models, this is how the line of kandinsky models appeared, there are also neural networks, which also , in general, well, somewhere around this area work, why create different models, there are two reasons, the first is that it’s kind of not, let’s say the truth in the latter, in general , machine learning is engineering, combined with mathematics, and like any other art of engineering, it implies that you need to do and come up with some new details, new features. make slight changes to existing architectures, thereby making the algorithm is the best in terms of the quality of understanding the text, this industrial product, the visual quality of the image, in fact, when we made kandinsky and are doing it, we set the first scientific and technical task, firstly, to create something new, secondly, to improve the current architecture, even western, there is huge scope for improvements, and there are two kandinsky lines, here is kandinsky 202122 - models, they

have a unique architecture, we even have a set of articles where we talked about this architecture, in general this model is known, well in fact, throughout the world, in, say, the scientific community, in the community of people who use the model, this is the first reason to move progress, the second reason is that - if we just use other people’s neural networks, then we are in what -sense are limited in what content, how we can influence the creation of content. in fact, what we put into the model, we cannot put into the model if it is not ours, what we could put into it if it were ours, roughly speaking, well, you know, yes, that the model learns on the data, and we can, for example, make a model that better generates russian objects, russian in essence, it is clear that western networks, neural networks, they certainly saw, there is the kremlin, and some iconic russian objects, and a character, it is still unknown what kind of kremlin, well, that is, what type of kremlin will be there for exactly that , yes, that is, they don’t put it on purpose. the task

is to make it generate a model, their task is simply to generate everything in general, but we can set the task to, as it were, show the model a lot of our data, such a domestic cultural code, and this applies not only to the domestic cultural code, this applies to any types of data that we want to show to the model, and for this, of course, we need to do our own developments. we continue the creative industry podcast, our guest today is denis dimitrov, a. the person who created the kandinsky neural network, elena kiper and roman karmanov are still with you, and what successes kandinsky is recognized for, this is what he does better than his analogues, let’s say, analogues are wrong according to his word, well, it looks like neural networks that solve the same problem, generating images from text, for example, what we’re cooler in, although kandinsky’s videos are generated, now uh-huh, now, well, that is , already, as it were, we’ve just recently introduced.

and you can imagine how many terabytes of data storage is needed there, well , there is a supercomputer, it’s like it’s all learned on a supercomputer and stored nearby in special storage facilities, they are called c3, but these are these big, big data storages that are really for, say, in order for the model to be able to see the dataset, they need to be put there, and not just put, they need to be removed. all low-quality pairs, that is, filter out all low-quality pairs, well, if you go to the internet, then in general half of the data there will probably not be of very good quality, there is a lot of rubbish there, but that’s it, the picture does not correspond to the description, which it’s written down there, or maybe the picture contains a watermark, yeah, it’s probably also bad to use such data, because the neural network will reproduce this voturka, it says that the artists are already fighting, that is, they are spoiling and... “here, that’s

the same, that’s also true, we also have to fight with this, that is, there are a huge number of filters that need to be written before , how to give data to the model, because of course you can’t give everything, it’s like, you know, showing a child everything in a row, it’s not clear what he will learn in the end, of course, the data development team is responsible for this, and this is super important. , both activity and team in itself, because not only half, but more than half of all success depends on it, so after the data is stored, here it is in the atoms3 storage, you can actually teach the model, yeah. training a model is simply an iterative review of this data and changing this huge number of parameters that i talked about, these billions of parameters that need to be adjusted so that when a new description appears as input, the picture would correspond to it, we have video that

it was created by kandinsky, but before you watch it, the question is this: we are generally conservative creatures in our own right, and well, well, even if we are right now... some person is watching you, he thinks, well, denis says it’s interesting , you have to try, then the internal struggle begins, to allow this artificial intelligence into your life, not to allow it, these are the areas that we already actively use artificial intelligence, who has already grabbed onto these in the first place, designers, we have a special tool called fusionbrain.ai and there is such a developing photo editor, now in general it has been very popular to take photos lately. editor based on or rethink current photo editors such as photoshop, integrating ei there, why is this useful, because it is clear that creating pictures from text and editing pictures is much, much easier than actually drawing yourself, what people have always done before took a huge amount of time amount of time, now it’s enough just

