Hello, everyone! Uh, we're on ToTheMoon. Every time we go out. Technology sites and news from the Silicon Valley. The governor vetoed this law. A law that required very serious safety tests for a powerful artificial intelligence.
And if the law were adopted, it would have made it difficult for new companies to try new approaches, and it would have left America back. Yeah, that is, hallucinations are possible, but the compression process is very strong. Like a law that is enacted to give law enforcement agencies the opportunity, and, if they want to, to investigate. But, uh, it won't work.
We see, uh, a bunch of irrelevant extraditions, unobjective, inappropriate information.
I sometimes feel in the States that I'm already in artificial intelligence-based in health. Because there's no real explanation. You can't ask for anything else.
And today, I think, is an incredibly important subject.
And throughout the whole thing, we'll be able to deal with things that, uh, are often hidden or not visible, such in the shed and which have political statements. But it is important for everyone to understand what is happening now in terms of artificial intelligence, its development and regulation. Today, we will discuss the interconnection of artificial intelligence, which is the regulation of it in different countries, and this will show us all the way to think and move the whole market. That's a cool subject! Hello, everybody. Oh, and we're gonna start with Governor California. E, it is considered that California and the United States are such a state of excellence in terms of the law in the f-- in the area of artificial intelligence.
Oh, I don't take an example, uh, advanced Europe, which is very fast in the implementation of laws, but they're very specific. They seem not to me to be directed towards the development of artificial intelligence, but to a very strong restriction and deterrence of technological peace. And there's a lot of this mix, like development, and containment, and everything. And plus here are leaders, uh, the world of artificial intelligence. So the governor, uh, signed 18 laws, one information, 18, there, 17, no matter how to regulate artificial intelligence. And the most interesting thing that happened was a very long discussion over the past months, I suppose, this is the debate of the last six months to introduce and impose severe limitations on artificial intelligence and To establish accountability. The topic of responsibility for major models and companies was discussed. At the beginning, there were numbers, I think, of $30 million that were invested or what, $100 million that was invested. The governor vetoed this law. The law that required safety tests is very serious security tests for a powerful artificial intelligence.
He said, "No, we're not gonna ask companies that invest a lot of money." It seems to me that it says that, uh, green lights have been given. You're gonna have to move on, because the new thing we want is not yet known, right? As she did, uh, forget the name of this woman, I think she's Chinese. At the San Francisco conference, she said she was named a great parent. Uh, and she said--
Faye Faye Lee? Faye Faye Lee. Well, about her, right, talk?
Yeah, I guess. Uh, yeah.
Yeah. Well, she just has a book there. Yeah, she has a book, and she's Professor Stanford. It's usually, like, enough to be a grandma or Grandpa's whatever it takes.
Or they called her God. Yes, she was named god.
Or God or God, yes.
Yes, she was named by God and in artificial intelligence. She said I didn't know, uh, when, uh, who thought about when they used some words like AGI, for example, because everyone would have something in it and what they had, it's not clear.
I kind of taught, studied different terms and what we're going to do. Nobody knows what it is. So, about this SB 1047 law, which was vetoed and which, in fact, was very strong, could have had a very serious impact on the industry. It's just huge, isn't it? The absence of this law is simply with new investments for OpenAI, amounting to more than $6 billion. The biggest investment in the first place. It gives you the light of that. What's on the technology market, Sasha? That's when the law was blocked. What happens next?
Well, I think that means that, to date, I think that's, uh, that means that there's still gonna be investments in companies that are developing, uh, best models. I assume that if the law were passed, then many would be afraid that new companies would-- it would be difficult to enter the market, wouldn't it? And in principle, I think the consensus is that the transformers are not gonna end. I mean, just to increase the number of GPT-4 parameters would not lead to AGI, whatever it means. Well, that's not a way to succeed. And if the law were adopted, it would have made it difficult for new companies to try new approaches, and it would have left America back.
I mean, I mean, it's like rumors that China is making a very active investment in, uh, researching state space models. So, as they say, this is the next one, it's the replacement of the transformer.
These rumors, what are they on-- what are they based on?
