Kai-Fu Lee: China’s AI Large Models Lag Behind U.S. By 6-9 Months

Kai-Fu Lee: China's AI Large Models Lag Behind U.S. By 6-9 Months

On September 11, reports surfaced from CNBC highlighting insights from Kai-Fu Lee, the former head of Google China. He indicated that China’s artificial intelligence models might lag behind their American counterparts by at least six months, yet the application of AI in China could accelerate at a faster pace.

Lee discussed large language models, which are trained on vast datasets to generate and process text, images, and videos. He noted that while top Chinese companies may have AI models that are six to nine months behind those in the U.S., less advanced models in China could potentially be 15 months behind.

Speaking at the AVCJ China Private Equity Forum on Wednesday, Lee, a notable AI commentator and the author of “AI Superpowers: China, Silicon Valley, and the New World Order,” expressed optimism about the future of AI applications in China. He predicted that by early next year, the adoption rate of these applications will greatly outpace that in the U.S.

Lee pointed out that the cost of training effective AI models has significantly decreased. “China will inevitably create the best AI applications in the world,” he stated. However, he noted that it remains unclear whether these applications will be developed by large corporations or smaller firms.

His startup, 01.AI, is currently focusing on search applications. Lee mentioned that it could take five to eight years to elevate generative AI consumer applications to a new level—one that encompasses a multifunctional super app. He added that the industry may require entirely new devices, emphasizing that suitable devices should always be active and listening.

Chinese tech giants like Alibaba and Tencent have already introduced their own AI models and commercial products, while also investing in various AI startups. Beijing-based Shengshu Technology, backed by Alibaba’s Ant Group, announced on Wednesday a new feature for its text-to-video model, Vidu. This feature aims to enhance the consistency and integrity of key elements or characters in AI-generated videos, which can aid advertisers in creating promotional content for their products.

Vidu was launched earlier this year, with its basic tools accessible to the public, while advanced functionalities are available through subscription. Co-founder and CEO Tang Jiayu noted that several companies have shown interest in acquiring services from Shengshu, indicating that they are not just experimenting with the technology.

Exit mobile version