Artificial intelligence (AI) is to simulate and expand machines by means of machine learning and data analysis.
In recent years, driven by big data, algorithms and computer capabilities, artificial intelligence has entered a stage of rapid development.
Artificial intelligence market structure
Artificial intelligence empowers the real economy and brings revolutionary changes to production and life. As the core force of the new round of industrial transformation, artificial intelligence will reshape all aspects of economic activities such as production, distribution, exchange and consumption, and give birth to new businesses, new models and new products. From food, clothing, housing and transportation to medical education, artificial intelligence technology is deeply integrated and applied in various fields of social economy. At the same time, artificial intelligence has strong economic radiation benefits and provides a strong engine for economic development. According to Accenture's forecast, in 2035, artificial intelligence will promote China's labor productivity to increase by 27%, and the total economic added value will increase by 7. 1 trillion US dollars.
Comparison of multi-angle artificial intelligence industries
Strategic deployment: big countries compete, and the layout has its own emphasis.
On a global scale, China and the United States "side by side" constitute the first echelon of artificial intelligence, and developed countries such as Japan, Britain, Israel and France pursue victory and form the second echelon. At the same time, in the top-level design, most countries strengthen the strategic layout of artificial intelligence, upgrade artificial intelligence to a national strategy, and protect artificial intelligence from three aspects: policy, capital and demand. China, a rising star in Ran Ran, has made breakthroughs in some fields. China's artificial intelligence started late and experienced ups and downs. Since 20 15, the government has intensively introduced a series of support policies, and artificial intelligence has developed rapidly. In the early days, China's policy focused on the Internet field, and the capital investment was biased towards the terminal market. Therefore, compared with the industrial layout of the United States, China's technology layer (computer vision and speech recognition) and application layer are in the forefront of the world, but the core areas of the basic layer (algorithm and hardware computing power) are relatively weak, showing a "top-heavy" trend. At present, artificial intelligence in China emphasizes systematic and comprehensive layout at the national strategic level.
The United States leads the frontier research of artificial intelligence, and its layout is slow and strong. The US government is a bit slow, and the 20 19 National Strategy for Artificial Intelligence ("American Artificial Intelligence Initiative") came late. However, because the United States has the natural advantages of geographical location (Silicon Valley) and human resources (talents), it has always been in an all-round leading position in the competition of artificial intelligence. Generally speaking, the layout of the United States in key areas is cutting-edge and comprehensive, especially in the fields of algorithms, chip brain science and so on. In addition, the United States pays attention to the influence and change of artificial intelligence on national security and social stability, and attaches great importance to data, network and system security.
Under the guidance of ethical values, European countries have seized the commanding heights of norm-setting. In 20 18, 28 European member States (including Britain) signed the declaration on cooperation in artificial intelligence, which formed a joint force in the field of artificial intelligence. From the national level, due to the cultural and linguistic differences that hinder the formation of large data sets, European countries do not have the first-Mover advantage in the artificial intelligence industry, but European countries have seized the "opportunity" in the global AI ethics system construction and norm formulation. The EU attaches importance to the social ethics and standards of artificial intelligence, and occupies a leading position in the world in technical supervision.
Japan seeks artificial intelligence to solve social problems. Guided by artificial intelligence to build a "super-intelligent society", Japan designated 20 17 as the first year of artificial intelligence. Due to the scattered data, technology and business requirements in Japan, it is difficult to systematically develop artificial intelligence technologies and industries. Therefore, the Japanese government focuses on three areas with comparative advantages: robotics, medical health and autonomous driving, and focuses on solving national problems in the fields of old-age care, education and commerce.
Basic level: the technology is weak, and the road to chips has a long way to go.