It's like another, much more complex system of how, uh, how to describe the world. And it's a much more complex system than transformers., just a mathematical model that-- just a mathematical model that's built inside is harder. And, yes, and I read the reports that China has a lot of investment, because, well, they think that transformers are not coming to AGI. Well, I mean, I think the cost of starting a state space model study in the area is about $50 million. And then, well, it's just that China doesn't understand very much, that's how much it costs to them, right? If they have their irons, for example. Oh, well, the wages are much lower than in America, so maybe it's worth zero, right? So if the conditional or Ba or Tencent is very much of its iron, you know, for them to invest their irons in scientific groups or in the start-ups, it's practically zero.
And the companies in America do that? How many of these companies can do?
Yeah, yeah. I think there are some start-ups, two-three start-ups that do it. But I do, but I think it's very speculative. So we're on the baccalaureate, on the baccalaureate, and, unfortunately, we haven't seen any other kind of model winning-winning transformers. Until we saw it. It's a Haipoot launcher called Liquid AI. That they have Liquid's foundational model.
What does that mean? For the user, tell me what Liquid foundational model is.
Ah. I mean, it's just, well, they're inventing, I understand, and, uh, they're inventing their own neural network family. I mean, I just brought it up as an example of what really, uh, that, uh, that's, like, so far, how many of these kinds of hypes, right? So, there's an alternative to transformers. I mean, on the one hand, there are signals, serious signals that transformers are not at-- just rocking transformers won't lead us to AGI. On the other hand, which would lead to a specific alternative architecture that might lead to it, it was not yet
clear.
Can I ask the users? What is the fundamental difference between AI and AGI? I'm trying to understand and I'm not quite sure.
Nobody understands the end, Tanya.
Look, I mean, now, actually, look, remember, here we are doing the video production, and now we're doing the video production, right? Ia, for example, we have a task to generate videos for an hour. Oh, let's say, in a video of an hour's length, what do you have to do? Well, to-- what does that mean, well, we want to figure out what a watch video is. So, here's AGI, we can come and say, "Say, is this video or not?" And he'll understand what it looks like, "all will write on the counts and give his assessment. Now we need to figure out what it looks like. To translate this into sufficient technical tasks. Like, to identify all characters first, then look at how the cha-- look at everything, like, the appearance of a character in the video. Like, if he looks like a counselor or not. Under--- and so on, huh? I mean, now, LLM is generally quite simple, understandable.
Simple, understandable instructions. They tried it, but they tried to make the model self-conscious with the GPT model, uh, with the O1 model, and, uh, the O1 Mini. But it shows that the number of steps she's doing is very large, like, eighty. So, this is the generation that takes minutes, respectively. Here. It looks like a crutch. I mean, in the case of AGI, it looks like this costal that she has to make eighty requests herself, that's not gonna happen.
I wonder how it's done.
And AGI is expecting him to talk to you in response. So you're asking him if it's a video, it's normal or not? He'll start asking you, so he'll be proactive in finding out, you know, how much abrasion, right? I mean, it's normal and it's gonna be a little more proactive Ro-RoI play.
I mean, if you have to mention something.
He's got some subjectiveness, he's getting it, or it's gonna be--
No, it's just a synonym, but he's gonna try, well, waiting for AGI, that he's not just gonna be, uh, that he's gonna be, like, a partner. So he can engage in a dialogue, he may, he may be able to make his proposals, and, uh, his wishes.
Interesting.
Many companies have their definition of AGI. One of the most common, probably, often sounds, is what Sam Altman says, his OpenAI is accepted. It's that this system, which is the vast majority of economically meaningful, uh, action is at or better than, uh, specialist. Yes, so you're ChatGPT now, on a par with the possibility of translating the text from one language to another or some narrow task. And it's supposed to be the whole range of economical tasks we're taking, and that's what he can do. But again, I repeat, it's just that we've been talking about it a little bit now that there are as many companies as the people who say AGI have no one that everybody would be holding on, because I'm sorry, it's just one of the options. Sasha's the other one, isn't he? There's a lot of them. And so far, the experts have not determined what AGI is.
Yeah, I'm curious, because we have some stuff in there, too. Remember that interview with Oprah?
Take me to 2035.
I think it's hard to say where all this can go without sounding like a crazy person. Um.
Go ahead.
