[000:00:00;00] my guest today is the ceo of a venture fund, orbital capital partners. and for many years in the recent past, one of the leaders of the russian venture company evgeny borisovich kuznetsov evgeny borisovich hello, thank you very much for taking the time to come to the studio. hello, evgeny borisovich, you, ah, what is it called? a futurologist is partly a specialist in innovation. yes, you can use such a word innovative word and futurology. you foresaw expected feared. that one i don’t know the rise that has taken place over the past, there are six months a few months, uh artificial intelligence. after all, everyone was somehow used to it, and everyone spoke. well, nothing, no big deal. well, he solves private

problems. this is a thing that reveals statistical patterns, and suddenly there is a thing that solves problems for you, does things for you, talks to you anything and so on. it was possible to foresee the questions of urology of foresight, or was it a surprise for you, well, on the one hand. it was pretty expected on the other hand, of course it was enough unexpectedly how quickly it happened, but here now i would start with some definition and say that well, futurology is such a generalized one. what do they do when they talk about the future? and i'm a practical person, i'm an investor , and my job is to invest in technology and venture investments have the property that you invest in those technologies that are at an early stage. and in this sense , you simply have to survive the time, well, there for 5-10 sometimes for 15 years, because during the time you invest these technologies have time develop to the point where they are already visible on the economic scale, on the political scale, and so on

and artificial intelligence. in this sense , our forecasting tool lends itself quite well, because startups are associated with the development of artificial intelligence. send money 10 years ago and that's it, the previous decade. they were growing up, so it was quite obvious that a simple rule works that if talents and money start to deal with some kind of problem, sooner or later it will be solved and artificial intelligence. it has long begun to bring into our lives a lot of interesting applications and the question, when will he become so smart, what is it, but it will already be difficult. mother, it was only a matter of very close time that the gpt chat. uh, and , probably, more similar things will appear. he talks to you in a completely human voice and communicates and solves problems for you, probably, he says that, as always, artificial intelligence determined this. mm testing. it's an outdated thing that testing turned out to be a cool thing back when it was unattainable when it was difficult to imagine a machine that

one could not guess that it was a machine instead of him, which could, from your point of view, serve as criteria for the fact that there is no intelligence there , consciousness cannot be defined as you wish. you know, this can be compared to the situation when a person sat all his life in a closed apartment, then suddenly opened the door and saw that there was a whole huge world in front of him. here we lived in such a very narrow framework of such a cage of our ideas about artificial intelligence and invented it's a lot of interesting rules, but the same is testing. or the famous three laws of robotics are all outdated in one second, as soon as artificial intelligence. really grown up to, how to say the level, to which he has grown up and we are now facing a completely new phenomenon. we didn’t pro-we didn’t predict what it would look like, we don’t understand it. we are not even able to determine how he achieves the results that he achieves, so we are now in fact in an alien situation. contact, when there is another mind working

on other principles, it studies us. we are studying it very interesting. he studies us. eh, that's so loud. we can turn the switch for him from the socket , pull it out, and he will stop studying us. it ’s true, as long as we have full control over it, that key question is how quickly we will lose this control. nobody knows. please tell me what happened. why does this car not think before, now i started thinking, let me ask the question in a different way. er chat gpt is a language model. yes, that is, it's just the ability to speak well. no, artificial intelligence works this way. deep learning-based neural networks are designed in such a way that they independently look for relationships in huge data sets and try to extract some internal patterns from these relationships. and in principle, people in the process of learning. if you take our entire history, they are also involved. in the same way, they try to find some patterns, but we are limited by what, and the power of each individual

person is not unlimited. but as a collective creature. we need a language, culture, institutions, and a way to convey thoughts from my head, yours is very slow, absolutely . i have to write a scientific article; knowledge was not burned at the stake. in general, we learned how to work with artificial intelligence. no need. he does everything very quickly, and agrees on his own. he himself builds these hidden patterns and now the main problem is is that we would really like to peep into his head and understand, but what patterns he discovered there, and we cannot do this, because this is too complex a system and it is, in principle , opaque. it is so saturated with information that our brains will not perceive it, but is it true that what we have learned to do, uh, including due to increased computer power, but also , apparently, thanks to programming technologies. we have learned how to better teach these systems or teach them to learn,

