Application of artificial intelligence technologies in weapons of foreign states
Currently, artificial intelligence (AI) technology is widely used all over the world.
AI technologies are based on machine learning, artificial neural networks and Big Data technologies (very large arrays of heterogeneous data). AI can be used in any complex technical systems for various purposes. Its distinctive features are high adaptability and self-learning. The theoretical basis of AI is made up of: probability theory, mathematical statistics, artificial neural networks, fuzzy logic, psychology, linguistics and computer technology.
AI has already surpassed humans in solving problems that require intuition, including in relation to predicting the behavior of other people, since intuition turned out to be an unconscious pattern recognition . AI is indispensable for managing and monitoring very fast or too slow processes. Mathematical analysis also shows that there are problems that are fundamentally unsolvable using only computational algorithms .
In the commercial sector of the economy, AI technologies are widely used to solve the following problems:
- recognition and translation of multilingual speech streams in the presence of interference;
- text recognition, recognition of emotions in the text, prediction of the continuation of a phrase, translation of the text;
- creation of original texts in natural language from a large amount of data;
- cryptography (decryption of texts), decoding of genomes of organisms;
- merging two-dimensional images into three-dimensional (for example, cartography, tomography, radiology);
- face recognition, recognition of emotions from a photograph, recognition of the authenticity of a photograph, recognition of handwritten texts, recognition of images of the area;
- forecasting the strength of buildings and structures;
- intelligent training systems;
- financial planning, sales, stock market games, management of securities portfolios, assessment of the possibility of issuing a loan (assessment of the applicant's creditworthiness and the size of the loan);
- logistics (ordering and managing the movement of goods and components);
- analysis of complex data in medical systems, diagnosis and prescription of treatment, selection of medicines, prediction of mental states;
- search for malware;
- search for cyber vulnerabilities;
- games, including card games with the possibility of bluffing;
- diagnostics of technical systems;
- legal advice and criminal proceedings, including with a forecast of a judicial decision of a particular judge;
- autonomous and automated navigation (vehicle traffic control) in 3 physical environments;
- flexible regulation of traffic (traffic light control);
- proof of theorems, formation of hypotheses, formation of knowledge bases for expert systems;
- synthesis of complex objects: synthesis of new drugs, synthesis of complex organic compounds with desired properties, synthesis of genomes for new organisms.
The total number of software products for each position is very large. Table 1 lists some examples of commercial AI software products.
In general, all major global (Google, Facebook, Amazon) and Russian (Vkontakte, Mail.group, Yandex) IT companies have in their commercial products (social networks, online services) services with AI technologies (translation of texts, recognition of images, user preferences, spam and malware, as well as images of the area) that function explicitly or covertly from the user.
The prevalence of AI technologies in the weapons systems of foreign states is currently inferior to the commercial sector of the economy; nevertheless, in these systems, there is an explosive growth in the use of such technologies, including in control systems of air defense and missile defense systems.
The greatest adoption of AI technology has been found in the armed forces of the United States, Israel and the United Kingdom. For example, the US Department of Defense (DoD) has established a joint JAIC center for AI, and an AI task force, A-AITF.
Carnegie Mellon University has become the main developer of AI technologies for DoD.
Moreover, on November 15, 2014, US Secretary of Defense Chuck Hagel, in his statement on the Defense Innovation Initiative (DII), announced the Third Offset Strategy (CK-3) as ensuring military dominance in the world through large-scale the use of AI in weapons systems . The introduction of AI into the field of warfare is assessed by DoD experts in the same way as the invention of gunpowder and nuclear weapons: this is a factor that can completely change the paradigm of armed struggle .
AI is predicted to be able to break the link between a state's population and the strength of its economy, on the one hand, and the combat effectiveness of its armed forces, on the other.
Improving weapons systems through the introduction of AI technologies is the cheapest and most cost-effective way to modernize them: this approach does not include costs for materials, components, electronic component base (EEE), production of mock-ups and prototypes; testing costs are reduced, since less their volume. In addition, these technologies do not depend on foreign supplies of materials, electronic components and production equipment. The US experience shows that AI can give new qualities to existing weapons, and even outdated ones.
Almost all AI technologies used in the weapons systems of the previously named countries are based on artificial neural networks. Specialized AI based on them is used:
- in autonomous (unmanned) vehicles (for air, water and terrestrial environments);
- in control systems of missile defense systems (THAAD, Patriot) and in the developed systems of directed energy weapons, where AI makes it possible to oppose the hypersonic speed of targets with the speed of decision-making;
- at aviation simulators for training pilots, and there is a tendency of stable superiority of AI over pilots, especially in maneuverable air combat ;
- search for malicious software (software) and cyber vulnerabilities in weapons.
