in America, it's also in China, and I think it's in Europe. It's just that he has different dimensions. Accordingly, I feel more winning, of course, as China's strategy. But America's probably gonna be pretty high on the turns, too.
Well, in general, when we say what strategy is best, yes, here, like in the data science, we need to define the meter we're gonna be, or there's a target variable we're looking at. Because each of the approaches has its own advantages, and obviously has its own minus. I mean, if we're talking about this freedom that we've given in the US, and that's the only way we've had to lift all the restrictions, do whatever you want to do for the development of systems, it's probably gonna be a little bit of a little bit of a breakthrough. I'm not very happy about any other life. In Europe, on the other hand, this direction may not be very active, but on the other hand, the impact of these systems on people will be understandable and managed. Well, for example, we have news about the OpenAI and the Norwegian's suit to the company. What's the reason it came up? The user asked OpenAI, you know, ChatGPT, I mean, who he is. Yeah, like, what do you know about this man? And she started to hallucinate the system. I've been hallucinating that he's there, well, we're not talking, very creepy crimes have been committed. Part could do, part couldn't, convicted, et cetera. In Europe, it is perceived as a slander as the dissemination of slanderous information. And accordingly, the system has to do something about it. Yeah, and there's some trials going on. On the one hand, yes, it imposes restrictions on the creators of this kind of AI systems. On the other hand, there's probably some kind of guarantee that, well, there's users, people are gonna be protected somehow. Yeah, there's no social conditioning where you're either hired or hired on the basis of what LLM will answer. Maybe she's hallucinating. And some other stuff. Well, China has a separate story, right? So they're there if they find a point of application, they're probably pumping a huge amount of resources there. Not only money, but minds, and iron and everything that needs to be, and that's where the result is achieved. So this is the question, what kind of parameter we want to get the answer, right? If we have absolutely no idea what effect this will have on the outside of the company, it is likely that the US approach will yield a much faster result than that in Europe.
Europe is likely to be behind us, but there's a conditionally controlled result, that's all there is.
Well, I want to comment here that it's not like the U.S. that's right to do everything. Given the fact that nowadays, Meta has literally introduced this week, she announced that she would require political advertisers to expose the use of artificially inseparable drugs. Intel in your announcements. And for example, if there are sgenerated or edited fragments in the advertisement or image, for example, photos of a talking candidate created by neuronets, this will have to be clearly reported. I mean, you're gonna have to make it perfectly clear that this content is not real. And on that side, we already see a large number of different details and limitations on this subject, but it is not clear how this will work. So, on the one hand, systems are being developed that, like, can determine that you have some watermark in your image or in the video, you know, the markings that the system put in to make you feel like you're in a position to be in a position to be in a way that's a kind of a way to get a little watermark. Identify that it is, for example, a content, a generated artificial intelligence. On the other hand, how much global verification is going to be and why would that lead to? So if we take all our systems, what can we do? It can lead people to stop believing in content on the Internet, because they don't understand what real and what not. Or, for example, social networks. Or, on the contrary, people are very selective. They're getting some opinion. They know that, for example, a part of the content can be artificially intelligently sgenerated, part not artificial. And they choose for themselves what to watch and what to do. Yeah, some kind of trend is coming. I mean, people have it. Because I'm some skates, honestly, I see, and now there are many sippies. And you read the news periodically that people there believed in the video that someone was arrested. Yeah, there was a skate, I think it's known that Trampa was arrested. Well, some of the fun stuff that happened. And you're at the beginning when you've been looking at it fast somewhere, you know, you can't really figure out if it's a dipp-- it's a fake, really or not. Yeah, or some pictures show up there. You can't tell how you are, like, visually. Plus, quality is unbelievable and, accordingly, in China, Sasha once said very well when DeepSeek showed up that whatever the Chinese say how much they invested, and who said who did it. I did, yes, it's a black box, and it doesn't know anything. Yeah, and it doesn't even know where, and how it's going to be. And the same Europe, they sort of regulate everything, but how much do they really regulate? If inside, for example, someone starts using Chinese models, like, or even the sensors are using something. How do they regulate it? How will they follow it to the end and how will it end up with the rules and demands to hold? Yes, Eldar.
With the marks on the content of the sgenerated. It's also very similar, it might just be like, uh, a parole on a pack of cigarettes, yeah, smoking hurts your health or something. It's always written.
