Analysis of Cutting-Edge AI Technologies and High-Quality Development

Artificial Intelligence (AI) refers to the ability of machines to achieve perception, cognition, and action comparable to or surpassing that of humans. Based on the different levels of intelligence capabilities, AI is generally divided into General Artificial Intelligence and Specialized Artificial Intelligence, also known as Strong AI and Weak AI. Currently, there are two interpretations of General Artificial Intelligence: one is the general understanding of AI (General Artificial Intelligence, GAI) by the media and the public, referring to intelligence that can handle many tasks; the other is the precise name in the field of AI, namely Artificial General Intelligence (AGI), which refers to AI that reaches human-level performance in all aspects and can adaptively respond to external environmental challenges, completing all tasks that humans can accomplish.

For a long time, AI systems have been designed to achieve specific or specialized goal tasks, falling under the category of Weak AI or Specialized AI. Since 2018, large-scale pre-trained models (commonly known as large models) have adapted to a series of downstream tasks through training on massive amounts of unlabeled data relying on powerful computational resources, marking the dawn of General Artificial Intelligence (GAI), although true Artificial General Intelligence (AGI) has not yet been achieved.

(1) Development History of Artificial Intelligence

The concept of artificial intelligence was formally proposed in 1956 at the summer workshop held at Dartmouth College in the United States. Over the nearly seventy years of AI development, it can be roughly divided into three stages.

The first stage focused on symbolic logic reasoning proof. This stage mainly researched how to simulate human reasoning abilities through logical reasoning or heuristic programs based on formal representation methods, solving problems such as algebra application problems, geometric theorem proofs, and machine translation. The second stage centered on expert systems based on artificial rules. The research focus during this stage was on summarizing the knowledge of domain experts into artificial rules for decision support, leading to rapid development of expert system technologies. The third stage is driven by deep learning based on big data. This stage effectively integrated algorithms, computational power, and data, shifting the focus of AI research from how to “manufacture” intelligence to how to “acquire” intelligence. In 2006, Professor Geoffrey Hinton from the University of Toronto proposed the “deep learning algorithm,” laying the theoretical and methodological foundation for a new round of AI development. In 2012, the deep learning neural network model AlexNet proposed by Professor Hinton and his students achieved a significant advantage over other non-neural network models in the ImageNet image recognition challenge, marking the rise of deep learning. From 2016 to 2021, Google’s series of Go-playing robots, AlphaGo and AlphaZero, not only defeated world champions Lee Sedol and Ke Jie in competitions but also achieved an undefeated record in subsequent matches. Meanwhile, Google’s AlphaFold2 achieved near-human experimental analysis levels in predicting protein structures, solving the “protein folding” problem that has plagued biology for 50 years. In the past decade, deep learning models and algorithms based on big data have been widely applied, achieving great success in machine translation, intelligent Q&A, and adversarial games, marking an accelerated development period for AI.

Among these three stages, the main idea of the first two stages was to design new theories and algorithms to simulate human intelligence with machines. Despite progress in theoretical methods, the overly ambitious goals and insufficient integration with applications led to fluctuations in AI development. The third stage, based on deep learning technology driven by big data, has become the mainstream development path for current AI, achieving large-scale applications in fields such as computer vision, natural language processing, and intelligent speech. Among these, computer vision is the most widely applied field of AI technology. Computer vision refers to the process of using computers to process images or videos, enabling automatic recognition, analysis, and understanding of information within images or videos. In 2015, the deep residual learning network ResNet proposed by researchers at Microsoft Research Asia became a milestone representative technology in the field of computer vision. The emergence of ResNet allowed deeper networks with over a hundred layers to be trained more effectively, pushing the limits of deep learning technology. Currently, ResNet has become the preferred architecture for computer vision tasks, such as image classification, object detection, and image segmentation. In 2023, the four authors of ResNet received the Future Science Prize for their foundational contributions to AI.

(2) New Trends in Current AI Development

Since 2018, large models have first achieved breakthroughs in the field of natural language processing, with the phenomenon-level product ChatGPT marking the dawn of General Artificial Intelligence and triggering a new wave of AI development. The current development of AI has transitioned from the era of small models to that of large models.

