"If you want to succeed, you should focus on the research that really excites you and intrigus you," John Hopcroft, a professor at Cornell University and a Turing Award winner, said at Microsoft Research Asia's 2021 Theory Symposium recently. The mission of the university is not to train a person to find a good job, but to educate students to have a good life. That's why education shouldn't just be a narrow technical field. That's why we insist on the humanities and social sciences. Imagine if you worked in a government department or got promoted to the management of a company, but all you knew was a narrow technical area, and you would be trapped. ”
1986 Turing Award winner John E. Hopcroft
John E. Hopcroft was awarded the Turing Award in 1986 for his fundamental achievements in the design and analysis of algorithms and data structures. From 1966 to the present, china's Turing Award winner is currently only Yao Zhizhi, president of tsinghua university, as dean of the Institute for Interdisciplinary Information Technology.
Hopcroft's research focuses on the theoretical aspects of computation, especially algorithm analysis, automata theory and graph algorithms, and he has taught at Princeton University, Cornell University, Stanford University and Shanghai Jiao Tong University, and is a member of the American Academy of Sciences, the Academy of Engineering, and a foreign academician of the Chinese Academy of Sciences.
In addition to the field of computing, education is also an area of great concern for Hopcroft, who has investigated the development of computer science education in 50 universities in China.
"It turns out that one of the top nine universities in China has a higher quality of freshmen than because of Berkeley, Stanford or Cornell. But when I confronted these Chinese students again four years later, I found that Cornell had outnumbered them. This means that a university education in China is not successful. Recently, John Hopcroft, a professor at Cornell University and a Turing Award winner, told the surging news (www.thepaper.cn).
Hopcroft continues, "In the U.S., the education system takes up a lot less time in elementary and middle school, and children have more time to interact with other children, participate in sports, develop social connections, and so on, and a lot of education takes place outside the classroom. But it's hard for Chinese students to have time to explore the world and connect with other students. There is already real pressure to reach certain (grade) goals, which American students don't have. So I think Chinese students are just focused on solving something, but seem to be missing curiosity. ”
At the opening ceremony of the Fourth World Top Scientists Forum, Hopcroft received widespread attention for a speech that "Chinese universities attach too much importance to international prestige, and how much the international prestige of a Chinese university depends on the university's research funding and the number of papers" was widely concerned.
Chinese universities are more concerned about how to improve international prestige and its indicators – research funding and the number of published papers – and he believes that the key indicators should be changed to the quality of teaching for undergraduates. Some of China's top universities have been able to train high-quality undergraduates, but the number is far from meeting the current needs of Chinese society. Hopcroft said at the time.
Hopcroft divides research into two categories: basic research and applied research. Applied research is research done to solve specific problems; basic research is research done out of curiosity and does not need to consider its application problems. "Evaluating basic research is very difficult because most basic research doesn't pay off. So, I want to ask the question: Why do universities hire researchers? ”
"When you hire a faculty member, they probably have a career of 40 years, and they're going to evolve over those 40 years as the field of study evolves. But if you hire a researcher, he will have a curiosity about the basic research he does, and this curiosity will make them constantly curious about the cutting-edge dynamics of their field, which will inspire them to continue to innovate and eventually develop with the research field, and even push the boundaries of research. And when something new happens, the course they teach is also likely to incorporate the latest developments, thus keeping the course up to date. I think it's the basic nature of people who do basic research to get excited about what they're doing and bring that excitement to the classroom. So, that's why universities hire researchers. Hopcroft said.
But why would a company hire researchers for basic research? "Because the company hopes that researchers can participate in international conferences around the world on the basis of improving the company's own technical reserves, so as to obtain the most cutting-edge information in the world and promote their own technologies, products and ideas." Comparing Hopcroft's answers, we can think about the current dilemma of the Internet giant AI Lab.
Definition and significance of theoretical research in computer science. Hopcroft said that when I first started my research, theoretical computer science covered very little, including theories such as finite automata, context-free grammar, computability, and determinability. That's because when the discipline first started, it consisted only of theoretical computer science and programming, and there was no field of algorithms. But today's computer science has taken on a new look. If the discipline was originally concerned with making computers useful, now not only have computers become useful, but computer science has changed radically. So now we have a variety of applications, including artificial intelligence, big data, cryptography and so on.
Hopcroft argues, "Today, theoretical computer science researchers are people who are interested in the basic ideas behind an application. They want to know why AI works, which is very different from those who just want to use AI to solve problems. ”
The following is the original text of the speech:
Today, I want to talk about academic research on industry, and I'm sure there are many people who may not have paid much attention to this. First of all, research can be divided into two categories, one is basic research and the other is applied research. Applied research is research done to solve specific problems; basic research is research done out of curiosity and does not need to consider its application problems. Evaluating basic research is very difficult because most basic research does not yield returns.
So, I want to ask the question: Why do universities hire researchers? When you hire a faculty member, they may have a career of 40 years, and they will continue to evolve over those 40 years as the field of study evolves. But if you hire a researcher, he will have a curiosity about the basic research he does, and this curiosity will make them constantly curious about the cutting-edge dynamics of their field, which will inspire them to continue to innovate and eventually develop with the research field, and even push the boundaries of research. And when something new happens, the course they teach is also likely to incorporate the latest developments, thus keeping the course up to date. I think it's the basic nature of people who do basic research to get excited about what they're doing and bring that excitement to the classroom. So, that's why universities hire researchers. But why would a company hire researchers for basic research? Because the company hopes that researchers can also participate in international conferences around the world on the basis of improving their own technical reserves, so as to obtain the most cutting-edge information in the world and promote their own technologies, products and ideas.
