The major changes in the world today have accelerated, the impact of the new crown pneumonia epidemic is extensive and far-reaching, and the uncertainty of instability has increased significantly. Scientific and technological innovation has become the main battlefield of the international strategic game, and the competition around the commanding heights of science and technology is unprecedentedly fierce.
On June 19th, Yuanwang Think Tank and CITIC Publishing Group jointly held a sharing meeting on the new book "Science: Endless Frontiers", which specially invited Li Di, CEO of Xiaoice, Fan Chunliang, professor of the University of Chinese Academy of Sciences, Fang Hanting, professor of the University of Science and Technology of China, and Zhong Cantao, vice president of the Institute of Advanced Technology of Peking University.
At this sharing meeting, Li Di, CEO of Xiaoice, made a keynote speech. He talks about his understanding of the world's shift to Asia from the following aspects, and describes how he sees "how Chinese companies can create a promised land for scientific research." Let's comb through his insights, hoping to inspire you.
The center of gravity of science is shifting to Asia, especially China, and we have ushered in a major opportunity that history has given us
The development of the entire civilization of mankind needs to be continuously advanced through scientific means, and only science can truly enable a region, a country, and an institution to obtain better core competitiveness. One of the clear things we're seeing is that there's a global migration of science to Asia today, and it's actually happening for at least 20 years or so, and today it's coming to a final inflection point. The future of science is likely to take place in Asia.
Today we re-read Science: Endless Frontiers, a report from more than 70 years ago, is actually timely. Because from my personal point of view, history has some similarities. In the United States after World War II, if not many people began to observe that the original basic science research, when it has a penetrating effect at certain moments, it will be so powerful, including governments, enterprises, institutions, scientists themselves can get huge benefits from it, so that these people are more convinced that this is a thing they must do, then there will be no "Science: Endless Frontiers" this great report that has influenced so far, changed the global scientific research landscape, even if there is, It may not be so solidly followed up and implemented.
Today, more than 70 years later, in China, we have also begun to notice that basic scientific research is an important cornerstone of the future, and from top to bottom, this is gradually becoming a consensus. In this sense, today's history has come to a very coincidental key point.
01 Shift in the focus of scientific research
Since I don't know much about other areas, I can only explain this problem in the field of computers and artificial intelligence.
We can clearly see that the world's center of gravity, including scientific research, is shifting to Asia. Unfortunately, this is not a direct shift to China, but to Asia as a whole, including japan, India and other countries.
In a way, this is a big coincidence and gives China a chance. This trend is not something that can be reversed today with the introduction of an Innovation and Competition Act in the United States. Nor is it that we think that we should promote more basic research, and we can shift all the focus to China at once. It is actually the result of a combination of effects.
From my point of view, I think that the basis of innovation in the process of shifting to Asia, especially to China, has been underway since at least 2006, and there are many factors that promote the development of artificial intelligence, the most important ones are data. Around 2006, because we were doing search engines, we probably observed a significant spike in data around the world, and China was one of the big increments. In 2009, China's computer- and ARTIFICIAL-related infrastructure was gradually improving. If you look back, some domestic cloud computing vendors, such as Alibaba Cloud, have been completed at this time. By 2015, at various academic top conferences, China's scientific research institutions began to brush the list, began to have a variety of exposures, and then there was a surge in patent applications, and China's artificial intelligence and computer patents began to increase significantly.
In 2018, China further formed some through products. Since 2018, we have seen a lot of end-to-end products in China, and these end-to-end products have even begun to affect the whole world. As a research team whose main body is Chinese, we are very excited. A project like Tiktok was unthinkable before, a product from China affecting the world, which is a rare initiative, although there are many problems, such as the threat of algorithms, but from an end-to-end perspective, it represents the power of China.
02 The future competition is not a single point of breakthrough, but a competition of frameworks and indicators
The characteristics of basic scientific research that we are discussing today are important that it tends to cover multiple disciplines and fields, rather than a single breakthrough. In the fifties and sixties, the sixties and seventies, as long as the older generation of scientists and politicians in our country tightened their belts and tried all kinds of ways to eliminate all difficulties, we could also achieve very good results in scientific research at a certain point and bring decades of peace to the country, but this was a breakthrough in single-point technical science. The real way to grasp the endless frontier is to complete a breakthrough that systematically covers multiple disciplines.
