Sasha! Sasha, you know what I've presented? You're always working in parallel with us, and you're printing. Do I know what I've presented? I was in-- here, two hundred kilometres from here to the woods, sleeping in a house like that. There's one bed standing there and there's no wall. The wall is kind of electrically up. And the owner came and said, "We have good bears, not like Tae, they don't attack people, so we can keep the doors open." And with me, it means a baby. I was nine years old with Igorem. I, Sasha, imagined that if the bear came out, you would be sitting, programming, saying, "Yeah, yeah, yeah, yeah, yeah. Now, now, now I'm gonna turn the door on." Here. Yeah, now--
So it's...
With the bear.
It's like it is. The most exciting time in AI. I mean, you can't--
I agree with you.
...stop opportunities. It's really hard right now. Recently, we have discussed that these processes, even basic, accounting, sales, vendor management, are changing from generative AI, all of them. Anyway, you can't boilerplate all work, you can't do just boilerplate. That's how you get used to it. It really requires a real, smart effort to keep track of what's going on, very, very difficult. I-- and now we need to hire the first person to buy, for example. Well, man, like people who bought some Oracle and used to selling them Oracle, like they're not exactly the kind of people who'll figure out what series A, series B, series C starttap is, in fact, yeah? So it's a very, very active time now.
You've raised the most awesome subject, by the way, because most of the staff in most companies and most of the managers, everyone is very much blocked, constantly blocking the number of different new ideas, I'm telling you, you're gonna do some tests, and so on. And basically, what you said, you have to sit around and roll over forever. It's not a test question, it's always an endless use of everything and a lot of stuff. What, we, uh, on the canal and we're developing this basic idea, right? Hello, everyone! We're on ToTheMoon. Technology news and sites from the Silicon Valley. I think we're on the channel-- that's our main theme. Sam Althman said a few weeks ago that the data you used in the models were not safe.
It's a really big question, yes, because topics that limit human freedom and life.
In any case, this cannot be used as evidence in court. And the wave on the Internet has come up, that everyone has been bored on it.
Oh, sports food, diet, that's what I'm gonna figure out with, uh, GPT chat.
I, Sas, where do I want to start before we get back to GPT chat. And, uh, I want to discuss this discussion with Sam Altman and Ilona Mask about who the world is monopoly and who makes the system for himself. Uh, I want to raise the subject of what, uh, Sam Althman a few weeks ago, uh, head OpenAI, said that you should all understand that your data you use in the models are not safe. And, uh, he brought the next story. By the way, it's probably fun for the Russian-speaking market, but, well, it's kind of a pretty big, serious sign for the American market, but it's, uh, a pretty serious, serious sign. What did he mean? He said that if you went to your lawyer or a doctor and told him some information, that information was completely depleted. It means that this information is absolutely confidential, it cannot be used anywhere, it cannot be used in court. If anyone tells her, a doctor or a lawyer, it's just gonna be a crazy responsibility for them for their entire life, and they'll be working for you and living for you all their life. Uh, this is a little different from Russian-speaking in general, probably mentality there in Russia, Belarus, Ukraine and so on, Kazakhstan. And he said that now you-- the main-- a lot of people use GPT as a system, and, uh, as a therapy, to get through their own case, problems, to figure out, I don't know, there, business, personal, child-related, health care everywhere. And he said that the data were currently not a law that would prohibit the data from being transmitted, yes.
So, uh, what do you think? The audience is ours, please write about this story. How do you feel about that? Do you believe in any safety whatsoever? Because many people don't believe in security, say it's all bullshit, and they're all still telling the story. But, uh, I think this is an important topic. And Sam Altman, what did he call for? He said that he'd be doing as quickly as possible, uh, base, oh, legislative, which would have to be included, uh, these are all the correspondence. Uh, I don't know how they're gonna share them. How do you guys know that? How will they share the types of correspondence? Uh, what can you use in court what's not?
Well, I mean, it seems to me that this is actually a super important question, yes, because, basically, like topics that limit human freedom and life, it's just, uh, it's legal, yeah, question and medical, Yeah. And it is clear that this issue of confidentiality is of great importance. It's about confidentiality that people pay a lot. That's why lawyers and medicine in America are the most highly paid professions, yes. So it's a standard way that immigrants come, they work as they can, and then their children, uh, get an education that all in all of them is worth two million dollars end to end, yes, either in the field or in the field. jurisdiction or medicine. It is not a value in knowledge, but in skills, not just knowledge, but skills. And here's the secret, yes, that's the one that's being pursued. And I think it's actually just a big opening for private models, yeah, for private models. I mean, I'm pretty good-- , uh, I imagine that's what it takes, and, uh, models like, I don't know about deep, but, uh, no fucking open--- well, so maybe OSS model, yeah, that's the OpenAI. Oh, well, maybe that's one of the skies. Because, uh, you can still do that if... I mean, if the open source model is very limited, what information will remain in any log.
