Comment of China's former deputy minister for science and technology on AI

Recently, Li Meng, a member of the 14th National Committee of the Chinese People's Political Consultative Conference (CPPCC), the Committee of Education, Science, Culture, Health and Sports, and former deputy minister of the Ministry of Science and Technology, comment on China’s AI development.

If there's one tech field with the most disruptive impact in the future, it's gotta be artificial intelligence. General Secretary Xi Jinping has shown keen insight, accurately grasping the trends in AI development and strategically providing a series of important directives and deployments for AI research and application. In 2017, the central government approved releasing the National New Generation Artificial Intelligence Development Plan, establishing major projects in the new generation of AI, leveraging the plan to lead and accelerate research breakthroughs and industrial applications. After years of effort, China's AI development has shown a positive trend, with overall and systematic improvements in innovation capabilities, placing us in the world's top tier.

AI development has entered a new stage, with the wave of generative AI led by large models sweeping globally. This is primarily manifested in several aspects: (1) Large-scale pre-trained models have become the new mainstream direction for technological innovation, changing the research paradigm in various AI subfields like speech, vision, and natural language processing; (2) Problems being solved are gradually shifting from primarily perceptual intelligence to cognitive intelligence, with breakthroughs in intent understanding and logical reasoning capabilities breaking through ceilings; (3) The establishment of generative AI models has begun, expanding AI capabilities from analytical and decision-making to generative AI; (4) The process towards general AI has started, showing certain cross-task versatility and demonstrating the basic characteristics of general AI; (5) The first phenomenon-level AI products have emerged, with ChatGPT's release leading to user growth surpassing internet products like Twitter and TikTok; (6) AI governance has become a hot topic domestically and internationally, with security governance transitioning from advocacy of principles and content discussion to practical implementation.

The global impact of AI is becoming increasingly widespread and profound, driving a revolutionary upgrade in productivity and production efficiency, triggering disruptive changes in various fields like economy, society, culture, and military. In this wave of AI, China is not just a participant but also a leader. As one of the few countries capable of "competing on the same stage" with the US in AI, China has encountered new development opportunities and daunting challenges in AI.

Based on China's national conditions and development stage, promoting AI development should adhere to a "two-legged" approach. On the one hand, it strengthens the foundational innovation and capacity building of AI, elevating the multi-modal universality of our country's basic large models and fostering more intelligent capabilities with stronger cognition and interpretability; on the other hand, enhancing the multi-scenario applications of large models, especially in deep applications in industrial vertical fields, accumulating data rapidly through application feedback, iterating models, and innovating algorithms.

AI permeates and empowers economic and social development, leading to new quality productivity in our country. With a massive market and the most complete industrial system, China possesses unique advantages in AI in many industrial sectors, such as rich scenarios and massive data, providing vast space for the integration, application, and iterative upgrades of new AI technologies, which is an important path for many Chinese industries to surpass international advanced levels, with significant implications for driving industrial high-end, intelligent, and green transformations.

One is to accelerate AI-driven scientific research, discovering and nurturing new industrial growth areas. AI has shown transformative potential in accelerating major scientific research and knowledge discovery in the research field, becoming important in protein structure prediction, new material discovery, and driving a batch of new technology-driven emerging industries in new drug development, biological breeding, and other directions. We should pay more attention to the role of AI in research in fields such as mathematics, physics, chemistry, biology, astronomy, genetics, drugs, and materials, integrate projects, platforms, and talent resources, form policy synergy, support researchers to engage in relevant research and build a new paradigm of AI-supported basic and frontier scientific research. It is especially important to increase layout efforts in new drug development, material research and development, biological breeding, and more effectively discover new genes, new targets, new drugs, new materials, new varieties, and new compounds, driving the cultivation of new formats and creating new engines in biopharmaceutical, smart agriculture, and new materials industries, revitalizing various industries.

