From the capability breakthroughs of China's DeepSeek model to the Trump administration's tariff policies, the U.S. artificial intelligence industry is striving to advance into the next generation of AI frontiers while absorbing shocks from various directions.
Why will the flow of private investment become a key weather vane? Where lies the next critical battlefield in the U.S.-China AI competition? Senior researcher Chu Chen-Tso, currently a visiting scholar at Stanford University, shares first-hand local analysis.
From the private sector to government officials, maintaining leadership in AI research and development is a unanimous goal for the United States. However, there is little consensus on how to achieve it.
The breakthroughs demonstrated by China's self-developed DeepSeek-R1 and the earlier V3 have raised concerns in the U.S. Even if this does not signify a failure of the export control policies of the Joe Biden era, it indicates that existing measures are insufficient to consolidate the lead the U.S. hopes to maintain.
The Donald Trump administration officially revoked the "AI diffusion rule" proposed by former President Biden on the eve of its implementation. Meanwhile, a press release from the Bureau of Industry and Security (BIS) indicates that future priorities will focus on containing Huawei's Ascend chips and continuing to prevent U.S. chips from being used to train Chinese models. This reflects the significant magnitude of progress in Huawei's AI chip performance.
What constitutes an ideal technology policy has become the focal point of debate. But is the U.S. focusing its attention on the right place? Is its vigilance well-founded?
Did Export Controls Fail? The Gap in US-China AI Capabilities Narrows
The AI landscape changes rapidly, but to examine the entire panorama of the U.S.-China AI race, the 2025 AI Index Report, released in April by the Stanford Institute for Human-Centred AI (HAI), serves as an ideal source.
Chu Chen-Tso, a lawyer currently visiting Stanford Law School, noted that the report reveals that the U.S. and China each hold leads in different aspects of the AI race. Specializing in AI law, technical regulations, and data governance, Chu previously served as General Counsel for companies such as Sharp and Hongzhun, and is currently the Secretary-General of the International Research Foundation for Artificial Intelligence Law.
The area where the United States has established the widest gap is in private investment in the AI sector. In 2024, the U.S. invested $109.1 billion in private capital in AI, nearly 12 times that of second-place China ($9.3 billion), capturing a significant share of the global corporate AI investment total of $252.3 billion.
Frontier models are another area of U.S. dominance. In 2024, U.S. institutions released 40 frontier models. By comparison, China published 15, while Europe as a whole contributed three.
China, however, prevails in the quantity of AI research papers and patents, although among the top 100 most-cited AI papers over the past three years, U.S. papers still hold the highest proportion. Additionally, China's leadership in industrial robotics remains solid.
Chu Chen-Tso noted that what makes the U.S. particularly nervous is the rapid narrowing of the gap in model capabilities between the two nations. At least across several mainstream benchmarks, the disparity has significantly shrunk.
Taking Massive Multitask Language Understanding (MMLU) as an example, a 17.5 percentage point gap existed between the two sides at the end of 2023; by the end of 2024, only a 0.3 percentage point difference remained.
At the same time, the capabilities of open-source models are quickly approaching those of closed-source models. In early 2024, top closed-source models held an 8% lead over open-source models on the Chatbot Arena Leaderboard. By early 2025, that gap had narrowed to just 1.7%
.
Read the full article: https://www.gvm.com.tw/article/121313
SSL 256bit transmission encryption mechanism