Yeah. But look, it's all about it, too. Still, this information has been kept somewhere, and it can be abstracted in some way if access to the computer is obtained or not, or so on. Well, or you mean, you can, like, limit it, say that this information is stored 24 hours, and then everything is removed.
Well, anyway, it can't be used as evidence in court.
Information received-
Maybe. Why? Because look. So you just have to understand that there are two kinds of paradigms, right? Google search and emails are one thing. I mean, I understand if there's a warrant for anything, well, just a simple, okay, we're not gonna be passionate about it. Corporate, this is some corporate story, right? In corporate stories, for example, how it is arranged that when lawsuits are in progress, such as one social network of popular lawsuits that they use biometric data. There's a very interesting moment. First, all documentation related to this cannot be automatically removed at that time. But it is for companies that build a policy retention so that old, for example, the messages in the salad or the Gugle story are automatically cleaned. And often in the di-- in the deposition process, all these emails become public. Like OpenAI versus Elon Musk, yes, these famous emails about what Elon wrote that Greg Brockmann and Ilya are acting like boys or girls, I don't know how he meant specifically, but what, They're kids, they don't understand. These emails are publicly available. It's not a p-- not a fan people put out, is it? And because it's, I understand, part of it here, open versus closed AI argument. I think that's why it's just, I think, the concept of what Sam says that the OpenAI chats are, like, closer to this one, closer to a-la, closer to the e-mails, as far as they can be. used. I mean, if, by implication, a man uses the OpenAI to make a bomb. Well, Anthropic, okay. That's just a real example, right? , in the case of OpenAI policy, they may, on request of the court, disclose all data. Apple says Apple, well, they obviously have people saying they have bakdors. Ah.
Speak up here, Sasha, explain to the audience.
I'm not, uh-- I just kind of don't want to talk stories about my past work, right? But I want to give some meat, right? So I'm gonna say some weird phrases because wasn't prepared for it, that I have to repent here, but that's not why nobody's suing me, is it? Ah. So I'm gonna be so much abs-- sort of, yeah, just some stories. Well, for example, it's a lawsuit that the company dealers intentionally operated the teenygers, yes, they're sabotaging the product through the edital, right? And what kind of concern was, and what-- and the question is, respectively, whether it was made of good motivation or bad motivation, right?
So either the sales manager and there--
Either the sales manager and the general leadership, I don't know, CEO, they didn't know what was going on. And like, "Well, we keep squeezing the engagements, showing up there, the scary content of the kids, but we didn't know, we just didn't know, did we? It just happened." And then, well, as it could-- it's plausible deniability. Again, I'm not a lawyer, it's my understanding. And you can say, "Well, that's what happened, but we're not gonna do it again." Either if you knew and intentionally did that, there's no plausible deniability. And as it is, well, liability is getting more. The question is, how do you know if a man knew or didn't know? That's why we need to read his email correspondence, emails, some sort of paperwork, maybe Google search, maybe see if the salesman asked ChatGPT, like, what? We need to do the subway, raise it to 100x, and it's worth a meteor to be 100 million. But it is important, but it is very important that-- but it also says that we are breaking such politics there. What am I supposed to do, huh? And if these chat rooms are like a community-- well, as a public asset in a court, it's, uh, it's a liability increase. I mean, well, I'm the one who understands it. Again, I might be wrong. Just my experience, there, isn't very rich, like corporate, commercial, talking about this.
Well, he's right now, look, there's a situation.
Very interesting.
Sasha, yes.
And they were just watching these people ask what Google asked. We just-- we can imagine what kind of data is being asked in GPT chat rooms. I have a question: will we not be here to see that GPT chat room will be in advance... Look, if you're in China, look, if you write a phrase in China in WeChat, send a friend something bad, describing the government, just inside WeChat from one to two, you'll be coming straight straight straight away. Notice you're breaking the law. And my friend, who's a foreigner, lives for over ten years, said they'd come home to you. The police will be here with a lot of chance and will report it. I have a question: would there be no such thing in models that if a person writes something serious, there's a serious matter of level, someone seriously thinks a nuclear bomb should be done, someone seriously thinks, I don't know, in the discussion.
You're gonna be the one who kills the man, and that you're gonna be exposed to this data right now, like the police, that you're interested in it? However, the OpenAI system allows for an analysis of the amount that no one has ever had access to or access to any Google, so far, no search requests have even been matched, Yeah? And now she's analysing, has a volume to do with these data. She might have seen and pre-emptively see a person who could commit a crime, for example. Would it not be like this system would reveal this information to anyone ahead? Is that gonna happen? But clearly someone's doing these systems.