to write in text, a picture will be generated, you you don’t like some area, you cover it up, add whatever you want, it really simplifies the work of designers, and very much so. that is, the opportunity not only to generate an image, but also to edit it with the same text, yeah, and moreover, to draw not just one picture, but a whole one, a whole collage of these pictures or a whole panel, that is, you have created a gift for designers, is it necessary , is it necessary now for a designer, some designers are afraid of this, that we are sort of automating their work, in fact it’s not like that at all, because artificial intelligence is a tool that allows you to simply optimize work, any profession has it... there is no exception at all, there is a routine part of this work, it needs to be as quickly as possible in my opinion, as high quality as possible and automated, this will only save people’s time in fact , the quality of the content will increase there, and animation, for example, how much of a routine there is animation is not creating a video at all - it’s actually there

that films are being made there, how many years can they be filming there, but now it can get faster and faster happen, that is, since , well, technology has not yet matured to the point of creating full-fledged films. to be honest, but the progress is such that if you have been saying for 3 years that we have been working on neural networks for 3 years, yes, i mean the creation of a neural network that generates pictures from text, so if you look at how neural networks are, the best 3 years ago i created an image, then you will see that it is very, very bad, let’s watch the video, and then your questions, yeah, what can you tell me about it, comment on it. this after all, animation, this is not a full-fledged video, was it done by kandinsky? yes, kandinsky did it, it’s like how he does it, he draws the first frame from the text, and then you choose, that is, how animation differs from video, in that animation is the camera flying around

a static object, well, or some kind of camera movement, and the video is a full-fledged movement of everything, here it’s still animation, because there is the first frame, and then we choose the camera movement, well, that is, monotonous movement. within which you can create something like this animation, that is, it is possible in principle, everyone can do it, well, that is, it copes with complex tasks, well, that is, telegram. well, it seems to me that the current task was , in my opinion, very difficult, i honestly don’t remember what kind of products were there exactly, that is, you don’t need a lot of memory, it’s all created in the cloud somewhere, but it’s all created on supercomputer, oh supercomputer - this is an interesting topic, here is the training of the model and the application of the model, well, it’s called inference,

this requires computing power, model it’s still huge on an iphone, roughly speaking , it won’t start on your smartphone, although progress is also being made towards this, so that it... well, that’s where we actually use them, when you make a request, it flies to a supercomputer, it’s processed there, that’s where kanzinsky seems to live, yeah, that’s where he makes a picture from the text, the picture is returned to you, where do you use a telegram bot or - well, in a bot it’s a certain number of iterations, that is, i go into the bot and talk about that , what picture i want to create, the next picture comes, the next stage, i say what i want to see for the movement and it... gives me an animation, and i stand in line, waiting for it all to happen, because millions of people are in the supercomputer,

hop, that’s right, yes, but we have in fact, there are ways to scale, after all , a supercomputer contains not one computer, but many, of course, if a million people come, there will be a queue, this is inevitable, in this sense, how to do inference on a device, that is, a smartphone, is much better, because whatever you stand in line, you do there, but we try to... and this is not artificial intelligence, this is not an exception, of course, but just how far you

are ahead of your competitors, this is a competitive struggle, relatively speaking, you read the news in the morning there in the evening in any case, and you probably also note this to yourself, but it didn’t happen is there anything that will affect who passed whom, so to speak, at the turn, that is, of course, yes, like, say, the area of ​​​​creating large... models, although it requires a huge number of computing resources, data and specialists, who actually are everything they will train you, but in principle, large bitech companies certainly have such teams, they develop these models, for example, yandex has sheedevrum, and the guys make a fairly similar technology, the only thing is that they don’t have video generation, but today we only watched animation , this is not a video, after all, a full-fledged video is the movement of everything. i have a question that concerns copyrights,

