In 2023, the general artificial intelligence model represented by ChatGPT has set off a new wave of development of the artificial intelligence industry in the world, and China's artificial intelligence large model market is showing a rapid growth trend of hundreds of "models" competing and changing with each passing day, and the market size is expected to reach 2.1 billion US dollars in 2023, a year-on-year increase of 110%. In 2024, the AI model will enter a critical period of empowering thousands of industries, giving birth to new models and new formats of future industries.
The artificial intelligence industry will maintain a rapid growth trend
In 2023, the scale of China's artificial intelligence core industry will reach 500 billion yuan, and the number of enterprises will exceed 4,400, of which the market size of artificial intelligence large models will reach 2.1 billion US dollars, a year-on-year increase of 110%. Looking forward to 2024, the artificial intelligence industry will maintain a rapid growth and development trend, gradually enter a new stage of deep empowerment of vertical industries and frontier fields, and vigorously promote the construction of new industrialization and high-quality economic development.
The scale of the artificial intelligence industry is growing rapidly. At present, the global artificial intelligence industry is in a period of rapid development. According to IDC data, the total global AI IT investment in 2022 will be US$128.8 billion, and the global AI IT total investment is expected to reach US$154 billion in 2023, a year-on-year increase of 196%。Looking forward to 2024, the development of the artificial intelligence industry will become the vane of global economic recovery. According to Sullivan Consulting**, it is estimated that the global artificial intelligence market will reach 615.8 billion US dollars in 2024, and China will exceed 799.3 billion yuan. Among the major subdivisions of artificial intelligence, large models, as a frontier hot spot, have the fastest growth rate. According to the report of the titanium ** international think tank, it is expected that the global artificial intelligence large model market will exceed 28 billion US dollars in 2024, and the market size of China's large model will reach 21.6 billion yuan, continuing to maintain a double-digit growth rate. The number and amount of investment and financing in the field of artificial intelligence will reach a new high, and the development trend will continue to improve.
The unveiling of the leaderboard has opened a new direction for the development of general artificial intelligence. In 2023, the unveiling of future industrial innovation tasks will be progressing steadily, covering a total of 52 unveiling topics in four key tracks: metaverse, humanoid robots, brain-computer interfaces, and general artificial intelligence. Looking forward to 2024, the unveiling units of various tasks of general artificial intelligence will further promote the industrialization process from basic research, technological innovation, product development to application landing in accordance with the requirements, and the integration and innovation of general artificial intelligence and other frontier fields is expected to play a key role, nurture and give birth to new industrial models and new formats in the future, empower new industrialization at a high level, and accelerate the cultivation of new quality productivity.
The training system for artificial intelligence professionals has been continuously improved. In 2023, major countries around the world have taken the cultivation of artificial intelligence technology talents as a key measure to enhance national competitiveness, and domestic universities have strengthened their deployment around core technologies and top talents, and the pace of building a new mechanism for general artificial intelligence talent training has been significantly accelerated. Looking forward to 2024, China's artificial intelligence talent training system will be further improved. In the interdisciplinary integration, the basic courses of artificial intelligence are integrated into the traditional curriculum system of other majors in gradients, and the key to promoting the deep empowerment of artificial intelligence in the industry-university-research ecosystem is to explore the integrated application from the perspective of different disciplines. At the same time, further strengthening the ethics and legal education of artificial intelligence, and cultivating artificial intelligence professionals with a keen sense of social responsibility are important guarantees for promoting the innovation and social development of China's artificial intelligence industry.
The trend of large models deeply empowering vertical industries and frontier fields is becoming more and more prominent. In 2023, domestic large models will show explosive growth for a while, and from January to July 2023 alone, a total of 64 large models will be released. According to incomplete statistics, as of November 2023, there are 188 domestic large models, including 27 general large models, and more than 20 large models have been recorded, most of which have been opened to the whole society. Based on the quantitative analysis of the relationship data of 2,200 AI backbone enterprises, China's AI has widely empowered 19 application fields such as smart finance, smart healthcare, smart manufacturing, and smart energy. Looking forward to 2024, large models will gradually expand to enable autonomous driving and embodied robots, and AI for Science will continue to empower scientific research, promote innovation in the field of science, improve research efficiency, and promote scientists to make new breakthroughs and achievements in exploring the field of "no man's land" and solving major problems.
The biggest challenge is that the barriers to entry in the industry have become higher
At present, the development of China's artificial intelligence industry is still facing the pressure and challenges of high industry entry thresholds, imperfect regulatory systems, low application rates in key industries, and disorderly competition in large models.
High demand for computing power + high cost investment" has raised the threshold for industry entry. On the one hand, to make AI models "bigger" needs to overcome computing power challenges and theoretical limitations, and making models bigger is not something that can be achieved by simply increasing the depth of neural networks and stacking artificial neurons. Models based on artificial neurons such as CNNs and RNNs need to adopt a serial structure, and the model training process needs to be executed sequentially, which cannot make full use of all computing resources. With the increase of the number of model parameters, the training time increases exponentially, and the convergence becomes more uncontrollable, making it more difficult to find the global optimal solution. On the other hand, the training cost of large AI models includes the cost of computing chips such as GPUs, server costs, standard cabinet costs, power consumption costs during the training period, and manpower input costs.