But that's good, Sasha? It's logical, I think, that it should be, or something.
I don't understand. One--
You think that information should be taken off?
Look, on the one hand, it's really adequate to find real criminals. On the other hand, uh, who's asking what a criminal rule is? The man who asked how to make a nuclear bomb in terms of scientific interest and to figure out how it is and what it is? Or, on the other hand, did he really want to do it? Yet, no one restricts the freedom and the human right to know what a nuclear bomb is made of. That's a perfectly normal question. Well, that's a good question. Uh, it's okay to get this thing straight and see, and have a wide range. And even, in terms of his own protection. The question is, why, why, why are we... how, how the system goes on, does a man know it or doesn't know it? What do you think? Elnar, what do you think?
Well, there are road cameras that are speeding, breaking the markings and other things that are controlled. Is that good or bad? On the one hand, yes, they're like, uh, a man thinks he's probably good at driving, driving his car, driving his car for 20 years, driving and all, and he can drive there, not for a period of 40 miles, but for a 60 miles an hour, Yeah? And it's okay, and it kind of restricts it. On the other hand, everyone else is comfortable driving, yes, and that's what it takes. So some rules are introduced that make society, you know, more comfortable, probable, less accidents and so on. And these, uh, rules have to be somehow controlled. Just put the sign on, like, well, not very effective, right? Putting a camera is a little more effective. If you remember, you and I were in Calcutta. In Calcutta, just on the road, there's a concrete block in the middle of it, just because speed limits on signs don't work, and a concrete block that needs to be moved slowly works, right? And just depending on what society is in, how it responds to the incentives that are, it will depend on that, and that control system will be, respectively. Uh-oh, and as for what happens or not, well, it's always just that when you do something, you're gonna come from what's gonna happen, right? Just that all you write there, I don't know, there, by mail, on the Internet, on chat rooms, on any chat rooms, yeah, wherever you are, always just keep your head in mind that, uh, at the limit, that's all that's always available to everyone. And here, uh, little one just commented. The subject is very serious, yes, but as a trigger that Altman's ma-- who's been pushing this subject up, maybe it's like that, remember, there was a story that was literally just a little bit lately, in my mind, plus-minus at the same time. It turns out that some of the Cups from the OpenAI that were sent to each other on the back, they invented Google and became available.
I was wondering if it was a Mashq trigger, for Altman to say that, like,
But it's a thrash!
- I need some base to defend myself.
But it's a thrash, to be honest. What happens, who hasn't heard this news, you know that chat can be screamed with anyone. And with that, uh, I think we're clearing the limits now. Even if you-- I was looking for, like, a neck with the person who's a plus, huh? I think we can make a big neckline now. Andy, these chat rooms that were in the neck, yes, they got access. I'll be honest, Illar, I came in later, I saw it later, a few days later, I came in, I had a bunch of chatty shearings, I thought, or if there's any chat room, I don't know, some of my personal documents or Some specific information I've asked and wrote because I'm also discussing in terms of some of my own, uh, highly specialized questions or episodes where we are being discussed by some small group. And all these chat rooms are open, they're gone. I wonder, in Google, I think Google then put a lock on these chat personally, because it stopped even showing up any index at all. He's kind of stopped taking this ID. I don't know, I've had enough time to... I've got resources that all index. I don't know if this side of resources has been analysed, but it's a big risk. And in fact, the idea is that the OpenAI should sue for it. I think the OpenAI should sue for it. Maybe we'll see a mass case soon, maybe not.
Well, that's why Altman started that there's no basis for, uh, legal, that's all that stuff, there, protected information, personal data and stuff. I guess I'm the one with the chain in my head.
It's getting a little bit of a hunch.
Yeah, yeah, yeah, cooks
the soil.
I will recall that the court is in, uh, New York, for the purpose of the action, I think New York Times v. OpenAI has made the OpenAI obligatory to keep all the logic, the logic of all the people of all the time unlimited. It's true for now. Except corporate chat, huh? It says except corporate chat, although I have a big question about corporate chatting, what's really going on there. It's a separate story that's still missing. And I will recall that in each system, you can at least check and disconnect the different scores that relate to the use of your data from the standpoint of the systems. At least. And in OpenAI, you can send a letter that will further and enhance the disengagement of your personal data, and they will put you in a separate extra piece. Well, at least I did. So who cares if you can add it again. But we have systems now a huge amount, and I'm not worried about OpenAI. I'm always worried about the middle type of company, because they're gonna get a leak at any point in time.
