（1） The overall development of the artificial intelligence market is improving

The scale of the artificial intelligence market continues to increase. The global scale of the artificial intelligence industry was 361.9 billion US dollars in 2021, and it is estimated to reach 423.8 billion US dollars in 2022. Compared to the relatively sluggish global economy, the artificial intelligence market continues to thrive. IDC predicts that the global artificial intelligence market will maintain a high growth trend, achieving a compound annual growth rate of 20% from 2022 to 2026, and the market size is expected to reach $1861 billion by 2030 (see Figure 1). According to CB Insights data, there are over 11000 existing artificial intelligence companies worldwide, half of which are startups established in 2017 and later. With the gradual maturity of the industrial system, technologies that are decoupled from commercial value will be eliminated from the market quickly, and the decline of the foam will also bring about the exit of a number of AI enterprises. According to statistics, the success rate of entrepreneurship in the field of artificial intelligence is less than 10%. Among them, the combination of high R&D investment, high talent demand, and slow industrialization speed causing cash flow rupture is the most important reason for the "death" of artificial intelligence enterprises. The high birth and death rates of the entrepreneurial ecosystem drive rapid industry iteration, keeping the artificial intelligence entrepreneurial ecosystem active at all times.

Figure 1 Forecast of Global Artificial Intelligence Market Size from 2022 to 2030

The capital market has reignited its enthusiasm for the field of artificial intelligence. Around 2018, investment and financing in the field of artificial intelligence briefly cooled down, but in the following three years, the capital market regained enthusiasm for the field. In 2021, the total investment in the field of artificial intelligence, including private investment, public offerings, and corporate mergers and acquisitions, was approximately $176.5 billion (see Figure 2). Among them, private investment in the field of artificial intelligence has the largest amount (about 93.5 billion US dollars), followed by corporate mergers and acquisitions (about 72 billion US dollars), public offerings (about 9.5 billion US dollars), and minority equity (about 1.3 billion US dollars).

Structurally speaking, the financing stage of artificial intelligence has significantly shifted backwards, with a noticeable increase in financing after Series B, with growth projects such as intelligent healthcare and intelligent driving receiving the most attention. Meanwhile, investment concentration has increased - the average private investment transaction size in 2021 was 81.1% higher than in 2020, but the number of newly financed AI companies continues to decline, indicating increased investment certainty. From the listing situation, more than 40 companies with artificial intelligence as their core business went public through IPOs in 2021, and the industry entered the capital redemption period.

In terms of technological achievements, the global publication and patent application of artificial intelligence papers are showing a rapid growth trend. According to the 2022 Artificial Intelligence Index Report, the number of global artificial intelligence papers continued to grow from 2010 to 2021, reaching approximately 330000 papers in 2021 (see Figure 3). During the same period, the number of global artificial intelligence patents also grew rapidly, with approximately 140000 global applications in 2021 (see Figure 4). In terms of content, medical image segmentation, occluded face recognition, intelligent dialogue and other fields related to the COVID-19 epidemic are hot spots of technological innovation, and research topics such as in-depth fraud recognition and credible AI have also attracted extensive research attention. From the perspective of research subjects, companies in Europe, America and other regions are the main contributors to artificial intelligence papers and patents, while in China, patents and papers are more from universities and research institutions.

Figure 4 Global Artificial Intelligence Patent Application Volume from 2010 to 2021

Artificial intelligence algorithm models are becoming increasingly complex, and large-scale models have become a development trend. From 2011 to 2021, the parameters of artificial intelligence models increased from tens of millions to billions. In 2020, OpenAI released the new generation text generation neural network model GPT-3, which can complete language generation and interaction tasks comparable to human intelligence with 175 billion parameters trained on 570GB of text. In 2021, the single model parameter values of "Source 1.0" in China reached 245.7 billion. Based on 5TB of Chinese corpus data for training, it can effectively complete tasks such as reading comprehension and logical judgment for Chinese. Based on the large model and combined with domain knowledge for optimization, rapid iteration and application in practical scenarios can be achieved. Currently, this seems to be the key path towards general artificial intelligence.

AIGC has become a hot topic in the industry, driving the rise of multimodal interaction technology. One sentence image generation "is undoubtedly the most" breakthrough "application of AIGC (Artificial Intelligence Content Generation) in 2022. The key technology behind AIGC, generative AI, has a long history, but in the past, it mainly focused on the generation and transfer of content in a single modality, such as text content generation for "text to text", image completion for "image to image", and image style transfer. In 2022, OpenAI released an upgraded version of DALL-E-2, followed by Stability quickly launching a fully open-source Stable Diffusion, Midjourney launching an AI drawing tool, marking a breakthrough in multimodal interactive generation technology - users only need to input short text (text) to generate professional level paintings (images or videos) in different styles. The winner of the Colorado State Fair Art Competition, the Space Opera House, was created by the creator using Midjourney. AIGC based on multimodal interaction technology not only improves creative efficiency, but also enables the creation of things that do not exist in reality, expanding the boundaries of human imagination in creation. In the future, multimodal interaction technology will generate richer and more diverse digital content and more realistic and natural interaction methods, and is expected to become the main tool for content generation in the metaverse.

（3） The steady advancement of multi domain applications of artificial intelligence

The global adoption rate of artificial intelligence is steadily increasing. IBM conducted a survey on the use of artificial intelligence in 7502 companies worldwide and found that 35% of the surveyed companies have already used AI in their businesses, while another 42% are exploring AI. The adoption of artificial intelligence is not only an important business trend, but also brings substantial benefits to enterprises. According to the "Future Intelligence" report released by IDC in November 2021, among the companies with the highest scores on the Intelligence Index scale, 60% achieved significant improvements in decision-making efficiency, and 47% of enterprise customers increased by more than 10%. Although there is still a gap in the application of artificial intelligence compared to expectations, the large-scale use of artificial intelligence has become a global consensus. According to IDC's forecast, by 2024, artificial intelligence will become a core component of enterprise workloads, with 75% of enterprises and 20% of their workloads adopting AI. According to statistics from HAI at Stanford University, since 2018, the cost of training an image classification system has decreased by 63.6%, while training efficiency has increased by 94.4%. Training a modern image recognition system, which cost $1100 in 2017, now only costs $7.43, which is only about 1/150 of the original cost. Obviously, the improvement of artificial intelligence performance and cost reduction are important reasons for its wider application.

