Intelligent Communication, Concept of Domain Junction, and Continuing Discussion on ChatGPT and Warfare
Issue 42, 23 April 2023
Hello and Welcome to China Tech Dispatch!
In this newsletter, I bring fortnightly updates on “informatized” and “intelligent” warfare, new and emerging technologies, and discussions on these themes from official and state-sponsored sources in China.
Happy reading!
I. Military and Warfare
A. Continuing Discussion on ChatGPT and Warfare
ChatGPT is seemingly on the minds of PLA scholars this April. Many articles continue to explore the effect of ChatGPT and larger-scale generative language models in the military field. Key questions these writings attempt to address are:
What changes will ChatGPT bring to the military field? (Looking at the military application of artificial intelligence from ChatGPT)
Will AI surpass humans?
How large-scale generative AI models will contribute to the future of warfare? (Military Application of Large-scale Model Technology) (How Generative AI Can Affect the Future of Warfare)
How far is generative artificial intelligence from being fully applied in the military field? (How far is it from comprehensive application in the military field?)
Some key arguments that emerge from these articles are as follows:
AI can be useful to
Improve battlefield awareness.
Promote human-computer interaction.
Boost the autonomy of command and decision-making.
A new model of logistics support.
There is a limitation to AI's application and development since the AI is essentially based on data that is used to train it.
AI cannot replace humans and will not surpass human intelligence, as of now.
At the current level of development, it is difficult for AI to "break the routine", let alone "create something out of nothing". This means that AI cannot deduce the unknown from the known. The capability to innovate will be the greatest advantage of human beings.
This further means that the ability of human commanders to break conventions and innovate tactics is irreplaceable.
Humans, not AI, still dominate the battlefield. This is reinforced by the ceiling of the development of AI technology. This limitation was put forward by "Gödel's incompleteness theorem." Simply put, AI based on current computer technology may never be able to surpass humans. (Note: Gödel's incompleteness theorem is a little more complex. However, my own understanding of this topic is limited. I hope this short video explainer might help those who are interested.)
The following quote from an article written by Mao Weihao summarizes the case of humans over AI,
"The real world is different from the digital world simulated by computers. The real world is uncountable, and the ratio of pi and the natural constant e can be calculated endlessly. AI operates in the digital world, while human beings live in the real number universe. Using AI based on current computer technology to simulate human intelligence is like using a language system to describe human feelings—language is countable, but feelings are uncountable, so some feelings can only be understood, but not expressed in words. It can be deduced from this that before the arrival of the real technological "singularity", AI can only keep approaching human intelligence, but it will never surpass it.." (而真实世界与计算机模拟的数字世界是不同的。真实世界是不可数的，圆周率π、自然常数e都可以无穷无尽地计算下去。AI运行于数字世界，人类则生活在实数宇宙。用基于目前计算机技术的AI来模拟人类智能，就像用语言系统来描述人类感觉——语言是可数的，而感觉是不可数的，所以有些感受“只可意会，不可言传”。由此可推导出，在真正的技术“奇点”到来之前,AI只能不断逼近人类智能，但永远无法超越。)
Some scholars argue that the full application of generative artificial intelligence in the military field seems to be relatively far away. It is primarily because:
Need for massive open source data to train generative AI and a lack of large-scale open military data to train AI. One of the articles mentions,
"at the level of military applications, due to the lack of actual combat experience in modern warfare, a large amount of data comes from daily training and exercises, which directly affects the effect of artificial intelligence training; in addition, the training data has not yet been manually labeled, and the availability and interpretability of the data are poor, making it difficult to Effective support for reinforcement learning pre-training." (然而在军事应用层面，由于缺少现代化战争的实战经验，大量数据来源于日常训练和演习，直接影响到人工智能训练效果；此外，训练数据尚未开展人工标注工作，数据的可用性和可解读性差，难以有效支持强化学习预训练。)
The biggest contradiction in the military application of artificial intelligence is the ethical issue and low trust in human-computer interaction. Internal algorithms of the large-scale generative AI models are like a black box, and it cannot be guaranteed that they will not attack or harm users. Hence right now these applications are limited to civilian low-cost mutual trust areas. The author also notes,
"The comprehensive application of generative artificial intelligence in the military field requires legal ethics, data base reserves , Computing capacity building, multi-source interactive environment and other aspects of efforts, it is expected to make it better serve human beings." (生成式人工智能在军事领域的全面应用需要在法律道德规范、数据基础储备、计算能力建设、多源交互环境等多方面发力，厚积薄发，才有望让其为人类更好地服务。)
B. Generative AI and Cognitive Confrontation
The article by Chen Dongheng and Xu Yan (陈东恒 许 炎), affiliated with the Military Political Work Research Institute, Academy of Military Sciences (军事科学院军队政治工作研究院) explores the role of generative AI in cognitive confrontation. Key points from the article:
Cognitive habit shaping (认知习惯塑造)
This involves the formation of habit through long-term exposure to the products to achieve a credible, approachable, and undefended cognitive convergence later. The authors give an example of how initially the market for the internet was shaped by low-cost free access. This made people addicted to the use of the internet.