That's because in the middle companies, what's going on there, there's a traction, mostly always. I-- let's get this one. I think it's important to people. So Elon Musk issued a statement from the perspective of the pattern of behaviour of market leaders. Elon Musk has issued a statement that Apple assholes and they've been running out of the OpenAI. So ChatGPT is moving up in AppStore for the first place. And you see, well, maybe he wanted to say that Gropca had seven hundred million users, too. I just don't understand how. Because the Xx has three hundred, how many threes or millions, huh, the lasers?
And or activity, which means, and then, it means, the OpenAI is in agreement. What Sam Altman wrote to Elon Musk: "Well, we all know you used, and, uh, X to, uh, increase your own interests in your companies and make your chat even in the top, and that--
So your chat could be in the tope and read them all over the world. What, uh, did he write, uh, Ilon Mack, did you not get a hold of? Uh, your post read three million people. You're lying. By the way, this post was already read by 14 million people last night. Here. And what Sam Altman wrote to him was, uh, yeah, 100 percent, it's true. And there was one person who asked Grok who was right by Ilon Mack or Sam Altman. And Grok wrote, of course, Sam Altman's rights.
Well, it's actually blown up, uh, people's field. I'm the subject of this story. Uh, there's a lot of things in here that you laugh at and discuss or not, but you can go deeper to the question of how much companies really do do some manipulation. Well, that's what Sasha was actually telling you from the beginning. I mean, how many companies are building algorithms to their own interests. Well, for example, Apple Store, under his own interests, there, or, uh, Ellon, Mask, his own systems, yes, under his own interests.
Well, Sasha's gonna say more professionally now, right? I'm like a shoeman. Look, about the fact that the OpenAI has more active users than the Grok, but it does not guarantee it a higher position in App Store. Yeah, the same DeepSeek, he just got a hack and went off on top one without even a small share of the OpenAI users. Aaah, uh, in response to what, uh, Mask says Apple has claims, there's a joke that you don't have a Google Play claim, where you're too much lower than, uh, ChatGPT. Like, well, before Apple, he can dig up, and to Google, well, Google Play can't do it. Although, like, there, he's pretty much losing. That's it. That's a little comment.
Interesting. Sasha, what do you say on this subject?
I understand there's a thing like that that they only have Grok, and doesn't even have a Google Play, like Twitter, but there's no Google Play application. It's like a douche, Ilona. Okay. Ah, it means that, um... It's really just Apple. Apple's just got this problem, not a problem, just a trend, that they let third-party techs now, right? And that's a completely new opportunity for a lot of people, right? Go and demand something from Apple because we've set a precedent. And now, in fact, from the point of view of Ilon, logically, in terms of business calculation, it's you-- it's logical.
And explain, Sasha,
a little bit.
Well, Epic Games won the trial, as I understand. Well, or you didn't, I don't know how to say it. We won, it must not be that way. I think they've made a positive decision, so they're in trouble, which is why they've solved third-party, a-a--a--a--a--the---the-band-band-band-band-band-band-band-band-bags that can take less than the Apple. Apple's taking a big commission. I mean, it's a 30 percent chance, yeah, they're taking. And why am I saying why it's really important? Because, for example, we're gonna look at AI now, AI, right? In AI there are these manufactured crusades, they're in the area of 30, 40 percent. Real grass-maris is to take into account free-of-charge generation, and these are all fritrailes, and so on. If B-b-- take this into the cruss-bag, then the r-b-b-b-b-b-b-- take it 10 percent, right? Iia, and-- how would that be-- just at Apple you can't make any normal AI apps very successful because there's no brand. Che, there's a negative marathon that's where business is to be hurt.
Well, you mean, if you give 30 percent.
Yeah, yeah, sure, yeah. I mean, well, actually, now is a good time just from the point of view of these people, all this dynamic of sticking in them.
Well, in terms of the system in its own interests. Is there any big companies that don't use the system in their own interests? Well, I think all companies use the system in their own interests. And someone says, well, we're gonna do something better, uh, I don't know, above, uh, tweets, over there, or we're gonna do something better, or we're gonna put some extra algorithm in. That's a question people like about it, don't like it, whether they're gonna use it or not. I have a-- is there a-- do you see that problem? I don't know, for example, big systems of cool people who don't stick to, don't use their own interests. They say they help develop something. Well, that's how you get Google, Google the commercial. Or this is the Yandex ad. Does anyone think this commercial is about helping business? She's determined to make the best money, isn't she? Why is she conditional or, uh, why is Amazon, uh, so badly recommended? I'm not gonna tell you that. I was told by my friend Amazon when I came to America, so cool recommendatory system, but a totally shitty, recommendatory system. What's she like, a cool recommendation system? She's always offering me stuff that's totally unnecessary. But I understand her model. She's interested in selling, uh, I'm as much as possible, right? It's not right for me to rationalize, uh, budget, make some recommendations, and then, and offer me a whiff. Have you seen any system? Imagine you buying a basket, she usually says you forgot to put something else. Can you imagine the system? You buy a basket in the store, she tells you, I think you've got too many eggs. I think you need less water. I think you need less milk. I think you bought too much candy. So, this system, means that a real person is helping. So all systems are in manipulation. I'm just from the stand here, like-- someone right in terms of Sam Altman or Ilona Mask. I don't think Sam Altman ever needed Elon Mask to answer anything.
