1 Introduction
Artificial intelligence (AI) technology emerged in the 1950s and has developed into one of the most cutting-edge high-tech fields. It is one of the main pillars of the fourth technological revolution and has become an important aspect of the future comprehensive national power contest. The United States, Russia, the European Union, Japan, and other countries and regions globally attach great importance to the development of AI technology, actively formulate relevant strategies and plans, and strengthen technological research and development in this field [1–4].
In recent years, China has launched numerous strategies and policies to strengthen the top-level layout so as to direct AI development, and to support the rapid and healthy development of AI technology. During this process, it is necessary to follow the objective laws of technological development and to seek an AI technology innovation path that is suitable for national conditions. The major current challenges that urgently need to be resolved are the development and innovation model as well as the optimal path of AI technology under the national system in the new era. Furthermore, the exploration of compatible investment policies, talent mechanisms, project management systems, and performance evaluation systems is required.
As AI is a disruptive emerging technology, systematic studies thereon have not been carried out regarding its development rules. To this end, this study uses comparative analysis methods to compare relatively mature nuclear energy, photovoltaic technology, and AI technology. These technologies are discussed according to the three aspects of basic theory, technology development, and the technology market application mode. The unique development laws and innovation path of AI technology are established. Furthermore, measures and suggestions are presented to promote the development of AI technology innovation in China.
2 Definition of AI technology and industry
2.1 Definition of AI technology
The Standardization Administration of China defines AI as follows
[5]: Theories, methods, technologies, and applications that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results.
According to this definition, AI can be understood as allowing machines to imitate human intelligence through specific technical means. Based on the maturity of the imitation ability, AI technology can be divided into different stages, including weak AI, strong AI, and super AI
[6]. Weak AI refers to AI that is good at a single aspect and basically operates at the level of intelligence in computing and perception. Strong AI refers to AI that can design plans, solve problems, conduct abstract thinking, understand complex ideas, and perform deep learning; it is similar to human processing in all aspects. Super AI refers to AI that has more processing power than humans in almost every field, such as scientific innovation, general education, and social skills. At present, AI technology is in the weak AI stage [7,8]. To develop AI technology from the weak AI stage to the strong AI stage, major breakthroughs need to be achieved in basic frontiers and core technologies.
The core technologies of AI include the following: machine learning, computer vision, knowledge engineering, natural language processing, speech recognition, computer graphics, multimedia technology, human–computer interaction technology, robotics, database technology, visualization technology, data mining, and information retrieval and recommendation. Benefiting from the progress made by universities, scientific research institutions, and enterprises in the area of AI core technology, AI technology has gradually assimilated the characteristics of intelligence, versatility, efficiency, and globalization; thus, its role in economic and social development will become increasingly important. It should also be noted that certain problems remain in AI core technology, such as insufficient flexibility, weak interpreting ability, and poor robustness, all of which pose a severe challenge to the future development of AI technology.
2.2 Definition of AI industry
Industry is a complex economic system that includes management, technology, personnel, production, market, resources, and information, among other factors
[9]. At present, the concept of the AI industry is usually defined in both broad and narrow senses, in which the meanings and manifestations of the related elements differ. From a broad perspective, the AI industry refers to the in-depth and extensive application of AI technology to promote technological integration and business model innovation, promote intelligent product innovation in key areas, and promote intelligent upgrades in key industries, thereby forming an intelligent drive, human–machine collaboration, and the cross-border integration of new industrial development forms. From a narrow perspective, the AI industry refers to the collection of a series of economic activities, such as the research and development, production, and sales of groups, teams, and individuals aimed at AI basic theories, technologies, systems, platforms, and products and services based on AI technology
[6], and representative nodes and applications have been produced in the civilian market. For example, the Webot 1 robot developed by Japan in 1973 was the first humanoid robot to run on intelligent software that could play music; the Webot 2 robot developed by Japan in 1980 could communicate with people, read sheet music, and play an electronic keyboard. The Deep Blue developed by IBM in 1997 defeated the chess champion, and in 2006, the French Aldebaran Robotics Company developed the intelligent robot Nao. From the perspective of demand, given the particularity of demand scenarios in the defense field, such as less data, uncertain boundaries, complex environments, and high real-time response, several problems remain in the application of AI technology in that field. Overall, the market application of AI technology differs from those of nuclear energy and photovoltaic technology and is still dominated by civilian needs. Limited by the level of technology, the application of AI technology in the civilian field will be more extensive than that in the military field for a long time in the future.