Large models are the product of the combination of “big data + big computing power + strong algorithms” and possess at least three characteristics: First, they are large in scale, with neural network parameters exceeding billions; Second, they exhibit emergent properties, generating unexpected new capabilities, which is the most significant new characteristic in the development of AI in nearly 70 years; Third, they are generalizable, capable of solving various problems.

The GPT (Generative Pre-trained Transformer) series of large models developed by OpenAI is currently a leading representative in the international large model field. In November 2022, OpenAI released the AI conversational model ChatGPT, demonstrating astonishing levels of intelligence, capable of engaging in long, natural conversations and writing high-quality content of nearly any type, completing many tasks requiring creative thinking. Upon release, it garnered widespread attention from global users, becoming the fastest-growing consumer application in history, triggering the “iPhone moment” of AI.

The outstanding performance of ChatGPT can be attributed to several key technological and strategic aspects. First, it utilizes a large-scale training dataset, particularly internet text data, to capture rich knowledge and language patterns. Second, the model is based on an efficient Transformer architecture, which effectively processes the relational dependencies of input sequences through self-attention mechanisms, making it highly suitable for natural language processing tasks. Third, ChatGPT improves its generalization and generation capabilities across diverse problems through multi-task learning. Fourth, the model has also undergone fine-tuning for specific tasks to better adapt to and solve problems in particular domains or scenarios. Fifth, by utilizing reinforcement learning and other techniques for model optimization, ChatGPT’s outputs on specific tasks are closer to human habits, further enhancing its performance. The integration and application of these technologies have made ChatGPT a leader in text Q&A tasks, igniting public imagination about the future development of strong AI.

In addition to language capabilities, large models are rapidly expanding their abilities in vision, hearing, embodiment (intelligent agents capable of interacting with their environment), and action, gradually entering the real world and developing physical intelligence, triggering the next wave of AI development.

However, it should also be noted that while large models and other general AI technologies bring tremendous opportunities for global economic and social development, they also pose various unpredictable risks and complex challenges. Large models are highly complex AI systems characterized by unpredictability. The progress made so far has largely been achieved through empirical patterns, and the intelligence emergence mechanisms behind large models remain unclear. The international community knows very little about how to build a safe AI system. Currently, general AI (GAI) represented by large models has revealed a series of risks related to ethics, data security, etc., necessitating enhanced safety regulation. In the future, as the arrival of Artificial General Intelligence (AGI) may trigger existential risks for humanity, stricter preventive measures are required. Strengthening governance of general AI has become a common issue faced by countries worldwide.

As humanity’s exploration of the path to general intelligence becomes clearer, the world is on the eve of “AGI” (near-strong AI), in a state of uncertainty. In the future, in-depth research is needed on the foundational principles of large models, safety and value alignment, and risk control strategies for AGI, promoting AI technology for the benefit of humanity.

(3) Major Countries Accelerate AI Strategy and Policy Deployment

Currently, artificial intelligence has become a new focus of international competition and a new engine for economic development, with major developed countries viewing AI development as a significant strategy to enhance national competitiveness and maintain national security, and actively formulating AI plans and relevant policies to strive for dominance in the new round of international technological competition.

Countries are strengthening their AI layout from a national strategic level. The United States has successively issued the “National Artificial Intelligence Research and Development Strategic Plan” and other related strategies and policies focusing on AI R&D and national security, aiming to consolidate its world-leading advantage; France released the “National Artificial Intelligence Strategy,” focusing on promoting intelligence in health, transportation, environment, and national defense; the European Union has identified achieving intelligent growth as one of its three growth goals since 2010. In April 2018, it released the “European Artificial Intelligence” strategy, systematically proposing the EU’s AI development strategic plan. In the same month, the “Artificial Intelligence Cooperation Declaration” was released, marking a new stage of cooperative development in European AI; Germany proposed to become a global leader in AI research through the “Federal Government’s AI Strategy Highlights”; the UK released the “National AI Strategy,” building a strong AI country from aspects of data acquisition, talent cultivation, scientific research, and industrial application; Russia released the “National Development Strategy for Artificial Intelligence by 2030”; Japan proposed to build a “super-intelligent society 5.0,” aiming not only to enhance industrial competitiveness but also to achieve intelligent living for its citizens.