What's up? It's always written. It's so close to seeing that nobody reads it.
Yeah, yeah, yeah.
There's already some scary pictures of diseases that have started to be on the spices, but nobody cares. Same thing if they write somewhere that the content is being generated, well, you're gonna have all the content, just stop paying attention. So the blind spot will be in this place.
One hundred-two is like small, little signatures are all hanging out on the ad.
Yeah. Yeah, yeah, yeah.
Who, who sees it, yes, and who's looking at it?
How do you know how to control volumes like that? I think it's hard to control that volume. I want to touch the topic today, the theme of robots. I used to give an example that robots show a lot of how they walk around, do something. I've always said that the world has not yet invented, uh, a hand that could raise any object and manipulate them. And this, the brass-- this, I think, is a very important accent that my partner once gave me because he was very busy with food, he was looking at how hard it was to put it in a warehouse. a robot who understands what to take from the viewpoint of a glass bottle or an apple, there, or banana, et cetera, yeah. Well, let's not mention any other stuff. So, now, there's a, uh, Canadian launch. I'm gonna tell you more about NVIDIA, yes, from the robot perspective. There was a Canadian launch of Sanctuary AI, who reported that they had made progress in the manipulation of small objects and that robot had learned to spin and move objects to his fingers. And that's the video we've been playing together right now, you'll be able to see, yeah, that's in-hand manipulation. It's a very difficult task for advanced robots. So, we often see robots that move, or drive or they say something, but in fact, this is not the task. I think models should help solve this story. Well, that's very strong. In fact, the technician training, that's all that artificial intelligence, the vast amount of this data, I think, at some point, should make it possible to accomplish this task. And it's gonna be a little bit more moving, a breakthrough. And here next thing, Jason Huang was at a meeting on, uh, NVIDIA, and he's a big accent, and he's done what they have... He said he said he was declaring that the next wave of artificial intelligence was robotics. And at the end of his life, he's been telling me about two hours about all the chips, all the chips, that this year's gonna come out, and next year it's gonna be so chipy, and in the twenty-seventh year, it's gonna be so chipy. And he took a little Blu robot on stage, who was salutely there, doing different teams. And NVIDIA has annulment of a number of novels for the era, there, different open neurosets, models that allow humanoid robots to be managed, yes.
And right now, they're putting a very strong bet on it and they're gonna move in robots. What do you say in terms of robots? We were doing a great video of the trend. We're there, I think, 12 trendes have shown that they'll be moving to the twenty-fifth from the twenty-fourth year. Several trends have already proved very important. There's a Trampa's appointment, for example, that's where investments are different, large from the standpoint of OpenAI and so on. Even the quantum computer we've got. Including the robots we were paying attention. From the robot perspective, we'll see if people are waiting for anything interesting at the end of the year, in terms of robots, more or less in mass, mass use or not?
The humanoid robots, I don't think, clearly. The challenge, first, is too complex, secondly, totally unnecessary. So, for a robot to go and do something with his hands, I think, well, it's not necessary. There are, well, other forms of de facto that are likely to be of some benefit. These are robots, which is just a manipulative hand that's conditional, so you're getting a package or you need some sort of a stiff set up or something in the store needs to be put down the shelves and stuff. That's what I believe more than in a humanoid robot that walks, picks up, picks up or something. I mean, he might have some manipulators, they might even look like hands. But, you see, the package came, he took the package, he climbed up on the elevator, walked up there, there, before your door, left the package somewhere near your entrance. I'm still ready to believe in some of these systems. And in the contingent optimums that Mask shows, I don't know, I don't believe in this story very much, to be honest.
Look, this is what you're showing up now, it's obviously already undergoing tests in China, but you see there and a lot of new automobiles that cross-exhaust, that's a test-- yeah, which are overloaded in general by different manipulators, very much by the steep straight manipulators, which are new quality, yes, quite new. I mean, you've shown it now, it looks like standard. I remember, Elinar, you're on, and I remembered how I was doing a robot soph in China 20 years ago, on Taiwan-- not in China, on Taiwan, on a robot sophth, and there was a similar hand that was there that was moving in. different planes. And it seems like it was 20 years ago. Well, now that's a completely different progress, it's another, different level, yes.
It looks like robots that do the surgery. No? Somehow they're called, I forgot, Leonardo or something.
There are things like that, yeah. But look here, unlike he can learn. So he has some computer block.