3.4 Conclusions of comparative analysis
According to the comparison with nuclear energy and photovoltaic technology in terms of the basic theoretical research, technology development, and technology application market (Table 1), it is evident that AI technology exhibits the following laws of development:
(1) AI technology is relatively lacking in terms of basic theoretical research
[16]; therefore, it is very important to lay a solid foundation for theoretical research as soon as possible to promote the mature development of AI technology in the future. Innovation is at the core of information technology development, and the rapid and continuous innovation of AI technology is the only solution for its development.
(2) As with nuclear energy and photovoltaic technology, AI technology also offers the potential for applications in both the military and civilian markets. The coordinated development of the military and civilian markets is an inevitable method for future development. Owing to the limitations of the insufficient development of high-end AI technology, and significantly different demands for AI technology between the military and civilian markets, the unique and innovative development path of AI technology in the application of the military and civilian markets is the key development direction for AI technology.
Table 1. Comparison of nuclear energy, photovoltaic, and AI technology research.
4 Path of AI technology innovation
4.1 “Evolutionary” development is inevitable for the development of AI technology
Owing to the insufficient development of basic theories, the current AI development remains at the weak AI stage; although there exists a huge civilian market and a large military market demand, practical application of AI technology in the military fields is rare; therefore, a unique innovation path must be adopted for the development of AI technology in China. To this end, the concepts of “involution” and “evolutionary” development are adopted for discussion in this article. Involution development refers to a method for increasing the total output by investing a large amount of existing technology in a limited field
[17]. Involution development is generally represented by the duplication, extension, and refined development of technology, but there exists the dilemma of limited efficiency improvement in the development process. Evolutionary development places emphasis on the technology development and the achievement of developmental breakthroughs, and results in new social benefits through expanding the application fields.
AI technology is currently following the involution development model, in which conventional technology is quite advanced, high-end technology remains underdeveloped, and the technological value and influence have not been adequately demonstrated. Certain AI technologies (such as language recognition and image recognition) have been used extensively in the fields of social media, translation, security, and medical treatment, but the low robustness and interpretability have not been improved, and the application scenarios are constrained. This phenomenon relates to the difficulty of making breakthroughs in the technology itself. Moreover, owing to the huge scale of the domestic civilian market, the use of only weak AI in certain specific markets and market segments can benefit technology owners significantly, thereby resulting in a lack of motivation for in-depth research on AI technology. The high returns of involution development are short lived and the subsequent market competition will weaken the returns
[18]. AI technology can only maintain its strong technical vitality by strengthening innovation and creativity, and adopting an evolutionary technology development model that focuses on non-specific markets and “non-comfort zone” market applications.
4.2 Integrated development promotes the evolutionary development of AI technology
New external environmental drivers are required to promote the development of AI technology to move beyond its “comfort zone” for transformation from an involution to an evolutionary development model. Compared to the civilian demands, the military demands for AI technology are more high precision and sophisticated, and the market characteristics differ significantly. To achieve a breakthrough in the field of strong AI and super AI, it is impossible to rely solely on civilian market demands; thus, the industry must be developed according to military demands. Military scenario applications require AI technology to exhibit features that work in environments with battlefield characteristics, such as boundary uncertainty, strong antagonism in games, effective real-time response, high environmental complexity, and information incompleteness. The corresponding equipment should meet the demands of reliability, maintainability, testability, security, and environmental adaptability, which forms the development pathway for advancing AI from weak AI to strong AI, and subsequently super AI. It can be considered that the realization of the integrated development of AI technology based on military needs has become the preferred means for overcoming the difficulties of AI technology innovation and ensuring innovation vitality.