Artificial intelligence applications are moving from single point applications to global intelligence. In 2021, the adoption of artificial intelligence was divided by industry and function, with high-tech/telecommunications product and/or service development being the most adopted (45%), followed by service operations in financial services (40%), high-tech/telecommunications service operations (34%), and risk functions in financial services (32%). Due to factors such as high implementation costs and complexity, difficulty in connecting supply side data, and incomplete overall ecology, current industrial intelligence still focuses on solving fragmented needs, mainly concentrated in intelligent sorting, equipment health management, defect inspection, supply chain optimization, and other links. The resilience demonstrated by artificial intelligence during the epidemic has made enterprises pay more attention to the value of industrial intelligence. Coupled with the advancement and popularization of digital technology and the investment in new infrastructure, these factors will jointly promote the rapid transition of industrial intelligence from single point intelligence to global intelligence. Especially in manufacturing industries such as automobiles, consumer electronics, branded clothing, steel, cement, and chemical industries that have a good foundation in information technology, the global intelligent application of enterprise production decision-making loops that run through various links such as supply chain, production, assets, logistics, and sales will emerge on a large scale.

The deep integration of artificial intelligence and scientific research is disrupting traditional research paradigms. Different from traditional theoretical research paradigms, artificial intelligence is accelerating the speed of scientific research based on deep mining of massive data. In November 2020, DeepMind's AlphaFold 2 deep learning technology provided a solution to the protein structure prediction problem that had plagued the biological community for 50 years. Artificial intelligence is also revolutionizing protein design. In July 2022, scientists from the University of Washington and other institutions published a new artificial intelligence software in the journal Science that can draw structures for proteins that do not yet exist in nature. In the long run, with the iteration of new AI algorithms and breakthroughs in computing power, AI will effectively solve the problems of long vaccine/drug development cycles and high costs, such as improving research efficiency in compound screening, establishing disease models, discovering new targets, discovering lead compounds, and optimizing lead drugs.

（4） Innovation factors determine the upper limit of artificial intelligence development

The scarcity of innovative elements in terms of talent and computing power, as well as the uneven regional distribution, will persist for a long time and become important factors determining the upper limit of artificial intelligence development.

Artificial intelligence talent has become the primary resource that countries compete for, and the density and height of talent determine the ceiling of artificial intelligence innovation level. On the one hand, the global demand for artificial intelligence talent continues to rise. According to the AI Jobs report released by UIPath, the annual growth rate of global demand for artificial intelligence talent is as high as 78%. On the other hand, artificial intelligence talent is scarce and unevenly distributed. The total number of AI talents in the United States is 1.6 times that in China, of which technology development talents are about 2.5 times that in China, and basic R&D talents are about 14 times that in China. According to the AI 2000 ranking released by Aminer, the United States has the highest number of selected talents, with an average of over 1000 people per year, and holds a global leading position in the field of top academic talents in artificial intelligence. China ranks second in the number of talents selected for the list, but there is a significant gap compared to the United States.

The explosion of computing power has created a huge market demand for high-performance chips, and the shortage of computing resources and regional imbalances have triggered a barrel effect in the development of artificial intelligence. Computing power is the core productivity of the development of artificial intelligence. According to the analysis results of the 2020 Global Computing Power Index Evaluation Report, from 2015 to 2019, for every 1 point increase in the computing power index, the country's digital economy and GDP will grow by 3.3 ‰ and 1.8 ‰ respectively. The explosion of artificial intelligence data and the increasing complexity of algorithm models have put forward higher requirements for computing power. In terms of data, according to IDC's calculations, the global data scale will reach 163ZB by 2025, of which 80% to 90% will be unstructured data. In terms of models, according to OpenAI data, the growth rate of model computation far exceeds the growth rate of artificial intelligence hardware computing power, with a gap of tens of thousands of times. The trend of large models is driving the continuous growth of computing power demand.

Currently, the AI chip industry is mainly classified into three types based on technology architecture: Graphics Processing Units (GPUs), Field Programmable Gate Arrays (FPGAs), and Application Specific Integrated Circuits (ASICs). In addition, there are new computing architectures such as neuromorphic chips and quantum computing, but overall they are in the exploratory stage. From a regional perspective, the global computing power market exhibits a high degree of imbalance: the GPU market is dominated by NVIDIA and AMD, while the FPGA market is dominated by Xilinx and Intel. From a technical perspective, the GPU/FPGA technology threshold is relatively high, and China has a 2-3 generation technology gap compared to the United States. The ASIC market is relatively dispersed, with a small gap between China and the United States. The "2021-2022 Global Computing Power Index Assessment Report" points out that there are currently approximately 600 hyperscale data centers in the world, of which about 39% are in the United States, four times that of China. In 2022, the Center for Strategic and International Studies released the report Cutting China's Path to the Future of Artificial Intelligence, which clearly pointed out that the United States is using its monopoly advantage in computing market to strangle China's artificial intelligence industry by regulating the use of high-end AI chips and chip design software made in the United States.

The tense geopolitical situation has given rise to different paths for the development of artificial intelligence in countries and regions, but the cybersecurity, social bias, and trust issues brought about by artificial intelligence have raised global concerns, and the need for governance of artificial intelligence has become a universal consensus. The complexity, autonomy, and inherent technological flaws of artificial intelligence bring many risks and problems to society. The 2022 Artificial Intelligence Index Report released by HAI at Stanford University shows that the biggest risk associated with adopting artificial intelligence in 2021 is cybersecurity (55% of respondents), followed by compliance (48%), interpretability (41%), and privacy risk (41%) (see Figure 5). With the increasing complexity of AI training model frameworks, the performance of AI systems continues to improve, but at the same time, the inherent biases in data and systems are further amplified. Compared to the 117 million parameter model developed in 2018, the toxicity induced by the 280 billion parameter model developed in 2021 increased by 29%, and the proportion of images of black people incorrectly classified as non-human is more than twice that of other races.

