ChatGPT is not a product, it's an experiment, and we're not OpenAI clients. Their secret weapon is Sam Altman. Someone told me the story about him that Paul Graham once said, "One of the smartest people I've ever talked to is Paul Graham, the man who founded Y Combinator."
Gemini is just a technological solution or is it a product?
What about Anthropic and their culture?
Hello, everybody! We're on ToTheMoon. Technological news, Silicon Valley sites around the world. We have a special episode today. I remember we're going on Sundays, every Sunday and Wednesday. Sometimes we have a special occasional release. And we have a special special special-- special special special special one today, so meeting Maxim with my close friend who... Max, how long have you been, 13 or 14 years in San Francisco?
Well, it's the thirteenth year.
Yeah, the 13th year, which means he lives in San Francisco, very deep in the artificial intelligence, in AI, technically, in general, knowledge, contacts. And he and I have been discussing the impact of the culture of different companies on the products they produce. And just to discuss why, for example, OpenAI episodes certain decisions and some things are not being produced. Or why the gui does what it does. We've come to this subject of culture and we've decided to launch, write a special edition. I think you'll be very interested in him. We don't forget to support the channel, to slide and to Max write his own hello.
One of the fundamental questions I had, why OpenAI for the Post--Care, even not in the last two years, has not produced a number of certain products in the last year from the perspective of such classic products. decisions. For example, they take, there, they move the slide of the model infinitely, yes, or they launch Pulse, which is not developing or producing the Deep Research version, and then they don't develop it either. Or they release Sora, then forget it, then they release, they re-operate, then they forget about it again, then they release GPTs, then they say it doesn't work, then they release the plagins, they release, then they don't work, Then, then, they say you can make your own applications, then you release agents, then forget them, give me reminders. I mean, OpenAI made a huge number of different decisions, but it's like it's not getting them to the end. Yeah? Shoping let me out, yes, the shopping didn't finish either. And why is that happening?
Well, that's not a unique situation. Someday Google was just as sworn that Google was running some stuff all the time and then shutting them down. But here, as we've discussed with you, I think it's typical of the use or typical, uh, typical situation described by the Conway Law.
Conway's law is, well, it's not really the law, it's the surveillance of one engineer who, uh... The law states that any organization that develops a system is inevitably designing a design that replicates the organization ' s communications structure. So the organization, in fact, is in the form of some product it works on, it creates a copy of itself. As a team is built, it's gonna be a product. Yeah? And if we're gonna look historically at the fact that the company here in the Valley probably built not just in the Valley, I'm just a much better familiar with the technology companies that are here. I can say that the law is consistent with the reality of 100 per cent. I mean, if we look at how Apple builds food, right? And knowing how Apple is a little inner, I can tell exactly that the iPhone is the exact replica of the organization that built the iPhone. Here. So, Apple is basically the manager, the company is the manager of the sales manager. Not engineers, not managers, not people on top, as everyone thinks. No, it's a sales manager, who, of course, people hired on top of something, bosses, huh? In fact, the culture has defined the person they hired, and then the sales manager is the company, and they are therefore building the product. That's why if you see the product look very coherent, it's very unsuccessful, isn't it? Because the sales manager agreed on how it's gonna look. And if something doesn't match the view to any one of the things they agreed on, they'll fix it, right?
Then let's see Google. Google's got a different situation. There are a few products that just technology-- everything just gets so high. That's the same email. No one can build a Gmail another, yes, so far. How old is Gmail, huh? He was two thousand fourth or two thousand third year of graduation. For years now, no one can repeat the success of the Gmail. But as much as the product itself, it's not that stained. And again, even if you see how Gmail releases, when was the last time you saw Gmail's release? What's your Gmail on now? Everyone knows we have an iPhone 17 now, right? Here's September, Tim Cook comes out in September and the new product version is rolling out. What's the Gmail version now? Nobody knows.
She's got some kind of internal bill, yes, but no one has any idea, because Google's not the lead sales manager. They're not food people. Google is led by engineers who are essentially technology-related. They have an idea how to do something. They've got this idea. This thing is installing a Gmail or in search, or something else, right? And as a product, it's not like that. It's a living, uh, engineering-living engineering system. There's a manager's product, but I worked, uh, looking for Google, yeah, I just saw how it was going. We had 50 PM on engineers. That's what was doing there, some kind of color of the button. But, in fact, all decisions were made at the level, so, the in engineered leaders, yes, the technicians, and so on. That's what was then, respectively, heading the entire product. Well, we're right now to what's going on inside, aaa, AI corporations now, right? And if you look at them through Conway's rule, yes, it's exactly the same picture. Aah, who founded OpenAI? Five directors.
These are the people who were in Google reserver AI of the model. Aah, people who are standing at the origins of the company are recording her culture. So they crystallize, as they create, uh, a culture that determines what the company will be working on, what it has, people will have a s--- a-a-- the best interests. The most important thing is who they'll hire. They hire people like them. They hire people who match, uh, their values, who think the same way that the world sees. They've learned a lot of other directors, and I can tell you, I have a lot of friends there. Aah, this company is run by the directors. OpenAI is a separate, elite organization that essentially commands, uh, the pace of the rest of the organization, the whole big corporation. Although it's already being mad, maybe 10 per cent of the company. Here. Aaah, and now, if we look at the product, huh? And this is an experiment. ChatGPT is not a product, it's an experiment. And we're not OpenAI clients. We, if you pay them money, or even if you just use ChatGPT, you're just sending data for this experiment. Well, if you're delivering money, you're putting some more resources, right? Uh, nobody wants to create a better chat-bot. And even more, O-- the leaders of OpenAI say that, they just don't listen to them.