And, uh, that's what it's like. She's always in America broadcasting public opinion like this. And if it was to be straightforward to one suggestion, she said that AI was now like a big date analyser, yes, that's some kind of data. And there's a human dimension, yes, if AI compares to AGI
or not?
No, they use the terms, they use, obviously, in AGI terms, that it's supposed to be an intelligence that works at the human level, and that there's an alleged next story after AGI called ASI, there, Artificial Super Intelligence, which can surpass a man and so on. But all that is just words, because, in the idea, artificial intelligence always called what we saw in movies, in movies, when it means space, a bunch of fucking supercomputers. Well, I mean, serious, very serious things, yeah, it was called artificial intelligence. I just think what happened at 2,000 twenty, uh, what's it like? Twenty-third year, huh? In the two thousand twenty-third year, when the world swallowed, uh, swallowed AI words, and until that-- and until the 20,000 twenty-third year, AI, I think it was even easier than the current models. Right, Elnar? That's what people were taught in faculties, right, there? Learned, yeah, what they taught there--
Dogs and cats distinguish between each other and train objects on video.
And they called it AI, didn't they? Now AI is what we call it. But it's time when, I think, artificial intelligence is, uh, intelligence. Well, that's the same thing now, and it's starting to compare. It's like a man with a dog, right? Or there with the tiger. Here, uh, time comes, and artificial intelligence, in the idea, if something serious, real, yes, real, will be called, uh, just artificial intelligence and everything. So far, it's just a certain terminology, especially in the future, unrealistic. For ordinary people now, I think it's very unrealistic, uh, simple things that happen and simple skies. And today, I think that less than a percentage of the world actually uses things like ChatGPT, well, in reality, and they know that they're even, not abstract, uh, not abstract, somewhere. They're either allowed to do some questions or something.
I mean, uh, they understand all these skies are real.
If there's a little more to talk about analogy and metaphors, Andrei Carpathic, I hope I call the right name, the man who worked at Tesla and in OpenAI, there's a GPT-2 and maybe GPT-3, too. I was still in the process. He has, uh, articles and videos that he explains, and what is GPT? That's the word that's conditional. And there's a very interesting one. He says, remember, there are archivists, there's some big data, you're there, Winrar, Winzip archived, like we've all done. And you're losing your volume, but you can archive and get the same amount of data without losing it. Yeah, well, I mean, how many archived, so much archival. And if we're talking about transformers and this about GPT, there, Generative Pre-trained Transformers, yes, there's a much larger volume that can be squeezed, but you don't have a way out of here, either. That's exactly the same amount, right? I mean, hallucinations are possible, but the compression process is very strong. So we take the whole Internet, basically, there's a lot of petabytes there, yeah, or more, I don't know how many. And you're compressing it all in, like, billions of parameters, but it's still unlike the original volume that was. So, basically, you've buried him, and then you can actually get information from him in some way, right? I mean, there's more work to do, but somehow, that's basically a archive of the knowledge that is now. And the fact that there are some sort of trends in the discussion is interesting, but it is not yet true that transformers can close this story. And that what Sasha says is looking for new architectures that might be a little different to all of this, is, in fact, moving right this way, as I see it, again.
Yeah, it's an interesting move.
I mean, here to close this subject, it's really important to add that there's a research. So, standard, is that what was the historical background in the typing? What's-- what happens if the model needs to be trained, there, speaks ru-- like Spanish, right? She only saw Spanish, and then she could give her some examples of English and sk-- talk- and see if she'd get in touch and speak English. I mean, it's like a ch-- it's like that, right. These models, which can be summarized, i.e., data they did not see just in a few examples, were considered strong. And the transformer was considered to be such a strong model from the beginning. And now there are, uh, artificial examples, as long as toys are, that when, when, in abruptly speaking, these models can understand and study the new programming language they haven't seen. Oh, but for now, it's just toys. I mean, serious, uh, like the Carpathovas's talking counterarguments until I've got toys.