rather the second. actually. this is machine learning. we can all the time improve the formats of this machine learning, improve its quality or change e, the types of data preparation or the type of submission of this data, for example, now the artificial intelligence competition is going on between systems that have colossal computing power, but not quite efficient ways, and processing information, well, the type of tokens of the famous a and those. who has less computing power, but has a more sophisticated system learning. and in modern, let’s say, competitions are still expected that the system, which is less powerful in terms of hardware , will catch up with these champion systems due to the quality of training. if this happens, then artificial intelligence will be in every teapot, please tell me, huh? the knowledge that demonstrates offers me in response to my questions there this or that chat i don’t want to call the same one every time, but hmm this is

a human level of knowledge. this is exactly the same kind of knowledge in terms of structure, as a specialist would say, and then the question. where did he get it from? whether he took it from an article by some specialist and simply compiled it, or he knows how, in an incomprehensible way, just as people create new knowledge, this question is extremely difficult to answer on the one hand. everything is traditional. they say that of course not, this is a statistical model. she just processed three-quarters of the entire internet and compiled that data at your request, though. in fact, artificial intelligence, able to produce knowledge of an inaccessible person, was, for example, like this, but wonderful experiment. it was more than two years ago, when artificial intelligence was set on a huge database of proteins with proteins. what's the problem? this is a volume doll and its volume determines its property linear and it collapses. and so, how this folding happened, people didn’t even know there was no theory, that is, not that there were complex formulas, but only a computer can,

no, there was not even a theory, there was no artificial intelligence with alphagol and deepmin, more precisely, facebook managed to do it in 2 months to solve this problem completely today - this scientific problem has been solved. and this is practically the theory of relativity, only it was not invented by einstein, but by a computer. and here is the question that you have already partly raised einstein explained and the creator of quantum mechanics. they explained what formalism, what needs to be used, and everyone who has and who has trained, who has the opportunity, can sit down and, in principle, repeat these calculations and find out with what force the sun speaks, how a radioactive atom decays. this can be very difficult. it is, in principle, possible to make some other history here. here the story is different, but not entirely different, but we all remember that einstein did not radically accept the quantum. exactly, because he radically disagreed with its principles in this sense, formalism is informalism, but there was a certain set of images that did not coincide on the other side. and what kind of formalism is taking shape in these numerous layers

of the neural network? we just don't know something there, maybe similar to formulas. something like theory. it's just that the problem is that for us it's a black box and in this sense we cannot blame artificial intelligence for not having any, and i want to reassure myself about the formal model by which it makes certain forecasts. and you don't have any way. eh, this is a fact to confirm that there is something there, like a theory, something like an image of a formula, for example, and so on somewhere inside this is such a general statement that we cannot neither confirm nor refute it. we know for sure that every time convolutions occur and e, systems of coefficients, systems of relations arise, some kind of metamodels arise, but how are these models arranged? even scientists who do this understand enough. for example, you know, on the one hand, it seems that from this colossal statistical processing. there are three quarters of 9 dozens of the internet no matter but really follows. well

, such a certain average taken statistical such a thing. and that new knowledge can not be theoretically created in this way. hmm. yes , but how is it possible? well, we don't know that's whenever we talk about artificial intelligence. we must honestly accept. yes, at least the fact that we hit a place where no human has ever set foot. we don't know we can, but we say yes. this knowledge is taken from all the data accumulated by people, but at least from a comparison of different knowledge from different fields and different subjects. you can already extract a bunch of interesting things because this thing. it is actually by definition between disciplinary. if so, you tried a linguist with a biologist by wife. they have already got married there and are already producing digital children, eh. it is quite possible that and this system will be able to step further the whole question in just one, how quickly, she will start to receive new data. look here. a familiar young man, desperate to find