They say we're not here to build chat-boots and SORA. We build, aaaa, A-AGI, a common artificial intelligence. Here.
Yes, general artificial--
They build a common artificial intelligence. This company has stated it from its very foundation. They said that. They were originally going to, uh, be like, aah, zo-open, if that's what you're supposed to call it, right? I mean, uh, OpenAI was founded as a company that would build AGI against Google. Because Google has, in principle, started working on these, uh, these models, and so on. And the other few directors realized it would be dangerous if Google was the only company that would have this AGI. And to make a counterbalance to create a counterweight that is accessible to other people, not Google, but for that technology they went and founded the OpenAI. It's-- they weren't gonna build a chat-bot, they weren't gonna build a video remediation, they weren't gonna produce any products. It was, it was-- it was a lab that was doing a study on artificial intelligence to build a common artificial intelligence. They're still doing it. So nobody's listening to them. They're talking about it. Yeah? Et-- they don't have a task to release some product. And therefore, when they don't produce another product, it's not worth surprise. All--
Who-- don't be surprised at that. All you-- we see is the experiment, it's part of a big experiment. These are micro experiments that are part of a larger experiment.
the things to do-- some clarifications to ask here. If they built AGI, the reservoir is over?
Uh, yeah! Because one of the basic properties of AGI is that he, uh, self-sustaining, will continue to behave. Once you got a self-sustaining system, aah, you don't have to do it anymore, she'll improve herself much better than you'll improve.
And they're going to do it, yeah, other research, build, there's three new ASI letters, everything, and so on. Or they're real AI, whatever. Or are they what's gonna happen? Is that gonna be a product? AGI will be a product for, like, people? If they build AGI, they're researching or because they're researching, it's gonna be, uh, the same research product.
Is electricity a product?
Up... not clear. I don't understand.
Yeah, that's--
I understand, you're just asking.
She's so infiltrated, uh, in the world for us at this moment, yes, we don't see without her like a life. So she's so everywhere that if you're out of electricity, you don't know what to do with you, do you? You can't really-- you can't do anything, you can't cook food, you can't do your regular job, probably, yeah, for most people. Here. And, uh, AGI is likely to take a very similar place in our world. I mean, it's gonna be a thing that's gonna let us handle some of the tasks, yeah, but it's gonna be so intrinsic, that's, like, the world around us, that we're not even gonna think about it
as a product.
Look, it's very interesting. And if you go back a little bit to Google. You've been doing this great. We're going to go to Gemini right now. I want to say Gmail. You've raised a very interesting subject. You say, look, there's a friggin' product, just unrealistically cool about world use, right? And they couldn't make more of it. Well, for example, we couldn't make it, uh, some infrastructure, I don't know how weChat did, right? Or couldn't get a massager on her base-- near him, or he couldn't get a video conference near him-- he couldn't even get a hold of them. It's still a separate annex. So we couldn't do something on that application, did we? He's staying with this old, unfamiliar app. I have the main box on Google, although I use it, there's Apple's interface. I don't like postal interfaces at all. I, uh, feel like they haven't changed since I made the first post, in the '90s. I mean, for me again, I have some of these things so and you've made it sound clear now, I guess. Aah, they had this job, Gmail?
No, it's not. So the Google team has always been technological. Aah, when they decided-- Gmail was an experiment, too, but it was a technological, not a scientific experiment. Uh, the idea was, if you remember, Doo-- if people are as old as me, uh, watching it, they probably remember that before Gmail, uh, free mailbox, usually, well, in it, they could keep it, There, there, 25 megabytes, maybe 30 megabytes of mail, right? And Gmail was such a technology experiment. Can a free mailbox be made based on, uh, aaa, a google infrastructure that will allow people to never remove mail? They first gave one gigabyte space, which then seemed to be just some giant, giant space at that point. People haven't sent omegabite postal deposits yet. The Internet wasn't so fast, but, uh, then two gigabaytes were added. They finally made ten hygabytes. Now, even the size of this mailbox can be expanded for money. That's what the experiment was about. Make a mailbox and make an email mailbox that would allow you to never remove the mail. This experiment has been a complete success. Again, Google, no one in Google was gonna build any food. It was a technological challenge to build scalable, such a scalable, very well-scaling pro-e-e, a technological solution that would allow people to never remove mail. And by that, some interfaces were built around this case. Then they've done a good fight against spam and everything. Gmail basically killed innovation inside email because they made such a good decision, yes, which-- which was impossible to compete, in fact.
They've decided on the main task, so they've decided on
the main challenge of the man.
