哈佛老徐：2026年初再看咱们与漂亮国，AI的胜负真的在大模型吗？

Core Viewpoint - The article discusses the competitive landscape of AI technology between Eastern and Western countries, emphasizing the importance of a comprehensive evaluation across five layers of the AI technology stack rather than focusing solely on large models like ChatGPT and DeepSeek [11]. Group 1: Energy - The energy capacity in Eastern countries is approximately double that of Western countries, which is crucial for supporting the rapid growth of AI technology [13]. - The demand for computational power in AI is growing exponentially, outpacing traditional power grid capabilities [13]. Group 2: Chips and Software - There is a significant gap in chip technology and underlying software ecosystems, with Western countries currently leading in hardware but facing challenges in energy supply and software compatibility [14][16]. - The software ecosystem, particularly NVIDIA's CUDA, is critical for maximizing GPU performance, and replicating this ecosystem is a monumental task [19]. Group 3: Infrastructure - Eastern countries have a distinct advantage in building AI infrastructure, capable of establishing data centers much faster than their Western counterparts [23]. Group 4: Large Models - Western countries lead in closed-source AI models by approximately six months, with major players like OpenAI and Google at the forefront [25]. - In contrast, Eastern countries have a stronger position in open-source models, which are essential for fostering innovation and accessibility in AI [27]. Group 5: AI Applications - Eastern countries are perceived to be ahead in AI applications due to higher societal acceptance, a larger pool of AI researchers, and faster implementation of AI technologies [29]. Group 6: Robotics - Eastern countries excel in robotics, meeting all three critical criteria: demand, AI capability, and mechatronics engineering skills, making them unique in this field [31]. Group 7: AI Risks - The primary concern regarding AI is not job loss but the transformation of tasks, with many tasks being automated while the overall job roles evolve [34].