From a policy perspective, artificial intelligence governance is moving from soft constraints such as ethical principles to a comprehensive and actionable "hard law" stage. Stanford University HAI's analysis of legislative records related to artificial intelligence in 25 countries shows that the number of bills containing "artificial intelligence" passed into law has increased from only 1 in 2016 to 18 in 2021. On April 21, 2021, the European Union released the world's first draft of the Artificial Intelligence Act, which categorizes artificial intelligence into different risk levels and manages them accordingly. In 2022, the United States introduced the 2022 Algorithm Accountability Act. Local AI legislation in China has been implemented in succession. The Regulations of Shanghai Municipality on Promoting the Development of AI Industry and the Regulations of Shenzhen Special Economic Zone on Promoting the Development of AI Industry all make statements on AI ethics and governance.

From a technical perspective, network security, trustworthy artificial intelligence related technologies and applications have become the focus of research. IDC released data stating that the total global investment in network security IT in 2021 was $168.77 billion, and is expected to increase to $287.57 billion by 2026, with an average annual growth rate of 11.3%. In the field of trusted artificial intelligence, privacy computing, interpretability, and model fairness have become the focus. The market size of trusted artificial intelligence reached 4.4 billion US dollars in 2021 and is expected to reach 21 billion US dollars by 2030, with an average annual growth rate of 19% (see Figure 6). American tech giants such as Microsoft and Google have established AI ethics committees, while representative companies such as SenseTime and Tencent regularly release AI ethics investigation reports. At the industry level, artificial intelligence governance is taking enterprises as the main body, forming a diverse ecosystem covering technology, hardware, software, systems, and processes.

Figure 6 Forecast of Trusted Artificial Intelligence Market Size from 2021 to 2030

Artificial intelligence has the characteristic of "creative destruction" proposed by American economist Schumpeter, which means that technological innovation in artificial intelligence can continuously reform the economic structure from within, thereby promoting the country's position in the global value chain. Therefore, artificial intelligence has received high attention from countries around the world. According to data from the Organization for Economic Cooperation and Development (OECD), more than 60 countries and regions around the world have successively introduced artificial intelligence policies and priority development issues. Major countries and regions have formulated or revised national AI strategies and actively explored AI development paths that meet their own needs and advantages in accordance with the changes in the international situation in recent two years and the technical requirements of the COVID-19 epidemic for AI (see Table 1).

表1 主要国家和地区人工智能战略举措
国家和地区	主要动向
德国	2020年底更新《德国人工智能发展战略》，从人才培养、基础研究、技术转移和应用、监管框架和社会认同五大重点领域确定了未来的一揽子计划，至2025年，德国联邦政府对人工智能领域的资助将从30亿欧元增加到50亿欧元； 2021年出台《联邦—州联合促进高等教育领域人工智能发展的指导意见》； 2022年7月1日起，德国将以每年5000万欧元的额度持续资助多个人工智能研究中心，联邦和所在州各资助50%
日本	2021年6月发布“AI战略2021”，致力于推动人工智能领域的创新创造计划，全面建设数字化政府；人工智能产业化路线图：2020～2030年目标，铁路等交通工具的无人化操作和货物运输配送的完全无人化，利用人工智能控制家庭设备等。2030年之后，希望通过人工智能分析潜在意识和丰富可视化体验，使看护机器人成为家庭重要成员
韩国	计划到2030年在人工智能领域创造455万亿韩元（约合2.7万亿人民币）的经济效益
英国	2021年9月，发布国家级人工智能新十年战略； 2022年7月，英国数字、文化、媒体和体育部发布新的人工智能规则，以便人工智能在英国被迅速安全采用，促进生产力增长
美国	2021年1月，美联邦政府成立了专门的国家人工智能倡议办公室，作为未来美国整个创新生态系统的国家人工智能研究和政策中心； 2021年6月，拜登政府白宫科技政策办公室（OSTP）和国家科学基金会（NSF）宣布成立“国家人工智能研究资源工作组”，研究建立国家人工智能研究资源的可行性
中国	《中共中央关于制定国民经济和社会发展第十四个五年规划和二○三五年远景目标纲要的建议》指出，要瞄准人工智能等前沿领域，实施一批具有前瞻性、战略性的重大科技项目，推动人工智能发展。上海、深圳出台人工智能地方立法条例
欧盟	2021年出台全球首部《人工智能法案》（草案）
法国	2021年底出台并推进“人工智能国家战略”新计划，未来5年内将投入22亿欧元用于加快人工智能发展

Table 1: Strategic Measures for Artificial Intelligence in Major Countries and Regions

From a global perspective, the development of artificial intelligence has significant regional imbalances, with China and the United States having a clear leading international strength in the field of AI, and their development far exceeding that of other regions. According to CB Insights, there are currently 11000 artificial intelligence companies worldwide with a cumulative financing amount exceeding 250 billion US dollars. Among them, the United States has 4171 artificial intelligence companies with a cumulative financing amount of $1601.9 billion, ranking first in the world in terms of the number of companies and financing scale. China has 1275 artificial intelligence companies with a total financing amount of 47.07 billion US dollars, ranking second in the world. There are also hundreds of artificial intelligence companies in countries such as the UK, India, and Canada (see Figure 7). The global bipolar and multi strong artificial intelligence landscape is highlighted.

Objectively speaking, compared to the United States, China still has a significant gap in absolute strength in the field of artificial intelligence. However, the speed of China's technological and industrial development in recent years has aroused the vigilance of the US government. Since the Trump administration defined China as a "long-term strategic competitor" in 2017, the United States has frequently interpreted China's digital technology development from the perspectives of geopolitics and security threats, gradually giving rise to technological nationalism and a tendency towards technological decoupling in the field of artificial intelligence. After the Biden government came to power in 2021, the crackdown on China's AI field will be more systematic, including export control, supply chain control, software control, investment review, and pressure on allies to exclude Chinese technology. Nowadays, the decoupling of geopolitics and technology has become a new context for the development of artificial intelligence, and digesting and responding to external risks has become a new proposition that countries, governments, and enterprises must face. As a microcosm of the changes in the relationship between the two countries, the game between China and the United States in the field of artificial intelligence and the overlapping and interconnected differences in values, geopolitics, and other aspects have further spread to the economic, political, security, and other fields, significantly affecting the direction of global technological competition and even international relations. As the world's largest single digital market, the EU's choice of path and strategic design in the development of artificial intelligence has become the biggest factor affecting the "battle situation" of artificial intelligence. Therefore, this report first focuses on the specific analysis of the development of artificial intelligence in the United States, China, and the European Union, and on this basis, understands the artificial intelligence measures of the United States, China, and the European Union from the perspective of international competition.