They're doing their best, but they killed Gmail at the expense of that. Then-- killed, killed innovation inside email because Gmail is, it's a good monopoly, so powerful, pro-long, good monopoly. But since they made the product free, yes, it's hard to make money on it. It's a mono-- it's a very strange monopoly that creates a lot of value, but it's not even captured, is it? What's the company's job? To create some value, yes, value. And then part of this value should be planted inside the company. And accordingly, you multiply the value on the amount it settled inside this company, which is called a-- well, English-- it's called to capture it, which means you've been able to take some part of it. That, that value, right? And your success and size of the company can be measured by, uh, this. The values multiplied by the co--- the percentage she could, uh, grab a seizure, right? That Gmail completely created a lot of value, grabbed the whole market, right? And as if this one couldn't, this value will invade, in fact, any of it. It's very, very close to zero, uh-
Make it.
It's very close to zero, yes. I mean, the money is fine. But why? Because they couldn't, like, create a car to take this value. If you look for a search, for example, right? That's where the situation was completely reversed. I mean, uh, Google really built a very, very valuable product, but then with a commercial, uh, copying AdWords, yeah, then, uh, by burying other ad products, they were able to find this value, The value they created, the value they created, they managed to take it. And that's why Google became such a valuable company, right? That's not how Gmail did it, it didn't work with the maps, because, uh, yeah, actually, looking for it was a s-- it was just, uh, it was an accident. So, uh, when Larry and Sergei were going to build a company around the search, their job was to build the best search. They didn't think about how they'd make money on it. They said out loud that we don't know, we don't think about it now. Something-- we'll create something very valuable and then we'll see. And Google's got it in his blood, and that's, uh, the idea that we're gonna create something of a technological, very technological, and then we'll see. Here. Oh, and, uh, in some sense, OpenAI went to Google. Well, there's five googleers who have built a company, yes, those people who have left Google and who have created a fabulous culture company in a way, right? But it's not technologically any, but experiments, it's more like a scientific experiment. Here. These are people who were primarily focused on RL, reinforcement learning. And they've been looking at this task a little differently.
Now, naturally, AI is already a technological task, very complex, isn't it? But the whole story is that neither Google nor ChatGPT nor OpenAI, they are not, not food companies, they do not build food. It's a com-- it's a technology company. Google was built around a technology task, which was to indicate the entire Internet and then, and, uh, to run the algorithm PageRank, right? At that point, this task, it was not done. There was no technological solution for her. Uh, Google has gathered a large number of very good infrastructure engineers, and they have been able to achieve this. And that's the DNA that we know and love as Google today, right? Ah, ChatGPT is a research laboratory that still handles one of these big experiments and periodically produces sub-operatives that everyone's trying to wear in that, uh, product robe, Yeah, and everybody sees it as a product. But it's just an experiment.
Tell me, what if Gemini take it? Gemini is a, uh, just a technological solution or is it a product? They've got them, they've done, like, Gemini, and around the feeling that models, like Veo video generators or Banana's picture model, or how are the products there? You can download a large number of data from different, you know, and, uh, work within AI. Is this about Gemini or is this just a different technology thing again? And that's why the interface is so stunning in this regard.
Well, Google's curve interface was always, say that. Oh, so no Google product I don't know a good interface, but they're just, they're not talking about it. Here. I mean, it's not a-- it's more like, this, it's an area that Apple is doing very well. I mean, graphic interfaces, of course, Apple skirts are very well reassured, and like other companies, too. But it's not about Google. Oh, Gu-- the whole Gemini story and the related products is all the reaction, and, uh, the outside irritant of ChatGPT, right? Gu-Gogle had all the technology, Google had people who wanted to work on it all ten years ago, right? Uh, and a lot of people have been upset that Google doesn't work on it, and they're gone, including in OpenAI, including in Anthropic.
In OpenAI, including in Anthropic. Yeah? A-a-m, and what's happening inside Google, is primarily a reaction of naa, popularity, uh, these AI, AI experiments, AI products outside Google.
And Google realized that, actually, well, it's nice to see you like that, yeah, and a little, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, a little bit, like, ChatGPT? It's just the next tying of the search. It's just the way people look for information in the future. I mean, even if we don't get AGI, the value is a Gi-Giant. Instead of licking around and integrating some understanding of the problem, yes, you can just start asking questions like you have an intern who's already got it all over the question. I did and I'm ready to answer any questions. And once ChatGPT could, for example, integrate, uh, advertising somehow efficiently, and it's gonna happen sooner or later, of course, it's going to be slow. Here, uh, Google can start a problem with their original product, which makes the most money. Yeah? And, uh, it's a reaction first. So Google has been quick and started integrating, AI answers to the search.
Maybe before I had to, maybe later than I should have, I don't know. But, uh, integration was one of the first integration, and it must be the first to be afraid of both OpenAI and Anthropic. This is primarily a search integration. Because we're talking about real, um, billions of queries a day, yeah, and-
Yeah.
The big one...
And the biggest-- the biggest search for peace. And we're talking about the biggest point of the entrance that people used to.
It's, uh, the biggest technological product in the world, if you think so, right? There's nothing else. I mean, people must have, you know, with the interfay-operative operating systems.
Operations, operating system, yes.
