Artificial intelligence, the evolution of intelligence, the power of stereotypes and the contours of utopia
The intrusion of neural networks into human life is growing literally from month to month. Today we will consider a number of publications, mainly in the domestic press, covering the topic of artificial intelligence (AI) in philosophical, military-political and military-technical aspects. Some of these publications claim to have conceptual depth, while others provide specific examples of the use of AI in military affairs, including the conflict in Ukraine.
THE SELF-TAUGHT DEMON
Political scientist Andrey Kortunov, in his article "On the way to AI – the coming of the demon Laplace" (INF website), recalls a thought experiment by French mathematician Pierre-Simon Laplace at the beginning of the XIX century.
Imagine a being receiving complete information about the position and velocity of every particle in the universe. Such a creature (dubbed the "Laplace demon") He will be able to restore the entire history of the universe and predict all its further development, there will be no unknown secrets for him.
In the twentieth century, Albert Einstein indirectly supported Laplace, saying that "God does not play dice": The laws of the universe should be combined into a "theory of everything" that has universal explanatory and predictive power.
But the supporters of free will did not give up. And they found support in the form of quantum mechanics with its "uncertainty principle". This principle has been developed in a number of fundamental studies. Laplace's demon was banished from the microcosm. This means that he has nothing to do in the macrocosm.
At the same time, no one could imagine that Laplace's demon could materialize in physical reality. Firstly, it is impossible to collect all the information about the entire universe (especially if we proceed from the postulate that the universe is infinite). Secondly, it is impossible to process and systematize such a volume of data. And both tasks are immediately possible only to the Lord God of the monotheists, who, as is known, is both omniscient and omnipotent.
But if you reduce the task to one planet and humanity on it, and moreover reduce the requirements for predictive power, Laplace's demon will no longer seem like a speculative construct. The worldwide Network creates fundamental opportunities to know "everything and about everything" at any given time. And AI technologies allow us to process "a huge catch of heterogeneous data." At the same time, protein intelligence cannot be directly connected to the Internet today, but AI can. This creates his demonic power.
Laplace's classic demon was an indifferent observer. AI is already an active force today. Neural networks are used in marketing and journalism, in detective and medical practice, in financial transactions and political technologies. But the widest horizons for AI, according to the author, are opening up in intelligence – commercial, military, scientific and technical, etc.
What limited the effectiveness of intelligence? The lack of reliable information and the lack of algorithms for processing and verifying data. The struggle of the intelligence services was along the line separating information and disinformation, facts and speculation, statements and assumptions. And human experience and intuition often solved everything here. AI multiplied by the Internet can solve both problems and make intelligence an accurate science. The importance of the human factor is declining, the romance of Stirlitz and James Bond is becoming a thing of the past, intelligence is turning into a battle of technology.
(Theoretically, it can be argued that humanity on its planet is not a closed system, the Earth receives information from the outside and gives it to the outside (at least in the form of radiation). This means that the amount of data about this system at any given moment will be incomplete and flawed. So the AI turns out to be an unlucky demiurge who is unable to create a stone that he himself cannot lift.
And in practical terms, the neural networks fighting on different sides will not only comprehend enemy secrets, but also continuously produce disinformation. So intelligence is simply transferred to another technological sphere, that's all.)
OF COURSE, ONE MUST HOPE FOR GOD
"Whoever owns the information owns the world." The authorship of this phrase, Kortunov continues, is attributed to Nathan Rothschild. And it was said in the same era when Laplace gave birth to his demon. Rothschild was the first in London to learn about Napoleon's defeat at Waterloo on June 18, 1815. And in one day he earned 40 million pounds on the London Stock Exchange, which was only slightly inferior to the military budget of the British Empire for this year. Rothschild received this insight with pigeon mail, the most operational communication of the time. He could not know how each soldier or officer moved on the battlefield, what was the trajectory of each projectile or bullet. Rothschild had enough binary ("yes – no") knowledge of the results of the battle.
(Note that Rothschild turned out to be more far-sighted than Bonaparte in terms of intelligence. When Napoleon met Laplace, he was interested in only one question: where is the place for God in his constructions? Laplace, according to legend, replied: "Sire, I did not need this hypothesis." Napoleon had already won many victories and had reason to believe that Providence itself was guiding his strategy. And the question of the pragmatic use of Laplace's idea did not occur to him – although Napoleon knew a lot about finance and could not help but understand the importance of communication and intelligence.)
And now imagine devices for processing infinitely large information arrays, the author continues. The endless information field is no longer processed with a hoe, but with a high-performance combine harvester. "The neural network does not miss a single spike, does not lose sight of a single grain, ensuring perfect crop safety. Now Rothschild could count not only on the answer to who won at Waterloo, but also on accurate data on the losses of the parties, on the remaining reserves, on the positions of the units... Availability of such information... it would make it possible to predict the further course of the 1815 campaign and even the future of Europe after the Napoleonic Wars."