（1） Strategic measures for the development of artificial intelligence in the United States

The United States needs to be prepared to take comprehensive and nationwide action to compete in the era of artificial intelligence.

In the global game of artificial intelligence, the United States holds the first mover advantage in the development of artificial intelligence. In order to occupy an absolute hegemonic position in the field of artificial intelligence, the United States has expanded the strategic layout of artificial intelligence through policies and laws, and constantly adjusted policy tools with the progress of artificial intelligence development in other countries. In January 2021, the United States enacted the National Artificial Intelligence Initiative Act, which aims to maintain its leading position in the field of artificial intelligence by increasing research investment, acquiring computing and data resources, setting technical standards, establishing labor systems, and collaborating with allies. In June 2021, the US Senate passed the US Innovation and Competition Act (USICA) with a total investment of $250 billion, listing artificial intelligence as a key area of national innovation and security. Compared with the previous emphasis on promoting technological innovation and industry development policies, the bill clearly states the need to compete with strategically competitive countries in the field of artificial intelligence, greatly enhancing the competitiveness of policy tools.

The US government's policy of supporting artificial intelligence innovation and competitiveness has gradually formed a strategic framework with strong integrity.

From the perspective of institutional development, in January 2021, the United States established a dedicated National Artificial Intelligence Initiative Office, which serves as the National Artificial Intelligence Research and Policy Center for the entire innovation ecosystem of the United States in the future. It is responsible for supervising and implementing the national artificial intelligence strategy, and coordinating artificial intelligence research and policy-making among the government, industry, and academia. In June 2021, the White House Office of Science and Technology Policy (OSTP) and the National Science Foundation (NSF) of the Biden administration announced the establishment of the National Artificial Intelligence Research Resources Working Group to explore the establishment of shared national AI research resources and infrastructure to support AI research and development.

In terms of the action plan, the National Security Council on Artificial Intelligence (NSCAI) of the United States submitted the 700 page Final Report: Artificial Intelligence to the President and Congress in March 2021. From two aspects of national security and technological competition, it introduced in detail the strategy of the United States to win the era of artificial intelligence, and described the more detailed action blueprint that the United States government should take to implement the recommendations.

From the perspective of strategic objectives, it can be seen that the policy measures of the US government to support artificial intelligence innovation and competitiveness can be mainly divided into three types: innovation promotion, industrial governance, and competition suppression.

The United States adopts various policies to directly stimulate innovation and competitiveness in artificial intelligence, including increasing investment in the field of artificial intelligence, supporting research in artificial intelligence, and strengthening talent cultivation in artificial intelligence.

Increase investment in the field of artificial intelligence: The United States aims to maintain its leading position as a strategic goal and continues to increase investment in the field of artificial intelligence. In 2015, the United States had only about $1.1 billion in unclassified research and development investment in the field of artificial intelligence. With a deeper understanding of the strategic significance of artificial intelligence, the amount of investment in artificial intelligence in the United States has significantly increased. In 2021, the US AI non defense budget will increase by about 30%, reaching a total of US $1.5 billion. The final report of the National Security Council on Artificial Intelligence in 2021 pointed out that "the government should spend at least 1% of GDP on R&D every year", emphasizing that the government should give priority to investing in artificial intelligence R&D in important areas that "support future national security and economic stability". The US Innovation and Competition Act prioritizes artificial intelligence in the US research and development budget for fiscal year 2022, with a total investment of $100 billion in research and development work in multiple fields including artificial intelligence in the future.

Supporting artificial intelligence research: The United States is also increasing its investment in scientific research. In 2019, the National Science Foundation of the United States collaborated with agencies such as the Department of Agriculture, the Department of Homeland Security, and the Department of Transportation to promote the National Artificial Intelligence Research Institute project. The project invested $140 million in the first round of funding for 7 artificial intelligence research institutes in 2020, and an additional $220 million in funding for 11 newly established national artificial intelligence research institutes in 2021. The 11 newly established research institutes will each receive approximately $20 million in funding over the next 5 years to support their research in areas such as human-computer interaction and collaboration, advances in artificial intelligence optimization, artificial intelligence and advanced network infrastructure, artificial intelligence in computer and network systems, artificial intelligence in dynamic systems, artificial intelligence enhanced learning, and artificial intelligence innovation in agriculture and food systems. The United States also plans to increase the National Science Foundation budget to $18.3 billion by 2026 for the construction of 10-15 "global key technology research, development, and manufacturing centers" covering artificial intelligence.

Strengthen artificial intelligence talent training: In order to ensure that the development of artificial intelligence in the United States has sufficient talent reserves, the United States has launched a number of measures to strengthen the cultivation of artificial intelligence talents in an all-round way. ① Strengthen the cultivation of STEM talents. The American Innovation and Competition Act proposes to increase investment in STEM science and engineering education through legislation, and promote the cultivation and development of interdisciplinary and multi skilled professional R&D talents through scholarships and research grants Enhance the artificial intelligence literacy of the entire population. The American Innovation and Competition Act clearly stipulates the introduction of computational science into primary and secondary education. The American Association of Artificial Intelligence (AAAI) and the Association of Computer Science Teachers (CSTA) jointly launched the American AI4K12 program to provide resources to help teachers teach students AI knowledge. The non-profit project AI-4-Al in the United States has developed free online courses to help people understand the working principles of artificial intelligence and create more opportunities for vulnerable groups in the field of artificial intelligence.

The United States emphasizes prudent regulation to promote innovative development, and has intensively introduced policies and regulations to shape the domestic artificial intelligence rule system and regulatory environment (see Figure 8).

2019 Policy Action: In May 2019, the United States released the Artificial Intelligence Initiative Act, which included the development of AI governance standards as a key focus of AI development, explored the establishment of standards for testing AI algorithms and their effectiveness, and planned an annual budget of $40 million for this work. Subsequently, in July, the National Institute of Standards and Technology (NIST) of the United States released "Leadership in Artificial Intelligence: Federal Government Participation in the Development of Technical Standards and Related Tools", aiming to promote the development of trustworthy artificial intelligence through standard building. In October 2019, the "2019 Algorithm Accountability Act (Draft)" entered the legislative process of both houses of Congress. The 2019 Algorithm Accountability Law (Draft) requires an impact assessment of "high-risk" automated decision-making systems to prevent discrimination against consumers/users caused by automated decision-making.