Yeah. But, uh, uh, oh, oh, no one thinks about the operating system. Nobody thinks I'm gonna be uploading to Windows right now. Everybody just thinks I'm gonna turn on the computer now, right? It's a subscription, too. Microsoft, of course, missed a lot of opportunities in this sense, too, but it's a different conversation. But, in other words, Google responds to, uh, in fact, the threat to their root business.
Yeah? This is a search. So ChatGPT really took out where they didn't want to--
Okay, where are they going? Here they are, look, they raised the threat, they raised the space product. Well, from a quality point of view, right? They're not a concession-- they haven't given up and failed to give up OpenAI. Okay, they've lost a big piece of view there, there, weekly use of everything. Although we understand that if they--
Mm-hmm.
They have, they still have a search, they still have a search, they still have a search, and yet the OpenAI has not yet taken over a huge portion of the search, there, the commercial, that's all. Where are they going? Gemini, where are you going now? They're just doing something, too?
No, well, in general artificial intelligence, uh, everyone believes. Google is very good, too, very faithful. Uh, and if someone has a chance, I think, it's one of those two players. I mean, why? Why OpenAI? It's because they have Sam-- they have Sam Altman, their secret weapon is Sam Altman. Uh, someone told me the story about him that Paul Graham once said, "One of the smartest people I've ever talked to is Paul Graham. It's the man who founded Y Combinator. Here, uh, he, he, uh, he had Sam Altman working for a while at Y Combinator, and he said that he was a man who, if he got on, uh, an island with some natives who don't speak English, through He's gonna be their leader for a week. In a week, huh? It's a man who can sell, snow of the Eskimos. It's a man who can sell water to people sitting in boats, Amazon. Anyway, it's their secret weapon. And by the, uh, Sam Altman, they might be raising enough resources to move forward. Because at the moment, it's a pyramid, in fact, right? I mean, uh, OpenAI re-- demon-demonstrates some kind of re-consciousness, not financial, mostly. The results are mostly technological and ideological. Here. And then, on the basis of these results, they raise the next round, yes, the next amount of money. Oh, Google still has a very serious business they can rely on, right? And invest, reinvest these money in, uh, building these more powerful models, huh? And there's a common art in mind. They already believe, yes. Uh, second, you're still, uh, secondly, you can think about how, uh, models are synergizing with the Google business. We were just talking about the search, right? Is that how much better it is, uh-huh, doing these, these are the genius models? The search has become a much better product. So they want to use it more now, right?
For some things that people haven't been useful yet-- they haven't used the search before.
You mean the quality of the routine interface or the new interface? Or are you talking about the old interface?
No, well, I can spend 20 minutes licking around when I do some reservoir or I can make one request, and I have a whole reserver right in the search panel, free and so on. And then, accordingly, the ads are tied to it, and it's better than they used to. Now, because I know exactly what you're up to, yes, uh, what you're wondering, I'm getting a lot better at you. But when you think about the synergy in other products, uh, uh, I was talking to a friend who's been doing video production the other day, huh? And he, and he says that, uh, that the need is unbelievable for a Ha--ca-ca-ca-ca-ca-ca-ca-ca-ca-ca-ca-ca-ca-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- Because these people are doing now, uh-oh, a platform video like TikTok, for Instagram, for YouTube, right? Ah, they see the possibility of spending much less money on the content. Look, you and I are sitting here, talking, spending time, and so on. And that's a very cheap content in a way. And if we could get all this straight why, why, why, why, why would you and I sit and build, uh, videos first, then sit, wasted, wasted.
Oh, and Google has the biggest video platform in the world. Yeah? So, uh, you know, if they're-- uh, synergi-- for, uh, synergizing aVEO, yeah, and you're gonna make a veop in YouTube? It's an incredible opportunity.
I mean, they're moving very fast, and they're already making a huge amount of internal decisions. For example, they started videos to cut into small rolls, generate. They'll be totally built. The amount of money they can make available. And they obviously have e--- they're obviously they have money.
With money, there's a story of interest with money. About ten or fifteen years ago, everything was fine and talked about, uh, a huge amount of cash, a cash, the cash that was sitting on these big technology corporations. There, Apple had something, $400 billion, just in the... in the kash. Well, that's a probative, of course.
It was, it was.
The rest of this is it. It's not like this, it's not that big s-- uh, big pool with gold coins, where, Zuckerberg, or, uh, Larry jumped and bought in it like Scrooge Macduck, right? Well, in some kind of liquid money, yeah, that was parked somehow. Here. But this, it wasn't reinvested money. And it was about what Dolina didn't know what to do with the money. They didn't know, they didn't have ideas about what to build, how to develop. And all, uh, everybody was looking at Amazon at the moment, which had the residues almost zeroed at all. And everybody was wondering, what's going on? And Amazon reinvested, Amazon reinvested all of this money, they built a giant logistics system, and now they're reinvesting the fruits of these reinvestments. And, for example, Apple and Google had no idea what to do with that money.