Because it is difficult to innovate at the basic level and the technical and financial barriers are high, the underlying basic technology and high-end product markets are mainly monopolized by a few international giants such as Europe, America, Japan and South Korea. Limited by the lack of technology accumulation and R&D investment, China is relatively weak in the basic field. Specifically, in the field of AI chips, international technology giant chips have basically built an industrial ecology, while China has not mastered the core technology, and the chip layout is difficult to compete with giants; In the field of cloud computing, core technologies such as server virtualization, network technology (SDN) and voice development are in the hands of a few foreign technology giants such as Amazon and Microsoft. Although domestic technology companies such as Ali and Huawei have also begun to invest heavily in R&D, the accumulation of core technologies is not enough to lead the development of the industrial chain; In the field of smart sensors, Europe (BOSCH, ABB), the United States (Honeywell) and other countries or regions have comprehensively laid out a variety of sensor products, while in China, Ding Hui's fingerprint sensors and other products have appeared, but the overall industrial layout is single, showing obvious shortcomings. In the field of data, China has a unique advantage of data volume, which is helpful to upgrade the computing power of the algorithm and land in the industry. However, we should also realize that China still has a long way to go in terms of data disclosure, international data exchange and building a unified data ecosystem.
"No chip is not AI", and the computing power based on AI chip is an important measure of the development level of artificial intelligence. We will analyze AI chips in detail, so as to grasp China's competitiveness in the basic layer of artificial intelligence more carefully and accurately.
According to the deployment location, AI chips can be divided into cloud chips (such as server terminals such as data centers) and terminal chips (application scenarios cover electronic terminal products such as mobile phones, automobiles and security cameras); According to the functions undertaken, AI chips can be divided into training chips and reasoning chips. The formation of training parameters involves massive data and large-scale calculation, which requires very high algorithm, accuracy and processing power, and is only suitable for deployment in the cloud. At present, GPU (universal), FPGA (semi-customized) and ASIC (fully customized) have become the mainstream technical routes in the AI chip industry. Different types of chips have their own advantages, showing the trend of parallel development of multi-technology paths in different fields. We will analyze the global competitiveness of China AI chips from three technical routes.
The design and production of GPU (graphics processing unit) have matured, occupying the main market share of AI chips. GPU is good at large-scale parallel operations and can process massive amounts of information in parallel, so it is still the first choice for AI chips. According to IDC's forecast, in 20 19 years, GPU will occupy 75% of the cloud training market. On a global scale, NVIDIA and AMD form a duopoly, especially NVIDIA occupies 70%-80% of the GPU market share. The GPU Tesla V 100 and Tesla T4 products promoted by NVIDIA in the cloud training and cloud reasoning market have extremely high performance and strong competitiveness, and their monopoly position has been continuously strengthened. At present, China has not entered the cloud training market. Because foreign GPU giants have rich experience in chip design and technology precipitation, and strong financial strength, China can't shake the market structure of GPU chips in a short time.
FPGA (Field Programmable Gate Array) chip has the advantages of hardware programming, high configuration flexibility and low power consumption. The technical barrier of FPGA is high, and the market is duopoly: Xilinx and Intel together occupy nearly 90% of the market share, of which Xilinx has a market share of over 50%, and has always maintained the dominant position of FPGA in the world. Domestic Baidu, Ali, Jingwei Qi Li are also laying out the field of FPGA, but they are still in the initial stage, and there is a big technical gap.
ASIC (Application Specific Integrated Circuit) is a customized chip designed for specific user needs, which can meet various terminal applications. Although ASIC needs a lot of physical design, time, money and verification, its performance, energy consumption, cost and reliability are better than GPU and FPGA after mass production. Different from GPU and FPGA, ASIC is just a technical route or scheme, focusing on solving outstanding problems and management requirements in various application fields. At present, the competition pattern of ASIC chip market is stable and scattered. The gap between China's ASIC technology and the world's leading level is small, and some fields are in the forefront of the world. Overseas, Google TPU is the dominant player; Domestic start-up chip companies (such as Cambrian, Bitland and Horizon) and Internet giants (such as Baidu, Huawei and Ali) have also made achievements in the segmentation field.
Generally speaking, Europe, America, Japan and South Korea basically monopolize high-end cloud chips. The domestic layout is mainly concentrated in terminal ASIC chips, and some areas are in the forefront of the world, but most of them are start-ups and have not yet formed an influential "chip". Platform? The ecology of "application" does not have the strength to compete with traditional chip giants (such as NVIDIA and Xilinx); In the field of GPU and FPGA, China is still catching up, and high-end chips rely on overseas imports.
