AI platform to boost scientific research

Fudan University, in collaboration with Alibaba Cloud and China Telecom, has launched the largest cloud-based research and intelligent computing platform among Chinese universities.

Launched on Tuesday, the platform, Computing for the Future at Fudan (CFFF), combines public cloud technology with a multicluster system to provide supercomputing capabilities and support complex scientific research, according to the university.

"Fudan University is fully committed to nurturing innovation and promoting scientific research, and the CFFF platform is at the forefront of our efforts in research and innovation," said the university's president Jin Li.

"Using new AI technology, CFFF allows scientific productivity to be liberated to accelerate the discovery of new scientific principles and technological breakthroughs."

CFFF consists of two parts, Qiewen AI for Science and Jinsi HPC(high-performance computing) for Advanced Research. While Qiewen, hosted by Alibaba Cloud's data center in Ulaanqab, Inner Mongolia autonomous region, features the largest multilevel data storage cluster among Chinese universities, Jinsi, deployed on Fudan's campus, is the only public cloud computing system associated with a domestic college, according to Qi Yuan, dean of the Institute of Artificial Intelligence Innovation and Industry Research at Fudan.

With Qiewen and Jinsi connected to form a truly "super "computer, CFFF is capable of parallel and intelligent computing that supports large-model training involving up to a billion parameters, Qi said.

Recently, a 4.5-billion-parameter large model, based on the CFFF platform, was trained within a single day.

"CFFF's two clusters together provide strong computing resources for AI-enabled scientific research and new ways of development, and promote the construction of basic and interdisciplinary subjects," Qi said.

The CFFF platform, on the one hand, aims to provide stronger, more stable, flexible, efficient, and low-cost computing power support for Fudan's basic scientific research, said Jin, the university president who is also an academician of the Chinese Academy of Sciences.

On the other hand, based on the CFFF platform, the university aims to establish a group of world-class scientific large models, such as those for life sciences, atmospheric sciences, material sciences and integrated circuits, Jin added.

"The application of large-scale AI models for scientific research can help discover more complex problems and rules, and greatly accelerate exploration and breakthroughs in basic scientific research," Jin said.

Zhao Dongyuan, a professor of chemistry at Fudan and an academician of the Chinese Academy of Sciences, is enthusiastic about CFFF's promotion of AI for scientific research.

"The CFFF platform's launch is like having a large scientific instrument at our disposal, giving research a quantum leap. If we can utilize the platform to design materials through big data, it can save a lot of time, and we would have a deeper understanding of the material," he said.

Looking ahead, the platform aims to improve its computing capabilities and expand into more fields of research, according to Qi. In the future, CFFF will open up and collaborate with other research institutions as well as from Fudan University.

A global scientific intelligence contest — the first World Intelligence Science Competition — was also launched to attract global researchers.

linshujuan@chinadaily.com.cn