The "control of intelligence and control of consciousness," characterized by "submission of will" and "subduing the enemy without fighting," will become the highest-level, most effective, and most deterrent military advantage in future military struggles. This article analyzes the military's demand for artificial intelligence by examining the characteristics of military operations and the advantages of artificial intelligence. Targeting the operational chain of perception, command, strike, and interconnection, it proposes directions for the application of artificial intelligence technology in the military field, exploring how to "effectively shape the situation, manage crises, deter war, and win wars" through the application of artificial intelligence in the military.
introduction
With the rapid development and breakthroughs in core technologies such as deep learning and machine vision, artificial intelligence has ushered in a new wave of development and entered a "golden age". At present, artificial intelligence technology has penetrated into civilian fields such as transportation, services, medical health, education, employment, public safety and protection[1], replacing "lazy people" to complete some physical and mental work. Currently, intelligent technology is constantly subverting military theories, combat rules and combat methods in the information age, powerfully promoting the transformation of the new military system, and gradually changing the form of future wars[2]. The US military regards artificial intelligence as a disruptive technology that "changes the rules of the game" and has already tried and applied artificial intelligence technology in fields such as unmanned combat platforms, electronic warfare, and auxiliary command and decision-making.
Accelerate the development of military intelligence and improve the joint combat capability and all-domain combat capability based on network information system [3]. Intelligent technology presents both challenges and opportunities for us. Faced with the threat situation under the new circumstances and the military combat needs of the new era, we must find our position, clarify our goals, and make great efforts to effectively deal with the risks and challenges brought about by changes in the internal and external environment, strive to seize the development opportunities created by scientific and technological progress, and achieve "leapfrog development".
The demand for artificial intelligence in the military field
2.1 Characteristics of Military Operations
According to Clausewitz's definition in "On War", war is nothing but an expanded struggle, an act of violence that compels the enemy to submit to our will[26]. Although the forms, methods and styles of war have changed over the centuries, the core essence of war has remained the same: to eliminate the enemy and preserve oneself. The characteristics of military operations mainly include the following three points.
1) Unfriendly, uncooperative, and uncontrollable. The success or failure of a war can determine the life or death of both sides, causing them to do everything in their power to deceive each other, conceal their true intentions, and exploit every possible weakness of the other. This makes it impossible for either side to fully and effectively grasp the true state of the war, leading to an uncontrollable battlefield situation. As Churchill said, "Once the first shot is fired or the first bomb is detonated, the political leaders lose control of the war, and the war itself becomes the dominant force."
2) High degree of uncertainty. Combat is a continuous confrontation between opposing sides. However, the complex battlefield environment, command decisions, and means of confrontation inevitably lead to uncertainties in elements such as combat space, combat forces, combat rules, and combat procedures. Therefore, commanders are required to be good at planning ahead, formulate meticulous plans, create opportunities that are favorable to themselves and unfavorable to the enemy, and be able to adjust combat actions in a timely manner based on rapidly changing battlefield intelligence data.
3) The laws of warfare are difficult to grasp. On the one hand, due to the "fog of war," the acquisition of combat data in the battlefield environment is often incomplete, fragmented, or even false, making it difficult for military equipment to learn and train itself, thus failing to grasp the objective laws of warfare and become usable equipment for the army. On the other hand, with the introduction of various reconnaissance and detection methods into modern warfare, all kinds of information flood the battlefield. The excess, overload, surplus, and expansion of data cause much valuable information to be submerged in the sea of data, leading to an exponential increase in unreliable, irrelevant, ambiguous, and contradictory information, thereby increasing the complexity of judgment. With the continuous breakthroughs in a number of emerging technological theories and the continuous expansion of their application scope, and the successive emergence of various new types of weapons, future warfare will be a full-dimensional war conducted on land, sea, air, space, electronic, and cyberspace, a contest of battlefield information processing capabilities, decision-making support capabilities, and rapid strike capabilities. The future of warfare is characterized by multidimensional space, diverse forces, varied styles, and accelerated pace. The demands on the ability to receive and understand battlefield information, assess and predict the battlefield situation, and respond quickly to combat operations will far exceed the thinking capabilities of combat personnel. This will inevitably require the reliance on machines with super computing, learning, and comprehension capabilities for threat assessment and operational decision support.