Technical level: Pursuing victory, leading domestic head enterprises.
The technical layer is based on basic theory and data, and is oriented to the development of subdivided applications. Mid-stream technology enterprises have triple barriers of technology ecology, capital and talents, which are the core of artificial intelligence industry. Compared with most upstream and downstream enterprises, it is easier to focus on a certain segment and expand the technology layer to the upstream and downstream of the industrial chain. This level includes algorithm theory (machine learning), development platform (open source framework) and application technology (computer vision, intelligent voice, biometrics, natural language processing). Many international technology giants and unicorns have made extensive arrangements at this level. In recent years, China's technology layer focuses on vertical research and development, and its technology is mature in the fields of computer vision and speech recognition. Domestic head enterprises stand out and have obvious competitive advantages. However, the core technology of algorithm theory and development platform is still lacking.
Specifically, in the field of algorithm theory and development platform, China still lacks experience and develops slowly. Machine learning algorithm is the focus of artificial intelligence, and open source framework has become the focus of international technology giants and unicorns. Open source deep learning platform is the source code that allows the public to use, copy and modify, and it is the core driving force for the development of artificial intelligence application technology. At present, the widely used open source frameworks in the world include TensorFlow of Google, Torchnet of Facebook and DMTK of Microsoft. The United States is still the country with the highest level of development in this field. China's basic theoretical system is not yet mature, and the algorithm frameworks of domestic enterprises such as PaddlePaddle of Baidu and Angle of Tencent cannot compete with the international mainstream products.
In some fields of applied technology, the strength of China is comparable to that of Europe and America. Computer vision, intelligent speech and natural language processing are the three main technical directions and the three largest commercial technical fields in China market. Thanks to the development of the Internet industry, a large amount of user data has been accumulated, and domestic computer vision and speech recognition are in a leading position in the world. The current market competition of natural language processing has not yet formed, but there is a certain gap between domestic technology accumulation and foreign technology accumulation.
As one of the most mature technologies, computer vision has a wide range of applications. Computer vision is to simulate the recognition, tracking and measurement functions of human eyes with computers. Its application scenarios are wide, covering security (face recognition), medical treatment (image diagnosis), mobile Internet (video surveillance) and so on. Computer vision is the biggest part of the artificial intelligence market in China. According to the data of iResearch, in 20 17, the market size of computer vision industry was 8 billion yuan respectively, accounting for 37% of the domestic AI market. Due to government market intervention, algorithm model maturity, data availability and other factors, the landing situation of computer vision technology has been divided. China's computer vision technology is mainly exported in the fields of security, finance and mobile Internet. The downstream of computer vision in the United States is mainly concentrated in the fields of consumption, robotics and intelligent driving.
The competition pattern of computer vision technology is stable, and domestic head enterprises stand out. With the gradual saturation of industrial testing in the terminal market, new application scenarios are still being explored. At present, the global technology market has entered a period of steady growth, the market competition pattern has gradually stabilized, and the technology gap of head enterprises has gradually narrowed. China is rich in technology accumulation in this field, and the combination of technology application and products is in the forefront of the world. In 20 18, in the most authoritative face recognition algorithm test (FRVT) in the world, domestic enterprises and research institutes took the top five, and China was in the leading position in technology in the world. The domestic computer vision industry is highly concentrated, and the head enterprises stand out. According to IDC statistics, in 20 17, Shangtang Technology, Yi Tu Technology, Defiance Technology and Congyun Technology occupied 69.4% of the domestic market share, among which Shang Tang ranked first with 20.6% of the market share.
Application level: competing with each other, the pattern is uncertain.
The application scenario has a vast market space and the global market structure is uncertain. Benefiting from the global open source community, the entry threshold of the application layer is relatively low. At present, the application layer is the largest level in the artificial intelligence industry chain. According to the statistics of the Chinese Institute of Electronics, in 20 19, the global application layer industry scale will reach 36.05 billion yuan, which is about 1.67 times that of the technical layer and 2.53 times that of the basic layer. On a global scale, artificial intelligence is still in the exploration stage of industrialization and marketization, and the richness of landing scenarios, user needs and market penetration of solutions need to be improved. At present, no monopoly enterprise in the world has an absolute dominant position, and the market competition pattern of many sub-sectors has not yet been finalized.