Now they know, now they know what to do with this money, and all these residues, they're in the right direction. Here. Apple wasn't so much of a certain Apple, of course, was wrongly sponsored, they're not the company that can build, uh, general artificial intelligence. They don't have it in their DNA. Now we're talking, starting to talk about culture, right? These are people who don't think, uh, the scale of experiments, they can't experiment. They have every product to be perfect at the time of the release, so they're so long... So, uh, they-- they're rumoring that Apple will finally release the warehouse phone, right? We can buy, uh, Samsung Fold7. Seven generations of warehouse phones went through while Apple was about to release the perfect. He's gonna be really cool in terms of, at least, hardware, right? Uh, Apple's warehouse phone will be perfect, proven, and so on. But it's not an experiment. They were waiting until they could make the perfect product. The same thing with iPhone was. People have been losing Apple for years and say, "Please, er, iPod Phone." There was a story like that, even someone was doing some jokes, uh, renders. They weren't called the renders at the time, but the pictures, to be exact, are there any kind of swings about what he might look like. If you googling, you'll find, uh, old pictures. Here. But Apple spent time. Uh, they've had a breakthrough, uh, quality, uh, tachoshrins inside the company. And then Steve Jobs said, "We're gonna build a phone around this." They spent time, they made good software for it, and so on. But it wasn't an experiment. They want to release the finished product. Uh, from the AI perspective, it is not clear how to make a ready product. Is that what a ready product AI is? All of it, it's probability systems, they're
systems. They'll always give something, with some probability, they'll be giving a wrong answer. And Apple can't produce that kind of product. They must have everything determinist, 100% perfect. Here. They, they, they can't release a product that is unintelligible. From their point of view.
Look, but it's good, but Apple, this is the Apple Intelligence story and the side button, and what they said this year in March, they're gonna be super cool, they're all screwed up. Well, look, you're still human, you're An-Android, so I'm Apple, right? It was just that for me... I don't know. It's certainly my life in Apple's infrastructure until I'm sure I'm in it. But it was a little strange to me. So they promised, they promised, they promised. They didn't. I didn't see that kind of failure. Why did that happen? They were scared of the market?
They were scared of the market?
In fact, this failure has been over 10 years. Ah, they once bought a company called Siri., this company, it grew up from, uh, Stanford, Stanford Research Institute. corporation that's been doing research in artificial intelligence for a long time since the 1970s. Here, and from this institution, a company called Siri grew up, and they bought it and named it their assistant., that's SRI Stanford Research Institute, SRI became Siri, that's the assistant, right? And, uh, that was a research project, too. The company was a research company, in fact, in a way, it reminded the OpenAI. There, uh, it was a doo deep learning model, and so people with, uh, trying to build things differently, it wasn't working very well. Uh-oh, and, uh-- I know the people founders of this company Siri, and-and-a-a-a, and I asked them once to do the best-if possible: "Why did you actually produce that raw product at some point?" Here's Siri when she did. If you remember, it was a totally useless thing, she didn't work.
Yeah, yeah, yeah.
I was surprised why you let it happen. And history sounds like that, uh-uh: Apple, when they got bought, contracted-- was bought by Steve Jobs at the time, and the deal with him was that we'd let Siri out when she was ready. So we're on our own, so the team will decide that the computer-- that the product will be ready, and then we'll release it., and, uh, it's been a few years, company, uh, that company was already integrated inside Apple and continued to work inside the reservoir, in fact, yeah? And at some point, Steve Jobs already, Steve Jobs is gone, he's gone. Uh, some big boss came to see them and said, "Ribya, everybody, we're going to start with the next version of the iPhone, we're launching Siri." They say, "Well, we're not ready. This thing doesn't work." Yeah? They say, "We-- we don't care. You have 12, 13, 14 months, right? You're doing this thing now as a product. Think about what chips you're going to start and so on, because we don't have a feature for the next iPhone, and we need a launch feature. And we decided that Siri will launch feature." This Apple can't... Apple had a patter that they needed some kind of launch feature. I mean, it's some new thing that sells the next iPhone.
Think about why I'm gonna go with the iPhone from 16 to 17?
Yeah, yeah, yeah, 100% needs a feature.
We need something.
I need some kind of thing.
Yeah, last year, it was definitely the thing that was a pgreade like that, that was definitely it.
Yeah, yeah, sure.
Although it's kind of like a pitch this year, but they're doing a really good sales this year, right? As a matter of fact, the question is, uh, but is that their mistake? What's this rate? Here, uh...
They, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, uh, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, uh, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, they, uh, they, they, they, they, they, they, they, they, uh, they, they, they, uh, they, they, they, they, they, uh, they, they, they, they, they, they, they, Here they were bought, April 2010, 15 years ago. Apple already understood the importance of artificial intelligence in her products, right? Just again, I repeat, Apple has no DNA from a cultural that will allow, uh, working on such systems. Once again, the--AI products, they're like, by the way they work, at least by the way we build, it's, uh-a-a, probability systems, they're pathetic, which means they're not guaranteed to give the right answer. That's when you press the iPhone button, it's definitely an annex Settings, right? It's not cracking, it doesn't open another app, it's guaranteed to open that app. And Apple used to build products that are like gambling. Yeah, it's got to work perfectly. AI is another product, it's just not gonna go through internal checks. If there's one of the PKs that, as I explained, basically runs the company, leads, uh, some kind of question, and that assistant's gonna answer it wrong, they're not gonna let him in, 'cause he's not. It works perfectly. They can't, uh, hang onto a product that doesn't... not working very well. Oh, uh, unlike, like, the OpenAI, which is a research lab, they don't even have a clue in their head that something's over. They're working on something, uh, something on the c-- something, right? That's their experiment. They'll continue working on it. They don't care how good he works now. Like the B-E-E-E, Google says Google has everything in-- is in a state of constant bay, right? Well, uh, ChatGPT isn't even a beta or an alpha, is it? Because alpha means that some kind of release will ever be. No, ChatGPT is an experiment in general, and we haven't even started to figure out how the product looks.