Countries are competing to increase investment in AI R&D. The US government invested over $2 billion in non-classified AI projects in the fiscal year 2017, with total investment reaching $24.9 billion by 2022, and projected to exceed $100 billion by 2028; the proposed defense budget for fiscal year 2021 included a total investment of $841 million in AI R&D, an approximately 8% increase from $780 million in fiscal year 2020. France plans to invest €1.5 billion in AI projects by 2022. South Korea aims to transition from an “IT powerhouse” to an “AI powerhouse,” planning to enhance its competitiveness in AI to be among the world’s top by 2030. According to the budget, if related measures are implemented, South Korea is expected to create an economic benefit of 455 trillion KRW (approximately 2.7 trillion CNY) in the AI field by 2030.

Countries are establishing new types of AI R&D institutions. The US National Science Foundation, in coordination with federal agencies, including the Department of Homeland Security, the Department of Defense, the Department of Education, and the Department of Agriculture, has jointly established 25 national AI research institutes; Europe plans to establish a world-class AI research institute and set up research centers in various European countries such as the UK; France proposed to establish new AI centers and form AI research networks; the UK is expanding the Alan Turing Institute and initiating the construction of data ethics and innovation centers, as well as establishing new AI technology colleges. Major global AI multinational companies and leading enterprises are also accelerating the establishment of AI R&D centers.

Countries are accelerating the establishment of AI governance systems. Since 2018, the United Nations has established the AI and Robotics Center to study the governance issues of AI; the US Congress has proposed the establishment of an AI Safety Commission to oversee the development and related technologies of AI and machine learning; the EU has signed the “AI Cooperation Declaration” and published the “EU Civil Law Rules on Robotics” to jointly address ethical and legal challenges posed by AI; since 2019, the EU has continuously strengthened its focus on AI applications and governance, issuing the “Ethical Guidelines for Trustworthy AI” in April of that year, setting an ethical framework for achieving trustworthy AI.

Currently, global AI development is transitioning from weak AI to strong AI, and AI has become a crucial support for national strategic competitiveness and a significant force driving technological revolution. In the future, “AI + high-speed mobile internet” will become a basic scenario in human social life. Looking further into the future, strong AI will bring disruptive and global impacts. The first to achieve breakthroughs will hold the dominant position in future development. If China falls behind in the new round of AI development, it will be at a disadvantage in global competition.

General Secretary Xi Jinping pointed out, “We should regard the new generation of AI as a driving force for promoting technological leapfrogging, optimizing industrial upgrading, and enhancing overall productivity, striving to achieve high-quality development.” In 2017, the State Council released the “New Generation AI Development Plan,” establishing a three-step goal for AI and elevating it to a national strategy. Since then, relevant ministries and local governments have been promoting the accelerated implementation of the “New Generation AI Development Plan,” while the technology, industry, and investment sectors have collaborated to propel China’s AI development into a critical period of leapfrogging.

(1) Breakthroughs in foundational theories and key technologies of AI, with deepening integration of AI with the economy and society

After years of sustained research and development, China’s AI technological innovation system has gradually improved, with the development of intelligent economy and intelligent society deepening, achieving significant results.

First, the foundational theories of AI have rapidly accumulated. In recent years, domestic scholars have made significant contributions in classic AI fields such as problem-solving, evolutionary computation, pattern recognition, expert systems, and intelligent control. In particular, extensive research has been conducted in emerging deep learning theories and reasoning algorithms. For example, Peking University proposed a deep cross-media learning method that significantly improved the accuracy of cross-media retrieval; Nanjing University proposed the “deep forest” model, which is the world’s first non-neural network and BP (backpropagation) algorithm-based deep learning method. Progress has also been made in brain-like computing, with advancements in brain-like chips, brain-like computing systems, and brain-like applications; the Chinese Academy of Sciences has made breakthroughs in brain-computer interface technology, developing the fastest scalp EEG brain-computer interface system currently in operation; Huawei has launched a meteorological large model, exhibiting advantages in several precision indicators and extreme weather forecasts of interest to meteorologists, demonstrating strong competitiveness and immense potential.