China pays attention to the industrial layout of application layer, and its market development potential is huge. The artificial intelligence industry in developed countries and regions such as Europe and America landed earlier, and technology giants led by companies such as Google and Amazon focused on building a vertical ecology from chips and operating systems to applied technology research and development to segmentation scenarios. The overall development of the market is relatively mature; The application layer is the most active field in China's artificial intelligence market, and its market size and the number of enterprises also account for the largest proportion in domestic AI distribution. According to the statistics of iResearch, in 20 19, 77% of domestic artificial intelligence enterprises were distributed in the application layer. Thanks to the vast market space and large-scale user base, China has great market development potential, and some enterprises have been at the forefront of the world in industrial application. For example, China AI+ security technology, products and solutions lead the global industrial development, and Hikvision and Dahua Co., Ltd. occupy the first and fourth places in the global intelligent security enterprises respectively.
On the whole, the complete industrial chain of domestic artificial intelligence has taken shape, but there are still structural problems. From the perspective of industrial ecology, China focuses on technology layer and application layer, especially the terminal products are rich in applications, and the degree of technology commercialization is comparable to that of Europe and America. However, compared with developed countries such as the United States, China lacks breakthrough and symbolic research results at the basic level, and the underlying technology and basic theory are still weak. Early domestic policies focused on the Internet field, the industry pursued speed, and funds were invested in terminal applications that were easy to realize. The development of artificial intelligence industry is relatively impetuous, which leads to the basic innovation with long research and development cycle, large capital investment and slow effect being ignored by the market. The development trend of "top-heavy" leads to China's dependence on foreign development tools and basic devices, which is not conducive to the ecological layout of artificial intelligence in China and the long-term development of the industry. In the short term, the input and output in the field of application terminals are obvious, but it is difficult to become the core driving force to guide future economic changes. In the medium and long term, the development of artificial intelligence is rooted in the breakthrough of basic layers (algorithms, chips, etc.). ) research.
Analysis on the Development Potential of Artificial Intelligence
Based on the development status of artificial intelligence industry, we will evaluate the development potential of artificial intelligence in 28 countries in China, the United States and Europe from three dimensions: intelligent industry foundation, academic ecology and innovation environment, and use entropy method to determine the corresponding weights of each index, and then use TOPSIS method to construct a comprehensive index representing the overall development potential of artificial intelligence.
From the perspective of the foundation of intelligent industry
Degree of industrialization: the growth is strong, and the industrial scale is second only to that of the United States.
China's artificial intelligence is still in the early stage of industrialization, but the market development potential is great. The degree of industrialization is a comprehensive index to judge the vitality of artificial intelligence development. In terms of market scale, according to IDC data, in 20 19, the artificial intelligence markets in the United States, Western Europe and China were 2 13, 710.25 and 4.5 billion US dollars, accounting for 57%, 19% and 65,438 respectively. The market size of China and the United States is quite different, but the rapid development of domestic AI technology has driven the rapid growth of the market size in recent years, with an annual growth rate of 64% in 20 19, which is much higher than that of the United States (26%) and Western Europe (4 1%). In terms of the number of enterprises, according to the statistics of Tsinghua University Science and Technology Policy Research Center, as of June of 20101year, the number of artificial intelligence enterprises in China and the United States (in 2028) is far ahead in the world, and the third place in Britain (392) is less than 40% of the number of enterprises in China. According to the data of Tencent Research Institute, 46% and 22% of artificial intelligence enterprises in China are located in the fields of speech recognition and computer vision. Horizontally, the United States leads China in the number of enterprises at the basic and technical levels, especially in the fields of natural language processing, machine learning and technology platforms. At the application level (intelligent robots, intelligent drones), the gap between China and the United States is slightly smaller. Looking ahead, with the support of policies, the enthusiasm of capital and the innate advantages of data scale, China's artificial intelligence industry will maintain a strong growth trend with great development potential.