We have not started to formulate and formulate as requests for what the product should look like. We're just experimenting.
So, in ChatGPT, the vast number of projects that even come out where they're agreed, actually, these arrangements and partners, they can at any point dissolve if they're a reserver. Going the other way.
I don't know. You should ask the lawyers, right? But, for example, if you see, uh, what, uh, how, uh, it went with Microsoft, which was agreed, uh, love to the coffin, right? was pretty sure Microsoft would eat the OpenAI at that moment, because Microsoft is really serious political players, right? And they have repeatedly agreed on a partnership with a large number of companies. And then these companies broke up. There, and there was one Microsoft in the end. Uh, can we remember at the beginning of Lotus 1-2-3, right? You can remember IBM et cetera, and that. Such giants are easy. That IBM was a giant, it was the computer corporation, right? Where's IBM now, where's Microsoft? Here. But, uh, Microsoft didn't-- he didn't take one thing-- it's Sam Altman. Sam Altman managed to overplay them somehow. I don't even know how it happened. Uh, I'm not in politics, so I'm having a hard time knowing what really happened in this company. But, uh, Microsoft is actually overthrowing his partnership with OpenAI, right? Uh, and they're no longer their main partner. And the same thing could happen...
Why didn't Microsoft do it?
Why didn't Microsoft do it, uh, model?
It's also, but they did something there, of course. But it's also a company that, uh... It's not-- it's not their DNA. Uh, the company's been a long time since becoming a corporate-- corporate software provider, right? They're very good at selling what they have, right? They're very good at working with corporations. They managed to, uh, convince the American army, uh, buy--- they convinced the American army to buy virtual helmets. was in shock! And yet these helmets were completely useless. So I used one paradise-- once, and I was completely uninterested, right? They somehow managed to convince, uh, the American army to buy some thousands of these helmets. Here. That's their DNA, right? They can't build the same thing, uh, experiments. Again, Microsoft, invested in artificial intelligence in a lot of years ago in the 2000s. Uh, they had an office in Cambridge, and they're a lot of people who were mas-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-e-money training, yeah, they were able to hire them. Here. The same as Google at his time was able to hire Hinton and, on the basis of that, make the first, uh, to work the first, uh, revolutionary models, right? Uh, just like Microsoft in the time, uh, hired, uh, people who, uh, lay down, basically, set the foundation, uh, this kind of skill training, which, on the basis of which all the artificial intelligence was built, Yeah? But it's not in them-- it was, it was, um, it didn't remember what it was called Microsoft Research, I think it was called an organization, right? They didn't do anything from that organization, but maybe some things for the compilators. I mean, uh, Microsoft is not the company that will build again, uh, modern, this scale. They're just, uh, if you look at them, uh, organizational structure, again, returning to the Conway Law, it's a company that builds very well-designed products. Yeah, well, if, for example, look at Word or Excel, it's so-- it's so uneven, not homogeneised, but very and very complicated products. If you look at any Apple product, it's gonna be so uniform, it's gonna be like it's gonna be, it's so flat, right? Everything works equally everywhere, the same interface, everything, it's very, very clear to users, but the complexity of these products is not very high. If you look at Excel, every menu is the world.
The menus is the world, right? It's the world. There, in Microsoft, in Excel, a team usually answers for substitution, a whole team, there, of tens of people who are all building separately, who are practically not talking to other teams. And that's why you go in another menu of Word or Excel, your hair just gets up on your head from the complexity that keeps it all there. But if you want to do something with the document, Word can do it. Not that you can find it. Not that you can understand how to do it. And there's Microsoft--
There's a ChatGPT now.
For that now, there's a ChatGPT or Copilot. There's a Copilot for that, yes.
Yeah, yeah.
But, uh, the Microsoft organizational structure, uh, it's what it represents, like, these teams, yeah, that are very isolated, and that are fighting for resources with each other. And that's what the products look like. It's like a lot of little locks, right? Is that how Word is integrated into Windows? I can't! Yeah? There was some story about some components that could be drawn from one application to another. How is Excel integrated with Word? I can't! There's something there-- there was some kind of attempt, uh, to do it.
Not normal. Not normal.
And, uh, Microsoft is a very political corporation, and they can't build anything like that just because of an internal organization. It's a need for an ex--- that's Google now doing a whole company-level experiment. It's a giant company. I spoke to one of Google's PM the other day, and he's very upset that, uh, inside the company, everything is stretching to, uh, that's the transformer based model. So we have, so Google now has a foundational model, right? It's Gemini, and everything's getting over it. That's what's supposed to work up her. He's, for example, very upset about the translation of Google Translate into her, because, first of all, he thinks it's very difficult to improve the quality, uh, translation of the work from that foundational. Models, because it's not the target of a team that builds the foundational model.