2.2 Advantages of Artificial Intelligence
From its inception, artificial intelligence (AI) has been entrusted with a noble mission: to replace humans in performing arduous, dangerous, and repetitive tasks. AI possesses significant advantages in these tasks, including greater speed, higher precision, and stronger resistance to fatigue. As AI develops, its capabilities in military deployment, battlefield operations, and decision-making will gradually surpass those of humans.
1) Artificial intelligence excels at solving complex information cognition problems. AI technology can disrupt existing operational rules, enabling machines to perceive complex problems like humans, accumulate experience, and solve them. Through the effective development of battlefield big data, commanders can improve their ability to discover and deeply understand intelligence across multiple battlefield spaces. Data mining and analysis methods can be used to extract high-value military intelligence from massive amounts of heterogeneous information from multiple sources, significantly improving intelligence analysis and processing capabilities. This allows commanders to grasp battlefield developments, predict changes in the enemy's and their own situation, and dispel the "fog of war."
2) Artificial intelligence excels at solving complex state-space problems. While inheriting the advantages of machines, AI technology possesses the ability to efficiently search for and optimize information for complex tasks, making it a powerful tool for resolving uncertainty and complexity. Go has 10,170 possible moves, more than the 10,800 atoms in the entire universe. However, compared to Go, warfare is far more complex and unpredictable. Warfare is characterized by greater battlefield openness, covert offense and defense, and multidimensional combat. Today, AI has breached the fortress of Go and is challenging even more complex games like StarCraft.
3) Artificial intelligence is good at self-learning and upgrading its capabilities. Artificial intelligence technology can achieve the purpose of self-improvement and optimization of system performance through unsupervised learning and machine game through the system background. Taking Go as an example, AlphaGo only took a few months to learn 30 million games played by humans. Based on learning and understanding human Go skills through a large number of historical game records, it trained itself and defeated top human Go players. AlphaGo Zero is fundamentally different from AlphaGo. It does not need to learn historical game records to master human prior knowledge. Instead, it only needs to understand the basic rules of Go. Through self-game and self-evolution, it can quickly improve its Go skills and achieve a hundred battles and a hundred victories against AlphaGo[27]. It is foreseeable that the application of artificial intelligence technology can greatly improve the combat capabilities of key processes such as observation, judgment, decision-making, and action in combat command activities. Artificial intelligence technology will become an important driving force for military transformation and will inevitably give rise to new war styles and accelerate the transformation of war forms.
3. Military Applications of Artificial Intelligence
3.1 Framework of Military Intelligent Technology System
In future warfare, we aspire to possess capabilities such as more thorough perception, more efficient command, more precise strikes, and more seamless interconnectivity. This necessitates addressing numerous challenges, including data interconnection between equipment across operational spaces, mission coordination, and the real-time processing of massive amounts of heterogeneous battlefield data. These challenges must be effectively resolved through deeper levels of intelligence, which will then act as a catalyst for perception, command, strike, and interconnectivity, forming an integrated intelligent combat chain and fundamentally enhancing the effectiveness of system-wide operations. The framework of the military intelligent technology system, as shown in Figure 1, comprises three aspects: the enabling system, the military intelligent system, and the combat system.
