On April 10, 2024, the National Industrial Information Security Development Research Center and the Electronic Intellectual Property Center of the Ministry of Industry and Information Technology released the "Analysis Report on the New Generation of Artificial Intelligence Patent Technology" (hereinafter referred to as the "Report").
The report illustrates the evolution of a new paradigm brought about by the new generation of AI from the "supply side", and the basic layer composed of computing chips and software frameworks, the model layer built by large models, and the new technology stack ecology formed by the construction of the application layer with "AI+" as the core will become the cornerstone of the development of the new generation of AI.
The report quantitatively reveals the latest development trend of the new generation of Chinese artificial intelligence technology patents.
First, the mainland has become an important competitor in the world's next-generation AI technology innovation. At present, the mainland artificial intelligence innovation chain is constantly improving, and the innovation of the AI basic layer and model layer is active, showing an innovation trend of multi-line development and ecological linkage. By the end of 2023, there will be 62,000 published patents in China for the new generation of AI basic layer (AI chips, AI frameworks) and model layers, of which nearly 20,000 are valid patents and 35,000 are under review. Since 2017, the average annual growth rate of patent applications has exceeded 43%, and natural language processing and computer vision account for the highest proportion.
Figure 1 Patent application trends and technology composition of the new generation of AI basic layer and model layer technology
Second, mainland innovation entities have shown a positive trend of innovation, and platform companies with a new generation of AI full-stack innovation capabilities, such as Baidu, Tencent, and Huawei, have become the key forces leading development and building a "smart base". According to the report, the number of new-generation AI patent applications filed by mainland applicants in China has formed a leading advantage, and the new generation of AI full-stack enterprises such as Baidu, Huawei, Tencent and Alibaba, which can effectively aggregate high-quality data, large-scale computing power and advanced algorithms, are becoming the "main force" of AI innovation layout. For example, Baidu relies on the Wenxin model and the mainland's first open-source, industrial-level deep learning platform (PaddlePaddle) to achieve an independent and controllable AI technology innovation ecosystem. At present, Baidu has about 1,432 patent applications for the new generation of AI, ranking first in China. According to the five indicator dimensions of "creativity", "protection", "utilization", "competitiveness" and "influence", the report analyzes the high-value of AI innovation subjects, and Baidu, Huawei, Tencent, Alibaba and other high-value patent reserves and innovation drivers are at the forefront.
Figure 2 Top 10 innovative entities (enterprises) in the field of patent application and authorization in the field of next-generation AI
Table 1 Patent Value Index of the Top 10 Innovative Entities of the New Generation of AI (Enterprises)
Third, the representative innovation subjects have shown strong strategic attention and innovation investment, and the overall development of innovation quality, innovation scale and innovation activity has been developed. Focusing on the development of the new generation of AI technology, the report integrates nine innovation factors to explore and construct the innovation potential measurement model of the main innovation subjects in order to explore the innovation potential energy and analyze the innovation driving form. First of all, in terms of innovation activity, Baidu, Tencent, Alibaba and other innovative entities have accounted for a very prominent proportion of patent applications in the past three years, showing a positive trend of innovation investment, and Huawei's overseas patents account for a more obvious proportion. Secondly, in terms of innovation scale, the four innovation entities all attach great importance to the quality of patent drafting, with an average of more than 17 claims, which is much higher than the industry average. The last four innovation subjects are first concerned about improving accuracy, followed by improving efficiency, and indicators such as "cost reduction" are still weak, which means that to promote innovation from fixed costs to "marginal", it still needs to be driven by application demand and led by innovation in basic theories and algorithms.
Figure 3 Comprehensive analysis of the patent innovation potential of representative enterprises at the basic layer and model layer of the new generation of AI
Fourth, AI native applications represent a new development direction for AI technology, and fields with high technology-driven utility, model training, and relatively good data foundation, such as generative search and intelligent documents, have become the first landing scenarios.
Generative search domains. Powered by a new generation of AI technology, information retrieval, mining, and recommendation are developing from traditional keyword matching to a new paradigm of generative search with "intent understanding" and "content generation". Generative search has the ability to integrate, refine, and concatenate information, which can better match user habits and respond to open-ended questions, which will greatly expand people's ability to obtain information. At present, Microsoft has launched the new Bing and new Edge browsers based on ChatGPT, Google has launched Google Bard, and Baidu has created Baidu search AI dialogue function based on the Wenxin Yiyan large language model. By the end of 2023, 1,837 patents have been published in generative search in mainland China, of which Baidu, Tencent, Alibaba and Ping An Technology have already deployed more patents, forming a relatively rich innovation reserve.