Great! We'll continue with the laws. We just said the law that we didn't take it and talk about what, uh, what's going on. And Sasha said that in a good way, failure to enact the law suggests that it is possible to develop a strong market and invest in it and finance it. Yeah, that's very important. That's a big funding. I just want to talk about what's been adopted, yes, and there are some zo-- as some of the lines, perhaps laws that, uh, are approved and that are given strong attention. Big law, we've been talking about him, I think we've mentioned him a few times. Uh, at least this topic was very serious. These are the rights of actors and their digital rejoin. And there's an interesting law, a-a-a-a-a-a-a-a-a-- that prohibits the use of digital rehearsal of the dead without the consent of the heirs. And they've made a law like that. He has an e-AB 1836, and the number doesn't have a strong meaning, but... And there's a law that's on the actors. This reinforces the requirements for actors to agree to use digital copies. And that's right, that's what's being taken seriously, yeah. I understand there's not only actors out there, I guess they're coming up, and there's probably stars. And there's a de-dec-- and it's probably gonna be expanded, right? And, uh, there's probably gonna be some famous people there, some politicians, and a bunch of different people, totally different people. And, in fact, uh, probably gradually going to be unused, uh, no, uh...
No, uh, content of another person. Because I wonder if the law was passed by the actors. So, you think I can do the line or what? And given what is happening in the world now, a huge number of benches and thefts when they fake their voices, the voices started to fake just unreal. I got a friend here a couple days ago, like a friend came to him with an audio message. It's so cool as chat, it's a telegram voice. And that voice still explains a complete, complete copy of a man's voice. It's impossible to tell. And I said to my wife, "You tell your mom to put everything, so, strong ecureths on Telegram and whatsApp, because they're two seconds--- and those-- and they're both self-aware. Because your mom, when your voice is coming, she can't diagnose that it's not you, right? Well, because the voice is totally perfect for now. Here, uh, interesting subject, of course, this one's on the side of the actors. See how she gets to people how they can manage a story with people and how it can be traced. I think it's technically difficult to do, unrealistic.
Guys, who are you looking at us, girls, please put in the comments now, write what you think about the law at all in AI. As you think, within your country, there will be an interesting place where you are, as it may not affect and in which zones. And you think that interference or non-interference, political presence, and new laws in AI are developing the industry, or, on the contrary, is narrowing it very much? And as far as you can see, what changes are going on? Write it, please. That's very interesting, very interesting.
You've got a lot of things right now.
It's such a very interesting subject. Oh, a little on that side. Now, literally, we've been watching a show here recently, and, you know, there's a thing, and, uh, you know, I'm sorry, you know, you know, you know, you know, you know, you know, like, this sound, and it's not like the voice of the original actor, right? Because in the language that the movie or the show came out, you might not be comfortable, right? Okay, English may still be, but some other languages are already very unclear. And there's a thought, and technically, especially what is now, there, and, uh, Advanced Voice Mode at ChatGPT and so on.
Hey, ChatGPT! I want to apologize to my grandma for being really late. Oh, that's sweet of you to think about apologizing. How about something like: "Grandma, I'm really sorry for being so late. I didn't mean to keep you waiting. Can I make it up to you?" Oh, um, she only speaks Mandarin. Got it. You could say: " 奶 奶 ， 对 不 起 ， 我 迟 到 了 。 我 不 是 故 意 让 您 等 这 么 久 的 。 我 可 以 怎 么 补 偿 您 呢?"
It is possible to make the film duplicated at the end in all popular languages at almost simultaneously. It's clear that there's a need to be further developed. But this law you're talking about is just building a calf like that, uh, bottom, that's where it's all going to move, right? So without that, we have one direction, and that's what makes it a legal form. But actually, I think in a few years, it's like a movie could come out right now or a series in a few languages, if there's a deal like that.
Well, this subject will definitely affect people. That's 100 percent. And she'll touch ordinary people, of course. And it was the law that was enacted, for example, that if, uh, social media contain some information, false information, they require the social media to remove deep fakes, there, on user complaints. Well, that's what happens. But there are things that are more specific. A law on transparency in data disclosure has been enacted, which obliges companies to disclose data for artificial intelligence training. I mean, who are these-- how do you use this? I mean, I wonder if these ordinary people can use or only use any special commissions or the FBI, or who, at all, who can access the data? I imagine I'm asking you, and I say, "Let me tell you, please, on the basis of what data you were trained to do, how can I study it when, uh, you're a number, Ildar, you're calling it that ordinary people never have. They'll take it, won't they? There's a hundred thousand zeros in that information, some crazy amount. How can you possibly understand that?
