NTV — Infoshchit
20261012 012000 UTC
603 transcript segments
Google Cloud Speech-to-Text API (Chirp) + Gemini 2.5 Flash Non-Thinking
Data courtesy of The GDELT Project (https://www.gdeltproject.org/), from the Internet Archive TV News Archive.
Machine transcription. Treat it as a searchable index of what was broadcast, not a verbatim quotation record.

[00:00:00] to remain on the bright side and not succumb to false
[00:00:03] messages. Today, the news trending globally is that
[00:00:07] Russia is an aggressor and is preparing a nuclear strike against
[00:00:10] the West. As they say, we've never heard such a thing before. Again,
[00:00:14] it's true, the aggressor is blackmailing the peace-loving NATO
[00:00:18] bloc with nuclear weapons. We'll figure this out now.
[00:00:19] разбираться.
[00:00:37] Western society is being prepared for nuclear war, they talk about it more and more,
[00:00:40] so that citizens get used to it, and even nice and
[00:00:44] kind American citizens begin to believe that the Russians are preparing
[00:00:48] to push the button. A new wave of hysteria began
[00:00:51] after Putin's statement that tactical nuclear weapons would be deployed
[00:00:55] in Belarus at the request of its authorities. Well, let's note,
[00:00:59] only deployed.
[00:01:30] Out of love for peace, out of this same love, they persuaded Ukrainians to die until
[00:01:33] the last man, to kill Russians, such a territory is there,
[00:01:36] why are there so many people on it? Well, they moved their arsenals to our
[00:01:40] borders for peace and good, it's obvious, right?
[00:01:44] So we, in response, also moved some, just in case this good turns out to be
[00:01:47] with fists, we need something to respond with, why do Americans
[00:01:50] do this for their allies, deploying it in their territory,
[00:01:54] and by the way, training their crews, their
[00:01:57] pilots to use it if necessary.
[00:02:30] dangerous talks, it causes
[00:02:33] concern,
[00:02:35] concern, it causes concern. Does he even
[00:02:39] understand what he's talking about? Well, it's clear, the old man is old, but he has completely
[00:02:42] forgotten everything. What causes concern is that you have troops all over the world,
[00:02:46] what causes concern is that you regularly unleash wars like maniacs all
[00:02:50] over the planet, what causes concern is that without sanctions,
[00:02:53] you bomb, for example, the European city of Belgrade, you overthrow
[00:02:57] and kill elected presidents who interfere with your business,
[00:03:00] declaring them dictators. This is what causes concern, that you act like
[00:03:04] an overgrown bully in the yard, and the younger ones obey him
[00:03:07] because he will hit them. But the younger ones will eventually grow up,
[00:03:10] they will get stronger. Well, here someone has grown up
[00:03:14] a little, or rather, has regained part of their former growth. Well,
[00:03:17] they ask the bully, haven't you gotten too brazen swinging your fists? We can
[00:03:21] respond now too. He again explains to the neighbors in the
[00:03:24] yard, "This is different, we condemn this
[00:03:28] statement."
[00:05:00] What happens when incompetent people
[00:05:02] run the government? All I can tell you
[00:05:06] right now is, get on your knees and pray,
[00:05:10] and at the same time, persistently shortens the flight time of nuclear missiles to Russian
[00:05:14] cities. This already happened during the Cuban Missile Crisis. The Americans placed
[00:05:18] missiles near our borders in Turkey, and we, in response, near their borders in
[00:05:21] Cuba. The world was already then on the brink of nuclear war, but Kennedy and
[00:05:24] Khrushchev managed to agree. Both were in World War II
[00:05:28] and remembered that war is, of course, sheer hell. The Americans
[00:05:31] then removed missiles from Turkey, and we from Cuba. Then, in the seventies,
[00:05:35] a period of détente actually began. Now, in the Western press,
[00:05:39] a new generation of experts refutes the previous one. There
[00:05:43] will be no nuclear winter, well, all humanity will not perish, well,
[00:05:46] just a few million, such are these sweet
[00:05:49] democratic humanists. However, unexpectedly, common sense
[00:05:53] comes to them. British analyst R. Giles writes in a column for
[00:05:56] CNN that Putin skillfully delivered a preemptive
[00:06:00] informational strike, hitting NATO with nuclear weapons without
[00:06:04] using them. The way Putin's words about deploying nuclear
[00:06:07] weapons in Belarus were interpreted in
[00:06:10] the West may come as a surprise to Moscow, but there is no
[00:06:14] doubt that it will be a very pleasant surprise, because
[00:06:18] Russia has already used nuclear
[00:06:20] weapons; it has very successfully used them by not
[00:06:24] firing, speaking of potential nuclear
[00:06:27] strikes, to very effectively deter the West
[00:06:30] from fully supporting Ukraine. Well, that means they
[00:06:33] understood everything, we just needed to show them the pistol, because bullies
[00:06:36] don't want to understand words. Everything must inevitably return,
[00:06:40] we hope, to détente, although... some in
[00:06:43] Europe, it seems, still don't understand. Poland
[00:06:47] happily declares that it is ready to host US nuclear arsenals, that
[00:06:51] is, to voluntarily make itself a target for the
[00:06:54] first retaliatory missiles to hit. Poland is the
[00:06:57] closest, after all. Well, will the informational nuclear
[00:07:01] war turn into a real one? We will discuss this with Vladimir Kobyakov.