2020 Policy Action: On June 29, 2020, the Office of the Inspector General of the US Department of Defense released the "Audit of Governance and Protection of AI Data and Technology in the Department of Defense" report, aimed at auditing the Department of Defense's AI governance framework, standards, and cybersecurity measures. In August 2020, the US Senate released the National Biometric Information Privacy Act, putting an independent personal biometric information protection law on the legislative agenda.

2021 Policy Action: In May 2021, the United States released the "Algorithmic Justice and Online Platform Transparency Act", which sets out the obligation requirements for algorithmic transparency from three main dimensions: users, regulatory authorities, and the public. In the same month, US Senate members proposed the "Artificial Intelligence Capability and Transparency Act" to implement the recommendations of the final report of the National Security Council on Artificial Intelligence, enhance the government's artificial intelligence capabilities, transparency, and accountability. In July, the US Government Accountability Office released the "Artificial Intelligence Accountability Framework" to ensure fairness, reliability, traceability, and governance of AI systems. In December, the National Institute of Standards and Technology in the United States released the "Artificial Intelligence Risk Management Framework" to guide the development of AI risk management frameworks.

Since 2021, the United States has taken the competition in the field of artificial intelligence as an important position to consume China's competitiveness for a long time, and the action orientation of suppressing China's artificial intelligence development is more clear. After Biden took office, the US Congress passed more than 400 China related bills, using export controls and entity lists to prevent Chinese AI enterprises from acquiring equipment and technology through technology exchange and trade. From 2018 to 2022, the number of Chinese companies on the entity list increased from 130 to 532. Leading Chinese chip companies, supercomputers, and software and hardware suppliers are all included in this list. On August 12, 2022, the Bureau of Industry and Security (BIS) of the US Department of Commerce disclosed a new temporary final rule on export restrictions, imposing export controls on EDA/ECAD software used for computer-aided design of 3nm and below chips. On October 7, 2022, BIS updated the contents of the Export Administration Regulations to restrict the export of 18nm DRAM, 128 layer NAND Flash memory chips and 16/14nm logic chips to China, and stipulated that Americans should not provide support and advice to some enterprises and bases engaged in semiconductor production in China without obtaining a license.

The United States not only directly attacks China's AI development through entity lists, investment reviews and other measures, but also carries out "multilateral cooperation" within the scope of allies to achieve the purpose of suppressing China, and also constantly exerts pressure on allies to exclude Chinese technology. In September 2020, the United States formed an alliance with NATO member countries such as the United Kingdom, Canada, and France, as well as 13 non NATO member countries such as Australia and Japan, to strengthen cooperation in the field of artificial intelligence. In November 2020, the US Department of Defense launched a new plan aimed at strengthening the interoperability of artificial intelligence technology with allied countries, which is largely a "check on China's approach in artificial intelligence". The United States and its allies have a strong tendency to cooperate in the field of artificial intelligence, but overall it cannot go against the values and interests of the United States.

（2） Strategic measures for the development of China's artificial intelligence

China is going through an important stage of transition from "Internet plus" to "Intelligence+", and the development of AI industry is an important driving force and guarantee factor. Overall, the deployment of artificial intelligence at the national strategic level is more systematic, including technological innovation, scenario cultivation, computing power optimization, talent cultivation, standard construction, ethical norms, and many other aspects of work (see Table 2).

表2 中国人工智能政策文件
时间		政策文件
2017年	7月	《新一代人工智能发展规划》
	10月	十九大报告
	12月	《促进新一代人工智能产业发展三年行动计划（2018—2020年）》
2018年	3月	《2018年政府工作报告》
	4月	《高等学校人工智能创新行动计划》
	11月	《新一代人工智能产业创新重点任务揭榜工作方案》
2019年	3月	《关于促进人工智能和实体经济深度融合的指导意见》
	3月	《2019年政府工作报告》
	6月	《新一代人工智能治理原则——发展负责任的人工智能》
2020年	8月	《国家新一代人工智能标准体系建设指南》
2020年	8月	《国家新一代人工智能创新发展试验区建设工作指引》
2021年	3月	《中华人民共和国国民经济和社会发展第十四个五年规划和2035年远景目标纲要》
	5月	《全国一体化大数据中心协同创新体系算力枢纽实施方案》
	7月	《新型数据中心发展三年行动计划（2021—2023年）
	9月	《新一代人工智能伦理规范》
	12月	《互联网信息服务算法推荐管理规定》
2022年	8月	《关于加快场景创新以人工智能高水平应用促进经济高质量发展的指导意见》
2022年	8月	《关于支持建设新一代人工智能示范应用场景的通知》

In terms of technological innovation, as an important guiding document for national development during the 14th Five Year Plan period, the "Outline of the 14th Five Year Plan for National Economic and Social Development of the People's Republic of China and the Long Range Objectives Through 2035" directly lists the new generation of artificial intelligence as one of the seven cutting-edge fields that need to be tackled. It proposes to continuously promote the research of basic theories and algorithms in the forefront of artificial intelligence, develop dedicated chips, build open source algorithm platforms such as deep learning frameworks, and innovate and iterate applications in the fields of learning reasoning and decision-making, image graphics, speech and video, natural language recognition and processing, and other specific work.

In terms of scenario cultivation, in August 2022, the Ministry of Science and Technology and six other departments issued the "Guiding Opinions on Accelerating Scenario Innovation and Promoting High quality Economic Development through High level Application of Artificial Intelligence", proposing to create major application scenarios around "high-end and efficient intelligent economic cultivation, safe and convenient intelligent society construction, high-level scientific research activities, national major events and major projects". On August 15, 2022, the Ministry of Science and Technology issued a notice on supporting the construction of new generation artificial intelligence demonstration application scenarios, initiating the construction of 10 demonstration application scenarios including smart farms, smart ports, smart mines, and smart factories.