Let me give you an example. In the field of artificial intelligence, the computer voice angle has a very important task is how to make the sound of artificial intelligence fit people, which sounds like a small thing, but there are many algorithms behind it. More than 30 algorithmic models are enough for face recognition, while speech may require one or two hundred models.
Pulling back the clock back to 2013, in 2013, the sound of the AI that came out of the lab was a sound like Siri's, and it was very mechanical. At that time, the world's indicators for speech synthesis still stayed on speech recognition, that is, the content was accurately transmitted to your ears in the form of speech, so it was more of a requirement that this content should not be mispronounced.
At that time, we began to consider whether we could make a relatively natural sound, which could have better emotions, and there would be some emotional characteristics behind it. This also requires the creation of a new metric, which we call "speech naturalness". The so-called naturalness depends on how long you can listen to this sound when you hear it without feeling unbearable, for example, if you listen to it for 7 minutes, you can no longer stand this sound, then using this sound to drive a lot of things, all can not exceed 7 minutes. Three years ago, we achieved an average sound of 21 minutes of comfort through many technological innovations. So far, it can be achieved with a single point of technological breakthrough and a single point of scientific research breakthrough.
But then, we made a sound that was comfortable for 31 minutes, and the sound couldn't be done with a single point breakthrough, it required a lot of interdisciplinary technology. Because in response to this sound, what we do more is to start to restore human flaws, human inadequacies, and bad places of human beings on this version, which will actually bring more natural results.
If we further engineer these technological achievements, we can use 15 minutes to simulate a person can say that the real voice, we are actually very cautious, because when the artificial intelligence technology is constantly breaking forward, and forming a penetration, new indicators may produce new effects, but it may also produce a lot of new problems. For example, such a voice is difficult to describe since then, we say that there are 7 kinds of evidence in the legal aspect, audiovisual materials are one of the evidences, then it is difficult to evaluate whether the person's recording still has the evidentiary value of the legal sense, which is a new situation that often occurs when basic scientific research is transferred to application, but its premise is that you can really penetrate, you can propose new indicators.
A voice like this one means that we already need to come up with some new metrics, and the whole industry needs new metrics. Whoever put forward this indicator, who has more first-mover advantage, we originally said that first-class enterprises do standards, and now first-class enterprises do indicators, which is very first-mover advantage. So the importance of a system is self-evident.
We are very happy to see that China has been developing in a systematic way for so many years, no longer pursuing a single point of technological breakthrough, that is, using a technology to solve a specific problem, using a research and development method to solve the development of a new drug, but as far as possible with a general method to solve the development of a series of new drugs, or with a more general framework to solve the development of a variety of drugs, including new drugs, including the past traditional drugs, including the entire testing process, etc., this is the importance of the system, It's what America used to have, and it's what we need.
From this point of view, China actually has a lot of advantages, because our country has a particularly good depth. We have first-tier to sixth-tier cities, we have a lot of different data, we also have a lot of a variety of application scenarios, these scenario-based applications plus the rotation of data can make basic scientific research continue to accelerate in the middle of this roulette wheel, and then apply it to different scenarios, which is very important.
03 Innovation of ideas is equally important
Another is the innovation of ideas. Before the whole world for the dialogue system, for the natural language processing crown of the jewel, everyone's understanding is the same, it is either as omnipotent as a warrior, or as omniscient as Einstein, but all these two aspects of research are facing philosophical dilemmas, which causes that the previous dialogue artificial intelligence is not very well extended.
Let me give you an example. When we were at Microsoft, we wanted to do a smart assistant project, so we interviewed a human assistant who did a good job, hoping that the AI assistant could learn from people. We envisioned a scenario in which the assistant was given the order: Order me a hamburger with fries at noon. We thought the best way to do that at first was to order the thing right away with a round of conversation. It's a little better to predict the next day's situation, because you have a meeting at noon, so it directly helps you prejudge and book for you, which is very smart. Further, if you give it authorization, it can use your wallet, you don't have to say anything to it, it will directly help you order, which we think is the smartest.