Second, key AI technologies have reached world-leading levels. In the early stage of the current AI technology explosion, China kept pace with the world in areas such as Chinese information processing, biometric recognition, machine translation, intelligent processors, autonomous driving, and intelligent robotics, achieving breakthroughs in several key AI technologies. Rafael Reif, the 17th president of MIT, has commented that China leads the world in fields like facial recognition and speech recognition. The main achievements include the Chinese Academy of Sciences developing the world’s first commercially available deep learning-specific processor, the “Cambricon” chip, which significantly outperforms central processing units (CPUs) and graphics processing units (GPUs) in terms of performance and power efficiency when running mainstream intelligent algorithms; SenseTime’s image recognition technology, iFlytek’s speech recognition and synthesis technology, and language translation technology are currently among the world’s leaders and have received international recognition.

Significant original innovations have also been achieved in computer vision. Peking University has rewritten the nearly two-century-old principles of exposure imaging, inventing the pulse photography principle, approximating high-speed photoelectron flow with bit sequences, and developing ultra-high-speed visual chips and cameras, achieving ultra-high-speed, high dynamic, and blur-free continuous clear imaging. It has established a pulse vision algorithm system and developed an ultra-high-speed system that uses conventional optoelectronic devices and chip processes to achieve continuous clear imaging and real-time tracking recognition of supersonic processes. Its patents have been authorized in China, the US, Europe, Japan, and South Korea, and are expected to reshape the technology and industrial system of computer vision from the source.

In terms of foundational software and hardware for AI, Huawei has released two AI chips based on the Da Vinci architecture, the Ascend 910 and Ascend 310, striving to create a complete ecosystem from foundational algorithms to application development based on chips, providing new options for global developers and enterprises, as well as security guarantees for domestic enterprises. On this basis, Pengcheng Laboratory launched the “Pengcheng Cloud Brain II” and is developing the next-generation facility “Pengcheng Cloud Brain.” “Pengcheng Cloud Brain II” is built based on Huawei’s domestically produced AI chips and is the first fully autonomous controllable E-level intelligent computing platform in China, boasting internationally leading AI computing power levels, having achieved championships on several international rankings. This platform shares approximately 70% of its computing time externally, supporting nearly a thousand domestic AI model training tasks and AI algorithm releases, becoming one of China’s most important open, shared, and autonomous controllable AI large model training platforms. The next-generation facility “Pengcheng Cloud Brain” will be an intelligent tool platform for 6G ultra-broadband communication, designed with a super-large scale and high-performance computational architecture, expected to be completed by 2025. The “Pengcheng Cloud Brain” large scientific device will further promote the development of China’s domestically produced autonomous industrial ecosystem in AI and is bound to become a significant scientific infrastructure supporting the innovative research of the new generation of intelligent network communication.

(3) Accelerating Integration of AI with Various Industries

In intelligent manufacturing, efforts are being made to promote the construction of intelligent manufacturing factories, achieving personalized product customization; in intelligent healthcare, AI medical imaging products have been developed for early screening of esophageal cancer, with detection rates exceeding the average rates of doctors using endoscopes; in smart cities, the “City Brain” has been applied to traffic management in Hangzhou, effectively reducing regional travel time; in intelligent logistics, AI technology has been applied to improve logistics systems, with sorting efficiency exceeding that of manual sorting by more than ten times; in intelligent transportation, the Capital Airport has adopted AI technology to complete the parking space arrangement for 1,700 flights within 50 seconds, reducing flight delays and increasing parking space utilization by 10%; in intelligent security, Guangzhou has utilized facial recognition technology to help discover and capture criminal suspects.

(4) Initial Construction of an Innovative Ecosystem for AI Development

The Ministry of Science and Technology has established national-level open innovation platforms in areas such as autonomous driving, city brain, intelligent healthcare, intelligent speech, and intelligent vision, supporting technological innovation in small and medium-sized enterprises, promoting technological progress and industrial upgrading in various sectors. A certain intelligent speech platform has over 800,000 developer teams, forming a complete AI industrial chain covering technology R&D, foundational platforms, IoT, and intelligent hardware; a certain autonomous driving open platform has more than 120 partners, forming the world’s largest autonomous driving ecosystem, covering all links of the industrial chain including vehicle manufacturers, parts manufacturers, mobility service providers, startups, communication companies, universities, and local governments. Meanwhile, co-creation spaces, incubators, and accelerators in the AI field are rapidly developing, and the entrepreneurial incubation system is gradually improving.