That would be a very strong position.
And, Sasha, and yes and no. It is obvious that this is the time at the generative AI now for such consumer companies. Consumer companies, they have to have some drama around them, they have to be on top of people mind. Otherwise you're not a consumer company. I mean...
So we talked about it.
Yeah, yeah, yeah. I mean, even we're gonna look at the most popular brands in America, it's Liquid Death, there's some very simple CPG, AG1, but they're also, they're just everywhere. And, as MrBeast would, it's a kind of thing that creates, and, uh, a drama around his personality. So I'm saying that all about what I think is just a hip, just a hip chasing, right? I mean, actually, this bif, I mean, I think that's a bif in court, which is what I'm not sure is useful to these two, for Ilon and Sam, but any Twitter bif is very useful to both of them. But I'm not sure the court's good, dear. Now, I think, given the location of the industry, there are no alliances, all play dirty.
With the courts, I think they're going-- hundreds of ships, every infinite, yes, or thousands. I don't know how many ships they've got. Thousands of endless ships against each other, against everyone.
Oh, although there must be companies that don't go to the courts much. I-- let's talk about ChatGPT-5. So it's been a week with a ChatGPT-5 extraction, and there's a lot of articles that say ChatGPT-5 is full of failure and fiddeck, and negative. And there are a lot of articles that say that, well, there were some glue in the early days and a great deal of superiority, and very much is being used. I've seen some of the programming articles. By the way, I've seen an interesting aspect of even programming, where there's a very large number of positive feedback. What do you say in terms of ChatGPT-5? I was, by the way, doing the show, I was making a graduation on Wednesday. We have a special edition on Wednesday, and there was a ChatGPT-5 issue. Look who's watching, you write off and remember to support our channel, and then you can describe your case of use and use of artificial intelligence in the commentaries. And plus your questions. We are gathering and trying to include such briefcases in our ongoing consideration. What do you say about ChatGPT-5 in a week of use? Tanya didn't say anything last time. Ah, Tanya said she didn't feel so much. How are you? And you said, "Oh, fuck, shit!" And that's it. What changed in a week?
Look, it's a matter of expectations. I mean, when you're waiting for GPT-5, you think he's gonna be on his head right now, he's not gonna get better on his head. Well, in my experience of using. Has it been worse? Yeah, it's getting worse. I mean, for example, in some deep research or in his answers now, for example, if you ask him to write a Russian, he can write a part of the words, there, the beginning of the word in English, the end in Russian. That's the first time I've been on three, I don't remember three. Yeah, it's a little too often. But there are positive signs, because, in general, it's like GPT-5 is more about free users than paid ones, right? When we sort of rolled up, you know, impolite, coolest models, just because we don't know what's under the hood is for all users and free, too. And that, to some extent, I think, triggered other companies a little bit of that. The same Elon Mask, for example, has raised quite a lot of laps--
For free users. And now Grok's got a little bit of a lot of it, like, with some limits, but it's available to-- to spin, try it. I was even interested in Grok, and I was riding this avatar for the first time in my life. Well, pretty busy. Ah, it's not gonna be useful, well, it's about communication. But, Tanya, then you remember, saying that when the personality comes, it's gonna be fun. Try Grok, talk to the personality is pretty funny, just in some kind of fun experience. The story is that OpenAI has access to, uh-a-a, the executive branch in the United States for $1 a year, thus providing access to its system. In a while, Anthropic runs out, says, "We're a dollar, there, three branches of government, all access." And there's a dispute on the other side, like, but you have users who pay you money, you don't do anything for them, and you give it a dollar for a year, yeah, whatever. What's the matter? So this is all it came up with the GPT 5. I see more of this than internal changes in the system, yes, as if internal changes in the system were quite minor.