Yeah, yeah, yeah.
He's watching, you're showing that you have to get this thing out of here, put it here, and he's actually starting to understand, repeating you. So you don't have to program it directly from scratch, do you? So he has the possibility of some kind of pathetic moves, some kind of manipulation to repeat after you.
Yeah, that's great.
I mean, the robot is as simple as possible. It's not the Optimus that Mask shows, yes, there's demonstrations or something. But because Vision's computer is developing, because there's some LLM algorithms, there's a probable, these transformers and stuff, it can start to bring a little more useful. And I have a feeling that we might be able to see.
And not only at the factories, but also on the basis, there's a conservative service in some HQ that might appear there, I don't know, it's a parole service.
There's a fun series of robots showing Helix. These are robots, robots there are two, as the world would have seen for the first time, that for the first time, a multi-brained collaboration has been successfully implemented, right? Where they make manipulation and make these manipulations, it's shown by the example, we'll also show a video on the fridge, how they collect the bags, and they're taking these packages together. So two robots work together at the same time and place things in the fridge. And there's another start-up Figure AI. It's good on the video. You see that, too, that robots first see those eggs or this pack of cookies and know how to treat them, right? And this Helix was developed by this company Figure, Figure AI, as I understand, and these are Felix robots, right? And they were experimenting with the OpenAI model for a while. Well, they kind of had a better decision. Well, I don't know what that decision is, how they make it. And they've had a very high score, by the way. They had a new round of $1 and a billion, and an estimate like they're gonna have $40 billion. That's a very interesting subject.
These robots look very good.
It's just that robots are unrealistic, some unreal videos from China. What they do is they do, they can repeat human movements. I'm honestly... On the other hand, all these robot videos are easy enough to do CG because, well, we don't get as good as people, we don't know robots, we don't know robots to distinguish robots. Well, I think we should just go to China more often, and I have nothing to comment on. I mean, yeah. I mean, of course, that robots are used as dummys, and that they're, like, interactive dummies, I've seen in China, that's cool. And they have some superpool shop down there, something like Nike they have in the locals. There, there, yes, these robots are used. I might find Twitter, try to get out, try to find out, try to get out. And I'm very, uh, very hard to imagine what adoption curve is gonna be, very hard to imagine.
Now, it's important, now, the next wave of big fandraises is the collection of robots from different skies like, well, more industrials, with a chain of sale, with a chain, well, without China. I mean, Vietnam can be used, but mostly, yes. I mean, trying with a focus on America. I mean, I think it's gonna be a big--
Yeah, but given what Trump is doing, the next four years, there's gonna be a strong fit in the robotics.
Big, huge. Yeah, yeah, yeah. I mean, again, because, well, at least America is competitive with AI. So no one can say America's behind, right?
Maybe not ahead, but not behind. And in robots and drones, probably not much ahead, probably behind. That's why it's a big subject. It's really so interesting that we have been very slow in these areas for a long time, right? And as much as the computer vision, as the top-all of the com-- for a long time.
As if the top companies were all the com-- very long top companies that were computer visions, it was just Waymo and Tesla. In general, the computer vision was considered very, well, under appreciated, shorter, very long, people were rolling from computer vision into text processing because the text was more commercially used. I've been watching this very hard, yes, because as a computer vision specialist, I've seen that, as a matter of fact, I'm in there as a specialist in ten companies in the world, you know, in America, right? Very few people who are serious- very few companies, especially small ones, who were not companies that made computer visions serious. And now as we can see the fruit, as a result of this, this trend, that America is behind us. Well, until 2022, there was no fun computer vision of the launch, except for this Mag Seven and Tesla with Waymo, where it would be fun for a professional to work computer vision.
And even if you don't take such super-cool companies in the middle of the day, if you watch a job and a lot of jobs in NLP and even Classic ML, and CV is a very and very low story.
Yeah, yeah, yeah. Well, we just see, that's just, well, we know that, yeah, computer vision and consumers are really strong in America, like, uh, under, well, how underfunded. And therefore, we see, respectively, that in the computer vision and in cons-consumer, China is very circumvented, yes. I mean, well, it's like a generation that I think is going through a generation. But I wouldn't have, uh, on all the robots, I wouldn't have gotten into it because maybe Amazon has a cool robot in the warehouse, and we just don't know anything, but, in general, yeah. I mean, it seems like they're going to be two to three years, for a whole generation.