(The abundance of metaphors is alarming here. The harvest field insensitively turns into a battlefield, and the landscape after the battle into a clearing of world history and futuristic forecasts. The sinister charm of the Apocalypse is based on such effects. Meanwhile, every field (in a military way, in every theater) has its own laws. A rural tractor driver can be trained as a tank driver, but a diligent player in naval combat is unlikely to become a major naval commander. Not to mention that the pragmatist Rothschild would simply consider such an abundance of information unnecessary – although he would be able to benefit from many specific plots.)
Imagine devices that hack into any information security system and will soon make any coding of it meaningless, Kortunov insists. The priorities of intelligence will shift towards penetration into information systems isolated from the Internet, and the priorities of counterintelligence will be to stop such penetration. "The intelligence race is increasingly turning from a duel between intelligence and counterintelligence into an AI battle. Countries with modern neural networks gain significant advantages. This means that attempts to impose a moratorium on the development of new AI systems or to agree on rules for its use are doomed to failure."
(All this is very interesting, but do not forget that the more technically sophisticated the system, the more vulnerable it is. One nuclear explosion (more precisely, the electromagnetic pulse caused by it) nullifies both the Internet and neural networks. The self-taught demon is insulted and leaves. And if there are many such explosions, the planet remains at the mercy of rats, cockroaches and naked diggers (these are such underground animals that are extremely resistant to radiation). Of course, they also have their own intelligence – but that's a completely different story.)
"FOG, FOG, GRAY VEIL..."
Nikita Solovyov on the same website presents an article "Dangerous consequences of the use of AI technologies in military affairs."
According to Li Kaifu, head of the Chinese company Sinovation Ventures, AI technologies have become the "third revolution in armament" after gunpowder and nuclear weapons. At the same time, there is a well-established practice in the West: for many military equipment based on AI, civilian prototypes are first developed and then adapted for military needs, which saves time and money.
An example is the development of programs for data analysis and intelligence work by the American corporation Palantir Technologies. The Palantir AI complex provided to Ukraine is aimed at eliminating the "fog of war". It includes a secure communication channel for the exchange of intelligence; access to a joint database of satellite images; calculation of the location of enemy troops, etc. The downside of this system is "the possibility of using this technology due to its universality" to violate international humanitarian norms.
In 2014, Palantir Technologies developed a system for accessing emails, phone call records, text messages and surveillance camera data for the US National Security Agency (NSA). In 2011, Palantir Technologies opposed the WikiLeaks project, conducted cyber attacks and disinformation campaigns to discredit Julian Assange.
Another example. "Speaking in the US Congress," Solovyov writes, –former CEO of Primer Technologies, Sean Gourlay, said that AI is able to increase the effectiveness of the US Department of Defense operations, and allocating 5% of the entire American defense budget to AI can provide Washington with a significant advantage over opponents such as Moscow and Beijing." According to Primer Technologies, their solutions meet the need for "situational awareness", which includes strategic analysis, threat detection, information operations, audio recording extraction and analysis, training and deployment of AI models. AI algorithms from Primer AI were specified for the needs of Ukraine and provided free solutions to Kiev for such tasks as interception and collection of audio data; neutralization of extraneous noise on the air; decoding and translation of Russian speech; highlighting key phrases related to the situation on the battlefield; recognition of objects mentioned in speech, identification of persons, organizations and locations; preparation of a report based on the processed data.
The American company Clearview AI has been known since 2020, when its founder Hoan Tong-Ta became the hero of a scandal. The Clearview AI software solution "represents the most striking example of the ambiguous attitude of society towards AI technology." The program deals with biometric identification of a person by photo based on a database obtained mainly from social networks. In 2020, it became known about the gigantic scale of her illegal use of personal data. The program's users included 2,200 entities in 27 countries, including the FBI, the U.S. Department of Justice, large banks, retailers, and mobile operators.
In Ukraine, the Clearview AI program was used to identify dead Russian servicemen in order to identify their relatives and exert psychological pressure on them. Clearview AI's cooperation with law enforcement agencies may lead to the introduction of technology into a new generation of lethal weapons capable of selectively eliminating targets, which, in turn, may cause the threat of creating a new type of weapon of mass destruction.
In general, the use of AI is still in the "gray zone" of international legislation on information security, Nikita Solovyov concludes. Here it is: the local "fog of war" with the help of AI may be accelerating, but visibility in global "gray zones" is only getting worse. It is time to deduce the general law of conservation of the "fog of war" in nature (by analogy with the physical laws of conservation of energy, momentum, etc.).
