China has launched what is likely the world's largest decentralized AI computing pool, alongside a high-speed data network that has been planned for more than a decade, according to a state-run newspaper.
The computational efficiency of a 2,000 km (1,243 mile) wide computing power pool formed through this network could reach 98 percent of that of a single data center, Liu Yunjie, a member of the Chinese Academy of Engineering and chief director of the project, said in a report in Science and Technology Daily on Thursday.
Behind the Nexperia crisis and high-tech tensions between China and the Netherlands | China Future Tech Webinar
Behind the Nexperia crisis and high-tech tensions between China and the Netherlands | China Future Tech Webinar
China's best computing centers are sparsely distributed across the country. This technology allows the two companies to work together seamlessly to rapidly develop the most powerful AI models and other cutting-edge technologies.
“The significance of this dedicated data highway is revolutionary for scenarios with extremely high real-time demands, such as AI large-scale model training, telemedicine, and industrial internet,” Liu told the media.
The Future Network Test Facility (FNTF) is the country's first major national science and technology infrastructure project in the field of information and communications. After more than 10 years of construction, it officially began operations on December 3rd.
“Training large models with hundreds of billions of parameters typically requires more than 500,000 iterations. With our deterministic network, each iteration takes only about 16 seconds. Without this capability, each iteration would take more than 20 seconds, potentially extending the entire training cycle by months,” he said.