Well, in my opinion, at least. A lot of people complained about GPT-4o being much more emotionally involved. Yeah, he might have hallucinated, but I was talking to him like a friend. And GPT-5 is dry, nothing answers, there's no imagination of the wood. But in the end, there's a deposit now in the choice that we can, uh-a, deprecated models, yes, these old models are. And GPT-4o was returned, at least for the paid. I don't remember. Well, there's no free choice. Oh, we've been praised, now's the last point, - we've been praised, like a good thing, that automatic choice of models, and now it's almost harder because you have GPT-5, have Thinking, Fast, there is Pro, respectively, there are deprecated models in which they can also be chosen. And it was like we should have left, and it was like we came back. Well, that's my latest experience with Grok. Grok, uh, the same story, just easier. There's a car choice there, too. Or you can hang out there. I was looking at a free version, but this is a busy experience. We're also recommending our audience. While you have a limite while you're out there, while you can hang free, try and see if you like any of it too.
I have a little mark. Yeah, I was very impressed that we were so critical, I think, fouro, and that kind of Internet wave went up, that everything bored on it, that, well, the
usual user liked fouro, I'm sure they're right, give us four.
Oh, yeah, actually, reminded me of that, too. Well, GPT-4o's missing because of some kind of psychological involvement. I was wondering. I think I'm there, uh, sports food, food, that's what I'm gonna figure out with, uh, GPT chat. And there was a post, uh, Doronicev, which is where he, uh, got his, uh, 42-year-olds, and he got a little food on his own with a GPT chate instead of a personal coach. And that's it. I think you should give me some interest for a stiff. Well, I'm in Russian with him. And now he's starting to get me started, that, uh, the coals have to be taken, uh, on the day of practice, and so on. Well, I mean, like, like, dry-to-community, but he calls carbons coals, right? It says, "We'll pull your retractions and, in a straight and portable way." I mean, he's trying to joke. So there's some of these things that I've never noticed before in O3 or O3 Pro.
That's by the way, there's one. He started using such words periodically.
Sleng is like that.
Yeah, it's a slang like that, and it's a Russian version. Where am I, where--
And I've got it too, and it's annoying.
...where I got hit, yes, where I got hit. He literally called me in a subject of my name yesterday, and I think I was telling you Wednesday in the graduation that he called me Cadillac a stupid system, uh, in a plan, well, that she did... I still think, "no way." Oh, my God! I was just angry at what--
Yeah, I get it.
...the electrical charge is made with a stupid interface, right? I sometimes am impressed by people who program, I would fire them, close them somewhere, which means they do X and UI. But he just started, he started to finish his words, and I got word "too" again last night, so, like, like, like, like, something like that.
I think he's trying to copy a little, uh, his user because I, for example, have me on the queries often when something's--
What are you implying, Tanya?
I'm going to see him.
Yeah, yeah, yeah. It was a shame, Tanya.
That you use the word "doo" very often. Sasha, you just don't see yourself.
No, I know I'm using the word.
And I often do mikes in English and Russian words. And he started to answer that. I'm so annoying. I want him to answer me or in Russian or English. Well, I think he's still getting a little bit of a little bit of a braid.
I can't stop using the word "doo." I'll tell you, so we're just in, uh, a few. I'll tell you now. And the X--- to Sasha tell you about ChatGPT 5. So, uh, I have neighborly kids, they're Italians, calling their bird, uh, Tonca. And my kids came and started calling my dog-- our dog Russian beast, we can call it some different numbers from time to time. My son called her Tonca. Tonca, Tonca, Tonca, Tonca. Even I started calling her that. Re--- just for some kind of fun. And we say, "What's that word?" And says, "This is some Italian word."
What a beautiful word.
Okay, that's what the bird was called.
So, last weekend, we're going to Tana for our birthday. Well, it's a long road, two hours. Pauline goes into ChatGPT and says GPT chat: "Look, kids call-- neighbor kids, Italians, they call Tank's bird. Iia, they said it was an Italian word. What does that mean Italian? And we want to figure it out and understand." ChatGPT says, "This is stupid." So my wife almost fell in a car stitch. And we realized that just the kids called a bird stupid. And we kept using it on the dog.
You go, and you have questions for GPT chat.
Maybe he, by the way, maybe he took the word tonco? Sasha, what do you say about GPT-5 chatting? Did your opinion and your love for OpenAI this week change?
No, no, nothing. Nothing changed. Nothing good.
You've got nothing good, right?
Nothing good.
Nothing good.