Technological innovation ability: many patents are not excellent, and the overseas layout is still lacking.
The number of patent applications is the core factor to measure the innovation ability and development potential of artificial intelligence technology. On a global scale, patent applications for artificial intelligence mainly come from China, the United States and Japan. From 2000 to 20 18, the number of AI patent applications from China, the United States and Japan accounted for 73.95% of the total global applications. Although China started late in the field of AI, since 20 10, the patent output has surpassed that of the United States for the first time, and it ranks first in the number of applications for a long time.
From the field of patent application, deep learning, speech recognition, face recognition, robotics and other hot areas have become the key areas of layout in various countries. Among them, the United States leads in almost all fields, while China has obvious advantages in speech recognition (the correct rate of Chinese speech recognition is the highest in the world), text mining and cloud computing. Specifically, most domestic patents were applied after the upsurge of AI technology, and focused on the application side (such as intelligent search and intelligent recommendation), while key areas and frontier areas such as AI chips and basic algorithms are still mainly controlled by the United States. This reflects that China's AI development foundation is weak and there is a structural imbalance of superficial prosperity.
The quality of domestic AI patents is uneven, and the overseas market layout is still lacking. Although the number of patent applications in China far exceeds that in the United States, the problem of "many but not strong, specialized but not excellent" needs to be adjusted urgently. First, China's AI patents are mainly domestic, and the number of high-quality PCTs is small. PCT (Patent Cooperation Treaty) is a treaty managed by WIPO to protect patent inventors on a global scale. PCT is generally considered to have high technical value. According to patent protection association of china's statistics, the number of PCT applications in the United States accounts for 465,438+0% of the world, which is widely used in the world. The number of PCTs in China (2,568 pieces) is relatively small, only 1/4 in the United States. At present, China's AI technology has not yet formed a large-scale technical output, and the international market is lacking; Secondly, China has a high proportion of utility model patents and a large proportion of patent abandonment. China's patent categories include invention, utility model patent and design, and the technical difficulty decreases in turn. In China, most AI patents are practical new patents, and the threshold is very low. For example, in 20 17 years, invention patents only accounted for 23% of the total number of applications. In addition, according to the report of Cambridge University, due to the high patent maintenance cost, 665,438+0% of AI utility models and 95% of designs in China will be invalid after five years, while 85.6% of patents in the United States can still be effectively retained.
Talent reserve: supply and demand are unbalanced, and there is a big gap in top talents.
The quantity and quality of talents directly determine the development level and potential of artificial intelligence. At present, the global distribution of artificial intelligence talents is uneven, and the supply is in short supply. According to Tsinghua University's statistics, as of 20 17, the top countries in the 10 talent pool accounted for 6 1.8% of the global total. There are 43,064 artificial intelligence talents in 28 countries in Europe, ranking first in the world, accounting for 2 1. 1% of the global total. The United States and China ranked second and third with 28,536, 65,438 and 08,232 respectively. Among them, China's basic talent reserve is particularly weak. According to the data of Tencent Research Institute, the number of AI technical talents in the United States is 2.26 times that of China, and the number of basic talents is 13.8 times that of China.
The supply and demand of artificial intelligence talents in China is seriously unbalanced, and the gap of outstanding talents is large. According to BOSS direct employment, domestic artificial intelligence talents can only meet 60% of the demand of enterprises in 20 17 years, and it is conservatively estimated that the talent gap has exceeded 1 10,000. In some core areas (speech recognition, image recognition, etc. ), the supply of AI talents is even less than 40% of the market demand, and this trend is becoming more and more serious with the increase of AI enterprises. In the exploration stage of artificial intelligence technology and application, excellent talents play a vital role in industrial development and even affect the development of technical routes. The United States (5 158) and the European union (5,787) have gathered a large number of elites based on their scientific research and innovation capabilities and development opportunities, and the number of outstanding talents is far ahead in the world, while the proportion of outstanding talents in China (977) is still obviously low, less than that in Europe and America15.