That is when a law is enacted to disclose data for AI training, how can ordinary people re-evaluate it?
Well, I understand that this is more like a cluster, like, a parole of Vickipedia, a text to Redd, a library like that, that's all the books of this library, something else.
Yeah, but you and I know that millions of sources or dozens of millions of sources, yes. Sources, I think these, these sources are growing, they're gonna grow up, yes. Plus, how will we track what is already artificially intellect and is not a source, like what was artificially generosized in this few years of artificial intelligence in Wikipedia and then Go on to Wikipedia, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go on, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, go, How do I show you-- how is that even possible?
Sasha, I agree. I think it's kind of stupid. That is, that is, a law that is enacted to give law enforcement agencies the opportunity, and, uh, to investigate if they really want to. But, uh, it won't work. Quite simply, everyone who needs the world is all, and that's what all the organizations that need AI have already started to develop their models. These models already know a lot of information. I mean, they're just, they're-- that's, by the way, that's the entire generation of GPT 4 models that have already been trained, who needs to be in the world, that's what Ilnar said, these are these archives, yes, information. I mean, they already have them, and they can always be, like, and now, the e-mails, they can reproduce a lot of copyright information. Like books. There was a lot of court, and against OpenAI, they used, like they were rolling books in a pirate manner. But now the OpenAI should not-- OpenAI will, of course, not reveal it. They'll write a choo because they already have GPT4, which was pre-activated to this law, which already contains this information about all these books.
Yeah. And they'll say that's our archive, respectively. We've learned GPT4.
Yeah.
Yeah, yeah, yeah, yeah, yeah.
We missed it.
resigned and the archives have been removed. Remember that we were discussing something?
Yeah, yeah. Yeah, yeah.
I think that an important topic is still in the legal and human context, that is the introduction of watermarks, yes, that requires the mandatory use of watermarks for content created by the Convention. artificial intelligence. I'm back, I have some skeptical in this. I don't know what this word is "mandatory" is like, how it's gonna be controlled. People-- if you look at the current search that is in, uh, I don't know, Gugla, yes, or there's Yandex, we see a bunch of irrelevant extraditions, unobjective, inadequate information. And little, this information is often imprescriptible and it's not real, it's impossible to confirm it and verify it. Given that a large number of information is now generated by AI, as again, this obligation is likely to affect some claims and some details. But how will that be used? Not enough how to use it technically, how much is it to be caused by watermarks? Uh, that's a question. Although there's a little bit of a good thing about it, right? I mean, the idea is, you can determine if it's a fae, not a fairy, it's a real voice or not. For example, there is a law that prohibits, where synthetic-- not forbids, but requires marketing to show where it is used. Interestingly, the law obliges the use of synthetic votes in telemarketing. Oh, that means, uh, calling me a phone. I think I'm talking to the seller, and he's telling me at the beginning that this is what the robot says, right? That's what this isn't a real man saying. It's gonna be so...
What's the chance you're gonna keep this conversation?
Huh?
What's the chance you're gonna keep this conversation?
What's the chance I'm-- Of course! What's the possibility of me using it then, huh? I mean, if they put it... it's like a commercial. If you put too many water signs, you're not reading the ad, there are watermarks. And if they're not, well, at least there's something out there. And if they're not, how do people start to understand that? I mean, these are all things, of course, they're very specific. Not enough of who-- in which countries have tools, tracking interfaces, who's gonna track it? There's a lot of tracking in the U.S., and so much spam is going to be in. I'm doing business recurrently, and you're just starting to get spams right now. They learned to turn your spam around like you had a government document. There's a little signature somewhere, and they say, "As such a law, you have to file a report. So, you'll pay for two hundred dollars immediately, and we'll do it for you." And you're sitting. So if-- just me in business for a long time, I'm looking at all this, I think it's nothing? But it looks very serious, very real, yes, very straight. And for ordinary people, it's just a misleading introduction. If this misleading introduction is not decided, what will be done with real artificial intelligence, right? If they don't catch the frauds that they're sending over tens of thousands of dollars to them, how will they catch here? I think we're in here.