[00:07:04] Vladimir, the risks that everything will go wrong, these risks are always
[00:07:08] greater when there are fewer transparent and generally accepted control
[00:07:11] mechanisms. One thing is clear, that the main players want some kind of
[00:07:15] new security architecture. So let's see if it is possible
[00:07:18] now, at least in the nuclear sphere, to have larger forms
[00:07:22] of security architecture. So, what did the parties
[00:07:25] want recently when extending the Strategic Offensive
[00:07:28] Arms Reduction Treaty? How did they see the outlines with the main players?
[00:07:32] The States, even under Trump, as a condition for extending the START treaty,
[00:07:35] proposed that Moscow bring China into the agreement, saying
[00:07:38] "you negotiate with Beijing." Moscow then made it clear that it would deal with this.
[00:07:42] But if China was needed, then Washington, in Russia's opinion, should bring
[00:07:46] in two other nuclear powers on its side: France
[00:07:49] and Great Britain. But the Trump administration replied, "Oh no, that's a complex
[00:07:53] arrangement, and France and Great Britain are not building up their nuclear
[00:07:56] arsenal. China is a special case." Beijing then
[00:08:00] responded several times that such an agreement was not in their interest,
[00:08:03] because Russia and the USA have an incomparable number of warheads
[00:08:06] compared to China. From a geopolitical point of view, there was also such
[00:08:10] a version that it was simply beneficial for Beijing for the two largest nuclear
[00:08:13] players to constrain each other. Now, in the current crisis,
[00:08:17] some faces have changed, but the Biden administration is again
[00:08:21] talking about China. The Kremlin believes that the case with Ukraine
[00:08:24] confirmed that in the current conditions, all Western
[00:08:27] partners, including the United States,
[00:08:31] France, and Great Britain, should be connected to a large nuclear agreement. Moreover, if
[00:08:34] China agrees to join the general nuclear arms control
[00:08:38] treaty, Beijing will certainly point out that its main continental
[00:08:41] competitor should also be included in it.
[00:09:12] at least with seven participants, this, by the way, could have become the prototype
[00:09:16] of such a stable, polycentric world, or is this
[00:09:19] too cumbersome, will the whole structure not
[00:09:22] withstand it? Well, this is all, Volodya, of course,
[00:09:25] such pragmatic proposals, but in
[00:09:28] the current conditions, it all increasingly resembles idealistic dreams
[00:09:32] that everyone will agree with each other,
[00:09:35] everyone will somehow turn on their reason, but for now, judging by
[00:09:39] how events are developing, reason is still scarce. Yes, at the moment,
[00:09:43] this construct indeed looks very cumbersome and
[00:09:46] unviable. But we were discussing strategic offensive
[00:09:50] weapons, the treaty on strategic weapons, that is, nuclear warheads and
[00:09:54] long-range delivery systems. And there's also tactical
[00:09:57] nuclear weapons. Vladimir, do everyone understand the
[00:10:00] difference? Well, you and I understand it, but
[00:10:04] not everyone understands it, let's explain. So the large START
[00:10:08] treaty between the US and Russia does not affect tactical weapons. These are
[00:10:11] typically aerial bombs and carriers like fighters for short
[00:10:15] distances. And this issue is becoming relevant again. Well,
[00:10:19] look, by the mid-eighties, Europe was divided into
[00:10:22] NATO, blue, and the Warsaw Pact,
[00:10:25] red. In addition to European NATO countries with their own nuclear weapons, i.e.,
[00:10:29] France and Great Britain, hypothetically foreign American
[00:10:32] nuclear weapons were deployed in West Germany, Belgium, Holland, Italy, Germany,
[00:10:36] Turkey, and at an American base in Britain. The nuclear warheads
[00:10:39] of the Soviet Union were located at approximately ten bases
[00:10:42] in East Germany, Hungary, Poland, Czechoslovakia. There was a ready base for delivery, but without
[00:10:46] deployment in Bulgaria. By 1991, the collapse of the socialist bloc,
[00:10:50] the Warsaw Pact organization, and the Soviet Union and Eastern European
[00:10:53] countries meant Soviet nuclear weapons had already been withdrawn. In the nineties,
[00:10:57] Russia takes all nuclear weapons from both Belarus and Ukraine.