In terms of computing power optimization, according to the "Three Year Action Plan for the Development of New Data Centers (2021-2023)" released by the Ministry of Industry and Information Technology, China's computing power infrastructure construction will adhere to a development pattern of reasonable layout, advanced technology, green and low-carbon, and computing power scale that is compatible with the growth of the digital economy. At the same time, regional cooperation and coordinated development will be taken into consideration. In this context, in May 2021, the National Development and Reform Commission, together with the Cyberspace Administration of China, the Ministry of Industry and Information Technology, and the National Energy Administration, released the "Implementation Plan for the National Integrated Big Data Center Collaborative Innovation System Computing Hub", launching the "East Data West Computing" project. Eight national computing hubs were launched in the Beijing Tianjin Hebei region, the Yangtze River Delta, the Guangdong Hong Kong Macao Greater Bay Area, Chengdu Chongqing, Inner Mongolia, Guizhou, Gansu, and Ningxia, and around these eight computing hubs, 10 data center clusters were planned, including Zhangjiakou, the Yangtze River Delta Ecological Green Integration Development Demonstration Zone, Wuhu, Shaoguan, Tianfu, Chongqing, Gui'an, Helingeer, Qingyang, and Zhongwei, to orderly guide the demand for computing power from the east to the west and optimize the layout of data center construction. The sentence is:.

In terms of ethical norms, in 2021, four departments including the National Cyberspace Office jointly issued the Administrative Provisions on the Recommendation of Algorithms for Internet Information Services, which will be officially implemented on March 1, 2022. It requires algorithm recommendation service providers to adhere to the mainstream value orientation, actively spread positive energy, establish and improve the mechanism of manual intervention and user independent selection, and not use algorithms to affect network public opinion, avoid supervision and management, and monopolize and unfair competition. In September 2021, the "Code of Ethics for the New Generation of Artificial Intelligence" was released, which integrates ethics and morality into the entire life cycle of artificial intelligence, actively guiding the whole society to carry out responsible research and application activities of artificial intelligence. In terms of "hard law", the legalization process in areas such as network security, data security, and personal information protection is accelerating, providing legal protection for the ethics and security of artificial intelligence (see Figure 9). In addition, local authorities are actively exploring legislation on artificial intelligence. Shanghai and Shenzhen have successively issued local legislation regulations on artificial intelligence.

Driven by the policy and market, the scale of China's AI industry continues to grow. On August 15, 2022, the data released by the Ministry of Industry and Information Technology shows that the scale of China's AI core industry will exceed 400 billion yuan, more than six times the same period in 2019, and a complete industrial chain covering the basic layer, technology layer and application layer will initially be formed. On September 14th, IDC released the "Global Artificial Intelligence Spending Guide", predicting that China's AI investment scale is expected to reach $26.69 billion by 2026, accounting for about 8.9% of the global total, ranking second among individual countries in the world.

From the perspective of investment and financing, the capital market has positive expectations for the development of artificial intelligence in China. On the one hand, as primary market projects move towards the later stage, there is an increase in massive financing. In 2021, the domestic financing amount for artificial intelligence reached 244.22 billion yuan, with 821 financing transactions. The single financing amount for artificial intelligence was close to 300 million yuan, a year-on-year increase of 17.5%. On the other hand, artificial intelligence companies are experiencing a peak in IPOs, with a significant increase in the amount of fundraising generated by IPOs. In 2021, 11 domestic artificial intelligence companies successfully went public, reaching a historical peak (see Table 3).

表3 2021年中国人工智能上市企业
企业简称	所属领域	上市时间	融资金额
医渡科技	智能医疗	2021年1月15日	39.0亿港元
Appier	智能商务	2021年3月30日	未透露
图森未来	智能运载工具	2021年4月15日	13.5亿美元
利和兴	智能制造	2021年6月29日	3.4亿人民币
理想汽车	智能交通	2021年8月12日	115.5亿港元
海天瑞声	智能服务	2021年8月13日	4.0亿人民币
微创机器人	智能医疗	2021年11月2日	14.6亿港元
鹰瞳科技	智能医疗	2021年11月5日	15.7亿港元
志晟信息	智慧城市	2021年11月15日	未透露
迈赫股份	智能制造	2021年12月7日	8.8亿人民币
商汤科技	计算机视觉	2021年12月30日	55.5亿港元
数据来源：深圳人工智能行业协会。

From the perspective of innovative achievements in the field of artificial intelligence, 51.69% of global artificial intelligence patents are applied for in China, which is about three times the number of patent applications in the United States. However, compared to the increasing number of artificial intelligence patent applications and authorizations, the number of patent applications in China (87343 in 2021) far exceeds the number of authorizations (1407 in 2021), with only 6% of patents granted. The patent authorization ratio in the United States is about 39.59%, more than six times that of China. This indicates that China's patents are "numerous but not superior", and there is an urgent need to shift from quantity to quality in the focus of artificial intelligence innovation. In addition, according to third-party analysis, China is the only country in the world where universities and research institutions have a higher number of artificial intelligence patent applications than enterprises. Due to differences in innovation goals between universities and enterprises, the effective combination of domestic artificial intelligence technology innovation and market demand deserves attention.

Overall, China ranks second in the world in terms of the number of enterprises, total investment and financing, and the number of patents, second only to the United States. However, we need to be soberly aware that there is a gap between the overall development level of China's AI and that of developed countries. There is a lack of major original achievements in basic theories, core algorithms, key equipment, high-end chips and other fields. The innovation of scientific research institutions and enterprises is still a follower, with weak innovation ability, and the lack of systematic advanced R&D layout. The top AI talents are far from meeting the demand.

In terms of industry, the development of China's AI industry mainly follows the "best innovation" proposed by Li Kaifu, that is, "China should first focus on the development of AI application technology to quickly occupy the market; secondly, through the globalization system, purchase AI chips from leading semiconductor manufacturing technology enterprises to make up for deficiencies". This development idea of following the comparative advantage has formed a good momentum in the short term, opening up the situation of China's AI development, but also causing the problem of "incomplete development" of China's AI industry chain, that is, the industrial advantages are concentrated in the application fields with relatively mature technologies and clear scenarios, and there is insufficient willingness and ability to invest in the long-term, high-risk foundation and technology layers, which has weaknesses and weak links, leading to serious dependence on foreign enterprises (see Table 4). Under the tense geopolitical situation, the long-term development of China's AI industry is vulnerable to restrictions from the United States and other developed countries.