I think we're in the middle of a very strong technological failure, you know, a lack of control, that's not a presence of control, no control.
I understand that this is so much a voluntary story, yes, that there are companies that commit themselves that their models will be equipped with some additional data there. It doesn't guarantee that somehow it's gonna be possible to determine whether it's actually generosized or not. But there's something about Google, yeah, she's starting to notice, aaah, the generated photos. But it's written on how it starts to notice. If the metadata on the file says it's a generated file, it's gonna be marked if it doesn't, it won't.
Yeah, that makes sense.
I mean, if someone edited metadata, how the generosized picture would be ungenerated.
But there's a movement going on. But how do you know that it's gonna be so far, well, very easy.
There are two laws that have been adopted, which are also important for people who have adopted education. One, he's so, a little weird, they're called, but he's putting, uh, AIR. illiteracy in school. I mean, there's a certain accent to this that we need to bring literacy to schoolchildren in terms of artificial intelligence. And I've told you before, I'm curious, uh, my daughter's reaction is so. She said, "I don't need artificial intelligence, so I'm 16 years old." Then it's been a week, she comes, says, "Dad, how do I turn it on? I see it's on your plate." I said, "I'm just developeric, there's a beta-version. You want me to put you on, and you'll turn it on, too, right?" Well, the young people are very different. On the one hand, they say it's a good thing to be with. Well, of course, the Era TikTok is here, Snapchat. Of course they want all these things to be different and they want to use them. The second law creates a working group on the use of artificial intelligence in education. Well, it's a little weird, weird, again, name. The state and California, in particular, says, "Let's go very carefully, uh, to develop education in terms of artificial intelligence." I don't understand either. Actually, I think it's just microconscious thinking that don't really give me the specifics, yeah, the real details. Here. But attention! It's always interesting to see what the State's point of view is, what it thinks is.
There's a big block, like health care, yeah. A huge number of start-ups are being made in this part. And right, again, attention to health care. What needs to be disclosed if artificial intelligence is used in interaction with the patient and require medical supervision. Here. I have, of course, me, you know what I'd like to know? I'd like to see a doctor, so he doesn't tell me that the artificial intelligence is using and what he's trained to do, and that the doctor tells me what he's trained for. Why? Why would I be recommended by a doctor, so when, uh, viral infection, drinking antibiotics? That's where he went to school, this man, how to take him down, means everything he has to learn for his whole life that he doesn't have to do it.
I sometimes feel in the States that I'm already in artificial intelligence-based in health because there's no real explanation. Very clear procedure. You can't ask for anything else, it's all the insurance company's. I mean, I don't know if it changes anything. It's a sickness for me.
Sasha, how much do you have a friend now who's in health care? I have a part of the guys who do the AI in health care, but I see it all superficially, because the area is very well regulated, and everyone's supposed to want to clean up the insurance and make it easier for them. Citizens. But I think that with this lavina, it's trillion dollars, it's power.
Yeah.
That's, uh, that's actually, I think, a crazy drawing. How much can this topic work? Or just public companies, some big operators are going to move things in, not fax, not to use anything?
Yeah, I'd say that until now, just to solve this love-- the hardest thing in healthcare is to unravel a love triangle. I mean, if the hospital, the insurance company and the patient, everyone has their own kind of incentive systems. I don't think generative AI is helping to solve this triangle. I mean, some concuit applications where AI can help, that is, for preventive purposes, but not more. I mean, mean, there's gonna be, uh, the best systems that drive health there, there'll be systems that can proactively ask how you feel, physically, and mentally, give good advice there, and so on. Go on. Such systems will come. And, yes, I know there may be one or two strong teams who work in this field, but I don't think in principle that generative AI changes health care in America.
I guess in large, I think that there's not much to change in China, or in Russia, either, because very strong regulation, and it can only be introduced by authorized operators. And in the States, I think, given what you said rightly, everyone has their chips and their dolls, and many of them are so old in terms of approaches. Not much they're not interested in showing information simple, transparent. So they're not interested in you having a 40-page plan or a hundred pages of your insurance plan, they're just pointing out. Why would they do that? Otherwise you'll get it straight and figure out some details where you're absolutely pointless, meaningless to be served. There are clear things where, where we know, this-- this is not working here. My wife was the one who knew about this, and she was buying dog insurance for 15 years, and all 15 years when she was wanted to do something just for a dog. They said, "It's not insurance." And when emergency happened, emergency moment, they said that the insurance guy was the type that was in emergency.
