全球都用上中国免费大模型后，美国AI该怎么办？

Core Viewpoint - Eric Schmidt, former CEO of Google, expressed concerns that due to cost issues, most countries may ultimately adopt Chinese AI models, following Nvidia CEO Jensen Huang's statement that "China will win the AI race" [1][3]. Group 1: AI Model Landscape - Schmidt highlighted a "strange paradox" in the global AI landscape, where the largest AI models in the U.S. are closed-source and paid, while China's largest models are open-source and free [3]. - Open-source AI models allow free and public use and sharing, making them attractive to governments and countries lacking substantial funding, leading them to adopt Chinese models not necessarily because they are superior, but because they are free [3][4]. Group 2: Open Source vs. Closed Source - The early development of large models favored open-source as the mainstream choice, with even OpenAI initially releasing GPT-1 and GPT-2 as open-source [4]. - Supporters of open-source argue it promotes rapid technological development and offers significant cost advantages, while proponents of closed-source models claim higher security and advanced capabilities [5]. - The rise of Chinese open-source models has diminished the perceived security advantages of closed-source models, as open-source can be deployed locally, and performance gaps are closing [5]. Group 3: Chinese AI Model Advancements - Chinese models like DeepSeek, Alibaba's Qwen, and others have embraced open-source and consistently updated their large models, gaining popularity and raising concerns about the U.S. AI competitive edge [5][6]. - MiniMax's new open-source model, MiniMax-M2, ranked in the top five globally, while Kimi's K2 Thinking model reportedly surpassed GPT-5 in performance with a development cost of only $4.6 million [6]. - Chinese models are increasingly being adopted globally, with reports of Japanese companies using Qwen as a foundational technology [6][7]. Group 4: Global Implications - The cumulative download of Alibaba's Qwen surpassed that of Meta's Llama, indicating its popularity as an open-source model [7]. - The choice of a U.S. company to use a Chinese open-source model instead of its parent company's offerings reflects a shift in preference towards quality and cost-effectiveness [7]. - Concerns have been raised about the U.S. AI industry's reliance on closed-source strategies, which may pose significant risks if they fail [7][8]. - The rapid development of Chinese open-source models is reshaping the global AI competitive landscape, prompting fears that more countries may turn to Chinese models due to their advantages in openness, security, and cost [8].