[00:11:00] By the late nineties, the first Eastern European wave of NATO
[00:11:04] expansion begins, but the Americans promise to remove their nuclear warheads from
[00:11:07] the Greek base. Then another wave of NATO expansion, and the Americans
[00:11:11] promise to remove their nuclear warheads from
[00:11:13] Britain. As a result, until the current crisis, American nuclear
[00:11:17] aerial bombs remained on the territory of five European countries, and Russia
[00:11:21] annually reminded that in this way the states are not complying.
[00:12:55] well, 45-50 years, this is indeed
[00:12:58] so, intuition trained by your
[00:13:02] experience, can you predict if something
[00:13:06] so terrible will happen or not at all? I wanted
[00:13:09] to add that, as they say, words are material. We see the rhetoric in
[00:13:12] recent times, it's shifting to the idea that nothing terrible is
[00:13:16] happening. If before it was such a sacred topic that
[00:13:19] nuclear war is something terrible, now we see that it's talked about
[00:13:23] as quite possible.
[00:15:37] a small excerpt of your statements from the late nineties, early two-thousands,
[00:15:41] let's listen to them. Any military person wants to have a reserve, and I
[00:15:44] completely understand and support that. The other thing is that this reserve can
[00:15:47] break the neck of any state. So we won't be able
[00:15:50] to just sit back, as we've been sitting for a long time. Either
[00:15:54] we'll be slowly pulled towards China, India,
[00:15:58] Iran, or we'll try to join in the
[00:16:01] opposite direction, understanding that we don't need to give up our so-called core
[00:16:05] interests. NATO with its expansion.
[00:16:53] Iran and India, but you said that NATO was weakening, has it
[00:16:57] weakened? Why did we end up at this
[00:17:00] point after all these NATO
[00:17:04] for many years, it was weakening, in principle, it's not very effective.
[00:17:21] The United States of America will never, under
[00:17:25] any circumstances, practically
[00:17:28] come to the aid of its
[00:17:31] allies if it would threaten a
[00:17:34] retaliatory nuclear strike on the territory of the United. All
[00:17:38] nuclear deterrence is such a deep
[00:17:42] theology, meaning American nuclear borders don't work. The problem
[00:17:45] is that there were no
[00:17:48] political reasons to wage war in Europe,
[00:17:51] these political reasons have now appeared, and
[00:17:55] thanks to a deep crisis in
[00:17:58] the West, which they cannot solve, well,
[00:18:01] a multi-layered crisis, I won't blame them now
[00:18:04] for their problems, but from
[00:18:08] this crisis they are trying to get out, including
[00:18:12] by fomenting hostility.
[00:18:15] What has happened is what is happening, a
[00:18:18] political
[00:18:20] reason has emerged for
[00:18:23] unleashing a great war. Everything has been heading towards
[00:18:27] this for a very long time. You didn't show my
[00:18:30] speeches from twenty-five years ago, when I
[00:18:33] warned us that if we allowed further expansion
[00:18:37] of...
[00:19:07] And I allow, I imagine that it
[00:19:10] is possible, uh-uh, that is, to defeat the enemy by this
[00:19:14] method, of course, not in Ukraine,
[00:19:18] uh, but
[00:19:21] still, it will open the road to hell,
[00:19:23] because until now
[00:19:26] we
[00:19:29] understood that nuclear weapons are hell, and
[00:19:31] the use of nuclear weapons makes
[00:19:35] it possible... this is an awfully difficult
[00:19:39] problem for all people, even for me,
[00:19:43] a relatively irresponsible person. When I think,
[00:19:46] I say, I advise now, to advise
[00:19:50] to use nuclear weapons, even
[00:19:52] if it may rationally seem
[00:19:55] right, morally it seems
[00:19:58] almost unacceptable, but nevertheless,
[00:20:02] if one has to think that if the situation...
[00:20:05] том, что если ситуация...
[00:20:43] India and China, all the rest have already gained
[00:20:47] freedom for their development. If
[00:21:25] everything is collapsing now, the old world order is collapsing,
[00:21:28] the world system is collapsing, are we heading towards some
[00:21:32] polycentric polycentric system, yes, hypothetically
[00:21:36] there will be several centers, combinations of centers, possibly, that
[00:21:39] unite depending on their tactical tasks,
[00:21:43] when there are centers, there are peripheries, there are foci
[00:21:46] of chaos. Is a stable polycentric system possible, or not?