Didn't my wife tell you that? No, it's just our story alone.
No, no.
That and I could have told you.
Yeah, there you go. Afterwards, 15 years, yes.
It's a common problem.
It's a common problem, because a woman at school told us here in Stanford, and, uh, this one, and she-- they said that we just realized at some point that it was pointless. Why? We're just paying for a nice reception, there's some MRI in there if we need something else.
Yeah, but it's tens of thousands of dollars, as quiet as you can pay for the dog.
She's the sums that led her to, and in the end, you're finally getting caught. So you don't have, uh-oh, final win. If you count it together, there's no final win, right? I mean, you're actually getting in the wrong place. I'm just saying these details aren't described. I mean, you don't know what they're gonna do. There are some general--
Yeah, you can't predict your expenses.
You can't predict, yes. Fifteen years ago, you bought CASCO, car insurance in Russia, and you were paid in full. And next year, I didn't care about insurance. In the US, if you've been in a long time, if you have one insurance case, you're in the middle of an insurance policy. And the question is, do I have to file this insurance case or do I have a better chance of repairing myself?
Yeah, yeah, yeah.
But in the next five to ten years, not overpaid for insurance. I mean, the question is, people don't know what's going to happen in the future, yes, and how it's gonna affect the future. And I think that even though it is an understanding of the familiar laws and their directions that we are discussing, they are just to be, uh, well, to be adequate in the perception that, in fact, Now, it's in terms of artificial intelligence. So I've been recommending all the time that different programs be used every day for a perfectly simple life.
I mean, when you use them in a simple life, you, uh, you're just rolling around with new skies that are super useful, just unrealistic. Uh, Eldar, you were talking about this interesting thing today about not seeing this news, and this information that the OpenAI said stop making start-ups. Tell me about her, please.
Yeah, but it's working out...
Interesting.
Yeah, there's two news in one. One thing we discussed before was that they attracted finance. And there's information that they've attracted six and six billion investments, and now they have a value or an estimate of 100, fifty, seven billion. And in general, they've attracted more than 19, now, second, no, 17, nine billion, they've been attracted from the moment of the founding. And the second news inside was that they kind of asked their investors not to invest in competing companies. Well, there's no money in Anthropic, there's another company, no money to send to Mask. And it's a little shock for me. Like, can I? You just got the money, and you're like that, and you're not the other. And I'm curious, how is the invest market? So, Sas Masrabov, you might know that this practice is so exclusive or many companies, and it's a market rule, as you think?
Yeah, I'd love to say that. I mean, it's normal. I'll explain. I mean, uh, basically for good companies, top companies, don't talk, that's for companies that are leaders in their field, right? For them, capital is a commodity in general. I mean, they-- they can find capital. What is the question of what financial circumstances? But they can find capital, and in-- and it's an investment.
And through-- and it's gonna be an investment or a loan, but in general, they can always. And... Because we're talking about top companies, and in general, investors can't help anymore. I mean, well, I can barely imagine anyone telling Sam Althman how to build a company. And Sam Altman is already communicating with all the presidents of the country better, right? So he doesn't need that kind of help. Iaa investors can give money, they can help with the connies. And the question is how they-- and the third question is-- and the third is the most important aspect directly to Sam Althman. How will they-- will they continue to support OpenAI and are they working with or not? Some firms, some VCs, like Menlo Ventures, that invested in Fixwell, yes, here in the start-up I'm doing, they don't invest in the competitors here. Generally speaking. If they invest-- they invested in Anthropic, they do not invest, they do not consider the other foundational text of the model. There are exceptions to the rule. For example, well, I think Lightspeed, Andreessen, all the big funds, they often, you know, as big funds, they often get to be seen as ETF. Well, those who give them money, they give them, they think it's like ETF on generative AI in some way, just in-- just in the private market, right? Andreessen, of course, is a big fund, like Lightspeed, operating billions of dollars and Thrive, and Coat is coming, and they, they often invest in all the competitors. And, well, it's okay that Sam asks. He, I think, has only one small request to investors to invest in competitors. That's what he's asking. I think these companies are less successful at a more early stage, too, often so requested. They often ask and choose investors on the basis of how loyal they are, in blatant words.
