A recent analysis by the World Intellectual Property Organization (WIPO) reveals China’s preeminence in generative artificial intelligence (GenAI) patents and research publications. The “Patent Landscape Report – Generative Artificial Intelligence” indicates that from 2014 to 2023, China submitted 38,210 GenAI-related patent applications, outstripping the United States, which filed 6,276.
Increasing Patents and Publications
Over the past decade, the number of GenAI patents and scholarly articles has surged. The number of patent families increased from 733 in 2014 to over 14,000 by 2023. Similarly, scientific publications on GenAI rose from 116 in 2014 to over 34,000 in 2023 — with a notable quarter of all GenAI patents and more than 45 percent of GenAI scientific papers being published in the last year alone.
In the GenAI patent landscape, Chinese entities are prominent. Tencent ranks first with 2,074 patent applications, with Ping An Insurance and Baidu following. Additionally, the Chinese Academy of Sciences, IBM, Alibaba Group, Samsung Electronics, Alphabet, ByteDance, and Microsoft feature among the top ten. In overall patents granted, companies such as Tencent, Ping An Insurance, Baidu, and the Chinese Academy of Sciences are leaders.
Academic Impact and Citations
Although OpenAI ranks 335th in the number of published scientific articles with only 48 publications, it is the 13th-most cited source in AI research, reflecting significant academic influence through 11,816 citations.
Most GenAI patents involve image or video technologies, followed by text-related applications, and a category labeled “speech/voice/music.” Software/code patents are less frequent, with only 239 patent families represented. Since 2018, there has been notable growth in patents related to “molecules/genes/proteins,” highlighting increased interest in employing GenAI in the life sciences sector.
China’s dominant focus on AI research is in line with its national strategic goals, contrasting with the United States’ emphasis on AI safety and security. In 2023, U.S. President Biden signed an executive order addressing the safe management of AI risks, demonstrating a more cautious stance compared to China’s aggressive R&D approach.
Technological Focus
Generative adversarial networks (GANs) hold the highest number of patents, succeeded by variational autoencoders (VAEs) and large language models (LLMs). There has been substantial growth in patents for diffusion models and LLMs in recent years. Many GenAI patents deal with image/video data, essential for GANs, while text and speech/sound/music data are crucial for LLMs. Other data types include 3D imaging, chemical molecules/genes/proteins, and software/code.
Last Updated on November 7, 2024 3:42 pm CET