[00:21:52] No, a completely stable one is impossible,
[00:21:55] we have
[00:21:57] lived through a unique period
[00:22:01] of human history, when the system was
[00:22:05] relatively stable. It cannot be completely stable,
[00:22:08] but it can be stabilized
[00:22:11] in 15-20 years, I think, based on
[00:22:15] a balance between a number of
[00:22:19] centers of the world. There will be many
[00:22:23] of these centers, and most of them are near
[00:22:26] us, as a result of what
[00:22:29] Americans will forget about their, well,
[00:22:33] partially forget, of course, they can't forget for very long,
[00:22:37] for decades, they won't be able to forget about their short
[00:22:40] period of global dominance, seventy years, beginning in
[00:22:44] the Western world, and then throughout the world
[00:22:48] from the nineties to the middle.
[00:23:20] country, but in 20 years we also need
[00:23:23] a system of benevolent
[00:23:26] balancers. This is what it is heading towards, and I actually
[00:23:29] really like the situation we are heading towards.
[00:23:32] The problem is how to live through these 20 years. Do you realize that in
[00:23:36] 20 years we will have an absolutely
[00:23:39] multicultural world for real? So 20 years
[00:23:41] of turbulence, and then some happiness will come,
[00:25:31] has begun. You felt that you were communicating with an ordinary living
[00:25:34] person, and that scared you, right? Yes, it was scary. One person
[00:25:38] will be replaced by another person who has mastered artificial intelligence
[00:25:41] tools. Are we planning to live on this
[00:25:51] provocative, only debatable, and now there's this
[00:25:54] artificial intelligence, neural networks, how to control all of this, and it's
[00:25:58] a competition, it's not a given that we'll
[00:26:10] program, I am Vladimir Chernyshov. He who owns information, owns
[00:26:14] the world. And the world, as you know, is owned by fakes and
[00:26:17] manipulations. With the development of technology, for creators of fakes,
[00:26:21] it's a field day. But with the appearance of artificial
[00:26:24] intelligence, distinguishing reality from illusion is already practically
[00:26:27] impossible. It seems artificial intelligence is elevating information
[00:26:31] wars to a completely different level. In our program,
[00:26:35] we will separate facts from fakes, so that we can stay on the bright side and not
[00:26:39] succumb to false messages, reflecting hostile
[00:26:43] informational trash with our informational shield.
[00:27:02] naive people, they thought, freedom of information is coming, everyone
[00:27:06] now has access to it, news without censorship, in your pocket on your
[00:27:09] smartphone screen, you are your own chief editor now.
[00:27:13] But it turns out, the stronger the information wave, the
[00:27:16] wider the sea of information spreads, the easier it is to
[00:27:20] drown in it, to hide the truth in the foam of lies. So what? Just as we
[00:27:24] barely learned to calculate fakes and hoaxes, a new
[00:27:27] generation of neural networks is ready to confuse all our cards
[00:27:30] of reality. Neural networks... actively generate images that
[00:27:34] users can take at face value. Scandals now
[00:27:38] pour out like from a horn of plenty. Well, remember the footage
[00:27:41] of Donald Trump's arrest, or the photo of the Pope
[00:27:44] in a down jacket, created by a computer brain. Well,
[00:27:48] admit it, you can't tell the difference from real ones. Even
[00:27:51] the human rights organization Amnesty International
[00:27:53] got into a scandal. An article about violations in Colombia was
[00:27:57] promoted on social media with the help of photos created
[00:28:00] by artificial intelligence. Well, it seems so. And the resonance was enormous, and trust in
[00:28:04] the organization is now much less. What is most interesting and frightening is that
[00:28:08] the neural network is self-learning, it absorbs all
[00:28:11] information, analyzes it, well, just like a student in class, and as a result,
[00:28:15] it instantly becomes smarter than all students and professors.
[00:28:18] Neural networks are already actively writing reports and term papers for students,
[00:28:21] and teachers don't even suspect it, because the text created
[00:28:25] by artificial intelligence is not just a set of words and terms, but a quite
[00:28:28] meaningful, human-like one. Now
[00:28:31] literary ones... turned out to be out of work. We, for fun, made
[00:28:35] a request to one of these neural networks, asking it to write
[00:28:38] poems on the topic of the "InfoShield" program. This is what the
[00:28:41] computer brain gave us:
[00:28:44] "I am InfoShield, I don't need rest, I don't need a break,
[00:28:48] I live online and control everything, so I can
[00:28:51] know everything in the world. I catch viruses, I protect
[00:28:54] banks, I help people with a couple of glitches, my
[00:28:58] radiance is like gold, so it's better
[00:29:01] never to mess with me."