The balance between appropriate regulation and development promotion is challenging. The AI industry is currently in a period of rapid growth, and there are many uncertainties about its technological evolution and economic and social impact. On the one hand, AI has a strong innovative power and is expected to develop into a new engine of economic growth and greatly improve social well-being. On the other hand, issues such as ethics, safety, and negative externalities brought about by artificial intelligence have also frequently attracted social attention. What strength, in what way, and at what time to reasonably regulate artificial intelligence is a difficult problem that regulators need to focus on.
The application rate of artificial intelligence in key industries is low. On the one hand, the application of artificial intelligence to most traditional industries in China is still in a small-scale pilot, and the penetration rate of artificial intelligence application in top manufacturing enterprises is far behind that of Europe and the United States. On the other hand, there is a lack of typical application cases for large models in key application fields. At present, the initial exploratory application of large models in the industrial field is mainly focused on design assistance, quality improvement, equipment maintenance, etc., and there is no widespread implementation or formation of replicable industrial large models.
There is a risk of disorderly competition due to the excessive number of large models. On the one hand, homogenization of the training set leads to the homogenization of large models. At present, the training sets of many large models in China are generally publicly available English training sets, and the homogenization phenomenon is more prominent. On the other hand, the ultra-high training cost and the threshold for developing technology make it difficult for small and medium-sized enterprises to invest in such projects. The cost of high-quality training corpus and large-scale manual annotation determines that only large institutions or leading enterprises have the ability to develop corresponding large models, and the blind follow-up of growing enterprises will lead to a large number of investment failures and too many bubbles.
Create a cooperative ecosystem for the artificial intelligence industry
In view of the above problems, the research group suggests lowering the threshold for the use of computing power, optimizing the construction of the computing power system, accelerating the empowerment of thousands of industries to build an industry cooperation ecology, promoting the innovation of regulatory methods, improving the ability to respond to challenges, and reasonably regulating disorderly competition to promote the healthy development of large models.
Lower the threshold for the use of computing power and optimize the construction of the computing power system. The first is to strengthen the research and application of key core technologies such as distributed computing, quantization, memory optimization, and operator fusion, reduce the delay of large model inference, improve throughput, and reduce the demand for computing power. The second is to release the implementation plan of the computing power coupon to support the application of large models. Provide subsidy support for computing power coupons for enterprises, strive to help enterprises reduce the cost of using intelligent computing power, and fully support enterprises in key areas such as manufacturing to carry out the exploration and implementation of large models in the artificial intelligence industry. Third, it is recommended to promote the construction of intelligent computing power centers step by step, first pursue the universalization of computing power, reduce costs and improve utilization rate, and then gradually expand capacity.
Accelerate the empowerment of thousands of industries and create an industry cooperation ecosystem. The first is to guide artificial intelligence companies to carry out targeted cooperation with industry leaders. Based on the real business scenarios, data and real needs of the industry, industry enterprises develop core algorithms and pre-trained models, and jointly develop large application models. The second is to build a docking platform for artificial intelligence enterprises and industry enterprises. Build a docking platform for artificial intelligence enterprises and enterprises in manufacturing, medical, agriculture and other industries to help both parties realize the docking of technology, models, data, scenarios and other resources, and incubate application models in the industry. The third is to rely on the industrial Internet platform to create a large-scale model cooperation ecology between artificial intelligence enterprises and industry enterprises. Through the industrial Internet platform, the rapid docking of the two is realized, the online transaction service of artificial intelligence elements to ensure the security of algorithms, models and data is provided, and a standardized large model development environment is established for different industries.
Promote the innovation of regulatory means and improve the ability to respond to challenges. The first is to closely track the development trend of artificial intelligence technology, keep abreast of the latest progress and judge the social impact. The second is to encourage the synchronous innovation of relevant governance methods and technologies, and promote the innovation of artificial intelligence text classifiers and generative artificial intelligence detection methods. The third is to improve the technical literacy of regulatory talents. Fourth, establish a governance pattern of multi-party cooperation. Relevant competent departments should strengthen exchanges and cooperation with scientific research institutes and enterprises in the field of artificial intelligence.
Reasonably regulate disorderly competition and promote the healthy development of large models. The first is to explore the establishment of a large model training and filing mechanism that exceeds a certain scale of parameters, and guide the reduction of disorderly competition of homogeneous large models. The second is to explore the establishment of safety and reliability evaluation standards for large models, and put forward specific technical standards and evaluation criteria to ensure that all kinds of large models can operate stably and reliably in various application scenarios. The third is to classify and manage the application scenarios of large models, clarify the scope and limitations of the use of different types of large models, and avoid the negative impact caused by improper application.
Author丨CCID Think Tank Artificial Intelligence Industry Situation Analysis Research Group Editor丨Xu Hengmei Editor丨Maria Producer丨Lian Xiaodong.