(5) The US and China Lead Large Model Development, with China Closing the Gap with the US in Language Models, and New Generation Large Models in Vision, Multi-Modal, and Embodiment Expected to Advance Together

Currently, the international large model field has formed a pattern where the US leads and China follows closely. According to the “Research Report on AI Large Model Map” released by the China Institute of Science and Technology Information in May 2023, from the distribution of large models released globally, China and the US are significantly ahead, accounting for over 80% of the global total, with China ranking second in the number of large models.

The development of large models in China is flourishing. As of May 2023, 79 large models have been released, most of which are language models.

International foundational large models are mainly divided into categories such as language, vision, and multi-modal. In language models, OpenAI’s GPT series and Google’s PaLM 2 have formed a leading advantage. China has developed language models such as Zhiyuan’s “Wudao Tianying,” Baidu’s “Wenxin,” Huawei’s “Pangu,” “Pengcheng Mind,” and Alibaba’s Tongyi, but there is still a certain gap compared to top overseas levels. With the prosperity of the AI model open-source ecosystem, the gap between large models in China and the US is expected to gradually narrow.

In the fields of vision and multi-modal large models, China is expected to reverse the following situation and achieve parity with the US. In vision large models, Zhiyuan Research Institute has innovated the research path, pioneering core modeling ideas such as “contextual image learning” and “vision-centric” to understand, interpret, and output images, developing a universal multi-task model Painter. After optimizing the Painter model for object segmentation tasks, the world’s first universal visual model SegGPT that completes any segmentation task using visual prompts has been developed, becoming a key milestone in international visual models alongside Meta’s released foundational image segmentation model SAM. In multi-modal large models, Zhiyuan Research Institute has developed the first unified multi-modal pre-trained model Emu that connects multi-modal input to multi-modal output, surpassing DeepMind’s multi-modal large model Flamingo, refreshing eight performance indicators, and covering the generation of images and text as well as video understanding, capable of completing multi-modal tasks such as image-to-text and text-to-image. The tri-modal (image, text, sound) large model “Zidong Taichu” developed by the Institute of Automation, Chinese Academy of Sciences, currently possesses full modal capabilities, reaching an internationally advanced level.

(6) Overall Development of AI in China Has Entered the Global First Tier

The Information Technology and Innovation Foundation (ITIF) in the US released a report in 2019 titled “Who Wins the AI Race: China, the EU, or the US?” comparing the AI technology innovation and ecosystem building capabilities of China, the US, and the EU across six dimensions: talent, research, enterprise development, application, data, and hardware. In January 2021, ITIF released the updated version of this report, noting that the US still maintains a significant overall leading advantage, but China’s score has significantly increased compared to 2019, surpassing the EU and rising to second place, second only to the US. China’s application scenarios are rich, giving it a certain advantage over foreign countries. However, the report also indicated that China still has significant gaps compared to the US in AI research, talent, and enterprise development.

According to the 2023 Global AI Index ranking released by the British media organization Tortoise Media, the top three countries in comprehensive AI status are currently the US, China, and Singapore. Among them, China leads the US in operational environment and government strategy, closely follows in infrastructure, research, development, and business, but has a significant gap in talent.

At the meeting of the Political Bureau of the Central Committee on April 28, 2023, it was emphasized that the development of general AI should be prioritized, fostering an innovative ecosystem while paying attention to risk prevention. The development of AI in China must deeply grasp the international trends in general AI technology, conduct forward-looking technological research, and strive for leapfrogging. Additionally, it is essential to strengthen the construction of policies, talent, foundational software and hardware, and open-source ecosystems to create a favorable environment. Furthermore, risk assessment should be enhanced, actively promoting AI governance to ensure the sustained and healthy development of the new generation of AI in China.

(1) Continuously Improve AI Planning and Policy Systems in China

In light of the new situations, opportunities, issues, and challenges in the development of general AI internationally, the “New Generation AI Development Plan” proposes a strategic goal for AI in China to achieve world-leading levels by 2030. When implementing the plan in the new era, new changes should be highlighted, forming new planning task directions. Furthermore, relevant supporting policies should be formulated to address the weak links and development needs of general AI in China, focusing on technological research, resource openness, scenario construction, and talent development, to establish a supporting policy system for high-quality AI development.