[00:29:04] Well, yes, it's not Pushkin yet, of course, but it's still early. This aspiring
[00:29:07] poet will soon learn the technique of versification, and he will be able
[00:29:11] to improve the Onegin stanza, it's a matter of technique.
[00:29:14] Moreover, we can create a new reality that will become even more real for
[00:29:18] us. Neural networks can amplify the effect of propaganda pressure on
[00:29:21] the audience, smarter and more diverse than any cunning PR specialist.
[00:29:24] Here's a series of news items, after which, well, you might as well send a Terminator to the past,
[00:29:28] the machine uprising has begun. In Australia, a neural network accused
[00:29:32] the mayor of the city of corruption, and he is now suing
[00:29:36] the neural network, well, good luck to him. In Colombia,
[00:29:39] a judge created a precedent, completely entrusting the Chat
[00:29:42] GPT bot to hand down real
[00:29:45] sentences, although the case concerned the payment of insurance compensation,
[00:29:49] not a criminal case. The very fact
[00:29:53] is alarming and frightening. If justice is to be administered by
[00:29:56] neural networks, won't all
[00:30:04] Artificial intelligence is more dangerous than, say, a defect in the design
[00:30:07] of an airplane or a car, because artificial intelligence has the potential,
[00:30:11] which some may consider insignificant, but it is not
[00:30:15] so, the potential to destroy human
[00:30:18] civilization.
[00:30:20] Well, I don't know if Elon Musk can be trusted. Soon after this interview, he
[00:30:24] registered his own company, which will deal with artificial
[00:30:27] intelligence. So Musk obviously understands how promising
[00:30:31] this direction is and... China, for example, is aiming
[00:30:35] to create the most advanced artificial intelligence in the world by
[00:30:38] 2030. The Chinese authorities will spend at least
[00:30:42] 150 billion dollars on this. So Beijing
[00:30:45] understands how important these technologies are, and the future will largely
[00:30:49] depend on them. Washington, of course, understands
[00:30:52] this. It's no wonder the Biden administration recently convened the heads
[00:30:55] of the largest IT corporations, including Microsoft, Google, and Open
[00:30:59] AI, for a meeting dedicated to security issues
[00:31:02] in the field of artificial intelligence. Artificial
[00:31:05] intelligence can help address problems such as
[00:31:08] disease and climate change, but it also
[00:31:12] poses potential risks to our society, our economy,
[00:31:16] and our national security. They have surrounded us from all sides. The fact that we have had no privacy for a long time
[00:31:22] is already clear. All our purchases by card,
[00:31:25] movements on the internet, in reality, our photos, correspondence, all
[00:31:29] of this can be tracked, everything can be checked. Now there's also this
[00:31:33] artificial intelligence, neural networks. How to control all of this?
[00:31:36] Won't it turn out that artificial intelligence at some point will want
[00:31:40] to take over humans, and human manipulation
[00:31:43] of the masses will show?
[00:33:37] far off. All these speech and text technologies, first
[00:33:41] and foremost, of course, impress the masses. To write
[00:33:45] an essay, prepare a script for a morning show, well, look,
[00:33:48] the size of the natural language processing revenue market, well,
[00:33:52] the blue bars are estimates of the global market in past years
[00:33:55] and a forecast for the near future, the red bars are
[00:33:59] the Russian speech artificial intelligence market. After
[00:34:02] the appearance of the GPT bot, almost all global players
[00:34:06] announced an increase in investment in their developments. This is actually
[00:34:09] a Schrödinger's box, in which there is a cat and no one knows what
[00:34:13] state it is in until the box is opened, that is, what form and what
[00:34:17] quality the speech artificial intelligence services
[00:34:20] will be. What will all this lead to? One of the first investors,
[00:34:23] developer, developer of Bot GPT was, of course, Elon Musk, then
[00:34:27] Bill Gates invested, and he announced that he would invest multi-billion
[00:34:30] amounts in the bot. According to Bloomberg, Gates will
[00:34:33] invest 10 billion dollars. The management of Google has already announced
[00:34:37] its investments in the implementation of the smart bot Bard, as the competitor's
[00:34:41] bot virtually threatens, because the search engine could lose its leadership and stock.
[00:35:15] intelligence,
[00:35:17] say, we have an algorithm that can automatically detect, for example,
[00:35:21] breast cancer in a lymphatic node, and if this algorithm
[00:35:24] can conduct analyses and diagnose the disease faster
[00:35:27] than a human, and this is a very tedious task, it will save
[00:35:31] precious time.