Furthermore, it should be noted that the development model of AI technology for the military and civilian markets differs from that of nuclear energy and photovoltaic technology. The marketing model of nuclear energy and photovoltaic technology serves the national defense and military field first, and then drives the application in the civilian market. The technical barriers to transfer from the military market to the civilian market are relatively low, and it is easier to cross the military–civilian industry boundary and to realize the evolutionary development of technology in the military–civilian field. Owing to the notable differences in the military and civilian market demands, AI technology faces a greater obstacle in transforming back and forth between military and civilian needs. The evolutionary development of AI technology can be achieved by strengthening the traction of military demands and crossing the military–civilian industry boundary.
Based on national conditions, a three-step route can be adopted to achieve the integration and development of AI technology. First, the civilian technological achievements are matched with the demand application scenarios of the special needs of the military market to realize the direct conversion of civilian achievements into military applications. For example, the civilian market of smart security technology is relatively mature, but military scenarios have not yet been thoroughly developed. Through this transformation, not only can the application scope of the technology be expanded, but the maturity and reliability of conventional weak AI technology can also be initially tested, and the deficiencies of the technology can be determined rapidly. In the second step, emerging AI technologies, such as intelligent security technologies, can achieve breakthroughs in military–civilian collaborative innovation and joint research, thereby promoting the emergence of general-purpose AI technologies with strong AI characteristics and dual-use attributes. The third step is to address the high demand for military use. The military will also guarantee necessary investments for special needs, conduct high-end technology research that is difficult to carry out or faces resistance in the civilian field, and promote the development and commercialization of super AI technology. It will also transfer research results to the civilian field in stages and explore the formation of “military traction for civilian use”; for example, bio-intelligence and intelligent micro-systems are possible directions for technological application.
5 Countermeasures and suggestions
5.1 Top-level layout and construction of new-type R&D institutions
It is recommended that the State Council and Central Military Commission jointly guide and support the construction of a new type of R&D institution with leadership and resource deployment capabilities in the civilian and military fields. In view of the diversification of AI technology and the characteristics of “group intelligence,” and based on the principles of high efficiency and pragmatism, the project operation mode of “top-level coordination, military–civilian coordination, individual focus, and teamwork” should be adopted, so that the basic theoretical research of AI can be explored freely and a highly qualified team can conduct centralized research on the high-end technology. The military and civilian market needs should also be effectively coordinated, so that the operation of new R&D institutions will conform to the basic characteristics and development laws of AI technology innovation.
5.2 Broadening channels and increasing resource protection
The integration and development of AI technology requires stable funding support channels. It is recommended that the “research funding + market benefit” driving model be adopted. When civilian achievements are transformed into military use, a major national science and technology project should be established to guarantee resources through scientific research funding. In the stage of military–civilian collaborative innovation and military traction for civilian use, when relevant research results are transformed into the civilian field, state-owned or private enterprises may be introduced into the industry, with a special cultivation fund for civil products providing directional support.
5.3 Attaching importance to talents and promoting the construction of an integrated talent team
At present, the leading figures and teams in the AI field are concentrated in well-known research institutes and key enterprises locally and globally, but the integrated development of AI technology requires the construction of a multi-source integrated talent team to focus on scientific research more thoroughly. The construction of this talent team needs to rely on relevant scientific research institutions. During the early stage, the “dual-employment system” model can be used to introduce advanced talents and teams to clarify the rights and obligations of personnel. Once the team building gradually matures, a targeted training mechanism can be introduced to cultivate core teams and leading figures. Thus, the deadlock in the introduction of existing talents can be broken. Moreover, to achieve “equal pay for equal work,” military and local talents can be attracted to tackle key problems jointly, and full play can be given to their respective advantages to achieve strong military–civilian collaborative innovation. During the process of building a talented team, an evaluation mechanism that reflects differentiation should be established simultaneously.
5.4 Collaborative development and conducting ethical research
The research on the ethics of AI technology in civilian scenarios has been developed. However, as the application of military scenarios started later than civilian use, several deficiencies remain in the current research on the ethics of military scenarios. Thus, attention should be focused on the coordinated development of ethical norms in the application of AI technology in the military and civilian markets. Furthermore, research should be conducted on military-specific ethical norms, such as responsibility ownership, deterrence, and destructive trade-offs, based on the existing research results of ethical norms.
PDF
(506KB)