Look, I, I'll personally, of course, say the following. I, on the one hand, amazed, amazed again, as I was struck by a high-quality swing in the back of a video of how to release a release and then three days later, redevelop UX and UI and get more confused. I'm not sure how much users are going to get. We're even in the blog, we've done two episodes about GPT-5 chatting, and we're getting the interface we showed, he's already different. Yeah, it's a little weird. Okay, they'd be, uh, cool and upgraded to the system, like Hicksfield's updating every day. But that's not how OpenAI does it. So the cool ones don't add. They're coming and updating UX UI. They really had him confused. Why do you add two thinking? Up top of the model thinking and down there doing some thinking. At the same time. At the same time, my preliminary feeling is that when I use the GPT-5 Thinking model and the GPT-5 Pro model, I'll just say I don't use GPT-5. It's a totally stupid system for me, just like 4O was, right? And I don't use them. I didn't use 4O, and GPT-5 chatting is a periodic offer. And GPT-5 chatting is thin, 100 percent thin. So I use GPT-5 Thinking and GPT-5 Pro chatting. I'm in these systems, well, I like how cool they work. At least now. And I like some of the boxes they've got. I was just telling you Wednesday, see who didn't see. I was telling the agent box that I first started doing this personally, I was operating an agent inside OpenAI, and he worked very well through Agent Mode. Well, I mean, for the first time, I felt, well, I could say, I was a little bit of a quality-wise. I'm a little bit of a quality, and at least the system's gone. Again, I used the agent again, and I did all the tests on GPT-5 Pro, so I wouldn't go. I've got one for now. We'll watch, watch what happens. I think we'll see some more updates around the next week. Google's gonna be doing something again.
I'll tell you again, they're cooler than anyone. But I remember that Google Gemini's general rating is still in the top, right? I mean, now, if you look at the general rating of Humanity, it's the OpenAI that's missing, I think Gemini is higher, then Grok is some model and then just OpenAI.
Well, there's different benches, yeah. But Grok is very high, too.
Although, yes, maybe different benches. Yeah, I'm not in there much. But I think Gemini number one, right?
Well, we should watch.
I don't remember all the benches.
Oh, humanity, yes, we must look at humanity. Okay. Before we finish today's graduation, Sashi, what in the week of the innovations were the video generators?
In the video generators? Really...
What's so cool now, really cool for people?
Well, it's clear that Veo 3 was the first such a multimodal model, but it's impossible to understand.
model. Well, you can't use it, but it's fun. And as far as the reserver is concerned, it's a real breakthrough. Aaa and Chinese will make ten times cheaper in two months, ten times better with adoption, and there are millions of users. So, it's a synulous, not a singularity, it's loud. Social media is close. I mean, Chinese, Chinese, there's gonna be two or three months, it's not real.
What is this, Sasha, can you explain to people a little bit?
Look, you just, you just need to generode some actor that says something about the camera, right? And then you'll have to get a buckle-bag and so on. And to get audio straight, sound effects so that it's photorealistic so that you don't have to do this long post editing. That's what the Chinese will have in two months.
But does it mean that ordinary people can be so generosized and not understand very much where the man is real and where he is?
Yeah, yeah, yeah, yeah, yeah. It's been three of them since VEO. It's just that VEO three is very difficult to use and very expensive.
Look, I've been trying to make things up in different models and in the VEO, and in Runway, I've got it all like two thousand years. But it's all very...
Well, that's true. I mean, I mean...
I don't understand until I'm working. I tried again, and then in Runway, there were these effects, that the man is being released and so on. A little bit of use and realized I'm not a professional at all, that I need to be a professional man.
No, no, no. Only Chinese models, only Chinese models.
We're all gonna put it in Hexcell to the maximum extent possible. But American models are not competitive now, they're not competitive, yes.
And wait for DeepSeek R2. They say it'll be over by the end of August. It's gonna be pretty...
It's not real at all.
Mm-hmm. What do they say in DeepSeek R2?
One and two trillion parameters. I mean, twice as much as I was. Same mixture of experts. Aaah, they promise that, like, learning is a crazy cheap thing about what was at OpenAI for GPT. That they're all very cheap at school. They learn that, according to their statements, at least those that were on Huawei chips on their own, on Chinese.
Well, someone says they're bad, someone says they're good, but like DeepSeek R2, they say they're being trained, and the model must be very good. Let's see what happens.
Sas, and from Hexcell's perspective, you're very focused on practical case files. So you're not saying that we can endlessly create a lot of things like a VEO or Runway policy where they give a lot of stuff, you come in and you can't do shit. You have a lot of treasury, right? And you're giving live, lively, real workable cases within these treasury.
Yeah. So our job is to make these models as easy as possible so that we don't have to make prompt engineering and so on, right? Iia, that's what social media is, right? So we're having the best picture model right now, aaa, to generate these instagram--
Can you tell people that people, people, people hear what the best picture is?