Another example author gives is ChatGPT, which is freely available and is modeled as an ‘assistant’ or ‘caring friend’ ready to help users.
Authors point out that the generative AIs, which seem to be harmless to humans and animals, but are actually controlled by humans, once they become "close friends" that people trust and talk about, their influence on people will be subtle. Hence, there are traps in "being close to the people", and one must not be confused with the appearance of tenderness. (Can’t argue with that now, can we?)
Cognitive information penetration (认知信息渗透)
Authors point out that the USA and other Western countries export values through media platforms and cultural and creative products to influence the value orientation of people in other countries.
Similarly, the USA has a first-mover advantage in generative AI undoubtedly gives it new tools to infiltrate its opponents.
This means that seemingly objective data could be biased. Hence, it is necessary to strengthen the protection of strategic resources such as data and information and prevent technological rationality from replacing value rationality.
Cognitive thinking-frame (认知思维构陷)
This approach uses the “novelty entertainment hunting+ consciousness penetration” ("猎奇娱乐+意识渗透”) approach to stimulate primitive desires such as the pursuit of comfort, novelty. It is used to stimulate cognitive feelings, and influence people's thinking.
Generative AI such as ChatGPT aims at the social and emotional difficulties of modern society by appearing as a pleasant chat robot, which meets the emotional demands of many people.
Authors mention that this kind of technological application that focuses on catering to the characteristics of human nature can easily lead people's thinking and cognition into narrow, one-sided, and superficial traps.
Authors warn that we must maintain a rational approach while using new technologies, and avoid falling into artificial traps while enjoying technological advantages.
Cognitive standard hijacking (认知标准挟持)
This involves fusing technology with values.
The authors mention this practice is emerging in the United States and Western countries, which have absolute control over the Internet and use it as a weapon to promote Western values. The authors also mention The United States has suppressed voices and practices that are different from those of the United States and the West on certain issues by controlling the use of the Internet and the use of international payment systems.
Lastly, the authors mention generative AI represented by ChatGPT can replicate this "bundle" hijacking technique, by generating answers and providing "standard answers" that incorporate American values. Hence, one has to strengthen cognitive confrontation, strengthen strategic information system technology innovation and product research and development, and use technology to support cognitive confrontation with a strong voice.
C. Intelligent Communication in Cognitive Domain Operations
An article by Li Xiaoyang (李晓阳) affiliated with National Security College, National Defense University (国防大学国家安全学院) highlights the importance of the Cognitive Domain of Intelligent Communication. Key points from the article are given below.
The author explains 'Intelligent communication' is an activity that applies technologies such as big data, cloud computing, machine learning, and virtual reality to information production and dissemination. It can shape the value orientation and behavior speculation of the target audience. Intelligent communication has become a new platform and an important field for cognitive attack and defense.
Characteristics of the Cognitive Domain of Intelligent Communication
Everything is ubiquitously connected and any smart terminal can become a tool for users to receive information. It can become a channel for the platform to deliver content, and the capability of such intelligent terminals to push content is very powerful.
A mixture of fiction and reality. Using multi-communication patterns empowered by intelligent technology, information can be forged, altered, and presented as if it is true.
Human-machine two-way interaction. With the support of technology, intelligent media can obtain human-like perception, computing, and even emotional intelligence. With the help of human-computer interaction, it can reversely affect the user's way of thinking, and shape the user's behavior orientation through algorithm recommendations.
The Mechanism of Intelligence Spreading in the Cognitive Domain
At present, intelligent communication has become an effective means to dominate information content and control the direction of public opinion. The mechanisms to achieve this include the following ways.
Cognitive influence is exerted through situational immersion. The developments in communication are increasingly showing an immersive trend. Intelligent communication enables information to directly hit the sensory channels of the target audience in a three-dimensional way, thus giving birth to the "Sensory Touch - Psychological Touch - Emotional Resonance" (“感官触碰－心理触动－情感共鸣”) in order to realize the influence and control of its perception on the physiological level.
Influence cognitive orientation through virtual mapping. Although smart devices capable of intelligent communication are now placed in external environments, in the future they will increasingly form an inseparable relationship with the human body. Hence, the behavior, activities, and physical status of the audience can be mapped in the virtual world in multiple dimensions. However, the individuals who are digitally mapped would be more susceptible to being observed and manipulated by others.
Using human-machine fusion to strengthen cognitive intervention. Based on user identity and real-time interaction, 'social robots' can affect the public's thinking habits and behavior patterns to a certain extent, and have remarkable efficiency in spreading false information, manipulating public opinion, conducting social mobilization, and inciting public emotions. With the continuous improvement of technology, the 'social robots' will be richer and more textured, and can even imitate human emotions.