YOU CAN'T LAY STRAWS FOR EVERYONE
Chess players, like housewives, have the concept of "homemade preparations". (First of all, the opening ones, although there were grandmasters who brilliantly played endgames.) And today, a computer helps to develop these blanks in 100% of cases. But this does not mean that the opponent will not be able to find the right answer right at the board during the game. Or that the player who uses the novelty will not confuse the order of moves. The time pressure factor, that is, lack of time, may also interfere. In short, so many surprises can happen during the game that no AI will provide.
During the first months of Russia's special operation in Ukraine, all experts unanimously emphasized the advantage of the Ukrainian side in terms of communications and intelligence. Gradually, however, it was practically eliminated – people learned from their own mistakes almost faster than neural networks.
Asymmetric solutions also played a role. The advantages of the Armed Forces of Ukraine in terms of communication and remote control of drones were offset by Russian superiority in terms of electronic warfare. Similarly, the advantages of long-range artillery are offset by a simple shortage of shells and the density of enemy fire.
In addition, military experts are equally united in noting that traditional weapons (the same artillery) play an unexpectedly large role in the Ukrainian conflict. The advantages of Russia in terms of universal mobilization (economic, technological, psychological, ideological) are also affected. As for AI specifically, its decisions are based on stereotypes, so that asymmetric and paradoxical moves are quite capable of baffling it. By the way, it was the positional deadlock that the commander–in-chief of the Armed Forces of Ukraine Valery Zaluzhny recently discussed - apparently, sadly recalling the former Ukrainian advantages based on Western technologies. And if some kind of miracle weapon changes the situation at the front in favor of one of the parties, then it will not be invented by AI.
"Take space spoons..." Manipulating an online community or a group of AI customers is as easy as a kindergarten group. A shot from the movie "Gentlemen of Fortune". 1971
autumn. THE PATRIARCHS HAVE BECOME MORE ACTIVE
Two patriarchs of American politics, 100-year-old Henry Kissinger and 83-year-old Graham Allison, appeared in Foreign Affairs magazine with an article "The way to limit the military use of AI."
"There are many echoes of the past in the current proposals on ways to contain AI. Billionaire Elon Musk's demand for a six-month pause in AI development, AI researcher Eliezer Yudkowski's proposal to ban it, and psychologist Gary Marcus' demand that it be controlled by some kind of global governing body repeat the failed proposals of the nuclear era. Each of them will require the leading states to subordinate their sovereignty to supranational structures. Never in history has any great power, fearing that a competitor might apply a new technology, refused to develop this technology for itself."
Further, the co-authors develop an analogy between AI and nuclear weapons (NW). It is important to recognize their significant differences. The development of nuclear weapons was directed by national governments, AI development is led by private companies that tend to underestimate national interests. YAO products were physical objects and could be counted. The evolution of AI takes place in the minds, its deployment is difficult to observe. YAO is tangible, AI is speculative. AI is developing at a speed that eliminates long negotiations. Control over nuclear weapons has been developed for decades. Limits for AI should be set "before machines start setting their own goals." This requires new international relations.
While the United States is still creating an AI management system, it's time to "start serious conversations" with another superpower. "China's national leaders in the technology sector – Baidu (the country's leading search engine), ByteDance (creator of TikTok), Tencent (creator of WeChat) and Alibaba (leader in e–commerce) - are creating their own ChatGPT analog systems. And although China is still lagging behind in semiconductor manufacturing technology, it has everything it needs to move forward in the near future."
But most of the article is still devoted to the "lessons of the nuclear age." In this part, the co–authors have something to remember - and they offer old recipes. "If Biden, Xi and other world leaders act now to confront the challenges of AI, as decisively as their predecessors did in the fight against nuclear threats... will they achieve the same success? Looking at the growing polarization of the global community, it is difficult to be optimistic. Nevertheless, the fact that we celebrated the 78th anniversary of peace between the nuclear Powers should inspire everyone."
(Interestingly, the patriarchs are sure that the problem of AI control should be solved by China and the United States. Although revolutionary discoveries regarding AI can be made in any developed country. For example, in the era of Great Geographical Discoveries, breakthrough technologies in the maritime business (new types of sailing ships, mapping and navigation systems) were born in little Holland. And thanks to them, she became the leader of maritime trade and a major colonial power.)
THE DICTATORSHIP OF THE STEREOTYPE
Grigory Revzin, an art critic and architectural critic, publishes a series of articles about utopias in Kommersant-Weekend. A recent article is called "The Absolute stereotype. AI as the utopia of modernity."