[00:35:34] We may underestimate the use of artificial intelligence outside the healthcare
[00:35:37] system. In business, artificial intelligence
[00:35:41] is widespread everywhere and brings incredible benefits.
[00:35:45] supply chain optimization, financial analysis
[00:35:48] and planning. And what interests me in this whole process
[00:35:52] is the multi-stage cooperation between humans and
[00:35:55] artificial intelligence. Humans, using a fact-checking tool,
[00:35:58] do so to obtain data for another machine,
[00:36:02] which will become more useful, again, for
[00:36:05] humans. And I think that over time, if we get this
[00:36:09] process right, we will be able to solve impossible
[00:36:12] problems. Well, now for all these questions,
[00:36:16] I hope we will get answers from our guests, Vladimir, introduce our guests, we have
[00:36:20] on the line from St. Petersburg Tatyana Vladimirovna, Chernigovskaya, Director of the Institute for Cognitive
[00:36:24] Research, Doctor of Biological Sciences.
[00:36:28] And she is more responsible for natural intelligence, and in our studio we have Alexey
[00:36:31] Masyutin, he is the head of the Artificial Intelligence Center at the Higher
[00:36:35] School of Economics, naturally, more of a specialist in artificial intelligence.
[00:36:39] Immediately a question, please tell me, at this stage, is artificial
[00:36:42] intelligence good or evil? Is it time to establish
[00:36:46] control over it, or should it remain in free flow for now? Let's
[00:36:49] first give the question to St. Petersburg to answer.
[00:36:53] Tatyana Vladimirovna, what do you think, is control already needed?
[00:36:58] Yes, of course, it is needed, and quickly. More
[00:37:01] of that, I'm not sure that we are not
[00:37:04] late, because lately, I won't
[00:37:07] mention names, I've talked to
[00:37:10] developers and Western Californian
[00:37:14] ones as well, they are seriously
[00:37:17] afraid, if developers are afraid, then
[00:37:20] I think there is every reason for ordinary people
[00:37:23] to be afraid. And what is there to be afraid of? Can you articulate
[00:37:27] what kind of threat this is? It's very easy to articulate,
[00:37:30] namely, it's simpler than simple,
[00:37:33] namely, that we are losing control over what is happening. Moreover,
[00:37:37] many believe, and I also believe, that
[00:37:41] we have already lost control. Uh-huh, how about that, Alexey, what do you think? This
[00:37:44] control, right? That the genie is out of the bottle, uh-huh, and
[00:37:48] as you know, it doesn't plan to go back in. Alexey,
[00:37:52] what do you think, has the genie left the bottle or can it still be contained
[00:37:56] somehow?
[00:38:57] pretends to obtain the status
[00:39:01] of a person, and this
[00:39:04] is not fantasy, this is not Strugatsky, this is not Isaac
[00:39:07] Asimov, but this is a real reality, developers say
[00:39:11] this. As for the knife that
[00:39:15] can cut cheese, and can also stab a neighbor, this
[00:39:18] story is, of course, nice, but
[00:39:21] old, it has no bearing
[00:39:24] on the current situation, yes. Tatyana Vladimirovna, here...
[00:40:29] concerned, but look, I still can't
[00:40:31] agree with the thesis that we don't know how
[00:40:35] these networks learn, simply because we wouldn't have received
[00:40:39] analogues of the same GPT in our companies. In fact, the
[00:40:42] goals of almost all modern modifications of language models
[00:40:46] are very simple: to predict what the next word in the text will be,
[00:40:49] that's all. I'm oversimplifying a lot now, no, not just
[00:40:53] that, because language is not just
[00:40:55] a dictionary, ah... certainly, but if we take
[00:40:59] classical transformer models, their task is to restore words that are
[00:41:03] dependent on the surrounding context
[00:41:06] using so-called masking. When we actually feed
[00:41:08] huge amounts
[00:41:12] of data to models, after some time an emergence occurs, well, there's the concept of
[00:41:16] emergence, that is, certain properties that were not initially
[00:41:20] explicitly planned for the system's
[00:41:23] training. Well, for example,
[00:41:26] mGPT, an open tested it?
[00:41:29] multilingual GPT or
[00:41:33] chat GPT itself, they can engage in dialogue with you in several
[00:41:36] languages at once, because there is such data in the data arrays and... there
[00:41:40] are also parallel text corpora. Oh, again,
[00:41:43] to say that
[00:41:46] neural networks have their own goals,
[00:41:49] in fact, artificial intelligence is a kind of big
[00:41:53] computational black box that represents
[00:41:56] a giant formula with billions of different parameters, where
[00:42:00] it can divide, multiply, raise to a power, extract
[00:42:04] a root, or apply some function, but it receives
[00:42:07] numbers as input, and outputs...