The inflow rate and outflow rate of talents can measure the ability of a country's ecosystem to attract and retain foreign talents. According to the classification standard of elemental AI enterprises, China, the United States and other countries are anchor countries with low inflow and outflow rate of AI talents, especially the total number of artificial intelligence talents in the United States remains relatively stable. Specifically, the cultivation of artificial intelligence in China is still mainly local, and the number of overseas talents returning to China only accounts for 9% of the total domestic talents. Among them, the United States is the largest source of domestic AI talents, accounting for 43.9% of all returned talents. It can be seen that the attractiveness of domestic policies, technologies and environment to overseas talents still needs to be strengthened.
From the perspective of academic ecology
Scientific and technological innovation ability: the output of scientific research is strong, and the integration of industry and learning needs to be strengthened.
Scientific research ability is the driving force for the development of artificial intelligence industry. In terms of the number of published papers, from 1998-20 18, the European Union, China and the United States rank among the top three, accounting for 69.64% of the total published papers in the world. In recent years, China has actively deployed forward-looking technology, and AI has developed strongly, from 1998 accounting for 8.9% of global artificial intelligence papers to 28.2% of 20 18, and CAGR 17.94%. In 20 18, China ranked first in the world with 24,929 AI papers. China's active research activities are reflected in the huge development potential of artificial intelligence.
The influence of China's papers still needs to be improved, but the gap with Europe and America is narrowing year by year. FWCI index is the best method to quantitatively evaluate the quality of scientific research papers at present. We use FWCI to express the influence of 1 standardized papers. When FWCI≥ 1, it shows that the quality of the tested papers has reached or exceeded the world average. In the past 20 years, the weighted citation influence of American AI papers has been "dominant". In 20 18 years, FWCI was 36.78% higher than the global average. Europe remains relatively stable, comparable to the global average; The influence of AI papers in China has increased significantly. In 20 18, the FWCI of China was 0.80, which was 44.23% higher than that in 20 10, but the influence of AI papers was still 20% lower than the world average. Judging from the number of 1% papers before high citation, the output of high-quality papers in the United States and China ranks first and second in the world, nearly four times higher than that in the third place, Britain. On the whole, the output of domestic top-quality papers is equivalent to that of the United States, but on the whole, the influence of AI papers still lags behind that of the United States, Europe and America.
From the author's point of view, scientific research institutions and universities are the absolute strength of artificial intelligence knowledge production in China at present, reflecting the shortcomings in the transformation of scientific research results. However, the United States, the European Union and Japan show a trend of joint participation of enterprises, government agencies and universities. Scopus data shows that in 20 18 years, the proportion of AI papers signed by American enterprises was 7.36 times that of China and 1.92 times that of EU. From 20 12 to 20 18, the proportion of AI papers signed by American enterprises increased by 43pct, while the proportion of AI papers signed by China enterprises only increased by 18pct. In addition, artificial intelligence is closely related to market applications, and school-enterprise cooperation papers widely exist. However, the proportion of school-enterprise cooperation papers in China is only 2.45%, which is quite different from that in Israel (10.06%), the United States (9.53%) and Japan (6.47%). From the perspective of the combination of production and learning, the research of artificial intelligence in China is promoted by academic circles, and the participation of enterprises in scientific research is low, or it is difficult to achieve market orientation.
The number of artificial intelligence universities in China is actually in the second echelon, and its strength is equivalent to that of the United States. Colleges and universities are the core carriers for the supply of artificial intelligence talents and the output of papers. According to the statistics of Tencent Research Institute, there are 367 universities in the world offering artificial intelligence-related disciplines, among which the United States (168) is the first, accounting for 45.7% of the world. There are 20 colleges and universities in China tied for the third place with a slightly smaller number in Britain. In addition, the strength of colleges and universities in China has generally increased and their performance has been strong. According to the Top20 list of AI universities published by MIT in 20 19, Tsinghua University and Peking University in China took the top two places, rising 1 and 3 places respectively compared with 20 18.
From the perspective of innovation environment
R&D investment: The gap between China and the United States has narrowed.