Utopias operate on the same set of ideas. "The hero fell asleep, woke up in the future, and there everything is the same: the answer to the question of how to achieve unity of people in the name of the highest good." They offered recipes for eliminating contradictions, highlighting their three main sources – property, sex and power. In terms of property, equality was proposed, in terms of sex, external control for the purposes of eugenics, in terms of power, "the rule of those who are predisposed to it, for the benefit of all." But equality of ownership rests on the question of whether there are enough resources for all ("projects were proposed to eliminate competition, but where it was eliminated, there was nothing enough for anyone"). Sex is an intimate matter: control is ineffective here. "And it is not possible to single out those who are able to rule for the good."
However, Israeli historian and futurist Noah Harari, author of the bestseller "Homo Deus. A Brief history of the Future", paints a picture of human interaction with the Network and AI, in which utopian questions receive clear answers. It is not difficult to choose people who are ready to rule for the common good: The AI will easily assess the effectiveness of each candidate. Problems with sex are also nonsense – AI will find the optimal partner for everyone both in terms of bed harmony and in reasoning genetic compatibility. And "equality in property will turn out to be a problem of income differences – the Network itself will increase or decrease your income within the framework of the balance of stability of society." Moreover, all these processes have already been launched, we are talking only about the scale of implementation.
Here Revzin comes to a key postulate. AI is the intelligence of stereotypes. "He calculates a stereotypical answer to a question for a certain group of people, and the group is determined by the question you asked. And if the question is stereotypical, the answer to it is based on the analysis of a huge database and will be acceptable to a huge group of people." According to AI himself (the author asked him this question), "stereotypical thinking is when you are convinced of the truth of a stereotypical judgment and feel an emotional connection with it." It is very convenient for building a utopia.
"People do not want to live as they are prescribed. They do not want the state to control their income, sex, profession, establish unity for them and lead them to a good goal. They want to keep their freedom." But stereotypical AI responses have nothing to do with freedom or non-freedom. AI does not generate thoughts and has no beliefs. Its task is to provide background assistance and provide information based on already available data. People decide on specific issues: which movie to watch, where to go on vacation, with whom to have an affair. And they accept stereotypical (for their group) responses as strategies for their own behavior. In this way, society is moving towards a utopian ideal without noticing it.
The Internet today is an actual form of God, concludes Revzin. He is omnipresent and omniscient. He helps to earn daily bread and find earthly and heavenly love. He heals and hints at the prospects of immortality. It connects everyone. "Building an ideal society has become an engineering problem and is being solved automatically."
"BREAKFAST IN THE SENIOR GROUP IS CANCELED TODAY"
The power of stereotype and the dictatorship of banality is indeed great. This is the basis for the success of popular bloggers and other information gypsies. It's nice to read or hear a set of ideas and judgments that you've pulled from the world by a thread, but have learned so firmly that you consider them your own, you just don't know how to express them as smoothly.
Meanwhile, there is nothing new in dividing society into groups dominated by prescribed stereotypes. These groups are estates, classes, workshops, corporations, and denominations. And everyone has class honor, class prejudice, guild rules, corporate solidarity, religious intolerance. Moreover, stereotypes valid for one group have no power over another. Moreover, a developed culture becomes more complicated mechanically – and the number of groups with their own stereotypes in it is constantly growing. (The development of the Internet proves this as well as possible.) In the developed and sophisticated culture of large and influential groups – that of Protestant churches in the United States. And volatile and marginal groups, sects and network communities are impossible to count.
At the same time, the group is not difficult to manipulate. Like in the movie "Gentlemen of Fortune": "Breakfast at the kindergarten is canceled today! We will take a space rocket flight to Mars. Igor is appointed commander... Pick up the space spoons!"
But there will always be individuals who are sick of banal bloggers and fashionable pants, from stupid blockbusters and political manipulation. They are the ones who invent something new, breaking stereotypes.
AND YET IT WORKS
But we have moved too far away from the military sphere. Colonel G. Yablokov ("Foreign Military Review", 2023, No. 10) in the article "The development of military AI technologies in the UK" gives a number of examples of how neural networks work in specific types of weapons today.
The Viking multipurpose land drone, equipped with AI, is designed to deliver goods to combat areas, for reconnaissance and search for improvised explosive devices. During the tests, the Viking formed the route itself, overcame obstacles, and changed the trajectory of movement depending on the terrain conditions and the current situation. The test was deemed successful. The Viking will be put into service in 2025.
The autonomous boat Pacific-24 is used in the fight against maritime piracy and smuggling, in border control and ship security. During the test in October 2021, the boat performed patrols, detection and identification of targets in offline mode with data transmission to a checkpoint 16 km away. Upon completion of the tests, similar AI systems are planned to be equipped with aircraft carriers, patrol ships, etc.
The article also talks about the introduction of neural networks into small-sized UAVs, but in this part Britain is not a trendsetter.
In general, the revolution in military affairs with the advent of AI has not yet become obvious. But since everyone is preparing for it, she will certainly be punished. In psychology, this is called the "Pygmalion effect." But this is a different story.
Yuri Yudin