Next, I want to talk about the definition and significance of theoretical research in computer science. When I first started my research, theoretical computer science covered very little, including theories like finite automata, context-free grammar, computability, and determinability. That's because when the discipline first started, it consisted only of theoretical computer science and programming, and there was no field of algorithms. But today's computer science has taken on a new look. If the discipline was originally concerned with making computers useful, now not only have computers become useful, but computer science has changed radically. So now we have a variety of applications, including artificial intelligence, big data, cryptography and so on. Today, theoretical computer science researchers are those who are interested in the basic ideas behind an application. They want to know why AI works, which is very different from those who just want to use AI to solve problems.
Therefore, I think enterprise research institutions like Microsoft Research Asia need to focus on specific areas that can help companies evolve their technology. And a key part of this is to build a theory. If you're working on a particular problem, you have to spend most of your time trying to solve it. But if you're only spending 5% of your time, can we come up with a new theory in this area? Sure, you may find other ways to solve the problem and can solve it more efficiently, but this only solves one problem and does not explore the related problem in depth. I must admit that companies don't usually hire a lot of basic researchers because companies want to hire people who can solve problems. But there is one company that is different, and that is Bell Labs. This is a special case because Bell Labs is run by the government, which employs researchers who only do their own interests. I have to admit that Bell Labs' research results are very fruitful, and its model is difficult for most companies to replicate. Because many companies tend to focus only on short-term benefits. But it turns out that only by taking a long-term view and insisting on doing basic research can we have a greater impact on academia and society as a whole.
Basic research is also very important in today's university education. The university's mission is to train the next generation of talent, but occasionally there are faculty members who discover something with significant impact that creates a whole new field of science, creates millions of jobs, and promotes economic growth. I think the U.S. decision to fund basic research at universities is probably one of the best financial investments they've ever made. But it's hard to evaluate because about 100,000 researchers have done research that can't be measured by ordinary evaluation methods. And basic research is unlikely to pay off in three years. So, when your money can't support 30 years of research and targets only a small group of people, how do you measure it?
Therefore, it is relatively easy for university teachers to measure the quality of their teaching. But the other question is, will a college teacher have a career that allows him to grow and continue to explore at the forefront of the field? That is, even if there is no return, will this person still be enthusiastic about his career? It is difficult to make such a measurement. Currently, I am approved to evaluate computer science education at the top 50 universities in China. Our assessment method was also very simple, which was to ask students to observe a lecture to see if the teacher had students involved in the teaching, and we also measured how many students were concentrating in the class. We have a special scorecard that records the five metrics we need to evaluate. The results show that this survey is of constructive value for improving the quality of teachers' teaching.
I think it is necessary to mention the term vector space model. The term was coined by a faculty member at Cornell University, but at the time we didn't think it would be a very influential concept. But it turns out that things that you think might not be important tend to lead to entirely new areas, because that's exactly what makes Google successful and creates millions of jobs. This is a digression. Back to the point, I think there are a lot of important things to think about. For example, if we train an AI network to recognize cats, we need thousands of pictures, but humans can learn from a single image and make recognition. So what we have to do is study how the brain works, and it's important to note that the organization of the brain is fundamentally different from the organization of the computer, and the energy consumed by the brain is very low. What kind of structure allows the brain to do something very simple, while artificial intelligence cannot do the same task? This is the research currently being done on our artificial intelligence. Instead of focusing on what others are doing, or just focusing on your own research, trying to explore a completely different field can be critical.
I'll tell you another story, I'm interested in how the brain learns, and I was told decades ago that three years old looks old. I just listened to this idea at the time, because I didn't think it was backed by scientific theories. But recently I met a man who confirmed this idea again, and when I searched for this study, I found a lot of research on how the brain develops. As a result, the brain gradually learns how to learn in the three years after birth. One study shows that researchers in an inland U.S. city provided a stable intellectually rich environment for one group of newborns for three years, but not for another group of children. 30 years later, they compared the differences between the two. It was found that children with a rich intellectual environment tended to have a smaller chance of suffering from mental problems and had a higher level of education, high-paying jobs, etc.
The stories I mentioned above are all because they are always related to other things. Just as in computer science, we often use proxies, and in most cases proxies do give the optimal solution. So the question is, why does some kind of agent work? This requires us to give an answer from a theoretical level. To me, it's like an intellectually rich environmental research problem.
But one thing I want you to understand is that if you want to succeed, you should focus on research that really excites you and makes you curious. That's what I tell my students, too. Because I found that many students major in computer science because their parents told them that they could get a good job in this major. Once, I was chatting with a graduating senior. He said, "You know, I really hate computer science, I like music more. "So, I would tell students that you should choose a major that you like because your career will be an important part of your life and you should have an exciting life." The mission of the university is not to train a person to find a good job, but to educate students to have a good life. That's why education shouldn't just be a narrow technical field. That's why we insist on the humanities and social sciences. Imagine if you worked in a government department or got promoted to the management of a company, but all you knew was a narrow technical area, and you would be trapped.
Take a moment to think about what you really like, or what exactly you're curious about! Maybe it's quantum communication and quantum entanglement theory, maybe it's distributed computing, maybe it's big data. Computers are now used in medicine, finance, manufacturing, and more. This is the computer science and cryptography that they are today, and they not only provide security, but also take on a variety of important roles. But more importantly, let theory return to the industrial lab!
![Microsoft Research Asia announces the establishment of a Theory Center[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)