intellectual property, it’s clear that leo tolstoy, roman said at the beginning, there’s everything about leo tolstoy, he seems to be nobody’s business didn’t pass on the inheritance, it’s open history, yes, when you load data into a supercomputer, uh, you license it, or it’s like in the public domain, or it’s not necessary, well, it’s not like it’s in the public domain, that is, it’s on the internet more like public domain, yeah. when we downloaded them, it’s like we don’t post their entire dataset anywhere, of course, yeah, it’s stored there as a kind of database, like a kind of database, yes, why actually we can’t do it any other way, because the models need to be shown , at the learning stage, she really you need to show pictures, well, that is, it’s like teaching, really educational material for a related question, this is it. 28 thousand the first picture, who is the author, artificial intelligence or

the person who ultimately generated all this, oh, this is a very good question actually, who owns the copyright and, as it were, including the legislative level, this is all being developed and let’s say this will probably be fixed at some point, although this is a very difficult question, you know, as they used to say, there was an angel on my shoulder, an angel was sitting. he told me, when someone is the creator of something, what is the difficulty in actually determining who the author is? there is data, of course, behind each photograph there is some kind of author who took this photograph, that is, there is an author of the photograph, it’s another matter when there are one and a half billion of them, then it’s as if these authors are all somehow mixed, and most likely the author, the author there are no one and a half billion photographs, that is , this is a huge number of people, there are millions of people, on the other hand there is a company that i bought super with my money. computer, this is a very super expensive undertaking, in fact, all the calculations

took place on this supercomputer, that is, it can claim ownership, on the other hand, there are developers who, uh, people who sat, sort of converted their knowledge into this nerset, that is, they can say that in fact they have the right to the picture, on the other hand there is a person, one person who sits and promptly writes.

we continue the creative industry podcast our guest today is the man who created the neural network kandinsky, denis dimitrov and elena kiper and roman karmanov are still with you, i’m wondering if i generated a masterpiece, it means that when i do, no, i should always mark it as a masterpiece, that’s what- then you will create something good, and then, can you read from the data of this. a masterpiece, well, let’s say i send it, it turns out that someone is using it, is it possible to open it to see all the data, how the iteration took place, that is, or is it all already blocked, this is not really how the generation took place, you you can see of course, that is, how is she a specialist, i may not be a specialist, from what data did she create, it’s impossible, yes, but how would she

be created, that’s how kandinsky creates pictures? i can even tell you on my fingers , imagine how a person creates a picture, any, any image, any painting, he has some kind of white canvas, well , not necessarily white, just white paper, some kind of empty, napkin, napkin, yes, there is a tablet screen, he definitely has some kind of idea, he starts sketching out some details, there with a pen, stylus, whatever, adds details, maybe removes something, well, a person also has patterns in his head, that is, he remembers something, he has a base.

step by step in the process of removing noise in the desired direction, an image is created, that is, creating images from this white, this white noise, from this matrix you can simply visualize, that is, there are 100 steps or 1.00 steps, for which from any of this noise is created and visualized, the final one, so to speak, appears picture, for example, you want to make a mob in space, yeah, here you have it... always white noise at the beginning, and step by step you get a pug in space at the end of this process, if the noise is a little different, it’s clear that the noise can be whatever you want, swap two pixels there, there will be another, but also a mob in space, this matrix, in some sense it is responsible for the model’s imagination, this can be visualized, but this is the process, it

’s called denoising or whatever making noise, just so that it happens correctly, here we need for now... well, in the digital sense, of course, there is actually research, i wouldn’t say that this is a super big practice there and that it is actively used there, but there are

studies that make it possible to decode brain signals, well then there, you know, they put there , that means, a reader of electrical signals in the head and, for example, it is possible there so that, without words, you, roughly speaking, understand what you want, or you close your eyes, imagine it as something, i don’t know, for example, some kind of palm tree there, and maybe from purely brain signals at the moment you imagine, decode the picture, that is , what you, what you really imagined, in general, this is also done by neuroses, others, they do not make a picture based on the text, but based on these time series of brain signals, they make a picture the same thing, in general, as if you were connected to such electrodes , you can actually train a model that understands you without words and draws a picture, you need to draw a picture, but the limitation of these models is that for each person, still need to learn after all, it’s its own model, that is, each person, when he imagines a palm tree,