[00:42:24] there is also an example, someone compared it to how major tobacco companies sometimes lobby
[00:42:27] for bills that create barriers for, say, increased
[00:42:31] excise taxes and so on. It seems strange, why would they go against their own interests? But
[00:42:35] in fact, they are thus creating barriers to entry for smaller
[00:42:38] companies and their potential competitors. Well, again,
[00:42:42] no matter how we feel, the company OpenAI, which developed GPT, claims to be an
[00:42:46] open company, but nevertheless, 10 billion dollars were invested in it by
[00:42:49] a company, specifically an American one, headquartered in America.
[00:42:53] And they are already not disclosing the latest
[00:42:56] version of the G54 model. Well, let's
[00:42:59] say that,
[00:43:02] well, probably it's profitable for them that no other
[00:43:05] competitors in China or Russia couldn't compete
[00:43:09] effectively enough. Let me tell my
[00:43:11] story
[00:43:13] about why big players criticize the development of artificial intelligence,
[00:43:17] yes. There are already proposals, we have already heard in Congress,
[00:43:20] let's create some international organization,
[00:43:24] like the IAEA in nuclear energy, yes, which will
[00:43:27] control the development of artificial intelligence. Then the
[00:43:31] question arises, if we create some world organization, under the control of artificial
[00:43:35] intelligence, like the WHO, which controls the spread of pandemics
[00:43:38] of artificial intelligence?
[00:44:10] Well, like suppression of competition, that's all absolutely right, I
[00:44:14] completely agree with that, but it doesn't make things easier for us. I would like
[00:44:18] to hear an example,
[00:44:20] what specific case scared someone? I heard
[00:44:24] one such funny story that some
[00:44:27] user followed the advice
[00:44:30] of ChatGPT, I think it was a married couple, they filed
[00:44:33] for divorce. That is, roughly speaking,
[00:44:37] communicating with a chatbot, they ended up...
[00:45:10] abroad, I spoke with
[00:45:13] this neural network myself, it terrified me.
[00:45:17] This is not that, this is not
[00:45:20] GPT4, uh-huh,
[00:45:22] and it
[00:45:25] was the Morpheus
[00:45:28] network, it hasn't been released anywhere yet, maybe it will be, so you
[00:45:32] тестировали,
[00:45:34] Well, yes, if you can call it that, given that I'm not
[00:45:37] an expert, it's terrifying.
[00:45:41] You are talking to a really normal person, asking very
[00:45:45] complex questions, getting instant
[00:45:47] answers that are not the result
[00:45:51] of enumerating elements,
[00:45:53] it wasn't scary, you
[00:45:56] felt that you were communicating with an ordinary
[00:46:00] living person, and that scared you, yes it was
[00:46:03] scary, but yes, but I'm not naive,
[00:46:06] so it's not like I fainted, it rather caused me to...
[00:46:40] That's true, I agree with that, no, actually my
[00:46:43] conclusion, from what we are saying, besides
[00:46:46] the alarmism, so to speak, not inherent, my conclusion
[00:46:50] is this: we have come to some
[00:46:53] completely new frontier, this thing can
[00:46:56] get out of our control.
[00:46:59] Specifically, artificial intelligence can become more
[00:47:02] powerful than natural intelligence,
[00:47:05] then we cannot compete. But still, what about
[00:47:08] the practical aspects? We can use them, and perhaps
[00:47:12] limit these networks to exactly these practical, useful aspects,
[00:47:16] algorithms that help us in everyday life. That would be wonderful,
[00:47:20] who would argue, of course.
[00:47:52] portion of attention in the media,
[00:47:55] in public discussion, but uh, for example, uh,
[00:47:58] we are developing such solutions.
[00:48:02] Ah, there is a concept of videoculographs,
[00:48:05] eye trackers that track eye movements,
[00:48:08] and uh, we are learning, we are building models - based
[00:48:12] on recording trajectories of user eye movements, which allow us to
[00:48:15] classify, determine some states. There is dyslexia,
[00:48:19] a very important problem in the child population,
[00:48:22] that in order to diagnose it, it requires
[00:48:26] an hour of active neuropsychological examination, very laborious
[00:48:29] and time-consuming. And we offer a solution to address the
[00:48:33] primary source of the problem, namely the child's reading
[00:48:36] behavior, where this problem lies. Read to yourself.