China has high R&D investment and intensity, and occupies an important position in global R&D performance. From the perspective of R&D investment, the United States, China, Japan and Germany have always been the main forces of global R&D investment. According to IDC statistics, in 20 18, the total R&D investment of the four countries accounted for 60.77% of the global total. Among them, with its strong R&D strength, the United States has ranked first in global R&D investment for many years. In recent years, China's investment in R&D has shown a strong momentum. According to Statista, the domestic R&D investment in 20 19 years was $51920 million, second only to the United States. The gap with the United States is narrowing. From 2000 to 20 19, the CAGR was as high as 14.43%, while the CAGR in the United States was only 2.99%. Due to many reasons, such as economic weakness, the EU and Japan show a relatively slow upward trend. According to the increasing trend of R&D investment and intensity, China may replace the United States as the global R&D leader within 1-2 years. From the perspective of R&D intensity, China's R&D intensity is gradually increasing, and the increase range is large. However, compared with the United States and Japan, there is still a gap in the emphasis on the intensity of investment in innovation activities. In 20 18, the R&D intensity of China was 1.97%, which was lower than that of Japan and the United States 1.53 and 0.87 percentage points.
Capital investment: there are more funds and fewer projects, and the capital investment focuses on the terminal market.
China and the United States are global artificial intelligence "financing highlands". The development cost of artificial intelligence is high, and capital investment has become the main force to promote the development of technology. On a global scale, the United States is the leader in investment and financing of new artificial intelligence enterprises. According to CAPIQ data, from 20 10 to 20 19, the accumulated financing of American AI enterprises was 77.3 billion US dollars, which was 32 billion US dollars ahead of China, accounting for 50.7% of the total global financing. Especially since the Trump administration, the investment in artificial intelligence has gradually increased. As the second largest financing country in the world, China's total financing accounts for 35.5% of the world. Considering the existing pattern and recent changes, it is difficult for other countries and regions to shake China and the United States in scale. Judging from the number of new enterprises in artificial intelligence, the United States is still in a leading position in the world. From 20 10 to 20 18, there were 7022 new enterprises in the United States, about 8 times that of China (870). The number of newly-added artificial intelligence enterprises in China decreased gradually after reaching a high of 65,438+079 in 2065,438+06. In the last two years, there were179 (2065,438+07) and1565,438+0 (2065,438+08) respectively, indicating that China's capital market has invested in AI. On the whole, the growth of new artificial intelligence enterprises in China is slow, but the total amount of financing has increased rapidly. This situation of "more funds and fewer projects" is an early warning of the impending bubble in the industry.
Compared with the United States, China's capital investment focuses on easy-to-land terminal markets. From the financing point of view, the development of various fields in China is relatively balanced, and the application layer is a prominent field, such as autonomous driving, computer learning and imaging, voice recognition, UAV technology, etc., which all surpass the United States. The American market pays attention to the development of the underlying technology. According to data from Tencent Research Institute, chips and processors are the areas with the most financing in the United States, accounting for 3 1% of the total financing. At present, China attaches great importance to the artificial intelligence chip market, but due to technical barriers and high investment threshold, domestic chip financing is in a weak position.
TOPSIS method based on information entropy: comprehensive index evaluation
The data shows that the United States is absolutely ahead in comprehensive indicators and three major project indicators, while China is second, and 28 European countries are temporarily behind. Specifically, the United States has obvious advantages in artificial intelligence talent reserve, innovation output and financing scale. As a rising star, although China has caught up with the United States, its overall level is still backward, especially in terms of excellent human resources and high-quality patent applications. However, in terms of the number and influence of papers, R&D investment and other indicators, China has developed rapidly, and the gap with the United States is narrowing. From the specific analysis of various indicators, the research of artificial intelligence in China is mainly distributed in universities and scientific research institutions, with low enterprise participation, fragmented output and lack of systematic integration with the market, which will not be conducive to the development of artificial intelligence technology and the play of industrial advantages in China. In addition, China's scientific research output, the number of enterprises and financing fields are concentrated in the middle and lower reaches of the industrial chain, and the upstream core technologies are still subject to foreign enterprises. In the future, if the domestic underlying technology field still fails to achieve a breakthrough, it will inevitably lead to the bottleneck of the development of artificial intelligence industry.
prospect
Transferred from the Special Committee on Collaborative Innovation of Informatization.