“I have a lot of affection for Shanghai.”
This is the first sentence of the Microsoft Research Asia Dean Zhou Lidong in an interview for more than an hour. 
Zhou Lidong, Dean of Microsoft Research Asia
Graduated from Fudan University with a bachelor’s degree. After receiving a bachelor’s degree in computer science, he furthered his studies at Cornell University in the United States. Researcher of Microsoft Silicon Valley Research Institute, Principal Researcher of System Research Group of Microsoft Redmond Research Institute, Executive Deputy Dean of Microsoft Research Asia; In July this year, he was promoted to Dean of Microsoft Research Asia and returned to his alma mater Fudan University non-stop and was hired as a part-time professor , Zhou Lidong was most grateful that Fudan professors worked hard to bridge the gap in domestic computer teaching and scientific research.
“Now all computer fields are related to AI in a sense.” Zhou Lidong said that Shanghai’s AI development has obvious advantages, and AI applications are becoming more and more “grounded” gas”. The top priority of artificial intelligence research is cross-industry innovation. Cross-industry and cross-discipline is a general trend. You can’t just do pure AI research in the ivory tower.
During the Microsoft Research Asia (Shanghai) Open Day a few days ago, he told News (www.thepaper.cn) that if AI is to truly land, it must be in line with the industry. get through. And those great AI innovations are fundamentally reflected by accelerating innovations in disciplines such as physics, chemistry, and biology. “DeepMind may have solved some problems and taken the first step towards’great achievement’. They gave us a great inspiration that we must go out of computer science, integrate with different disciplines, and cultivate Cross-border talents, this is a long process.”
Zhou Lidong believes that computer science and artificial intelligence will become basic subjects like mathematics in the future, computational thinking, data Everyone needs to master the ability to drive, just like we want to be literate. Regarding research, he believes that it is not about how many papers have been published, and that the original meaning of research cannot be sacrificed for short-term goals.
In September 2018, Microsoft announced the establishment of Microsoft Research Asia (Shanghai) in Shanghai. In the past two years, there have been more than 20 researchers and engineers in the Shanghai Research Institute.Improve innovation ability in one step. Today, Lidong Zhou is fully responsible for the research work of Microsoft Research Asia in Beijing and Shanghai, as well as the cooperation with academia and industry in China and the Asia-Pacific region.
Regarding the development of Shanghai Research Institute, Zhou Lidong said that Microsoft Research Asia is investigating and studying the possible positioning and opportunities of Shanghai Research Institute in various aspects, and is starting to promote Shanghai research The institute and other Microsoft teams in Shanghai have formed greater cooperation. At the same time, based in Shanghai, with the help of the unique attraction of the Shanghai area, it will attract more international cutting-edge talents.
You can’t just do pure AI research in the ivory tower. The most important thing is cross-industry innovation.
The cutting-edge technology of computer science is developing rapidly. Where is the artificial intelligence business model in full swing?
“When AI first started to emerge, everyone was very enthusiastic and felt that AI was going to completely subvert every industry. Now I think everyone is slowly calming down.” Zhou Lidong Said that this is not a simple matter of subversion, but to find the true value of AI in these problems to be solved by various industries.
He told the news (www.thepaper.cn) that there are two business models for AI. One is empowering applications to improve the efficiency of existing industries, such as Artificial intelligence optimizes the logistics industry; second, disruptive applications in the future. For example, the artificial intelligence program AlphaFold 2 has made great progress in advancing protein structure prediction. “With these technologies, many new business models may be born, including Genomics and other fields will have a great impact.”
“While we are solving existing problems, we must also consider which directions may have problems in the next 5 years. Now we have a completely different model. We need to be forward-looking.”
Zhou Lidong said that whether it is past successful practices or future development directions, we are all concerned about artificial intelligence. Technology puts forward new requirements. Explainability, security, privacy protection, fairness, etc. are all frontier issues in AI research. Only when these issues are resolved can the capabilities of AI be truly released.
And the top priority of artificial intelligence research is cross-industry innovation, and cross-industry and cross-discipline are the major trends. Zhou Lidong told the news (www.thepaper.cn) that if AI is to truly land, it must be harmonious.Get through the industry. And those great AI innovations are fundamentally reflected by accelerating innovations in disciplines such as physics, chemistry, and biology.
For example, DeepMind announced in the journal Nature that it has predicted 98.5% of human protein, and plans to increase the number of predictions to 130 million by the end of this year, which is known to humans. Half of the total protein.
“DeepMind may have solved some problems and has taken the first step towards the’great achievement’. One of the great inspirations they gave us is that we must Going out of computer science, integrating with different disciplines, and cultivating cross-industry talents is a long-term process.”
Zhou Lidong said that the research institute cannot “only Do pure artificial intelligence research in the ivory tower”, “If we don’t understand the actual problems and don’t have real data, then we won’t have a very deep understanding of the problems we really need to solve.” Therefore, we must first be able to let go The body is “grounded” and understands the actual problems facing the industry and which difficulties can be improved with new technologies.
At the same time, in the field of artificial intelligence, there is no shortage of AI talents who can adjust parameters and make models. Crossover talents are truly scarce.
Zhou Lidong said that talents who understand algorithms, systems, and compilers, as well as computer architecture and hardware, are increasingly in short supply. In addition to computers, if you understand both artificial intelligence thinking and physics, chemistry, or biology, you can play a key role in the exploration of cutting-edge scientific research.