The second is to promote the landing of large model technology and shape scene-driven new industrial advantages. The development of large models has become the main direction of current AI research, application, and governance. The comprehensive penetration of AI into the next generation of industrial and living systems relies on the scenario-based application of a series of technical systems. Relevant departments and regions have made deployments, with the key being implementation in specific enterprises and industries. In the past two years, relevant departments and regions have made layouts in directions such as smart farms, marine ranches, intelligent ports, smart mines, smart factories, autonomous driving, smart education, smart finance, smart supply chains, and smart manufacturing, achieving remarkable results. Continuous improvement, promotion, and priority should be given to smart manufacturing and smart mines. Encourage leading enterprises to propose scene-driven overall AI solutions and accelerate the exploration of commercial models for the industrial landing of large models. Support many small and medium-sized enterprises to form specific arrangements in segmented fields, using new technologies such as content generation, autonomous design, and knowledge assistants to reshape intelligent business processes. Play the role of national AI demonstration zones, accelerate the creation of a batch of replicable and promotable AI demonstration scenarios, fully unleash the first-mover advantage of AI empowering economic and social development, and reshape various industries.

Third, improve scene-driven ecosystems to promote innovative AI applications. Drive the integration of supercomputing, intelligent computing, and quantum computing to enhance computing power. Scientifically plan and build public computing platforms and networks. Accelerate the supply of affordable computing power for young people in universities and research institutions, encouraging their participation in large model training and entrepreneurship. Developing various computing centres and networks should prioritize accessibility for small and medium-sized enterprises, researchers, and entrepreneurs. Open up public data organised, promote the sharing and utilization of industry data resources and basic scientific datasets, and create more high-quality training datasets. Support the opening of AI application scenarios in local and traditional industries, encourage the organization of large model innovation application competitions to explore more application areas and specific directions, using innovative AI technology applications as a driving force for technical breakthroughs and continuous improvement. Innovate policies and institutional mechanisms, strengthen collaboration and technological integration in scene research and development, accelerate AI technology breakthroughs, equipment upgrades, product development, optimization of industrial chains, and industrial cultivation.

Fourth, focus on developing humanoid robots to usher in the "iPhone moment" for robotics. Develop intelligent carriers, promote the integration of robots with multimodal large models, implant brains with stronger logical reasoning capabilities into robots, and accelerate the advancement of China's robotics industry. The emergence of intelligent humanoid robots will bring about the "iPhone moment" in the robotics industry.

Fifth, the focus is on developing intelligent green technologies and green intelligent solutions, creating a synergy between intelligence and sustainability. Some say the ultimate goal of artificial intelligence is nuclear fusion, but in reality, it's about the massive energy consumption AI entails. Issues like wind, solar, and water energy being wasted due to grid limitations are prevalent in developing various forms of renewable energy sources. The development of green and intelligent technologies faces certain constraints in reality. It's crucial to prioritize the development of green computing, establish a green computing system, optimize models, algorithms, and software, design green computing models, and reduce energy consumption. Initiatives like algorithm competitions can encourage young tech enthusiasts to develop eco-friendly algorithms, reducing the energy consumption of cloud and edge computing users. Managing various renewable energy sources in green and low-carbon development is challenging. There's a need to develop larger-scale, smarter, and more resilient smart grids to increase the integration of renewable energy sources. Enhancing the intelligence levels of power generation, transmission, and consumption while reducing energy consumption in power production is essential.

Sixth, there's a need to enhance ethical governance capabilities to keep pace with the rapid advancements and ensure a balance between innovation incentives and effective regulations. The rapid adoption of new AI technologies, like large models, poses challenges to societal ethics and legal regulatory systems, leading to increasing calls for strengthened ethical governance in the field of AI. Enhancing governance capabilities and aligning them with scientific exploration and industrial applications is a shared responsibility for regulatory bodies, businesses, and the scientific community, especially for leading enterprises and publicly traded companies. Establishing an agile governance system that adapts to the characteristics of AI development comprehensively and cautiously is crucial to promoting healthy and orderly R&D and industrialization. Implementing classified governance for different risk scenarios and accelerating the improvement of ethical norms, safety, and standards systems in key areas like intelligent diagnostics and autonomous driving will provide businesses with clearer and more stable regulatory guidance, thus promoting high-quality industrial development efficiently.