Yeah, yeah, yeah, yeah.
They have a common purpose, and the quality of translation for them is not-- not even a 30-purpose, right? Now that you get that, uh, user, and that's downstream of this model, right? You got it thrown out, and you're gonna have to get her to work. You have no way of building something and making--
Yeah, you can-- hard, actually, you can, on the one hand, like, improve the product, but on the other hand, you can lose the product, you can screw it up very hard. That was an interesting aspect. I-- we updated Alexa for two months or two months ago, and there was a renovation, there, closed users. They've got AI inside and they've updated. I had kids coming in in a week and said, "Dad, throw out this stupid system, can we get the old one back? Because she's slowing down, gluing, she's not answering time." I left it purely for my own experiment. But I can say that putting the alarm was hard. So you're saying, "Step the alarm" - you have a alarm instantly. Now you say, "Start the alarm," she might not give you a chance to-- don't answer or answer later, right? And this is a problem. Well, that's a problem. And the same thing in translation. You're in translation, you still want to get a super straight standard quality. And people get used to getting words, these are examples, there, showing some kind of variation that is. Yeah, it's a little short of translation, same as ChatGPT or inside Gemini, but there's a standard interface. It's interesting how they'll decide.
Yeah, well, I'm saying, but Google's ready to put the farm on it, right? So they're willing to move the whole company now to the tracks of these foundational models because, uh, this--- Google in their time, they had an idea about, uh, moon shots. So they even set up a sub-organization that basically shot the Moon, so-called, right? So they, uh, built experiments that could do something big. They were not interested in experiments that could be successful with a lot of probability, but something small to do. They were interested in experiments that, uh, would do something big, but with a little probability, right? That these are risky experiments that can make a big difference. Now Google is willing to risk the entire company, in fact, now looking, advertising, uh, a bunch of other products, dragging on this foundational model, waiting for the foundational model to improve so much and that You can, uh, run them so cheap, it's all gonna end up in five, ten years, right, right, right? Uh, well, that's a question, because, for example, the same cost of starting requests for these models, yeah, it's still much higher than the same-- the same transfer or search, right? That's the amount of money Google is spending on the generation, and, uh, on creating these genetically-generative answers in search, right? They're not getting back at all right now. It's an investment.
Sure.
It's an investment--
There's a Microsoft that never will take a chance. Apple, maybe. Apple took a few chances. It was at Steve Jobs. Tim Cook is going to retire, and he's not Jobs, is he? At Steve Jobs, the company was set up for a few times. It happened a few times, and he won every time. And by that, he got this respect and even the kind of imagination that's still a little left in the Valley. Here, uh, Google took a chance. Uh-oh, Google was this, uh, Mac-McOS moment or, uh, iMac moment, uh, when they decided to put the whole company on the line at some point in connection with these models.
Uh, it was-- they were responding to an external threat, but I think they've made the right decision to respond to that threat, so, uh, a va-bank, basically.
What if you take Anthropic, what about Anthropic and their culture?
Anthropic is a part of the box that was moved from OpenAI. It's the same exact culture, right? But it's a culture of early OpenAI. OpenAI culture changed the inner line when the money went up there. I'm, uh, this whole nonprofit thing, and this is, uh, human-oriented, a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- They still want to build, uh, this general artificial intelligence AGI, but they want to build it for another purpose. They want to make money., Anthropic is, uh, a company made by people who still keep some, uh, and-and-and-idea-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- And they still want to build AGI to, uh, secure the future of the human race, because it's not the fact that we're gonna need it. Uh, sure, that's a lie argument, right? Uh, if the audience doesn't know, the Ludiths are a movement, uh, humans, uh, ex-cloths, uh, at the end of the eighteenth century in Britain. Then the first, automatic buckle that created the tissue automatically, yes, based, uh, steam power, on the basis of the pa-paro engines, huh? And the tepisodes, uh, got bored. We don't need it anymore. Why would people want to do anything around us? Oh, and that was written, uh, some science fiction of the book, right? , the world was just as imagined that everything would be replaced by machines. That's the kind of culture that's coming from, right? Ah, it turns out that the lie-- the liars lost their war, of course, didn't they? And, uh, we're surrounded by cars now, but still people have enough to do. There's a modern idea, I think, that she's very much like the idea of the lies, saying that after AGI comes here, people will be useless. Uh, I don't believe it, yeah. I don't believe AGI would look like it all feels like it. Secondly, I don't believe that, uh-oh, people are gonna be so...
useless. Here.
Mm-hmm.