(2) Strengthen Research on Foundational Theories and Key Technologies of General AI

Proactively lay out research on cutting-edge technologies in general AI, exploring foundational principles and new architectures of large models, and conducting research on the next generation of large models in vision, video, multi-modal, and embodiment, utilizing large models to solve major scientific problems, and forming an internationally influential original theoretical system for general AI.

Lead innovation in key core technologies of general AI, focusing on breakthroughs in distributed efficient deep learning frameworks, large-scale cognition and reasoning, controllable content generation, and efficient low-cost training and reasoning algorithms, establishing China’s technological innovation system for general AI.

(3) Strengthen the Foundational Software and Hardware Ecosystem for AI

Promote breakthroughs in domestically produced AI chips to comprehensively support the computational power needs for large model training, multi-modal processing, and scientific computing, exploring new architectures such as reconfigurable chips, integrated storage and computing chips, and ultra-spec high-performance intelligent chips to provide computational power guarantees for AI development in China. Strengthen research and development of autonomous open-source deep learning frameworks, enhancing core capabilities in distributed training of large models and multi-end multi-platform reasoning deployment, and developing tools for the entire process of model development, training, compression, and reasoning. Support extensive adaptation and optimization of AI chips and deep learning frameworks, creating a deep collaborative ecosystem of domestic foundational software and hardware for AI.

(4) Strengthen the Gathering and Sharing of Resources such as Data and Computational Power

Establish a multi-level data openness system. Relevant government departments should introduce policies to promote publishers, magazines, libraries, museums, archives, and other copyright data or public data institutions, as well as internet platforms to orderly open data for AI technology research and development, breaking down data barriers. Establish a long-term mechanism for constructing large-scale high-quality Chinese datasets, integrating and gathering large internet enterprises, large model R&D enterprises, data service enterprises, major publishers, libraries, mainstream media, and industry organizations to build large-scale high-quality language, speech, image, video, and multi-modal datasets, as well as industry datasets in healthcare, transportation, etc., providing foundational guarantees for the long-term healthy development of general AI in China.

Consolidate the infrastructure for computational power. Strengthen the construction of intelligent computing centers, gradually increasing the localization rate of computational facilities to provide high-performance computing resources and services for large model research. Promote the construction of the China Computing Network to achieve interconnectivity among national-level supercomputing centers, intelligent computing centers, and data centers for “East Data West Computing,” realizing coordinated scheduling and efficient computing of large-scale computing resources nationwide, integrating cloud, network, and computing resources to form a national-level computing infrastructure and unified large computing market that supports digital economy development and provides inclusive computing power for AI technology innovation and industrial intelligent transformation.

(5) Strengthen Risk Assessment and Governance System Construction for AI

China is at the forefront of international governance of general AI, having issued the world’s first normative policy document on generative AI, the “Interim Measures for the Management of Generative AI Services” in July 2023, serving as a reference for other countries in formulating related policies. With the rapid development of general AI technology, China should adhere to the principle of balancing development and safety, establishing and improving a governance system that meets the needs of AI development in China. First, strengthen the assessment of safety risks in the development of AGI, adjusting China’s development strategies for general AI according to risk issues in a timely manner. Second, establish an agile governance system with Chinese characteristics. Given the rapid development of general AI and the emergence of new applications and models, an agile governance system suited to China’s economic and social development characteristics should be established, maintaining policy flexibility and allowing for institutional development space to ensure the long-term healthy development of technology. Third, conduct research on risk prevention technologies, focusing on technical supervision technologies, conducting in-depth research on foundational principles of large models, safety and value alignment, and risk control strategies for AGI, promoting AI technology to benefit humanity.