an error in his own script. he's programming there. i don't know what, but thought i'd ask, uh, dpd chat got the bug fixed instantly. it's like it's all just enough, it's clear, because and machines in the language are quite formalized systems. and uh, their artificial intelligence. should understand a person much better than a person. oh good. and then ask him m-m. write for me war and peace, but not as war and peace. i think it's a matter of time to copy writing styles or and he already knows how to create some plot moves and combine them. i think he also knows how remained the question. just a really big idea. so far , really, what remains with a person and what is not yet a bunch of gpt. this is the very design of the system only answers questions. she creates what you ask for, but she has not yet

learned to ask her. roughly speaking, she has no fantasies. and this is for now the only thing that leaves us. at least temporarily at the top of the food chain. i understand correctly that she has. in addition, you said a very important thing, there is no fantasy of a request for knowledge. well, because everything that she reads is open to her. you read everything she doesn't have. is it true that she has no emotions. and we, damn it , consider it a human property. this is possible, but it is quite firmly stated that the artificial intelligence system of the current generation has emotions. no, although they can mimic emotions because they read wonderfully, er, all literature, where these emotions are full. and in this sense, portray yourself as a psychopath girl for artificial intelligence. this is in general, the task is there, how to send two bytes, as they say correctly. but, when i see that the artificial intelligence in which i communicate is emotional. how to say it in russian, but expresses emotions. i'll tell myself, but this is an emulation. it's just an emulation. and you will tell me, but when

ivan ivanovich expresses emotion to you. and sympathy is probably also an emulation. and this is a very difficult question. what is really going on in our head and this question is connected with what, thinking about how these chats are arranged, what we learn about our own intellectual e, the side of the thinker and the intellectual country, the physiological structure of the brain, and the functional device touched upon not educationally interesting topic, that the entire history of digital civilization and somewhere around there, well, in the narrow sense, 30 years in the broad sense, there are 680. and we actually look at ourselves in the mirror, that is, you take all sorts of things. this is actually an attempt to look at themselves as a reasonable machine and an attempt to present their consciousness thinking as a reasonable machine, and until recently, these approaches were extremely limited. now we have received a huge boost and huge opportunities to move on. if we talk about emotional intelligence, then for the last 30 years. we realized that really the talent of making decisions is the talent or the talent of believing and

the talent of moving forward. this is largely the work of emotional intelligence, and in modern, for example, management, it is believed that for a successful career and for a successful main management of large companies. emotional intelligence. much more important than the usual far. not everything is rationalized. and emotions are designed for survival originally arose in our species , and even in modern conditions they help us to achieve success, if they successfully applied e quite right. plus, emotions have one more key task. they play a role as the basic foundation of communications. we are, after all , collectively thinking beings. this, by the way, also became clear relatively recently. that when people attributed to themselves brilliant the discovery of thought or creativity, they sinned a little against the truth, as they said, not that i was standing on the shoulders of all this, yes. we are collectively thinking beings and, accordingly, and they are machines. in a sense, they only accelerate the process of collective thinking, and in this sense , gpt chat, which can make a squeeze out of e products of collective

thinking made before us in 5 seconds, present it to us and we we're going to play with this. it is simply the rate at which our collective thinking process is accelerating. but what brings us immediately to the question of the network, and not the human network of communication is much faster than the computer network. and whether it’s being discussed, i don’t know whether the interaction of different specialized artificial intelligences is reasonably unreasonable. and it does such a task. and, relatively speaking, he will ask this one for me. here and here you can solve this problem, and then. if so, then this could be it. how to say super intelligence. so in artificial intelligence in a strong sense, because he is not trained to beat at go it’s trained to keep up the conversation, and as if all together they can do everything, this is what kind of person has already been described, one of the methods that are now building complex architectural solutions for teaching artificial intelligence, because in some cases it’s cheaper to train artificial intelligence, not labeled data, because such labeled data

means that people have to process with pens, there are video texts of the pictures and feed them to artificial intelligence for training. even now there are really specialized data, but a system that deals, for example, with marking up data for before learning with me, a disaster for us , a disaster for us, because it will be an exponential exponential growth machine remember how it started with alpha-zero, one machine taught another she taught her. so, i took it out for everyone in general, and at first there was zero knowledge. and here. eh, in general, we need to talk a little about the danger beyond intelligence, i’ll talk about this a little later. as long as there's absolutely no danger in it, because a system with learning e-cars to each other only works in fairly standardized situations, for example, in this way a number of car companies reduced the gap from the leaders by teaching the artificial intelligence of a car. they just didn’t drive on the roads, but drove in simulators, but in cars and thereby sharply reduced the gap, but it turned out that yes

