Strategist from China Proposes Development of AI Software to Reduce Dependency on Western Countries and Outperform NVIDIA’s CUDA.

Scientist Li Guojie believes that China needs to develop its own AI software technology instead of relying on Western technologies like NVIDIA’s CUDA. While Chinese companies like Huawei, Hygon, and Cambricon have AI chips with capabilities similar to NVIDIA’s, Li argues that China cannot completely stop relying on NVIDIA. China must create software systems that surpass CUDA, not just hardware development but also suitable software integration.

Additionally, Li questions OpenAI’s reliance on scaling principles for AI efficiency, where better data training and increased computing resources lead to improved performance. This concept is not scientifically proven but is based on risky assumptions, as evidenced by the delayed introduction of GPT-5.

Li suggests promoting the use of AI models and chips in everyday devices like smartphones and computers to increase China’s self-sufficiency in AI technology. This will help the AI industry in China grow further.

TLDR: Scientist Li Guojie advocates for China to develop its own AI software technology and promote the usage of AI models in everyday devices to reduce reliance on Western technologies like NVIDIA’s CUDA.