Original Headline: Professor Zhou Zhihua, Active Both in Cutting-Edge Research and Science Popularization
Artificial Intelligence: How much do you know about it?
Speaking of artificial intelligence (AI), people may have visions of smart robots in those blockbusters. With the development of science and technology, AI is making a growing impact on every aspect of our life and is gradually changing our understanding of what it is.
Professor Zhou Zhihua, of Nanjing University’s Department of Computer Science, has been working in the area of AI for more than 20 years. During these years, He has not only ascended to be one of the most influential scientists in this field around the world, but also witnessed the progress and development of AI science in China.
AI: How it is looked at differently between scientists and the public
Sitting in front of a large bookshelf in his lab, Professor Zhou Zhihua explains his daily work: forming an idea after discussion, analysis and hard thinking, then trying to realize it by algorithm and programming, and finally testing it by running the data.
“There’re two specific areas in the AI study,” Zhou said. “The first is called artificial general intelligence or strong AI, which can often be seen in science fiction movies. The study in this area aims at making robots that are as smart as and even smarter than human beings.
“The second area is called Weak AI or narrow AI, and it tries to make robots imitate certain human abilities and become a help in people’s daily work. This second topic is at present a major study of research by scientists.”
At this point, Zhou gave a metaphor: 100 years ago, people saw birds flying in the sky and wondered whether we could invent something like a bird that could help people fly. Later, the airplane was invented with the help of aerodynamic research. However, do airplanes really fly “better” than birds? It is hard to say. Although airplanes can fly higher and farther than birds, they are less flexible and cannot give birth to “little airplanes.” AI is to some extent doing similar things by making robots imitate birds’ behavior, and it can be seen as a kind of advanced bionics, the study of imitating the functioning of biological systems to solve engineering problems.
“Of all branches of computer science, AI may be the one that interests the public the most,” Zhou continued. “However, because there hasn’t been enough science popularization, people often expect a lot from it and have some unrealistic fantasies.”
For example, in the 1980s, Japan put forward a “Fifth-Generation Robot Plan” to develop its AI science. Although many of the so-called predated goals set in the plan at that time have become reality today, the plan encountered many difficulties in those years since technology was not so advanced as it is now. As a result, people’s high expectations turned into questioning and the whole of the research stagnated.
Now, AI is everywhere in our life, from Internet search to voice interaction, to entrance guard systems and to urban transportation dispatching systems. However, Zhou pointed out, controversies come along with AI’s rapid growth. People ask many questions:
“Does it mean that robots have already surpassed humans with AlphaGo’s continuous victories?”
“Will more and more people lose their jobs if we have ‘robot writers’ and ‘robot office workers’?”
“Actually AI research has never been intended to threaten or replace human workers,” said Zhou. “Every science deals with certain theoretical issues with certain technology and reach certain goals. Researchers only explore ways whereby machines can help people.”
The airplane flies much than human beings, but it does not mean that running cannot still be a popular sport. People take part in sports just to improve their physical and mental health, according to Zhou, who continued with his metaphors.
During the industrial revolution, machines liberated people from exhausting physical labor, but at the same time, they created new jobs; for instance, car drivers replaced coachmen. In the future, machines will be more and more involved in liberating people from repetitive mental work so that people can be engaged in higher-level mental work.
Path of research: Witnessing how China’s science started and take off
Zhou was recently elected an expatriate academician of the Academy of Europe, according to the results of this academy’s 2017 cooptation.
This was not the only good news this year.
Last August in Melbourne, Zhou was elected chair of the steering committee of the 2021 International Joint Conferences on Artificial Intelligence (IJCAI), making him the first mainland Chinese to hold this position since the founding of this organization in 1969.
Last February, he was elected, with a well-known professor from University of Michigan, one of the co-chairs of the steering committee of the 2019 Association for the Advancement of Artificial Intelligence (AAAI). This made him the first non-European, non-American person to hold this position since the association’s founding in 1980.
It is worth pointing out that in 2016 Zhou was elected a fellow of the AAAI, Association for Computing Machinery (ACM) and American Association for the Advancement of Science (AAAS), respectively. Along with his fellowship in the Institute of Electrical and Electronics Engineers (IEEE) and International Association for Pattern Recognition (IAPR), he was the first Chinese person to have a Grand Slam in the all five AI-related associations.
