Artificial intelligence has become the core driver of the fourth industrial revolution. In the era of digital economy, the artificial intelligence industry plays a pivotal role in unleashing economic potential and promoting scientific and technological innovation, and artificial intelligence is the key engine leading the development of new quality productivity.
According to data from the China Academy of Information and Communications Technology, the scale of the mainland's core artificial intelligence industry will reach 578.7 billion yuan in 2023, and the number of related enterprises will reach 4,482. The AI industry chain covers key upstream and downstream links such as chips, algorithms, data, platforms, and applications. Across the globe, China has become one of the superpowers in the field of artificial intelligence.
As the backbone of China's economy, central and state-owned enterprises are leading and promoting the rapid development of the artificial intelligence industry. In February 2024, the State-owned Assets Supervision and Administration Commission of the State Council held a special promotion meeting on artificial intelligence for central enterprises under the theme of "AI Empowerment and Industrial Renewal". The meeting demanded that the central enterprises should speed up the layout and development of intelligent industries, and accelerate the construction of a number of intelligent computing centers.
Today, the opportunities presented by large models and generative AI are comparable to those of the Industrial Revolution, and every company and organization is thinking about how to evolve and use AI as a new technology to improve their competitiveness.
Chinese's artificial intelligence industry ranks first in the world
Data, computing power, and algorithms are the three key elements of the development of the AI industry.
According to the Digital China Development Report (2022) released by the Cyberspace Administration of China, China's data production reached 8.1 zettabytes in 2022, a year-on-year increase of 22.7%, accounting for 10.5% of the global total, ranking second in the world, and by the end of 2022, China's data storage capacity reached 724.5 exabytes, a year-on-year increase of 21.1%, accounting for 14.4% of the global total.
Computing power is at the heart of AI competition. From the perspective of computing power scale, the Digital China Development Report (2022) shows that by the end of 2022, the total scale of China's data center racks exceeded 6.5 million standard racks, with an average annual growth rate of more than 30% in the past five years, the total scale of computing power of data centers in use exceeded 180EFLOPS, ranking second in the world, the total scale of storage power exceeded 1000EB (1 trillion GB), the one-way network delay between national hub nodes was reduced to less than 20 milliseconds, and the scale of the core computing power industry reached 1.8 trillion yuan.
In July 2023, IDC, Inspur Information (000977), and the Global Industry Research Institute of Tsinghua University jointly released the "2022-2023 Global Computing Power Index Evaluation Report", which shows that among the 15 sample countries, China's computing power index ranks second in the world and is in the leading position, and China's overall server market will still maintain a positive growth of 6.9% in 2022, reaching 27 billion US dollars, accounting for 25% of the global market, second only to the United States and ranking second in the world.
In terms of the global pattern, the research report released by Zheshang Securities (601878) shows that global artificial intelligence companies show a "Sino-US dominant" pattern. As of the third quarter of 2023, there are 29,542 AI companies in the world, with China and the United States accounting for nearly half of the global total, with 9,914 in the United States (34%) and 4,469 in China (15%).
Gao Wen, an academician of the Chinese Academy of Engineering, believes that China has become one of the superpowers in the field of artificial intelligence. China has been at the forefront of the world in some key core technologies, such as face and voice recognition technology. At the same time, China's artificial intelligence development has a high degree of penetration in the combination of various industries, such as Baidu's unmanned driving, Alibaba's City Brain, Tencent's intelligent healthcare, iFLYTEK (002230)'s voice recognition, SenseTime's image and video processing and other open platforms, as well as Huawei, Cambrian, Hikvision (002415) and other real economy platforms have been created.
China Mobile (600941), PetroChina, China Energy Group, State Grid and other central enterprises are also continuing to make efforts to promote the application of artificial intelligence in business scenarios. On April 2, China Mobile announced that China Mobile's Jiutian AI large model can officially provide generative artificial intelligence services to the outside world, and it has also become the first large model developed by a central enterprise that has passed the dual filing of the national "generative artificial intelligence service filing" and "domestic deep synthesis service algorithm filing".
Most researchers believe that accelerating the development of artificial intelligence is an inevitable requirement for state-owned central enterprises to play their functional mission, seize strategic opportunities, cultivate new quality productivity, and promote high-quality development. On the other hand, the central enterprises have the advantages of large demand, complete industrial support, and many application scenarios, and the entry of central enterprises into the bureau is conducive to accelerating the improvement of the basic foundation for the development of the artificial intelligence industry in the mainland, promoting the formation of industrial demonstration projects, and providing direction for the innovation of other enterprises.