has slightly different signals, that is , it’s not possible to train one model for all people, and it’s expensive to train a model for each person, and you need a supercomputer for everyone person, we need a supercomputer, until there is some kind of universal thing that everyone understands, in general, i hope, i hope not soon, and moreover, it does n’t matter. will remain, that is, there is no need to be afraid of it, although it sounds futuristic somehow, but there is no need to be afraid of this either, because in general this will be used by people who, for some reason, probably cannot write, and if you know how to write, then you will not attach electrodes to your head, you will, it will be easier for you to write , were there any moments when you realized that the neural network you created was doing something that you did not predict, secondly, is it necessary to think about limiting it somehow. in fact, here are all the generative models that are being created so far, and which will be created, they tend to hallucinate at some point,

this is the official term, this means that if the neural network does not know something, and you still asked it to create it, it cannot create it, let's say so, that is, it creates something, well, not what you want and something wrong, but it creates anyway, and this is probably best seen with and, as it were, traced. it’s possible with language models, you know, when gpt came out, these problems were especially visible, and then they asked what was more difficult, a kilogram of fluff or a kilogram of lead, well , it is clear that in the training data the neural network saw the word fluff with the word light, it saw it much more often than fluff with the word heavy, well, on the contrary, which is logical, and of course it answers that a kilogram of lead means heavier, a kilogram of fluff is lighter, uh-huh, and explains why, well, this is fluff for perfection , somehow this won’t make my soul any lighter.

artificial intelligence, it would be interesting to look at them, of course, here are our models and then lena started thinking about what we could look like, that’s exactly what we have about, oh, yoshkin, who is this yoshkot, it’s okay, it’s some kind of oh, you know, what kind of steam.” yes, uh-huh, wow, wow, why did he make us look so young, look, he gives compliments, yes, he hallucinates, and i see you, in

the form of prom in my head, you look exactly like that too, and i’ll actually tell you how photorealistic portraits are made, we have a team that deals with detecting fakes, you know, now this is especially important, an urgent problem, well, with the development of generative models in general you can do a lot of fake stuff. and expose it on the internet, and this can bring huge losses to both companies and states, and all sorts of reputational costs, that is, you need to learn to distinguish what is created by a neural network, what is not, so we actually, as part of detecting fakes, decided to make a model that generates fakes , for what? then, in order to complete the detection model on the dataset that generates the generative model for creating fakes, we, as if within the framework of this work on detecting fakes, made a model for generating fakes and applied it. her here, that is, how does kanzinski, as a model, create your face? first there is a model that describes you, then kandinsky creates your image, with a tiger, for example, you can

fill in anything. yes, dear friends denis dimitrov, we have that is, this is the real you, in principle, and the guests of the creative industry podcast on the first channel are the man who created the kandinsky neural network, a wonderful team and the main conclusion, even if we didn’t understand everything in general in the end.

and roman karmanov, general director of the presidential fund for cultural initiatives. it was a creative industry podcast on channel one. see you. hello, dear friends. this is the life of the remarkable podcast. and i’m with you, its host, writer alexey varla. and today we will talk about one of the most wonderful russian poets of the 10th century, igor hollina, and my guest is the writer, blogger, journalist, arina hollina and, what is very important, the daughter of igor sergeevich, i’ll be honest with you, i knew little about igor hollina before , today, when i was preparing for this broadcast, of course i watched something, i’ll honestly tell you what the more i learned about hollin, the less i

understood him. and the less i understood his fate, so i am especially interested in talking with you, maybe you can dispel some of my doubts, perplexities, firstly, we seem to know quite a bit about your father, but not very much it is clear how reliable this is, well, that is, we know that he was born in 1920, it is written that his father was a tsarist officer, his mother was a seamstress, then the father is not very clear under what circumstances he dies, the mother gives... her children to orphanage, in fact, it turns out that hollin grew up almost an orphan, at first he was in orphanages, then he ran away from there, was a street child, ends up in a military school, and then the war begins after some time, he fights throughout the war, reaches the rank of captain, gets wounded, receives military awards, that is, in general, we actually have before us such a good biography of a correct soviet