The Means of Intellectual Diffusion in the Cognitive Domain
Intelligent communication has increasingly become a "brain control weapon" and gradually developed a set of cognitive manipulation tools. These tools can:
Create information fog and attack cognitive black holes. Intelligent programs represented by 'social robots' have the ability of automatic recognition, intelligent response, etc. They are tireless and can work round-the-clock. Such information bombardment forms high-intensity cognitive oppression. Once false opinions or falsified events form a public opinion environment, the individual reaction time will be greatly reduced, and it will be difficult for the audience to think clearly, and it is easy to be driven by the herd mentality
Implement targeted attacks and build a 'cognitive cocoon.' By grasping the target's online traces, shopping records, social status, etc., it can quickly collect cognitive data of different dimensions, levels, and modes, and provide support for efficiently grasping its values, public opinion positions, etc. It is possible to discover doubts, weaknesses, and needs of the target object's cognitive system, and to set content themes, narrative methods, and discourse frameworks according to their thinking preferences. This can create a cognitive cocoon which can then be used to change decision-making ability and influence users’ actions.
Penetrate psychological fissures and break down emotional embankments. With the help of tactile, sensible, and interactive intelligent communication, one can create a deceptive, confusing, and provocative information scene which strongly stimulates the user's psychological feelings. This can force users to get deeply immersed without realizing it. Once immersed, one can use common emotions and common values to incite emotions or impose moral coercion, destroy the users’ emotional dependence, destroy their value support, and then shock, occupy, or even polarize the mind of the target object.
D. Domain Junction: A Contested Field in Future Wars
An article by Yuan Yi, Zhang Yujun, and Li Zhifei (袁 艺 张玉军 李志飞) explains the concept of "domain junction"(域接合部). Key points from the article are:
The cross-integration of various combat domains will produce "domain joints" with important strategic status and great combat value, and hence give birth to a series of new operations.
The "domain junction" mentioned in this article refers to the connection between land, sea, air, space, electricity, network, and other combat domains.
Examples include
Coastal areas and offshore areas where land and sea are connected
Low-orbit and sub-orbits where air and space are connected.
Cognitive domains where network domains are connected to cognitive domains
Power domains are connected to electromagnetic domains.
Authors argue that in future wars, the competition for comprehensive control of the battlefield may first revolve around the “domain joints”, which will attract attention as the focus of the joint global battlefield.
The “domain junction” is an area where the capabilities of two or more combat domains overlap with each other.
Domain joints have become the basic direction for the multi-domain expansion of service combat capabilities. Domain junction operations are the touchstone for measuring the level of joint operations. Taking offshore landing operations as an example, it includes not only single-domain operations such as the landing operations of the Army and Marine Corps, the Air Force's air control operations, and the Navy's sea control operations but also anti-ship and anti-submarine operations of the Army and Air Force, naval guns and Cross-domain operations such as cruise missile land attacks.
Strengthening the capacity building of domain joints is an important part of future war preparations.
Integrate and coordinate the development of the capabilities of the domain joints of various services.
This includes developing the capability structure of each service in a coordinated and integrated manner.
The authors also emphasize the need to encourage and accommodate all military services and develop overlapping cross-domain combat capabilities based on actual combat needs.
Avoid a “one-size-fits-all” approach and avoided avoid blind expansion of service capabilities.
Innovate the cross-domain collaboration mechanism of domain joints
Learn from the experience and practices of the armed forces of powerful countries.
Strengthen military cooperation and joint training for different types of domain joints
Open up cross-domain kill chains
Innovate multiple cross-domain coordination mechanisms for different domain joints, such as land, sea, and air joint strikes against offshore surface targets coordination mechanisms.
Accelerate the development of domain joint combat technology and equipment
Explore cross-domain combat support technologies such as
Cross-water and air-media communication
Air-space shuttle flight
Heterogeneous network communication "middleware" to improve interconnection, interoperability
Cross-domain mobility capabilities of various combat domains
Promote the multi-functionality of combat platforms and precision munitions.
Accelerate the development of cross-domain combat equipment, such as
Cross-water and air domain aircraft
Cross-domain combat platforms such as space vehicles
All-area combat ammunition that integrates anti-ship, air defense, anti-missile, ground strike, and other functions.
II. Additional Reading
Lingel, Sherrill, Jeff Hagen, Eric Hastings, Mary Lee, Matthew Sargent, Matthew Walsh, Li Ang Zhang, and David Blancett, Joint All-Domain Command and Control for Modern Warfare: An Analytic Framework for Identifying and Developing Artificial Intelligence Applications. Santa Monica, CA: RAND Corporation, 2020. https://www.rand.org/pubs/research_reports/RR4408z1.html.
Grasp the key to the development of emerging military technologies
Innovation to drive science and technology to strengthen the military
Intelligent maintenance support: helping weapons and equipment return to the front line
What are the advantages of manned/unmanned aerial vehicle coordinated operations?
Megha Pardhi is a Research Analyst (China Studies). She tweets at @pardhimegha21. All views expressed are personal.