表4 中国人工智能产业相关情况
产业层级	发展现状	存在问题
基础层	人工智能基础层产业规模年均增速低于全球增速； 2021年上半年中国人工智能芯片中，由国外公司寡头垄断的GPU芯片占有90%以上的市场份额	智能芯片依赖进口；国产化算力基础薄弱，算力需求与算力能力之间存在鸿沟；开源框架受制于国外科技巨头
技术层	计算机视觉和语音识别等应用技术垂直领域不断突破，技术层产业规模年均增速高于全球增速；大模型、多模态等前沿算法领域处于跟随状态； Aminer 2022年人工智能全球最具影响力学者榜单中，美国在机器学习领域顶尖科研学者数量上占据绝对优势，位居全球第一，而中国仅居第四	底层算法理论和开发平台等基础技术领域发展节奏缓慢；技术方面缺乏系统的超前研发布局；尖端人才储备不足
应用层	2022年应用层产业规模达到161亿美元，成为我国AI产业链中优势最突出的部分；我国AI应用层在智能安防、自动驾驶领域相对成熟，同时也加快与制造业融合发展	场景碎片化导致算法复用率低，难以形成标准化、规模化产品，落地成本较高
资料来源：刘晨，《我国人工智能产业竞争力评估：国内格局和全球比较》，《宏观观察》2022年第41期。

To unleash the potential of artificial intelligence, we must find a European path.

The international competition strategy of the European Union for artificial intelligence does not follow the usual path of technological catch-up, but hopes to make up for the EU's shortcomings in artificial intelligence hard power based on its legislative capacity and market size, and strive for development space for the European artificial intelligence "technological sovereignty" path. Guided by this approach, the European Union has issued a series of policy documents and ethical requirements, including the European Artificial Intelligence Strategy, the Artificial Intelligence Coordination Plan, the Code of Ethics for Trustworthy Artificial Intelligence, the White Paper on Artificial Intelligence, and the 2021 Revised Edition of the Artificial Intelligence Coordination Plan. The aim is to lead the global legislative regulation and standard construction of artificial intelligence through principles such as "people-oriented" and "ethical and trustworthy artificial intelligence". In April 2021, the European Commission passed the proposal for the Artificial Intelligence Act, marking the transition of global artificial intelligence governance from soft constraints such as ethical principles to a comprehensive and actionable legal regulation stage.

As the world's first comprehensive artificial intelligence bill, the "Artificial Intelligence Act" strictly delineates risk boundaries around the reasons and methods of using artificial intelligence systems, dividing them into four types: unacceptable risk, high risk, limited risk, and minimum risk, and adopting different restrictive measures for the use of artificial intelligence systems with different risk levels. For example, artificial intelligence systems that pose a threat to human physical and mental health are classified as unacceptable risks and are explicitly prohibited; Introducing CE certification for AI systems with high risks, AI developers and users must strictly fulfill seven obligations to obtain certification, including: ① a comprehensive risk assessment system; ② Provide high-quality datasets to the system, minimizing risks and discriminatory outcomes; ③ Retain system logs to ensure traceability of results; ④ Provide all necessary information about the system and its purpose for the government to evaluate its compliance; ⑤ Provide clear and sufficient information to users; ⑥ Take appropriate human supervision measures to minimize risks (such as stop buttons); ⑦ High level of security and accuracy. In addition to strict pre review procedures, the bill also requires full process supervision and compliance assessment of high-risk artificial intelligence systems (see Table 5).

表5 《人工智能法案》风险等级划分
风险程度	描述	监管措施
不可接受风险	·威胁人的安全、生计和权利，包括违背自由意志操纵人类行为的人工智能系统（如鼓励未成年人危险行为）和允许政府使用社会信用评分的系统	·禁止； ·若违反，处以前一财年全球营业额最高6%的罚款
高风险	·重要基础设施（如交通），可能威胁人的生命和健康； ·教育或职业培训，可能决定某人受教育的机会（如考试评分）； ·产品的安全零件（如人工智能在机器人辅助手术中的应用）； ·就业、员工管理（如招聘软件）； ·基本的私人和公共服务（如信用评分剥夺公民获得贷款的机会）； ·可能干涉人的基本权利的执法（如评判证据可靠性的系统）； ·移民、庇护和边境控制（如合适旅行文件真实性的系统）； ·运用于司法和民主程序的人工智能系统	前置审查： ·完备的风险评估系统； ·向系统提供高质量的数据集，最小化风险和歧视性结果； ·留存系统日志以确保结果的可追溯性； ·提供有关系统及其目的的所有必要信息，以供政府评估其合规性； ·向用户提供明确、充分的信息； ·采取适当的人为监督措施最小化风险（如停止按钮）； ·高水平的安全性和准确性。全过程监督和合规评估严格执法和处罚
有限风险	·使用人工智能（如聊天机器人）时，使用者能意识到在与机器互动进而作出明智决定	·实现透明公开
最低风险	·允许自由使用人工智能的电子游戏或垃圾邮件过滤器等应用	·不作干预

The EU not only continuously improves its own laws and regulations in the field of artificial intelligence internally, but also maintains consistency and integrity in member states' actions. It also strengthens cross-border jurisdiction and establishes alliances and technological partnerships externally. In 2018, the European Union established the Advanced Expert Group on Artificial Intelligence and the European Artificial Intelligence Alliance, actively promoting the internationalization of artificial intelligence governance principles and standards. In April 2021, the European Union updated its "Artificial Intelligence Coordination Plan", calling for the establishment of artificial intelligence technology cooperation relationships between member states and between member states and other countries, and proposing a specific set of joint action plans. In June 2021, the EU-US Trade and Technology Committee was established, and both Europe and the United States reached a consensus on establishing credible and responsible governance principles, proposing that Europe and the United States should cooperate in developing and deploying trustworthy artificial intelligence systems based on common democratic values and formulate global norms. The EU is establishing various multilateral and bilateral cooperation relationships at the member state, regional, and global levels through institutional cooperation frameworks. On the one hand, this can reduce unilateral dependence on key artificial intelligence technologies and ensure "technological sovereignty"; On the other hand, the EU is also utilizing its advantages in multilateral international mechanisms to expand the influence of the EU regulatory system and artificial intelligence values in the international community, in order to intervene in the global rule making of artificial intelligence.