Figure 1. Military Intelligent Technology System Framework: The system leverages artificial intelligence algorithms such as machine learning, human-computer interaction, and computer vision to form an AI optimization algorithm engine for military applications, enabling the application of AI technology in the military field. Military Intelligent System: Utilizing enabling technologies and addressing military operational needs, the system relies on fundamental support as an AI algorithm "multiplier" to achieve intelligent OODA (Output-Oriented Development) operational links encompassing perception, command, strike, and interconnection. Operational System: In air combat, anti-missile and anti-adventure operations, space confrontation, and land-sea operations, combat units utilize the military intelligent system in collaboration with humans to enhance operational effectiveness and create an asymmetric advantage over the enemy.
3.2 A more thorough perception to achieve information advantage
In the field of detection and perception, functional technologies such as natural language processing, meta-learning, and random forests can be applied to target information acquisition and battlefield data analysis to achieve information advantage, as shown in Figure 2.
1) Applied to target information acquisition. By comprehensively utilizing various detection methods such as microwave radiation, visible light, multispectral, infrared, acoustic, and magnetic, the efficient and accurate collection and acquisition of battlefield target information can be achieved; by applying technologies such as multi-band-multi-system collaborative detection and multi-source data intelligent fusion, the multi-dimensional feature extraction of targets can be improved, the target position can be accurately calculated, and the target attributes, types, nationalities, identities, friend or foe can be quickly and accurately identified, so that the target information is what you see and what you get [28]. In 2010, the Defense Advanced Research Projects Agency (DARPA) launched the "Mind's Eye" project [29], which aims to develop a visual intelligence system to observe target combat information through unmanned combat platforms and provide timely countermeasures for combat personnel. The project mainly uses intelligent image processing and machine vision technologies to identify and analyze the actions and behaviors of objects in video information, and through accurate perception of the dynamic behavior information of objects, it can identify and recognize potential threats in complex combat environments. Coincidentally, the U.S. Department of Defense established the "Algorithmic Warfare Cross-Functional Group" in 2017 [30] to address the challenges of analyzing massive amounts of intelligence encountered by the U.S. military in its counterterrorism operations against ISIS in the Middle East. This project utilizes technologies such as deep learning and computer vision to replace thousands of intelligence analysts with several computers, thereby improving the efficiency and accuracy of intelligence extraction and supporting more timely and effective decision-making [31].
2) Applied to battlefield data analysis.
By comprehensively utilizing technologies such as big data, machine learning, and data mining, we can find the inherent correlation between the massive amounts of data generated during complex combat operations, quickly and efficiently analyze battlefield combat actions and situational changes, organically link the detected combat force distribution with activities and combat environment, enemy combat intentions and mobility, analyze and deduce the causes of events, obtain estimates of the enemy's force structure and usage characteristics, and infer future possible events through known events [32]. DARPA established the "Insight" project in 2011 [33], aiming to develop an intelligence analysis system that integrates the operator's knowledge and reasoning ability into the system, thereby improving the ability to quickly respond to network threats and unconventional warfare. This project mainly uses technologies such as heterogeneous information association and multi-source intelligent fusion to analyze and integrate multi-source sensor detection information and intelligence data from different resources, thereby assisting in enhancing the information processing and sharing capabilities of intelligence analysts. In 2019, DAR-PA established the "Knowledge-Based Artificial Intelligence Reasoning Model" project [34], aiming to develop a semi-automated artificial intelligence reasoning system that applies the knowledge base obtained through language and common sense reasoning to the understanding of complex real-world events, thereby solving the problem of multi-source information hindering event understanding. The project uses technologies such as knowledge graphs to quickly identify the correlation between different events by reasoning and predicting the internal components and timelines of complex events, thereby improving the ability to understand events.
3.3 More efficient command to achieve decision-making advantage
In the field of command and control, intelligent technologies such as parallel simulation and brain-computer interface can be applied to operational plan simulation and remote command and control to achieve decision-making advantages, as shown in Figure 3.