Some examples of AI weaponry designs being developed for DoD are shown in Table 2.
Possible areas of application of AI technologies in the CSA of air defense-missile defense troops
To date, the systems of automation equipment (KSA) of formations, military units and subunits of the air defense-missile defense forces (hereinafter referred to as the air defense-missile defense troops) use mainly classical computational algorithms. At the same time, there are a number of tasks that, in the course of the battle, are solved only by the personnel of the combat crews of the command posts of the aerospace defense or are not solved at all.
For example, target assignment tasks are not automatically solved taking into account the level of training of combat crews of lower military formations, target class recognition by its radar portrait, as well as target trajectory forecasting and target tactical designation (except for a number of "obvious" ones (for example, ballistic ones) goals). An indirect confirmation of this can be the fact that when working on control targets or training and combat firing at ranges, crew commanders very rarely make a decision about choosing an automatic mode of operation of the KSA and combat control points (PBU).
At the same time, combat crews are subject to stress, fatigue, and have a heterogeneous level of training, which does not allow ensuring the stability of the quality of solving the above tasks. In addition, AI systems, unlike military personnel, do not have problems of psychological compatibility with each other or with operators, and are also easily retrained.
It is worth paying serious attention to the fact that the United States and a number of other leading world powers are investing heavily in the creation of fully autonomous attack unmanned aerial vehicles (UAVs) with AI, capable of operating in the contested space without external support (Table 1). Thus, in the near future, the air defense and missile defense forces will face a situation where autonomous UAVs will combine the intelligence of manned platforms and the maneuverable characteristics of unmanned platforms, which will greatly increase the dynamics of anti-aircraft combat and complicate its nature. The combat crews of the command posts of the air defense-missile defense forces will not be able to make decisions in real time in such an environment.
Despite the obvious lag in the field of AI of Russian weapons, it is necessary to cancel the fact that the prevalence of AI in the weapons of the US Army today is also significantly inferior to the commercial sector of software products, which reduces the distance between domestic and foreign military equipment in this segment and potentially makes it possible to reduce this break.
AI could find its application in solving the following main tasks, unsolvable and not fully solved by traditional computational algorithms of the KSA of the air defense-missile defense forces:
- recognition of the class and type of target by signal signs, target identification against the background of interference (solving the problem of image recognition in the radio frequency range);
- recognition of the type and class of the target by trajectory features;
- recognition of the tactical designation of a target and a group of targets (revealing the intention of an enemy strike) by a combination of target signs, information about the terrain and defense objects;
- tracking actively maneuvering and separating targets, including those launching various decoys, traps, aircraft weapons, using "intellectual" interference;
- solving the problem of assessing the level of preparedness of combat crews of controlled fire and reconnaissance means (both on a real scale, that is, directly during the battle, and in cumulative scale - based on the results of previous combat experience);
- solving the problem of target assignment and target designation, taking into account both predicting the movement and actions of the target, based on its class and tactical purpose, and the level of training of subordinate combat teams.
Additionally, KSA with AI could solve the following auxiliary tasks:
- automation of the algorithms of actions of various calculation numbers in typical situations established by the governing documents (for example, actions of the operational duty officer when detecting an intruder aircraft, bringing to the highest degree of combat readiness, etc.);
- the creation of various training raids (assistance in the formation of a raid and the implementation of the strike plan), the implementation of the interactive (in relation to the actions of the trainee calculation) behavior of training goals;
- assistance to the calculation in identifying faulty units of the KSA itself, as well as in assessing the health of subordinate assets based on the results of their actions during combat work.
Such a KSA could effectively solve tasks not only during the preparation and conduct of anti-aircraft combat, but also in solving the tasks of combat duty in air defense.
At the next stage in the development of weapons systems, AI technologies could find their application in solving problems:
- target recognition against the background of interference;
- self-study and simulation (including self-study during simulation) of combat operations;
- automated extraction of knowledge (positive experience) during training battles and combat operations;
- accumulation and application of knowledge about the peculiarities of the enemy's tactics in the region of application of the KSA;
- solving problems in a common information space with a large amount of heterogeneous information (Big Data technology): data on meteorological conditions, time of day and year, terrain, engineering, radiation, chemical and biological conditions, non-radar (radio and radio engineering, optical, intelligence ) intelligence information.