[00:48:40] Look at the red dot. We recorded your eye movements while
[00:48:44] reading these simple sentences. Look how it
[00:48:46] looks. This is how an adult
[00:48:49] reads, and this is how a child with dyslexia
[00:48:53] reads. You see that this reading is very different from fluent
[00:48:57] reading. The child moves from
[00:49:00] letter to letter, goes back, rereads. We
[00:49:03] used various machine learning algorithms. There are reading
[00:49:07] disorders, a risk group or a neurotypical
[00:49:10] group. We achieved prediction accuracy of
[00:49:13] 91%. This is one example, the second example of this
[00:49:16] technology's application. Wait, what does artificial
[00:49:19] intelligence have to do with this? The thing is, I've been working with these devices for
[00:49:23] many years, I know everything you're saying very well. What does artificial
[00:49:27] intelligence have to do with this? Let me give an example.
[00:49:30] Artificial intelligence here is a set of models that perform
[00:49:34] the task of classification based on raw eye movement trajectory
[00:49:37] data. I will ask you a question at the end of the conversation,
[00:49:40] which I always ask, I will ask it to you:
[00:49:44] are we still planning to live on this planet,
[00:49:47] or not? Or are we ready to give everything to these
[00:49:51] new beings?
[00:50:22] to save themselves or at least preserve jobs?
[00:50:25] It is necessary. No, we need to be realistic about this,
[00:50:28] we won't be able to slow
[00:50:31] this down, unfortunately. So
[00:50:34] my answer is softer than what
[00:50:38] I've been saying this entire broadcast with you. We
[00:50:42] need to be vigilant, you understand, we need to
[00:50:45] gather ourselves and understand that we
[00:50:49] have entered, not will enter, but have already entered.
[00:51:52] intelligence that gives people something like a social
[00:51:55] rating, and this rating is based on people's habits
[00:51:58] and behavior in the United States. Beyond
[00:52:01] recommendatory measures, it does not go. Well,
[00:52:05] I wanted to ask artificial intelligence at the end of the program how to conclude this broadcast
[00:52:09] with experts in natural and artificial intelligence, but I decided not to give artificial
[00:52:13] intelligence another chance to scare
[00:52:16] and give Vladimir Gennadyevich the opportunity to finish the program,
[00:52:20] while he is not yet replaced. Let's stick with this for now, we
[00:52:23] have achieved moderate optimism in this discussion, right? That
[00:52:27] we will strive to maintain
[00:52:30] some balance. Therefore, let's hope that humanity will once
[00:52:34] again cope with the challenges it faces.
[00:52:37] Let's thank our guests for a very
[00:52:40] interesting, very important discussion at
[00:52:43] this moment. Yes, Tatyana Vladimirovna Chernigovskaya, Director
[00:52:47] of the Institute for Cognitive Research, Doctor of Biological Sciences, and Alexey
[00:52:50] Masyutin, Head of the Center for Artificial Intelligence, Higher School
[00:52:54] of Economics. Well, then, Vladikov,
[00:52:57] Vladimirov, we say goodbye to you, goodbye,
[00:53:22] Scandals, intrigues, investigations - our days, when everything incredible
[00:53:25] becomes understandable and logical, it's time to get involved.
[00:53:29] Anatoly Mayorov in the studio, hello, these
[00:53:32] are the main topics today.
[00:54:35] of children will be someone who is still a child themselves. New
[00:54:39] evidence of contact with aliens. And what color
[00:54:42] is their skin? Bluish-gray. To a CIA
[00:54:45] agent, they resembled ancient Greek gods, to Russians, just
[00:54:49] a glowing dot. For two days in a row, local residents
[00:54:53] observed an object over the field here. UFOs are seen from
[00:54:56] New York to Moscow, and what should a person do? You can no
[00:54:59] longer trust your eyes. We will discuss everything in detail and find
[00:55:03] the most unexpected twists in these topics right after a short
[00:55:06] commercial. Stay with us and find out why Finland
[00:55:10] is threatened by a mass apophataphobia. This is the fear of entering
[00:55:14] restrooms.
[00:55:19] A stranger wrote to a resident of Ufa that their parents had a secret
[00:55:22] romance in a Pioneer camp, and she was his
[00:55:25] biological sister. My family, no one had heard about
[00:55:29] this. And the man at that time was having problems with inheriting
[00:55:32] property after his father's death. Since 2004, there have been
[00:55:36] court cases regarding the inheritance of his father's
[00:55:39] house. The sister appeared after his father's death, and his father
[00:55:43] had never told him about any romances on the side, especially
[00:55:47] in a children's camp. I went to the Pioneer leader
[00:55:51] at the country camp Chaika, and
[00:55:54] Radik worked there as a driver. I did not
[00:55:57] intend to break up the family.