, before learning to a certain level in the simulator, it happens very quickly, but then you still need to go out on the road and learn from real data and in fact, here the peak of the possibilities of artificial intelligence is achieved, as a rule, well , the top 10% due to 90% of labor, and these 90% yes. they are just related to the most sophisticated data preparation. it needs people, and very highly adaptive ones. actually gpt chat , or rather, exactly for this purpose and publicly their program, because they taught taught him how they needed people as a closed system, so that yes people would teach, and artificial intelligence is actually these millions. users who started crawling, they are training this one here are the top 10%. that is, he lives in symbiosis with us, in some way, yes, with us and with everything that we have created and with all our data, which are now stored in this way, but, and dangers from super intelligence. which first before moving on to the danger is very short. um, i'll use my favorite metaphor to understand how artificial intelligence works. here we

are used to seeing all intelligence anthropomorphically, that is, there are people, each of whom, in principle, is a rational being, but they have really high intelligence and high mind through the cooperation of coordination. and on earth there are other examples of intelligent systems. we simply did not really study them, for example, quite a long time ago. it is clear that, for example , the forest mycelium forms a reasonable system. he is quite capable of understanding what is happening with the forest that grows there over his body is capable at the right moment. there stimulate the birth of trees in one place. or ditch the trees elsewhere. that is, he has some kind of cognitive. feature, well, how intelligent is he, there, how complex is he, uh abstraction. it builds we don't know, but the fact that this is a cognitive machine is a fact. so this is artificial intelligence. it is arranged partly like people, and partly like a mycelium that communicates with more than one. that is, she does not have one body. this is a huge single organism, covering the entire space, but like a mycelium, covering, uh, everything, humanity,

but it does not have subjectivity on the one hand on the other hand. he, while communicating with each e, is able to create some kind of quasi, individual subject from the moment of communication moment of communication. yes and this is the gap - this is another form of intellect, and another form of intelligence is potential. if we talk about risks, then, of course, the main risk, which scares everyone terribly. this is not at all a job loss, because job losses have been in the history of this issue more than once. just like uh, social systems are able to adapt. everyone says lawyers will not be needed, translators will not be needed yet when we were translating the walls of shinist graphics. yes, he didn't die either. yes, they had to retrain. it was hard and so on. a the only question here is that it will happen not there in 20-30 there for 40 years. and for uh 10 there is 15 maximum. well, or that's why people need to somehow retrain faster. but this is not a dramatic problem, everyone is afraid of the other. if, uh, artificial intelligence develops at such a pace, then on the horizon

of decades. here, it’s probably still very difficult to speak, to be more precise, but at first it will reach a level really comparable to a human one, not only in the production of texts, but also in the creation of some abstract models and a. possibly even by system from the very motivation, and in the next step, it is absolutely inevitable that it will exceed this level, and the fact is that if human development and intelligence have a certain physical limit, this limit is determined simply by the number of neurons in our head by the number of people on earth and the ability to communicate well, the amount . the people participating in the communications of the speed of communications, this whole, so to speak , set of variables in the equations he sets, a certain limit of the reasonableness of human collective thinking, artificial intelligence, developing at a much higher exponential speed, does not have such a limit, because moore's law is not going to stop. another 20 years. and this means that the cognitive machine will become several orders of magnitude smarter, there already in 10 years i

need this explanation for your words directly required, but it is implied that in some sense people do not have control over the level of development of artificial intellect. this is somehow difficult to understand on the one hand. we program it, but on the other hand. you say he will become he will become. we want it or he wants it. this is an unusually interesting process, where it is impossible not to talk briefly about economic and technological history. here is a wonderful slide. and i, probably, will dwell on it, but briefly, but it shows that in the history of mankind, and for a long time the dominant economies were india and china. they feel great until technological progress began in europe with a technological revolution, and from now on. how do you see europe and then the united states has simply completely turned the whole story upside down. this story has two beginnings - a formal start - this, of course, is the british age of industrial revolutions, then