Well, you just need to sgenerate UGC content. Like a youth photograph for Instagram. Like they do. Well, not-not-not-- not-- not-- take what fashion brands do, not-- not Louis Vuitton, I mean, but those who really hack into the youth. That's what you can do to make a stomatal, instead of doing this, you can go to Hexcell. It is clear that it is from the point of view of style, that it is a level of composing. It's clear that, well, as the AI can't pull that artistic, it can often not pull out the artistics that the top actors can give. But these photos don't even take off the top actors. It's a little more lucky than usual. I mean, in Instagram, I've seen a commercial that I've been down there, like, to make it sound more confident. I have a bold voice in there, I don't know, or there, remove the accent or something. And the people are telling their story. There is, obviously, like-- and this kind of positioning as UGC content, but with some basic light, it's been a long time since Hexcell. Now, that's a possibility, now, you can do more than a premium survey, for which professional actors are already hired, for which professional-- well--- vocational premises are being built, professional visualists. It's all possible to do with the AI, but it's still not a level there. Well, I understand there's still a great value from Sydney Sweeney's actress, because she has more than her appearance, it's a story, and so on.
Well, that's cool.
I think that's the key to everything. If you match the AI and the real life.
Describe our viewers who use whatever models to generate video or pictures and where you do, you've already tried and who likes what you have problems.
I'm seeing a lot of face distorting episodes in pictures and cool, and these are all the cool breakthroughs. Write down who Hexcell used, too, who other models used. Tanya, you're using pictures in a lot of ways now, or is it less than that?
It's getting smaller again. You know, I had a first break, a super inspiration when I was trying to visualize the tridash in a few seconds. But then, we were talking about this with our common friend Snick. I told him that somehow ChatGPT can't take geometry of space. He's not even in proportions, in size. There may be ten centimeters, he'll put a bookcase in there, something else. And he said that there was a major problem for the developers that they could not overcome it yet. But I kind of found a way to load the drawings, and he already knows how to do the drawings, and he's got three Ds on the skate. But still, you know, quality is so medium if you need to show a client quickly. But I was just saying that I was there on a course-- just on AI's video-card generation. It's so nice. There's a couple of escorts, and everybody shares their experiences, who's been generating every day. And so many positive feedbacks about Higgsfield. I read it every time, and I'm a little bit proud.
Of course it's like it's mine.
Yeah, yeah, yeah.
Even though Sashik is programming during the recording, but pride is still unlocking.
No, what's the programming? There's just a lot of things going on here. Yeah.
Well, it's humor, Sasha, what are you doing? It's love, it's good, it's good. I'm not gonna make that joke anymore. Yeah, yeah. I can hear about Higgsfield, too, but the bad news.
That's really fun, by the way. I don't know. I don't see a negative in person. I was on Sunday alone, there, on one party, and I heard good things, too.
There's actually enough negatives, yes. But you have to understand that-- you have to understand that negative comes from there, that there are the Midjourney and--
No, no.
Creators, creators, they, well, we understand that, from our perspective, yes, well, as from all perspective and from the point of view of product development and marketing, they need to compare to Midjourney because they have seven hundred and fifty. millions of revenue, sort of. And, well, as often as it is for these social media use cases, Higgsfield wins. And it's really funny that people write these paid posts, and that's, uh, almost always when I see those posts that aren't paid. I mean, there's no way we're even with these creatores. And that, well, yeah. I mean, to take out another market like it comes with some drama. I mean, I-- but it's just because the aggressive comparisons with Midjourney, but it's actually, like, like, in the world, this video, I'm all like scandals, intrigues, investigations, like, who, you know, can. Someone to move Midjourney and so on. So this is our own inner world, where, by imputation, Midjourney is the king of the mountain now and how Higgsfield tries to move them with help, there's more technology. Yeah, I see, because there's a lot of revenue. Well, the creators are also trying to hack and show that Higgsfield is better than Midjourney, because it's such a very popular tox.
But for me, Midjourney while I left it, I don't know, maybe I'll be back later, but I haven't been back to him in a while.
Yeah. I mean, I know you're saying Gemini's kind of worse out there, but Gemini's two points five deep sea-- deep thinking worse than three. I've always used three. Other models have stopped for me. GPT five. I'm going back to Gemini and, like Gemini, what he's looking for on the Internet is always better looking for, in my opinion, well, well, well, well, more useful. I mean, like me, GPT five is something that doesn't make sense.
Do you have Gemini for $250?
Yeah, yeah.
Well, I used it. I'm like the other one used Grok four, I used Grok four last month, but it's very hard for two sies--
Noo's very hard to use two systems. I think it's really hard to use two more systems. And you're getting something out of here. You're getting it now, re--? You're actually going to Gemini now? Everything moved?
Testing who tested Gemini, uh, the biggest version of the complex things you saw. We'll do a lot of tests on this subject, too. Cheek! Well, I'll see you in a week. See you later!