1) Applied to combat plan simulation. Through deep learning technology, the intelligent agent is trained to learn and simulate knowledge such as battlefield combat rules, combat command decision-making, and event cognitive reasoning, thereby improving the intelligence, real-time performance and scientific nature of the intelligent agent's cognition. On the basis of real-time sharing of battlefield situation, the battlefield data is intelligently processed, and the combat plan is simulated through parallel simulation to form an intelligent prediction of the opponent's next possible military action and the trend of battlefield evolution, and automatically match the best action strategy [35]. In 2007, DARPA arranged a system development project called "Deep Green" [36], which aimed to build an artificial intelligence combat auxiliary decision-making system. The system uses parallel simulation, dynamic game and other technologies, based on real-time battlefield data, to dynamically simulate the combat actions of both sides on the battlefield, and predict the trend of the battlefield situation, helping commanders to think ahead and shorten the decision-making time. In 2018, DARPA launched the “Compass” project [37]. The project mainly uses big data analysis, game theory and other methods to analyze battlefield data, build enemy combat action and path models, help combat personnel determine the enemy’s true combat intentions, and formulate and select the most effective action plan for our side.
2) Applied to intelligent remote command and control. The concept of "metaverse" is applied, and artificial intelligence technology is used to construct a virtual combat space parallel to the real battlefield. Intelligent human-computer interaction technologies such as voice recognition, gesture recognition, and brain-computer interface are adopted to give commanders and combat personnel an immersive experience and realize barrier-free communication between humans and machines, command units, precision strike weapons and information application systems [38]. In August 2021, at the U.S. Navy's largest annual event, the "Sea-Air-Space Expo" [39], the Naval Information Warfare Systems Command verified for the first time the development capability of the "Surrounding Environment Intelligent Conversation Interface" project, demonstrating how intelligent and natural interaction technology can realize future information warfare. The project aims to introduce the next generation of digital assistants for naval command and control. By using artificial intelligence and machine learning to understand who the speaker is and what the conversation is about, the conversation can be used by decision-makers as a direct way to obtain the information they need, helping decision-makers obtain timely and synthesized information.
3.4 More precise strikes to achieve a power advantage.
In the field of weapon strikes, intelligent technologies such as computer vision and multi-agent collaboration can be applied to achieve force superiority, mainly in areas such as autonomous operation of single weapon platforms and distributed lethality of combat formations, as shown in Figure 4.
1) Applied to autonomous combat on single weapon platforms. With artificial intelligence technology as the core, and a variety of weapon equipment platforms with embedded artificial intelligence algorithms as the means, it can achieve real-time precision strikes in multiple dimensions and realize the individual intelligence of the weapon [40]. When the US Tomahawk missile attacks the target, if the target or mission changes, it will hover over the battlefield according to the instructions, and then autonomously search for and reselect and determine the appropriate attack target. The US-developed "Wasp" missile is equipped with an advanced detection and control equipment, which can realize the identification of target camouflage facilities and the intelligent autonomous allocation of multiple mission targets, thereby achieving the maximum cost-effectiveness and hit accuracy.
2) Applied to distributed lethality of combat formations. Drawing on the intelligent swarm and collaborative technology of biological group behavior in nature, the system’s resilience and mission success rate are improved through decentralization; the overall effect and collective intelligence level of the system are improved through efficient information interaction between simple combat units, thereby ultimately realizing autonomous decomposition of missions, autonomous collaboration of combat units, autonomous planning of combat plans and autonomous strike of combat targets under complex battlefield conditions [41]. DARPA established the “Cooperative Operations in Denied Environments” project in 2014 [42], aiming to develop an autonomous collaborative combat system that enables one operator to command multiple UAVs. The project solves the problem that UAV swarms cannot complete combat missions under complex interference conditions through advanced algorithms and modular software architecture, and improves the ability of UAV swarms to complete missions. DARPA established the “Gree” project in 2015 [43], aiming to establish a reusable UAV combat swarm and realize a stable, reliable and cost-effective combat mode. The project completes large-scale rapid reconnaissance and deception interference of the battlefield area before the war by using integrated design, autonomous collaborative planning and other technologies.