China's development of artificial intelligence should adhere to the "two-legged walk"
From the perspective of the industrial chain, the upstream basic layer of the AI industry is computing power and data, including servers, chips, optical modules, switches, data centers, liquid cooling equipment, etc., the midstream technology layer is a large model platform built on the basis of computing power and data, with algorithms as the core capabilities, and deep learning, natural language processing, transfer learning and other key technologies, and the downstream is the application layer, including games, media, film and television, finance, office, medical and other industrial scenarios.
After investigating more than a dozen artificial intelligence companies, young economist Jia Ming found that in the long run, only from a technical point of view, the essence of artificial intelligence competition is the competition of computing power, and behind the computing power is the chip. Although the mainland has increased its R&D investment in the field of semiconductors, it is still difficult for China to make subversive breakthroughs in the field of chip manufacturing in the short term.
Data is an important basic strategic resource for the development of the artificial intelligence industry, and China is one of the most data-rich countries in the world, but there is also a lack of high-quality data.
Large models are the source of the core competitiveness of the artificial intelligence industry chain, and at present, major technology companies are actively investing, developing and launching their own large models. For example, in June 2023, the iFLYTEK Xinghuo cognitive model developed by iFLYTEK passed the first official credible AIGC large model basic capability (function) evaluation organized by the China Academy of Information and Communications Technology, and was certified for all functional items.
Kunlun Wanwei (300418) released the first AI search engine in China in August 2023 - Tiangong AI Search, which has now formed six AI business matrices: AI large model, AI search, AI game, AI music, AI animation, and AI social networking.
Sugon (603019) has developed a variety of servers and workstations based on domestic processors in the field of high-end computing, all of which have passed the product quality test of the national laboratory, and completed the design of storage IO modules for the field of cloud computing and artificial intelligence in terms of localized components.
With the rapid development of the industry, the related applications of general large models and vertical large models are being combined with scenarios to accelerate their entry into production and life. For example, the Shanghai Artificial Intelligence Laboratory has taken the lead in developing the AI weather forecasting model "Fengwu", China National Petroleum Logging Co., Ltd. and Huawei have developed an L2 application model for geological stratification, storage division, parameter calculation, and oil and gas identification, and Shandong Energy Group has jointly released the industry's first mine model, which has been applied in nine professional fields, including coal mining, tunneling, transportation, and safety supervision.
According to Liu Xingliang, a member of the Information and Communication Economy Expert Committee of the Ministry of Industry and Information Technology, at present, the application of artificial intelligence in most industries is still in the early stage or experimental stage. Looking at the world, if the research and development of large models in all walks of life is a long march, we have actually taken the first step.
Regarding the current development of the domestic artificial intelligence industry, Robin Li, founder, chairman and CEO of Baidu, said that AI-native applications are becoming a major trend. There are many large models in China, but there are very few native AI applications developed based on large models. When we look at foreign countries, in addition to dozens of basic large models, there are actually thousands of AI native applications, which are not available in the Chinese market.
"I think that the sign of mankind entering the AI era is not the generation of a lot of large models, but the generation of a lot of AI native applications. The large model itself is a basic foundation, similar to an operating system, and developers should rely on a small number of large models to develop a variety of native applications. Constantly reiterating the development of basic large models is a great waste of social resources. In the AI-native era, we need 1 million AI-native applications, but we don't need 1 million large models. If our industrial policies can encourage the native application of AI based on large models, we will be able to build a thriving AI ecosystem and promote a new round of economic growth. Robin Li said.
"On the other hand, due to the lack of intelligence emergence, the value of dedicated large models is actually very limited. Robin Li believes that many industries and enterprises are now buying cards, hoarding chips, and establishing intelligent computing centers, wanting to train their own dedicated large models from scratch. As everyone knows, the large model trained in this way does not have the ability to emerge intelligence. The industrialization model of the large model should combine the general capabilities of the basic model with the professional knowledge of the industry. That is to say, the large model sets the small model, the special small model responds quickly, the cost is low, and the large model is more intelligent, which can be used to cover the bottom. A powerful basic model will drive the explosion of AI native applications. China has a leading basic model, which is a solid foundation for the development of AI native applications and an underlying capability.