“We now also attach great importance to cross-disciplinary talents.” Microsoft Research Asia has recruited interdisciplinary talents in recent years, including physics, biology, psychology, Linguistics, etc., after enrolling undergraduate Chinese Department and then continuing to study computer natural language processing researcher, “Chinese background provides a unique perspective for his research, which is very remarkable.”
Zhou Lidong believes that computer science and artificial intelligence will become basic subjects like mathematics in the future. Everyone needs to master computational thinking and data-driven abilities, just like we need to be literate. Computer and artificial intelligence education will be more and more integrated into general education, whether it is liberal arts, science, or engineering, these abilities need to be mastered to a certain extent.
“I don’t think liberal arts or science students should switch to computer science, but heAfter we learn computers, they may bring great breakthroughs in their own fields. “
Artificial intelligence is also one of the three core industries in Shanghai. Regarding the development of artificial intelligence in Shanghai, Zhou Lidong said that there are unique advantages here.
On the one hand, cross-industry cooperation and innovation not only foster the artificial intelligence industry, but also integrate artificial intelligence research with all walks of life. This is an important way to break the boundaries. And each district in Shanghai The industry composition is different, and the positioning of AI is not the same, so we can try a new cooperation model in Shanghai.
On the other hand, the combination of industry, university and research is very important. The cultivation of talents and the judgment of the future require unique insights. In this regard, Shanghai has a very big advantage.
After taking office as the dean of Microsoft Research Asia, Zhou Lidong was at the end of September. Back to his alma mater, Fudan University, and was hired as a part-time professor. “We discussed with the teachers of the School of Computer Science of Fudan University how to use the strength of the School of Computer Science to really help other departments in the future. “
“Fudan, Shanghai Jiaotong University, or other universities, all have the opportunity to make some very important innovations in the talent training system and structure. For example, how to cultivate the next generation of cross-field talents, and how to turn AI and computers into a part of general education. These issues are already being considered by Fudan University, which is very remarkable. ”
Doing research is not about publishing many papers, and not sacrificing the original meaning of research for short-term goals
Zhou Lidong graduated from Fudan University with a bachelor’s degree in computer science After the degree, he studied at Cornell University in the United States and obtained a master’s degree and a doctorate degree in computer science. After graduating from his Ph.D. in 2002, he entered the Microsoft Research Institute in the United States. Researcher and executive vice president of Microsoft Research Asia.
After leaving Fudan and Shanghai in 1995, he teased that he “wandered” around the world for more than 20 years. Zhou Lidong said that when he studied at Fudan in the early 1990s, what made him feel the most was that there was a considerable gap between China’s computer teaching and scientific research at that time and the world’s top level.
“Remember that when we were working on assembly language, the computer room used a PDP-11 machine, and that machine was beautifulIt was produced by National Digital Equipment Company around 1970. Later, when I came to work in the United States, the first colleague I met was one of the designers who designed the PDP-11. At that time, the PDP-11 machine they used was already a 64-bit processor, but it was still used in our country. A machine with a 16-bit processor. So the gap at that time was still very large. “
Although lagging behind in equipment, in theory, “our level is completely close to other schools in the world.” In the first semester when Zhou Lidong entered Cornell University, there were two “devil” courses. One was computer algorithm design and complexity analysis, and the other was computer language theory. This is a key course for obtaining a doctorate. br>
But among the students from all over the world, Zhou Lidong is the only student who got A in both courses.
div>”In just half a semester, if the basic theory is not good, no matter how hard you work, you will definitely not reach this level. Zhou Lidong said that he was most grateful to the professors at Fudan University at that time. “In order to bridge the gap, they worked hard and spent a lot of time on textbooks, especially in computer theory, including discrete mathematics and mathematical logic, which are the basis of computer science. “
Looking back on the past 20 years, Zhou Lidong said that after work, he has gone through “several very interesting stages.” The period of Microsoft Silicon Valley Research Institute was a stage of self-cultivation. At that time, he learned how to do research, “It is not how many papers are published, but the inner passion to promote research. “
He has worked with pioneers who have done pioneering and ground-breaking work in the computer field, especially in the field of distributed systems. Distributed systems can be said to be the last ten years. The main supporting technologies behind the Internet, search engines, cloud computing, and artificial intelligence that lead the technological revolution.
” When graduating with a Ph. ‘Or publish a paper. When I arrived at the Microsoft Silicon Valley Research Institute, I was very touched. Some of the Silicon Valley Research Institute researchers, including the two who later won the Turing Award, did their research entirely out of interest, rather than publishing papers and winning. The purpose of the award, winning is just a recognition of them. “
He believes that doing research is a long-term pursuit of “truth, goodness and beauty”, so researchers also enjoy the research process very much. In the following career, Zhou Lidong I have always reminded myself not to sacrifice the original meaning of research in order to pursue some short-term goals.
System research is one of the most basic computer research fields. It is abstract and hidden behind various applications. Even if it is the same publication, the publication of system research papers is relatively more difficult. The field is the world’s top There are only a small number of academic conferences such as SOSP (Operating System Principles Conference) and OSDI (Operating System Design and Implementation Symposium) once every two years, and only 20 or 30 top papers can be selected by the conference each year.
< div class="contheight">