Of course, the way people spend their time is really bad, well, it's already changed. If, uh, before-- before ChatGPT came out, yes, how we spend our time, how we do our job, it's been very much changing over the past 20 years. It's been a little changed again. It's still changing. Of course, the way we work, the way we spend our time free, the way we find information, how we absorb it, it's re-examined. Yeah? But I would-- I wouldn't take the position of the lieaders that this is all about to stop immediately. Anthropic is the people who believe that what is needed is to build aGI. Not in one place, but the source-- the sources of the AGI technology examinations, they should be a few. Uh, and at least one of them should be focused on the interests of mankind, not just on the interests of the cap-capitalist system, right? And, uh, uh, uh, uh, uh, uh, uh, Peter Tile has a great book called Zero to One. From zero to one. She's very, very popular in, uh, the start-ups. And one of the main ideas of the book is
One of the main ideas of the book is that any company wants to build a monopoly in the end. What should-- she should build a monopoly. That's what any company's task is in the capitalist system, right? Build a monopoly. The idea of the people who run Anthropic is that in the case of AGI, it would be very dangerous. What if we have such a strong, uh, monopoly in the AGI sphere is very, very dangerous for mankind. Here. And they work, first of all, on the basis of the idea that we should avoid it. Here. And it's still a lab reservoir, and it's still a set of, uh, e-mm, nerd reservoirs that are primarily experimental. They don't build food, but they're a little more imaginative about getting out, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, like, a little bit of a imagination, like, that's what they do. So, their coding model, yes, they're more careful that these coding models generate.
Uh, and, uh, and, uh, and, uh, trust and safety, that's, uh, that's the safety aspect of models being released, uh, Cl-- that Claude answers to you, he's gonna be stronger, because it's part of their DNA, they're worried about That's it.
What about the xAI?
Oh!
To complete, add to the end the XAI culture. I think it's just a cherry on the cake.
It's a very interesting story. Uh, a little who knows, but one of the original investors in OpenAI was, uh, Mask. The mask was one of the first investors in OpenAI, and even in this regard, the court was when it was finally thrown out of the broth, right?
Yeah.
Uh, there's a few court-ordered litigations on this. Here. It didn't end. I'm thinking, "HAI" is a few things. First is, well, it's the richest man in the world, and he wants his toy, he wants his AI, right? The world's greatest AI. Here. And if the richest man in the world wants something, he does. Second, uh, we're still in the bubble, like you're cool, right? We're in a big, big bubble. And he saw the opportunity there. It's a man who sees the opportunity very well, right? He's proved it several times. He built some completely fundamental companies, yes, for the s-- for his industry. And he saw that, yes, it was, it was a really big opportunity, it was a big opportunity. And he could raise a giant money very easily under it. Yeah, he, he was able to gather a very strong team pretty fast. But like all the other companies, uh, Mask, they're tied to his personal-- uh, personality, yeah, and his, his identity. They're spinning around him. I mean, uh, speaking of the way I'm gonna... say it. If you don't like it, the door is there. Ah, and I know a few people who were, in fact, in the first ten. Ah, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, oh, That speaks a little about the culture of xAI. It's an opportunistic company, yes, which, in fact, has the opportunity, needs to be taken now. Here. They have no, uh, and-and-idealistic ideas, and so on. These are people who try to maximize the amount of resources and make them as fast as possible, right? To sell it all and make some money. I mean, it's, uh... I wouldn't even call it a product. This is PayPal, or not, is it? It was an opportunity at that time. There are people who saw that we need to create e-money. This is Mask's first company, right? That was X.com, by the way. The first Mask company was called X.com, frozen with PayPal. Here. That's what I meant. I mean, you need-- Internet needs e-money. How do we do that? Here. Ia, uh, it was just an opportunity, it was an opportunity. The mask is an opportunist. He saw the opportunity, built it. That's the xAI, it's the opportunist company. Well, they're not gonna make anything interesting, in my opinion. I think so. Because around a bunch of idealists, they're building a big vision, right? A-a-a-a-a-e-e-e-h-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-i-is is an opportunistic company. Grok is an opportunistic model, but it's not exactly what it looks like, and it's why it's not. Plus, they're a little off the rest of us.
Well, the only way they have is to introduce, uh, Tesla, right? Well, if they start robots out there, they'll let robots in.
How much Tesla is in the world, huh? Well, there's a million.
It's not--
Skok, scoka--
That's not her--
Yeah.
...it's not a billion installations, it's not two billion users there.
It's--
Yeah, it's not, it's not 2 billion or, it's not 5 billion, uh, android phones.
And while it's far away, and until it's a little shorter day sp-- well, not a time that a man spends all day behind his phone, is it? I mean, it's not, it's not.
Yeah.
A7, A8, A10, et cetera. And then they, they were able to move their computers that were available on them. This Apple Studio, for example, is actually the fastest, most accessible computer users can start a big model. If you're not an enterprise, if you don't have your date center and so on, the biggest computer you can, uh, launch AI model is a big Mac Studio. Because of that fashion-- because of the technical decisions they've put on the basis of these chips, right? And that, uh, that could test anyone, you could put it on Apple Studio, uh... Ha! Funny, huh? Apple Studio can put LLM Studio, download sufficiently large models and launch them very efficiently. Here. And that Apple doesn't use this superstition in, uh, computers that they're holding very seriously, right? What they have, uh, no, no solution that would work exactly on the device and use these models. It's just a loss, an incredible loss for-- and for the market, and for Apple, naturally, too.
Well, that's great! Uh, I'll see you guys on Sunday. You'll have to leave your comments, you'll be fine, you'll be there for the release. Bye, everybody.
Thanks for talking, Sas.