Song of laborers| Chen Yunji: the scenery of the climber's life

"Make computers smarter", "Make machines better serve humans"... This is the dream of Chen Yunji, the winner of the National May Day Labor Medal and a researcher at the Institute of Computing Technology of the Chinese Academy of Sciences.

Over the years, he has devoted himself to scientific research, and the results have been frequent; scientific research is not easy, but he enjoys it. He said: "The future of artificial intelligence is a sea of stars, and I will devote all my energy to exploring and struggling." ”

Chen Yunji showed off the world's first deep learning processor chip "Cambrian 1" developed by his team. Photo by Xinhua News Agency reporter Li Xin

Thin-rimmed glasses, light blue shirt, Chen Yunji in front of him, elegant and slightly funny, full of vitality. Unexpectedly, this "post-80s" young scientist has been working in the field of domestic processor chip development for more than 20 years.

Embarking on the path of scientific research begins with interest and curiosity. In his childhood, Chen Yunji was deeply attracted by the books on the shelves at home, whether it was engineering or history; in the third year of college, he held an interest in chip research, knocked on the doors of all the laboratories in the computer department, and finally won the opportunity to participate in scientific research, although he did some "chores", he deeply felt the great charm of computers.

Since then, Chen Yunji has become the youngest member of the R&D team of Loongson, a domestic general-purpose processor from the Institute of Computing of the Chinese Academy of Sciences. At the age of 25, he became the main architect of the 8-core Loongson 3. Loongson 3 is different from Loongson No. 1 and No. 2, and has evolved from single-core to multi-core. "In the past, one table of dishes was given to one table of guests, and now one table of dishes is served to eight tables of guests." Chen Yunji used an analogy.

In 2007, Chen Yunji (first from right) took a group photo with members of the Loongson team. Courtesy of respondents

The design difficulty is greatly improved. Under the condition of very scarce resources, Chen Yunji proposed a fast and complete storage consistency verification method through continuous exploration, and solved the problems in the design verification of multi-core processors.

"Scientific research is your own interest, ideal and belief, and it is also a lifelong pursuit." Chen Yunji said.

"What should the intelligent computing system of the future look like?" More than a decade ago, Chen Yunji pondered this question.

At that time, the scale of intelligent algorithms was growing rapidly, but the development of hardware systems was very slow. "Artificial intelligence applications such as face recognition and automatic translation require deep learning algorithms as support, which requires powerful computing power, and traditional chips are far from meeting the demand." Chen Yunji said.

To this end, he led the team to develop the world's first deep learning processor chip "Cambrian", its intelligent processing energy efficiency of nearly a hundred times the traditional chip, so that the deep learning processor gradually from theory to practice, has been used in hundreds of millions of mobile phone products, by Harvard, Stanford, Massachusetts And other global more than 200 scientific research institutions and chip companies tracking and citation. Chen Yunji once named the deep learning processor architecture "DianNao" and "DaDianNao" in his paper.

This is the world's first deep learning processor chip "Cambrian 1" developed by Chen Yunji and his team. Photo by Xinhua News Agency reporter Li Xin

The authoritative academic journal Science commented on chen Yunji's team's deep learning processor chip research results as "pioneering progress" and considered them "pioneers" and "recognized leaders" in the field.

"What kind of abstraction can make neural networks emerge stronger artificial intelligence?" The answer may be far away, but every small step forward will allow machines to better serve humans. Chen Yunji said that he is willing to take artificial intelligence as a lifelong struggle.

At the age of 14, he was admitted to the junior class of the University of Science and Technology of China, at the age of 24 he obtained a doctorate from the Institute of Computing of the Chinese Academy of Sciences, and at the age of 29, he was promoted to a researcher... In the eyes of outsiders, Chen Yunji is a genius who always "overtakes in curves", but he believes that there is no shortcut to scientific research.

Midsummer sweated profusely to write papers in an unair-conditioned dormitory, and repeatedly submitted papers to top international conferences and was rejected... Before the hardships and setbacks, he did not give up the dream of "artificial intelligence" and did more in-depth research.

Chen Yunji clearly remembers that when the first deep learning processor chip they developed finally came out and the test results showed that the indicators met expectations, he was particularly happy and went back to sleep well. The next day, I went to work as usual and started the next cycle.

"As a researcher, exploring uncharted territory is lonely, especially in 'no man's land' explorations that others have not set foot in." Chen Yunji said, but people with scientific ideals must have the courage to explore unpopular and unknown fields.

Chen Yunji works in the office. Photo by Xinhua News Agency reporter Li Xin

He often told students that doing scientific research is not to issue papers and projects, but to have the courage to get involved in areas that have future prospects but are rarely paid attention to now, although it is not easy to issue papers, but it has greater significance for economic and social development.

"Scientific research, such as mountaineering, will have one problem after another, and it is necessary to constantly try and work hard to solve the problem. Climb all the way up, look back, the scenery under the mountain is wonderful! Chen Yunji said.

Chen Yunji (first from right) discusses the design with members of the research team. Photo by Xinhua News Agency reporter Li Xin

Reporters: Zhang Quan and Wen Jinghua

Editor: Liao Yi