3.5 More free interconnection, realizing the advantages of the network
In the field of battlefield interconnection, cognitive computing, game-theoretic confrontation and other intelligent technologies can be applied to battlefield network resilient communication and network attack and defense to achieve network advantages, as shown in Figure 5. 1) Applied to battlefield network resilient communication. Artificial intelligence technology is used to agilely perceive the network environment, flexibly load communication waveforms, and autonomously manage network resources to improve the resilience of the battlefield communication network system. In recent years, in order to adapt to new military strategies and combat situations, the US military has been exploring how to ensure a flexible, agile and resilient guaranteed communication system in anti-access/area denial combat environments. The US Air Force Laboratory and the Canadian Defence Research and Development Center Communications Research Center have carried out the "Guaranteed Communication in Challenging and Confrontational Environments" project research [44], which is mainly aimed at the harsh communication conditions that future combat personnel may face, especially in remote and service-deficient conditions and dynamic and confrontational environments. By developing new concepts and technologies, flexible and adaptive spectrum access can be achieved to ensure robust and reliable communication capabilities. In 2017, DAR-PA launched the "Radio Spectrum Machine Learning System" project [45], which uses artificial intelligence to understand radio signals, improve and promote spectrum sharing technology, and enhance wireless communication capabilities. 2) Applied to network attack and defense. Using artificial intelligence as a weapon, malicious attack behavior can learn on its own and adapt to the differences in the target defense system to achieve the purpose of attack by mass-recruiting potential vulnerabilities. At the same time, the use of artificial intelligence technology can improve the current status of network security, and can identify known or unknown threats more quickly and respond in a timely manner. In 2017, Stanford University and the startup Infinite jointly launched an autonomous network attack system. The core processing unit of the system is a customized artificial intelligence processing chip [46]. This new network attack system can run in a specific network, complete the autonomous collection, learning and autonomous writing of attack programs, and can adaptively and dynamically adjust the attack programs, with strong concealment and destructiveness. In 2018, DARPA launched the "Advancing Autonomous Systems Against Cyber Adversaries" project [47], which aims to establish a secure and reliable network proxy and achieve effective countermeasures against botnets.
Containment. This project develops quantitative frameworks and algorithms to accurately identify botnets, infer existing vulnerabilities, and generate software patches, thereby reducing their adverse impact on systems.
3.6 More robust support to realize empowering advantages
1) Intelligent artificial intelligence systems provide “new impetus” for military intelligence. Traditional machine learning methods require training the system with datasets before system deployment. Once the training is completed, the scenarios and problems that the intelligent agent can deal with will be solidified, making it unable to deal with new scenarios. Retraining is inefficient and labor-intensive. When carrying out military operations, artificial intelligence systems need to be able to learn and improve themselves in the mission, applying previous skills and knowledge to new situations to deal with various combat scenarios [42]. In 2017, DARPA arranged a project called “Lifelong Learning Machines” [48], which uses target-driven perception to continuously learn and form autonomous adaptation to new scenarios, changing the current situation where intelligent agents cannot deal with untrained scenarios. 2) Low power consumption, strong computing power, and easy expansion of intelligent chips provide “new infrastructure” for military intelligence. As an important physical foundation of artificial intelligence technology, the current mainstream artificial intelligence chips have bottlenecks such as high power consumption, insufficient memory bandwidth, and solidified framework. In order to better support the application of artificial intelligence in the military field, the next generation of artificial intelligence chips should have the characteristics of low power consumption, strong computing power, and easy expansion. In 2020, Nvidia announced its AI chip for supercomputing missions[49], which boasts more than 20 times the computing power. In October 2020, Intel announced that it had been awarded a second-phase contract for a project with the U.S. military[49] to help the U.S. military produce more advanced AI chip prototypes domestically. This packaging technology can integrate “small chips” from different suppliers into a single package, thereby integrating more functions into a smaller finished product while reducing its power consumption.