Zhou, born in 1973, got his bachelor’s, master’s and doctoral degrees all at Nanjing University. He was promoted as an associate professor at age 28 and won the National Scientific Fund for Brilliant Youth and became a professor at age 29. At age 32, he was chosen to be a distinguished professor of the “Yangtze River Scholars Program.”
In computer science, Zhou has published a number of monographs and more than 200 papers on world’s top journals and conferences. His works were cited by researchers from over 60 countries around the world for more than 25,000 times. Besides, he holds patents to more than 20 inventions.
Speaking of the path of his research, Zhu said that he has witnessed and experienced China’s development in AI, particularly in the field of Machine Learning (ML).
“ML is the study of theories and methods of data analysis with the use of the computer,” said Zhou. Big data are like a mine whose value is to be unveiled through the data analysis technique, and ML provides such technique.”
His encounter with ML began in 1995 when he was an undergraduate and happened to read in the library a book titled Machine Learning: An Approach to Artificial Intelligence.
The book was published in 1983 and its content was old even at that time; however, it still fascinated him, at a time when the university had no internet and no research in the field had started within China. Even in 2005, when he talked about ML on some academic conferences, he was still asked: what do you want machines to learn, “picking cotton or picking grapes”?
China has made progress in AI research by leaps and bounds over the past 20 years, moving from having no access to international papers to having access to international papers at any time, from reading international papers to learn what others are doing to keeping abreast with international cutting-edge studies.
At the International Joint Conference on Artificial Intelligence in 2017, China surpassed the United States for the first time in the number of the papers accepted.
It is also remarkable that increasingly more and more Chinese scientists are elected chairs at top international academic conferences.
Zhou admitted that he is lucky to have seized the opportunities of our time.
As the rise of China and its enhanced economic strength, the government and enterprises are investing more and more in research,” he said. “We researchers are inspired upon seeing the report of the 19th CPC National Congress makes mention of artificial intelligence.”
Active in science popularization: letting more people learn about ML
In 2007, Zhou Zhihua founded the Institute of Learning and Mining from Data (LAMDA) at Nanjing University. After 10 years of development, the laboratory has now eleven teachers and 60 graduate students; more and more young students interested in AI are applying for this famous laboratory.
Over a span of more than 60 years, AI has experienced three stages of development, namely, the logical reasoning stage from the mid-1950s to the 1960s, the knowledge engineering stage and finally the machine learning stage since the 1990s.
Today, great achievements have been made in ML technologies. However, Zhou Zhihua believed that the current ML technologies still have technical deficiencies such as huge data demand, weak environmental adaptability, poor interpretability, and impractical experience sharing. In order to solve these problems, he proposed an idea recently, that is, learnware.
In this vision, learnware is composed of a learning model and a specification which describes the model. When users want to build their own ML applications, they no longer need to build them from scratch, but look for suitable pieces of learnware in the market.
Based on theoretical research, Zhou also focuses on promoting ML’s actual application.
We have been working with enterprises in many industries to provide solutions to the challenges they face in data analysis, said Zhou.
Now some scientific and technological achievements have been put into practical use, including product recommendation on the internet, financial analysis, and fault diagnosis and early warning for energy facilities.
Zhou is diligent and energetic in the eyes of his students.
Sometimes he finishes answering e-mails at two o'clock in the morning, gives revision suggestions on WeChat at 7 o'clock in the morning, and then at 7:30 asks us to go for a meeting in his office, said Ph.D. student Zhao Peng.
ML is a hot topic, but Zhou deliberately controls the number of papers and often encourages his students to improve quality of papers through quite, in-depth research, or sitting on the cold stool, as he puts it.
Zhou is keen to make ML reach out to more and more people. He has his Weibo, on which he comments on hot discussion spots with the basics of ML and analyzes the technologies behind AlphaGo for “technology aficionados.”
He is called a “best-selling author” because the ML textbook he compiled has been reprinted more than 20 times since its publication one year ago.
In school, he always offer courses for undergraduates. With simple, humorous language and rich examples and case studies, his classes attract not only students of science and engineering majors, but also many of the humanities majors and even some from schools out of town.
Zhou once complained on Weibo: I do not know what to say. The enrolment was 144 for the course, but a classroom of 240 seats with added seats could not hold them all. They students told the Office of Academic Affairs about the shortage of seats. Then, next week, the office relocated the class to a classroom of more than 300 seats!”
Page 16, People's Daily, November 15, 2017