On July 4th, 2024, at the World Artificial Intelligence Conference, the China Institute of Science and Technology Information, in partnership with Peking University, released the "2023 Global Artificial Intelligence Innovation Index Report." This marks the index's fourth consecutive unveiling at the conference.
The report reveals that the United States retains its leadership in global artificial intelligence development with a score of 74.71, while China ranks second with 52.69—nearly 15 points ahead of the UK in third place. Since the index's inception five years ago, the U.S. has consistently held the top position, while China has remained second since 2020.
Innovative technologies, particularly large models and generative AI, are fueling a new wave of advancements in AI research and industry. Despite a decline in newly established AI companies worldwide between 2018 and 2022, 2023 experienced a resurgence with a 21.5% year-on-year increase in new companies. Furthermore, the number of generative AI open-source projects on GitHub surged from approximately 17,000 in 2022 to around 60,000 in 2023, reflecting a 3.5-fold increase.
Zhao Zhiyun, Party Secretary of the China Institute of Science and Technology Information, emphasized that the report aims to offer a comprehensive overview of global AI innovation and governance trends while clearly defining China's standing. It evaluates 46 key countries using a structured three-tier indicator system across five dimensions: foundational support, resources and environment, and international cooperation.
Countries were categorized into four tiers based on their overall scores and rankings of primary indicators. The U.S. and China firmly occupy the top tier, significantly outperforming others. The UK retained its lead in the second tier for two consecutive years, yet the score gap with the U.S. expanded from 26 points in 2022 to 38 points in 2023, and from 9 points to 15 points with China.
Among second-tier nations, the UK excels in educational resources and high-quality research, while Japan leads in patents. Germany has strengths in AI venture capital and market scale. Notably, India and Saudi Arabia have ranked higher for three consecutive years, while Brazil and Indonesia made strides in scientific R&D and industry applications, particularly in AI open-source projects and venture capital.
Zhao identified four key trends in the global AI landscape. First, advancements in large models are significantly accelerating technological innovation, particularly in natural language processing and multimodal applications. In 2023, the number of machine learning models launched worldwide soared by 91.2%, marking the largest growth in nearly a decade. Second, the industry is markedly outpacing academia in model development, producing 176 independently developed models—three and a half times more than those created in academic settings. Third, the rapid growth of generative AI open-source projects illustrates the sector's momentum. Lastly, there is a notable uptick in the formation of new AI companies, reversing the decline observed from 2018 to 2022.
In terms of its AI landscape, China has successfully maintained its second-place global standing, achieving noteworthy progress in talent cultivation, research output, and industry development. The gap in high-level talent compared to the U.S. is narrowing, along with a promising increase in impactful research outputs, including AI open-source projects and investments. However, challenges persist, especially regarding data utilization and original innovation, particularly the lack of high-quality data. For example, Chinese datasets on Hugging Face represent only 5.1% of total open-source training datasets, significantly less than English datasets. Furthermore, China has contributed to only 68 of the 387 key machine learning models published globally from 2019 to 2023, compared to 262 from U.S. institutions.
Ranking