Thus, a CSA with AI can be trainable (in other words, it can additionally take on new combat experience obtained and systematized by other combat crews at other control points) and self-learning (that is, it can correct heuristic algorithms based on its own combat experience, taking into account the peculiarities specific terrain, enemy, interacting forces, subordinate means).
AI technologies will make it possible to automate the application and dissemination of the experience of the most experienced combat crews, the experience of real combat operations to all crews of the air defense-missile defense forces.
In the course of its development, KSA with AI will be able to help form fundamentally new tactics for dealing with an aerospace enemy, similar to how it happened in other areas of application of AI technologies, for example, in chess, which can also be considered as a simple model of two-sided combat. action.
So, for example, on December 7, 2017, Google's AlphaZero software won against Stockfsh 8, the 2016 world champion among computer programs . Stockfsh 8 had access to human experience gained over hundreds of years of playing chess, as well as data from chess programs over several decades. She could analyze 70 million chess positions per second. AlphaZero's computation speed was only 80 thousand operations per second, and the creators of the program did not teach her chess strategy - not even standard openings. In mastering chess, AlphaZero used the most modern machine learning methods, playing with itself. Still, out of a hundred games played against Stockfsh 8, AlphaZero won 28 and drew 72. AlphaZero took four hours to learn how to play chess and prepare for the match against Stockfsh 8. In four hours, the AlphaZero program, without any human help, made its way from complete ignorance to the pinnacle of mastery. In other words, AlphaZero has found more effective game strategies in 4 hours than all of humanity in several millennia.
The AI technologies created for the KSA of the air defense-missile defense forces and the very experience of their development would be easy to extend in the future to other control systems of various control levels of the arms and services of the Armed Forces of Russia.
Possible approaches to the implementation of AI technologies in the CSA of the air defense-missile defense forces
Due to the limited experience of Russian developers in the field of creating systems with AI, it would be advisable to start the task of developing AI technologies for use in the CSA of air defense-missile defense troops with an initiative R&D (or preliminary project) to determine (select) the technology for creating a promising CSA with AI.
It seems that at the first stage, to reduce the development risks, the optimal solution could be a CSA with AI based on heuristic algorithms with formal logic (an expert system with a knowledge base), built using the existing ECB. This approach would be cheaper and easier to implement.
Further development of KSA, with the development of special ECB and technical capabilities, would be the introduction of AI technologies based on artificial neural networks.
Within the framework of the proposed research work, it is advisable to set the following tasks for solving:
1. Collection of information from various sources (information networks, periodicals, books) on existing commercial and military foreign information and technical systems and on similar systems in the commercial sector of the Russian economy, where AI technologies are used. Systematization of information. Search for the correspondence of the tasks of the commercial and military sectors being solved by the AI to the tasks arising in the conduct of combat operations by the air defense-missile defense forces.
2. Building a domain ontology. Selection and justification of the optimal AI technology for the use of air defense-missile defense troops in the KSA. Selection and justification of the optimal level of automation for solving problems by artificial intelligence.
3. Selection and justification of the optimal method for building a knowledge base of CSA with AI.
4. Selection and substantiation of the optimal method for extracting the necessary knowledge and experience from the personnel of the combat crew, the method for structuring the existing experience of combat operations.
The scientific and technical results obtained during the implementation of this research work should be applied in the future in the course of development work on the creation of a new generation CSA.
In future armed conflicts, in the medium term, the center of gravity will shift to the confrontation between weapons control and reconnaissance systems, since in order to ensure military dominance in the world, the United States has chosen a strategy of large-scale use of AI in weapons systems, which, as shows historical retrospective, will force other countries to follow the same path.
It seems that one of the possible effective solutions for countering promising high-tech intelligent air defense missile systems of the leading powers of the world can be the introduction of AI technologies into the air defense missile defense troops.
The creation of such CSA should begin with the selection and implementation of expert system technologies from other fields of technology, in which tasks similar in their algorithmic essence are solved.
1. Brian D. Ripley. Pattern Recognition and Neural Networks. Cambridge: Cambridge University Press, 2011.
2. Roger Penrose. The Large, the Small and the Human Mind. Cambridge University Press, 1997.
5. Nicholas Ernest et al., Genetic Fuzzy based Artificial Intelligence for Unmanned Combat Aerial Vehicle Control in Simulated Air Combat Missions, Journal of Defense Man-agement 6: 1 (2016), 1-7.
6. Google's AlphaZero Destroys Stockfsh in 100 Game Match, Chess.com, 6 December 2017, URL: https://www.chess.com/news/view/google's alphazero destroys stockfish in 100 game matches, accessed 11.02.2018/XNUMX/XNUMX.