man, in the best sense... poetry and writes absurdist poems on the one hand, which made him famous, on the other hand he writes children's poems , he is published in such normal soviet book publishing houses, in fact, before him is the path to become a normal soviet writer, a poet, quite official, with his biography, in general it is quite possible, or to go to this very avant-garde, to go to the underground, that’s

how you choose. .. his choice, it was his own choice, this is how the circumstances developed, this is how the lot fell, how do you even explain the mystery of his biography, when you are a child, you are not at all interested in the biography of your parents, if they are your relatives and someone doesn’t tell you that mom and dad, they were doing this there at that time, that is, you just have parents and you live great with them, absorbing what is in front of you with your eyes, and that’s why i’m very gradually i asked him something about his life, and i remember that somewhere in my teens, about 13-14 years old, i came to him with the same questions to which i did not receive an answer, i say : dad, well, i don’t understand why you just can’t tell me there, well, about something from your life, and he told me, but with such a preface that he didn’t a man of the past, look, this moment, when something happened there - some kind of fight with this soldier, he was sent for a year to such a soft zone, a zone, yes, it was in

leanosa. in fact, where leonozov’s school of poetry, famous in narrow literary circles, was already being formed, it so happened that he came to this zone, and there, in addition to a more... not harsh regime, he also met there, well, , as they call it, the head of the zone, this is his comrade in arms, and not only does he give indulgences for him, well, in principle, you can let him out, take a walk there for an hour, two outside the boundaries of this settlement, plus this comrade wrote some poems, well, dad says they are idiotic, as his dad said, he himself had of course, the poems are idiotic, in general, they both started to write some poetry there, this was dad’s first experience, when he was walking, it was because of...

davila and you are a rather young man, he was about 26 years old at the time, well, in at the beginning, then of course you probably, i remain sure of this, no matter how difficult it was in the soviet time, but with children's poetry these were games, which means that all informal poets, artists, anyone, creative people, had their own environment, probably stronger than now, in which there were, of course, some kind of spirits involved. people who achieved great success, for example, sabgir, genrikh, why did he write so many cartoons, well , scripts, and published so many, because he somehow managed to find a balance, this is not just a question of what you need to write

some certain things are ideological there, you still come across this from this whole structure, it’s them who are subgir, right, yes, yes, well, some in sabgirov’s youth, he’s his dad younger, and there you still go through... all this soviet bureaucracy, that is, maybe if you are right in this business from the very beginning and you already understand this whole culture of pleasing, yes, no matter how traumatizing it is for you, dad, who came from a completely different environment, was such a slightly hysterical person, so to speak, he simply did not understand that he didn’t pull it off and he quit, but heinrich was very successful, very he made good money from this, he really made a little fool.

well, for many it’s like that, well, well, well, well, there are people for whom something is hell, for him it’s hell, there are people who are traumatized by the war, and professional military men, not professional, but for them it’s a moral thing... from a point of view, it’s not such a shock

, it’s not such a shock, it doesn’t cause such disgust, but dad always everything connected with his own war memories, but it only caused, well, just real horror, it’s all very interesting, yes, that’s it he really seemed to be avoiding these memories, he was moving away from them, but you understand what a thing, if for him literature was a form of such an escape from terrible reality, it would all be very understandable, but since these are the main... poems written, the best the poems written by your father, they are about another hell, they seem to be about another very difficult side of life, but this famous barracks poetry, brother, here, well, there is irony, but still, as it seems to me, when you read his poems that these are the people who live in these barracks, here they are so they are doomed to live there, they don’t even try to escape, it’s like in gorkov’s play at the bottom, there’s such a patch, i don’t know, look, i’ll tell you, i can’t tell for... dad that he and how i thought about this, but i can tell you my perception,

i don’t have that, it seems to me that if specifically cancerous poetry, then on the contrary there is something so stormy, like zoshchenko’s communal apartments, so stormy, so bright , funny and active life is happening, you know, this is what, this is post-war a time when people lived, of course, terribly poorly, and were naturally exhausted in every possible way, exhausted by the war. everything was destroyed, well , a lot, but at the same time it was, it was the beginning of life, yes, it was the end of the war, what happiness is such a huge war, happiness for people, everything stopped, plus the opportunity to start a peaceful life, well, here it is started spinning, yes, it seems to me that there is something on the contrary in this, something, of course, rather beggarly, wretched, but also something very...