Li Meng, director of the State Administration of Foreign Experts Affairs, believes that China should adhere to the "two-legged approach" in the development of artificial intelligence: on the one hand, strengthen the underlying innovation and capacity building of artificial intelligence, improve the multimodal generalization level of large models, and promote the emergence of intelligence with stronger cognitive ability and more explainability;
Improving the ability of AI to innovate is a top priority
"Improving the ability of artificial intelligence in the real sense of the original innovation is the top priority. Following the development model of contemporary AI hotspots is neither sustainable nor can it form a leading opportunity for the future. The mainland still needs to make up for its shortcomings in computing power and data infrastructure to promote differentiated development, especially to support innovative R&D institutions and start-ups to strengthen the original innovation of artificial intelligence. Speaking of Chinese's artificial intelligence industry, Zeng Yi, a researcher at the Institute of Automation of the Chinese Academy of Sciences and an expert from the United Nations High-level Advisory Body on Artificial Intelligence, said.
The core element of disruptive technological innovation is "people", which is also the most obvious shortcoming of China's development of artificial intelligence industry. According to statistics, the number of top AI talents in China is only 20% of that of the United States.
Zhou Luming, a former deputy director of the Shenzhen Science and Technology Bureau and now president of the National Offshore Innovation and Entrepreneurship Base for Overseas Talents (Shenzhen), has a similar view.
Recently, the results of artificial intelligence source innovation in the United States have come out one after another, following GPT (Generative Pre-Trained Transformer, generative pre-trained model), SORA (artificial intelligence research company OpenAI released artificial intelligence Wensheng video model) was born, each round has stirred the nerves of China's scientific research and industry, and also caused us to worry about the backwardness of core technologies.
Zhou Luming believes that the concern about the backwardness of core technology is partly due to our one-sided understanding of the formation path of core technology, thinking that the path model of achievement transformation is the only way for the formation of core technology, and this path dependence has dominated China's science and technology strategy policy designers, scientists and some entrepreneurs for decades. After the advent of GPT, there was a "100-model war" in the domestic artificial intelligence industry, but GPT soon had a tendency to be replaced by new things. If the development of Chinese artificial intelligence completely follows the United States, the result may be very bad.
"In fact, the vast majority of the core technologies of modern industries are gradually formed by enterprises to carry out research and development on the application side around industrial problems and market demand. In the past 20 years, in the development process of the electric vehicle industry, domestic industry, university and research have carried out industry-university-research collaboration around the real problems of the industry, solved various problems raised by the industry in the way of parallel division of labor, and formed a world-leading core technology in the key links of battery, electronic control and other systems. This is a very convincing case for correcting the path dependence of domestic core technology development. Zhou Luming said.
In Zhou Luming's view, the explosive development in the field of artificial intelligence is producing a series of disruptive changes, and the problem we really need to realize is that the real gap between China and the United States in source innovation is not technology, knowledge, and money, but a gap at the methodological level. This round of American artificial intelligence innovation has seen some new trends in methodology: it no longer organizes source innovation along the linear way of "basic research-applied basic research-development research", but basic research is directly aimed at industrial problems, and the whole innovation chain of basic research, applied basic research, and product research is directly compressed in one organization (OPEN AI), and industry-university-research is condensed in a product form to carry out collaboration. This trend is worth studying in China, and if our innovation strategy in the field of artificial intelligence is still organized by the model of achievement transformation, it will not only waste money, but also miss the opportunity for industrial development. China should give full play to the advantages of the application side, explore the real problems of the industry, and organize parallel collaborative innovation of industry, university and research.
From a technical point of view, artificial intelligence has become one of the most active areas in the development of information technology. Driven by the wave of innovation represented by large models, artificial intelligence technology, industry, application and other links will usher in a critical period of rapid iterative evolution and exploration and breakthrough. In this regard, Yu Xiaohui, president of the China Academy of Information and Communications Technology, said that at this stage, the development focus of the artificial intelligence industry has shifted from single-point technological breakthroughs in software and hardware to system collaboration, and more emphasis needs to be placed on the all-round coordinated development of applications, algorithms, key software stacks, and underlying hardware in the future. At the same time, it is necessary to make arrangements in advance in terms of policy guidance, standards and norms, and regulatory means to ensure that artificial intelligence technologies such as large models play a greater role in the high-quality economic and social development of the mainland.
Source: Enterprise Observer Author: Zhang Ning