Figure 4 Top 10 Innovative Entities (Enterprises) in the field of patent application and authorization in the field of next-generation AI generative search
Table 2 Patent Value Index of the Top 10 Innovators in the Field of Next-Generation AI Generative Search (Enterprises)
The field of intelligent documents is a natural fit with model technology, and it is an application field with relatively mature technology research and development and fierce competition. Domestic Internet companies have integrated their own general large models and document products, and have launched intelligent documents, such as New Baidu Library, Kingsoft WPS AI, DingTalk Slash, etc. By the end of 2023, China has published more than 3,343 patents, mainly in major fields such as content generation, text processing, data utilization, and human-computer interaction. These technologies will improve the user experience by enabling a more accurate understanding of user needs and improving the precision and accuracy of the use of relevant documents. Based on the analysis of patent value, Baidu, Tencent, Alibaba, Kingsoft and other entities ranked high in terms of high-value patent reserves and innovation driving forces.
Table 3 Patent value index of the top 10 innovative entities in the field of next-generation AI intelligent documents (enterprises)
Smart industry. As the "main battlefield" of the application of the new generation of AI technology, there is a huge amount of accumulated non-process knowledge (experience), unstructured text, images and videos in the industry, and the combination of the new generation of AI technology has great potential. As of the end of 2023, there are 14,485 patents published in the smart industry in mainland China, most of which are concentrated in cross-modal, prediction models, detection models and other technical directions, and in terms of application innovation, they are mainly deployed in application scenarios such as intelligent manufacturing, coordinated optimization scheduling, defect detection methods, and capacity forecasting methods.
From the perspective of the distribution of high-value patents, Baidu, Tencent, Alibaba, and Huawei have taken the lead in the application and innovation of next-generation AI technology in combination with the industrial field. Huawei's Pangu Mine Model combines AI technology with mine application scenarios to achieve a number of innovations, such as "separation of operation and management from intelligent production", "data does not leave the park", and "learning and analysis of small samples".
Figure 5 Top 10 innovators (enterprises) in the field of patent application and authorization in the field of next-generation AI smart industry
Table 4 Patent value index of the top 10 innovative entities in the field of next-generation AI smart industry (enterprises)
Smart Finance. As a data-intensive field, the financial field has accumulated a large amount of high-quality data from user portraits to transaction information, from financial data to market dynamics, built a complete and stable digital infrastructure, and built a solid foundation for data and computing power, becoming the first important scenario for the new generation of AI to land. By the end of 2023, there will be 33,301 public patents in smart finance China, and the effective patents of the top 10 innovative entities such as Baidu and Tencent account for less than half of them, indicating that this field is in a development trend where many innovative entities have entered the market and are actively innovating.
Figure 6 Top 10 innovators (enterprises) in the field of next-generation AI smart finance patent applications and authorizations
Focusing on promoting the innovation of a new generation of artificial intelligence technology, the report believes that we should seize the precious opportunities of scientific and technological and industrial transformation, accelerate the empowerment of independent innovation, and accelerate the formation of new quality productivity.
Through patent analysis, it can be seen that the attributes of the new generation of AI technology stack as "infrastructure" are gradually complete, and the innovation of the basic layer and model layer is gradually maturing. The report suggests that in combination with the strategy of building a strong country with intellectual property, we should further improve the whole-process and embedded intellectual property public service system, smooth the supply and demand channels of industrialization services such as patent licensing and transfer, value evaluation, and pledge financing, promote the transformation and application of AI patents, give full play to the "action" of intellectual property rights, continue to provide strong backing for technological innovation, and help the innovation and development of a new generation of AI technology.
This year's government work report for the first time put forward the "artificial intelligence+" action, with the help of policies, markets and technology dividends, the new generation of AI technology innovation will accelerate the development of wide coverage, deep penetration, and strong integration of inclusive empowerment. The new generation of AI technology has opened the prelude to the development of general artificial intelligence, which is an important driving force for a new round of scientific and technological and industrial transformation, and is of great significance for seizing the development opportunities in the digital economy era and promoting the formation of new quality productivity.
Source: Electronic Intellectual Property Center of the Ministry of Industry and Information Technology