First Published on Zhejiang News
On March 31, our reporters paid a visit to Zhejiang Lab. At the time of our visit, people at the Lab had fully restarted their work and already been working for nearly 50 days. Researchers were racing against time to make progress for key projects, contributing to the development of an innovative intelligent computing system through different kinds of research activities such as designing new algorithms to teach sensors to perceive information like humans, developing new types of processing-in-memory chips, and inserting endogenous security genes into the Industrial Internet.
In the process of containing COVID-19, Zhejiang Province fully leveraged the strength of digital information technology and obtained effective results through precise and intelligent control. Zhejiang Lab President Zhu Shiqiang said that to prepare for a smart society where everything is connected, we need innovative theories and methods, architectural systems, and technical standards for intelligent computing. As a novel research establishment, Zhejiang Lab utilizes its institutional advantages to work with powerful organizations such as the Chinese Academy of Sciences, Zhejiang University, and Alibaba Group for collaborative innovation. It aspires to achieve innovation in every aspect from computing theories and architectures to computing equipment, computing systems, and computing services, providing comprehensive support for the construction of Digital China and Smart Society.
In the field of intelligent computing, a number of key projects launched early by the Lab have already made progress in certain phases. According to Research Fellow Liu Qi, who is leading the project on processing-in-memory chips, the project for chips of new architectures was started in August 2019. The joint research team includes members from Zhejiang Lab, the Institute of Microelectronics of the Chinese Academy of Sciences, and Zhejiang University. It took a little more than six months for the team to quickly complete a series of research tasks such as the overall design and simulation optimization of processing-in-memory chips, which are scheduled for tape-out this year.
The research on brain-like chips is being carried out simultaneously. “We model the structure of human brain neurons to design the chip, expecting to replicate the abilities of human brains on our chip and achieve highly efficient computing with low energy consumption,” said a researcher of the project. So far, the research team has built a flexible and configurable low-power hardware model of neurons, aiming to make it possible to put millions of neurons on a single chip.
An innovative intelligent computing system not only includes efficient intelligent computing capabilities, but also requires a reliable computing environment and ubiquitous computing services. According to Zhang Ruyun, Deputy Director of the Lab’s Research Center for Industrial Internet, research teams at the Center are now focusing on two core technologies of Industrial Internet: efficient computing and cyberspace endogenous security. Researchers address technical challenges and develop engineering solutions, endeavoring to insert endogenous security genes into a future network capable of connecting everything, find green and reliable solutions to intelligent manufacturing, and support the healthy development of the digital economy.
The project team for the Dubhe Artificial Intelligence Open Source Platform, which was previously launched by Zhejiang Lab, is preparing to fully start the platform’s open source services. As the platform shows superior efficiency for large-scale parallel computing as well as a better and more applicable automatic parallel mode, it can help to build an innovative ecosystem for the effective and efficient development of artificial intelligence.