the european one, and so on, but there are even earlier starts, namely the age of discovery. here is my favorite example, that china swam to africa and somehow didn’t overtake it before the europeans. almost for centuries but what they did they burned their entire fleet and executed the admiral because it was all kind of against their traditions. after that, china began to gradually degrade technologically and socially. that is, he began to lose not only relatively, but also in the absolute level of income and by the seventeenth century. although they still remained larger than europe, they were no longer technologically and politically serious for europe. as a rival, europe itself , therefore, first captured india and then china, so now when everyone is talking, is it possible to stop the progress of artificial intelligence theoretically? yes, but this must be stopped. this should be done by europe and china and america and then everyone understands that in this game the stakes are so high that the one who stops first will actually shoot himself in the head. and this is a very interesting point, in fact, artificial intelligence. the one that has not

yet taken place has already caught people in such a logical dilemma in order to control the development of artificial intelligence. we must be a united humanity capable of negotiating the most complex issues. in the meantime , it doesn’t even smell like it, and it turns out that in order to stay in place. gotta run fast absolutely accurate and yet, as far as people may not want to. uh, consciously without consciously resisting the introduction of the emergence of artificial intelligence around the example, but self-believers, if unmanned vehicles are not very gone apparently. the feeling that people don’t really want to, one accident and already a scream that artificial intelligence has crushed someone, although living people do this much more often, what potential does such an everyday person have. this is something new to resist. hiera, i don’t know, there is theoretical resistance of the machine there is such a potential and, of course, we remember

the luthdites, who were quite noticeable by such a phenomenon in the world of industrial england nonetheless. we remember that somewhere we were suppressed, and technological progress will continue the opportunities that progress always gives in the end exceed the risks. this we know the whole history of mankind and, accordingly, when someone will slow down the development of artificial intelligence. someone will definitely achieve huge success by taking advantage of it and, well, the latter always defeat the former, this is pretty obvious, so now fear of artificial intelligence. he is in many respects really, and irrational. we're just so afraid. the fear of change is such a healthy conservative program for a person. in general, she insures here from approaching the abyss and jumping there. just wondering, this is a thing, but in matters of progress. and we simply do not have the right to stand. uh, on the spot, so to speak, because otherwise someone else will overtake us and then we will be in a very losing

position and in a very unenviable situation. it remains to wish everyone progress and productive and constructive interaction with what we are creating, so that it serves us as assistants and competitors in the sense that it would move us to a better greater and move forward. thank you very much. thank you. all the best,

goodbye.

we'll take care of this. it will be the honest

detective. the european union is ready to abandon ukrainian products, moreover, brussels is thinking of banning their imports, germany is refusing peaceful nuclear power in the country, the last nuclear power plants have been closed, now berlin has begun to speak for sanctions against rosatom about this not only further in the international review. hello on the air international review in the studio of fyodor lukyanov events and weeks of chronicles facts comments grain deal is out of fashion europe no longer wants ukrainian products

the agreement has lost its relevance here on the threshold of the civil war fighting is going on in the center of khatum the military are sorting things out. germany abandoned nuclear energy is not the best time for the country materials of our program. in this concrete cube beats the atomic heart of the reactor. energy transforms glass. a documentary film from the seventies about the construction of the ludmin nuclear power plant in greifswald. it was the largest of the two nuclear power plants in socialist east germany, five power units with a total capacity of 2,200 mw. closed in the summer of 1990, even before

the formal unification of germany, soviet technology was considered not to meet west german standards, and this week ended. the history of the entire nuclear power industry in germany was disconnected from the network the last three operating nuclear power plants from r2 not corvest him 2 and jr. in 1954, before joining nato germany signed the paris agreements , which eased the restrictions on its military industry following world war ii, the german government pledged not to produce nuclear weapons and use the atom exclusively for peaceful purposes. already in the mid-fifties, we’ll shine in the arche, the experimental nuclear reactor started working, and in the sixty-first , the first nuclear power plant was put into operation by today’s standards, its