Establishing a sound regulatory framework for AI-related laws, regulations, and standards is a critical component in ensuring high-level improvement and high-quality development of AI. There should be a gradual improvement of safety guarantees and ethical norms for AI, ensuring the safety and credibility of AI. First, formulate and improve laws, regulations, and standards related to AI, covering all aspects of AI development, use, and application, clarifying related responsibilities and legal consequences, and regulating the development and use of AI. Second, strengthen safety guarantees for AI, enhancing research and technical guarantees for AI safety, preventing malicious use and attacks on AI, ensuring the stable operation of AI systems and the security of data. Third, establish ethical norms for AI, clarifying the moral and social responsibilities of AI, and avoiding negative impacts and ethical risks brought by AI. Fourth, establish a regulatory mechanism for AI, including assessment, review, supervision, and monitoring of AI, strengthening regulation and governance of AI to ensure its safety and credibility. Fifth, enhance public participation in AI governance, involving social organizations, experts, scholars, and citizens, strengthening social and democratic supervision, and promoting the healthy development of AI.

(6) Strengthen the Role of AI in Supporting Enterprise Upgrading

As a highly penetrating and disruptive technology, AI holds significant importance for all aspects of the real economy and social life, serving as an important support for constructing a modern economic system and achieving high-quality development. Enterprises, as the basic units of social and economic activities, face the market and serve it, representing the most active innovative force. To achieve the transformation of traditional industries, the continuous growth of emerging industries, and the accelerated formation of a modern industrial system, a comprehensive approach is needed. First, leading enterprises should play a guiding role, creating an open, collaborative, and shared innovative ecosystem, particularly constructing a full-chain AI innovation ecosystem that includes foundational research to application promotion, facilitating the intelligent and high-end transformation of traditional enterprises. Second, increase support and investment in the AI industry, enhancing the capacity of specialized, innovative “little giants” through the establishment of AI innovation funds and supporting AI enterprises to go public. Third, establish open innovation platforms for AI, empowering enterprises, universities, and research institutes with platform resources and technology, accelerating the R&D and application of AI technology, and continuously improving technological innovation capabilities. Fourth, actively guide the promotion of data openness and sharing, facilitating data integration and intercommunication across various fields, forming a comprehensive data resource pool built and shared by society, and effectively advancing the deep integration of AI with the real economy.

(7) Strengthen AI Education and Talent Cultivation

The development of AI in China hinges on people. Cultivating high-quality talents with innovative and practical capabilities is crucial, which can be achieved by establishing AI majors, strengthening education and training in AI-related fields, and supporting the introduction of high-level talents, gradually constructing a talent cultivation system and curriculum system for AI, and enhancing the overall quality of AI talent in China. Meanwhile, the government should actively encourage collaboration between enterprises and universities, strengthening the practical aspects of AI talent cultivation, promoting deep integration of theory and practice. Additionally, a series of policy measures should be tailored, including tax incentives, R&D funding support, talent rewards, and high-level talent programs, to encourage innovation and entrepreneurship, providing strong support for the cultivation and development of AI talent.

(8) Strengthen International Exchanges and Cooperation in AI

China’s AI development should actively participate in the formulation of global AI standards and technological exchanges, enhancing cooperation with top international AI enterprises and institutions to promote global innovation and development of AI technology. By strengthening the construction of international cooperation mechanisms, establishing a transnational cooperation framework in the AI field, promoting the openness and sharing of AI technology globally, strengthening intellectual property protection in AI, establishing international AI technology standards and intellectual property protection mechanisms, and promoting the international application and exchange of AI technology. Actively participate in the formulation of international AI standards to promote the internationalization of AI standardization processes, enhancing China’s voice and status in international standard-setting. Strengthen international talent exchanges and cooperation, encouraging outstanding AI talent to study and exchange abroad and inviting international top talents to work and cooperate in China. Establish strategic partnerships with leading international AI enterprises and institutions to jointly carry out cooperation in technological R&D, application promotion, and talent cultivation, accelerating the formation of internationally competitive AI industrial clusters. Participate in international AI competitions and contests to enhance the international influence and competitiveness of China’s AI technology.

Author: Gao Wen, Academician of the Chinese Academy of Engineering, Director of Pengcheng Laboratory

Source: 2023 Issue 6 of “Central Committee Study”; “Current Affairs Report” WeChat Official Account

1. Welcome to Register for the 2024 Global Smart Education Conference

2. [Concept Document] GSE2024 Global Smart Education Conference V0.8

3. Call for Excellent Cases in Smart Education (2024)

4. Synthesis Report of Global Smart Education Conference 2023