I'm curious here, and the rest of the companies, they're gonna get in some way affected by investment or so many investors on the market that, in general, Anthropic, and the rest of us, too. Take the money? I apologize if it's a very, uh, diletant question,
but--
There's one question here, here, as far as it's clear, there's more, uh, strategics that give, uh, GPU, right? I mean, NVIDIA, they're directly giving, there's Microsoft, Google, Amazon. I think that these demands for strategics are a little weird, because strategies play a completely different game. In general, yes, as large investors, they have already been dispersed between-- large investors, they have already been distributed between the text of the foundational companies. I mean, his request, I don't think that somehow it's gonna have a real impact on the state of affairs. I think that this request, I think that just someone wanted to make a news story. I don't think there's a big fund that hasn't invested yet. That's right, he's got three billion of them. Well, invest in some of these companies, yeah, some sort of thing. I mean, all these companies have been taking some unrealistic assessments. So there's a lot of investment in there that's not just one billion, but more. Well, if you're serious, investing. And I think there are funds in there that have a potential billion or more, I don't think there's any more of those funds left.
Yeah. So, in support of your words about the Suzkever, we've already told him he's opened his company, and there was information that he's attracted more than one billion investments in the first round. That's interesting, of course. I mean, earlier, uh, the company that had a billion-dollar evaluation, it was a unicorn, and then you kind of started up and raised the money, so you don't have a billion-dollar estimate, and you're attracting a billion-dollar. It's a crazy thing to do.
Yes! Because I think everyone understands that they start too late, and that's why they need a lot of money. I mean, wow-- because they're even now in a lot of investments, like you said, 18 times less than OpenAI, and they're very hard on them. I mean, it's okay when there's some other business model or something, there's other models.
Yeah, well, you can start, then you can be competitive and start with little, smaller rounds. But in the case of Suzkewer, there's no other option.
I think I just wanted to add that if there is a statement, I just don't understand how often such statements are in open, how OpenAI showed no investment. It's like a little impulse to other funds that you're gonna have to spend even a little money, and you're not gonna get in that big game anymore. Because, in general, who is in the AI market, I think I set an example a few months ago of my closest comrade, very much in AI, namely, the inj-infector, the Santista. He said that you'd never do anything, you'd only have to be there, in top companies. Well, I mean, it's hard to do without access, I mean, without access to resources, without access to strong contacts. And that's why the money's been spent, I just think a lot of money is being charged. It doesn't mean the industry's not gonna develop. I'm sure there's more to go on, and more money. And the people who want to invest in it will be more. I think that people who invest-- first, a huge number of people invest even in projects that we don't see, yes, that go through the funds, but just when a man sits in his own. company. I know myself about a dozen businesses that in their companies raise some kind of internal softs and think they're gonna sell. And how much I hear that another one of the series did some kind of cool elevator for his CRM system, which in AI is doing something there. It's an innovation, a fucking thing, and we're gonna sell it to the world, right? Well, that's a lot of-- that's just-- and this market, I think it's much more than start-ups. Well, I mean, the number of these different refinements, changes. Because a lot of companies have come up, uh, just when people have made a personal saft, yeah, for their personal company. Such companies are huge, crazy. Not what they did on the side. Well, that's a great graduation today. Thank you so much. And, Sasha, we're waiting for you to return to the valley tomorrow. We'll meet tomorrow, right? Tanya, we'll meet in three hours, two hours in the ocean.
Two hours.
Here. We're online. Here, online. Well, thank you very much. We're going out on Sunday. We're always four people. Technological news, Silicon Valley sites. Sign on the canal, you support it very much and put the liki on it. It really does develop it. And even in the commentaries, when you just write the word "thank you." And as I said in this video, it is important to hear your opinion as well as what you think about all the laws and where they're going to go, and what laws are still missing, right? It would be interesting to see what laws were missing. We'll have a very interesting subject in the next edition. We'll wait for you in exactly a week. Bye, everybody.