Changes in Competition and Cooperation in the International Artificial Intelligence Field

（1） The United States is increasing pressure on China in the field of smart chips

The deeper the United States' understanding of the strategic significance of artificial intelligence and the stronger its desire to pursue global hegemony in the field of technology, the more difficult it is to reconcile its power struggle with the rising China. Since 2018, the United States has regarded the field of artificial intelligence chips as an important battlefield for long-term consumption of China's competitiveness, and has spared no effort to slow down China's technological development. The Bureau of Industry and Security (BIS) of the US Department of Commerce has implemented targeted export restrictions on high-end artificial intelligence chip related items, software, and technology to China through the "Entity List", disrupting the original order of chip technology flow and production supply. Restricted by the weak position in the high-end chip field, Chinese AI related enterprises have been severely impacted. From 2018 to 2022, the number of Chinese companies on the entity list increased from 130 to 532. At the same time, the US Export Administration Regulations include extraterritorial applicability, which means that chip products exported from other countries to China that contain more than a certain amount of US items will be listed as controlled objects. In addition, the United States is using political means to pressure third-party countries and restrict their sales of artificial intelligence chip related products to Huawei.

On August 12, 2022, the Bureau of Industry and Security (BIS) of the US Department of Commerce disclosed a new temporary final rule on export restrictions, imposing export controls on EDA/ECAD software used for computer-aided design of 3nm and below chips. On September 1, 2022, chip manufacturers AMD and NVIDIA China announced that they had received a notice from their headquarters that US officials had requested them to stop exporting high-end GPU chips to China and Russia. On October 7, 2022, BIS updated the contents of the Export Administration Regulations to restrict the export of 18nm DRAM, 128 layer NAND Flash memory chips and 16/14nm logic chips to China, and stipulated that Americans should not provide support and advice to some enterprises and bases engaged in semiconductor production in China without obtaining a license.

In response to the recent intensive suppression measures taken by the United States in the field of artificial intelligence chips, the Center for Strategic and International Studies of the United States released the report Cutting off China's Path to the Future of Artificial Intelligence, which directly pointed out that the Biden government is restricting the future development of China's artificial intelligence industry in terms of computing power by regulating the use of high-end artificial intelligence chips and chip design software, as well as limiting China's use of semiconductor manufacturing equipment and parts made in the United States. The naked suppression measures of the United States in the chip field will also have a chilling effect on many areas such as Sino US trade, finance, and technology cooperation, causing even more chaos and confrontation.

（2） Competitive Action under the Opportunistic Strategy of the European Union

As the world's largest single digital market, the EU's choice of path and strategic design in the development of artificial intelligence has become the biggest factor affecting the "battle situation" of artificial intelligence. In the strategic competition between China and the United States, the politicization of technological issues and the expansion of long arm jurisdiction by the United States have compressed the autonomy space for the development of artificial intelligence in the European Union. The EU also stated that it should change its long-standing approach of mainly viewing technological issues from an economic perspective, and prioritize following political logic, placing the development of artificial intelligence in the context of the EU's pursuit of "technological sovereignty" and geopolitics, and strengthening its opportunistic attitude between China and the United States.

From the perspective of US European relations, the common values of human rights and democracy between the two sides have promoted substantive cooperation in artificial intelligence technology and standards. In June 2021, the EU-US Trade and Technology Committee was established, and both Europe and the United States reached a consensus on establishing credible and responsible governance principles, proposing that Europe and the United States should cooperate in developing and deploying trustworthy artificial intelligence systems based on common democratic values and formulate global norms. The United States hopes that the European Union will participate in the boycott of China by exerting political pressure, but the EU is well aware that blindly following it will undermine the global supply chain and cooperation, have a significant negative impact on its own interests, and also hold a hesitant attitude towards the strong unilateralism of the United States, so it has not actively followed.

From the perspective of China Europe relations, cooperation and competition have long coexisted between China and Europe. The monopoly position of American technology companies in Europe threatens the EU's technological autonomy, so there is room for the EU to cooperate with China in areas such as antitrust and platform economy governance to balance large American technology companies. However, at the same time, the EU still holds a distrustful attitude towards China in the field of security, and has strengthened coordination with the United States in areas such as technical standards, secure supply chains, information and communication service security and competitiveness, export controls, and investment reviews, while maintaining a defensive attitude towards China in the field of security.

Currently, the decoupling of geopolitics and technology has become a new context for the development of artificial intelligence. China, the United States, and Europe are joining the war in different ways. The familiar thinking of "benchmarking" or "catching up" in peacetime may not be suitable for the current international competition in the artificial intelligence industry. Robert Carlson and Rick Weblin, associate professors at the University of Washington, believe that in the long run, the decoupling of American technology will only slow down China's development of technology according to its own strategy in the short term. The core of the competition between China and the United States will be the competition of technological strength, and maintaining technological innovation capability and industrial growth momentum is the biggest endogenous driving force for China to respond to American technological nationalism. In response to the United States' containment of China's development in basic technologies and cutting-edge disruptive fields, China must increase its efforts in technological innovation, seek truth from facts, and develop its own innovation capabilities in a down-to-earth manner, and overcome the technological gap in basic fields as much as possible. Intensify research efforts in cutting-edge fields such as 6G technology, quantum computers, and brain computer interfaces that may change industry rules, and participate in the formulation of relevant international laws and regulatory standards.

The technology nationalism policy of the United States has seriously impacted the world trade and technology flow order. Against the backdrop of delicate international competition and competition, China should actively engage in in-depth dialogue with other countries in the field of artificial intelligence technology, respect the reasonable security concerns of all parties, strive to maintain the stability of the global industrial chain and supply chain, and proactively avoid the global risk escalation caused by the tension in Sino US relations. In the field of artificial intelligence technology innovation, China should actively participate in and initiate international cooperation platforms and major science and technology plans, strengthen regional interoperability of artificial intelligence systems through the provision of international science and technology public goods, and cooperate in data security flow and datasets to build a broader community of interests. At the same time, we will strengthen cooperation with the European Union in the field of artificial intelligence through global governance issues such as ecological governance, trustworthy artificial intelligence, computing power "carbon footprint", natural disaster prevention, and climate change.