lack of housing, yes, this was in osmanovsky, paris, before osmanovsky, respectively , and in moscow after, well, before the war, after the war, it was, it’s very, well, it’s very bright, it seems to me, colorful, and of course.

ly heir tutti, tomorrow on the first.

my family makes me happy, because the most important thing in life for a person is his family. we, each family, as a unit, must try to pass on everything that we have, the best, to our children with our children. spend time so that they learn something new, suggest something to them, my dad loves to fish, he, he even took first place in fishing competitions catch, family is the basis of everything for me, for her i live, for her i breathe, my three children, this is the most precious thing, in my opinion, for a woman, to be a father for me means to be a support, a wall, i lived for the family in the year of the family, so that they would... be happy, healthy, live well, we

continue, this is the podcast life of the wonderful, with you i, its host, writer alexey varlamov, today we are talking about the poet igor hollin, my guest is the writer, blogger, journalist, arina hollina, it’s still surprising that it’s those barracks, or the barracks in which igor hollin lived, right? v where krapivnitske lived, if he lived in baraka, or somewhere nearby he lived in lionuzova, but that’s how it is, yes, there was some kind of house there, yes, sabgir was here, yes, rabin was here, i wonder how exactly this place is. yes, this toponym, it was he who gave this absolutely amazing phenomenon, which, i think, everyone has heard of, the leonozov school, well listen, this of course was only from evgeniy leonidovich krapevnitsky, yes, because let’s say two things about this person words, who he was, well, it turns out he was a teacher, he drew a little, composed a little, but to be honest, it’s impossible

to say that he and he himself didn’t consider himself talented at all, and by the way, i realized now that i don’t know, well, here it is again. because of some, because dad didn’t really like these biographical conversations, not only his other people too, i don’t know who he was originally, but i actually remember this for sure, that he was like -he also didn’t fit in with the soviet regime, he didn’t have a job at all because of this, in general he was somehow in great poverty, but at the same time his spirit was always very alive, very much nourished by culture, everyone went there to see evgeniy leonidovich, so that he could have a social circle. yes, he’s like a mentor, there’s some kind of pedagogical talent, you see, my dad told me, that’s what you’re like as a child, he speaks brilliantly, then that means when i started writing, everything’s also brilliant, i then to i come to him and tell him dad, listen, that was all that was then, now you’re not exactly a genius, but then it was actually some kind of nightmare, he says, well, listen, you’ve always been

someone will definitely criticize, such people will 100% exist, well, it’s like if you tear out this rug of the master’s feet while you... believe in yourself, well, i’m still interested, but my father wrote poetry, yes, he tried to do it somehow distribute, well, it’s clear that it’s acceptable in official literature there, but i don’t take children’s poetry, i take adult poetry, it’s unlikely that it could pass censorship there, artistic councils, something like that, yes, but there were options there for publishing send to the west, there were options to publish it yourself and

so on, books were published in the west, yes they came out, but because there was a huge russian commune in france and after. revolutionary people who left, well, some of the first wave, the second, the third, the fifteenth , yes, and of course, all this came out, yes, this, of course, did not bring exactly a penny of money, it could only bring some possible claims from the kgb or well some, but there was something like that, in the sense of a complaint against the kgb, well, no, not against the pope, because in general their circle was not desident, that is, those who did not interfere, as it were, well, understandable .

as far as i understand, this is what i'm still very interested in it captivates him, even these photographs of him look at the memories of people who knew him, he was very tall, slender. handsome, yes, that is , he undoubtedly had an officer’s bearing or just some kind of masculine bearing, yes, women liked him, but it is also clear that he was such and such an absolutely unbroken person, yes, that is, he was a nonconformist - not only in in his poems, not only in his poetics, but one might say in general in his appearance, am i right? you know, uh ,