When we talk to the human assistant, the human assistant says that this is not the case, he is likely to refuse, he will say I do not order you McDonald's, we were very confused at the time, because to iterate a system must make sure that the system completes the task to get full marks, now do not complete the task can get full score, but can not iterate. We subconsciously think that there may be higher intelligent requirements to achieve more basic goals, such as health, health is not more important than booking a McDonald's at noon, obviously this assistant is more intelligent, it also knows that there are healthy tasks to complete.
However, this human assistant laughed at us, saying that you really think too much, I just want to reject my boss and find all kinds of reasons. Because I want my partner to know that I am not a person with only hands and no brains, I am such a person with my own ideas, my own thinking, my own views. From then on, when I established a good relationship with him, not only for the simple tasks of ordering food, he would trust me very much, but we would be able to get along well in many other jobs in the future. So it doesn't depend on how any individual task should be done, but on how a larger range of things are done.
What I want to tell you is that this new idea is more observed in Asia than in Europe or the United States. When you observe this phenomenon, you have an opportunity to introduce this new idea, so when we build the dialogue system of artificial intelligence, the core idea behind it is no longer to complete the task of the moment, but to create a long-term relationship between people and artificial intelligence.
We have observed through a large amount of data in China and Japan that this long-term relationship can be maintained, and this relationship is fixed. We look at a lot of innovation in the United States, but we don't know that these innovations have already begun in China. When we do that long-standing concern, Amazon is offering a million dollars to people who can get its smart assistant Alexa to talk to people for 30 minutes; a project like Facebook once wanted to do a task-accomplishing robot like Jarvis, which has been shut down and is now doing a project called Facebook Blender, and they're all about our Chinese stuff. In fact, many innovative ideas have been carried out by our side.
How to determine that these ideas are reasonable, this is a point that our country needs to pay attention to, that is, we must learn to tolerate a variety of ideas. If we look at the moment when science: Endless Frontiers was born, I see that at that time, when the innovation of any kind of idea involved basic scientific research, there was a very large degree of tolerance in this country, and its researchers also had a very good tolerance in it, and it would try and discern all kinds of possibilities. Trial and error for a large number of frameworks and ideas, one institution is not allowed, but one country can, so we especially need a hundred flowers now, maybe some institutions bet on the wrong direction, but the fertile soil of science as a whole will not bet on the wrong direction, which is a very critical thing. We need inclusion more than ever at this time.
04 Some suggestions for scientific development
(1) The Chinese team needs self-confidence
We have to make one thing clear, China's team really needs to be confident, Chinese researchers, Chinese enterprises, Chinese institutions, China's public really need self-confidence, of course, not blind self-confidence. Why? First, today we mention that we should put basic research in China, and China will promote the world, which is not something that has just happened, it has actually happened for a long time. Second, we do have such soil today, we can have single-point technological innovation, we can also have systematic technological innovation, we can have indicators, we can also have ideas. And these are not the same as earlier, in fact, they are already being paid attention to by the whole world, especially by the United States, which is a fact. We need to avoid looking more at American innovation and then replicating it. As we all know, in the field of computers, including the Internet, for a long time we are all me too model, that is, there are things in Silicon Valley, we will have in China, China's own innovation we are not particularly concerned, this is not confident, maybe self-confidence came a little late, but it is needed.
(2) Pay attention to the role of negative samples
Scientific development, especially the basic scientific research part, is often important not only its positive sample, but also its good results, and its bad results are also particularly critical. Our country has a particularly good place, that is, our negative sample is particularly large, if we do artificial intelligence planning, or do image recognition visual things, we will get a lot of bad cases. When we run on the road, we get a lot of wrong data, China's traffic environment is generally bad, these are negative samples, these negative samples are very good. From other points of view, our country's doctors go to the United States, and our country's doctors are also very strong, because the United States does not have so many cases, a doctor does not have so many training opportunities, China has, of course, including failed training opportunities. So there's a key thing, that is, we need to share negative samples, and the whole industry needs to share not only scientific research results, but also failed cases, including failed data sets. Only then will these basic scientific studies accelerate.
(c) The popularization of science is necessary
The third very important thing, and one of the things mentioned in the report, is whether science should reach the general public. I myself think this can still be a great inspiration from Science: Endless Frontiers, which is one of the reasons I admire this person. The general public will argue that a certain science and technology is useful, or a certain science and technology is useless, and then the scientist will say that you don't care whether it is useful or not, you let me get it myself, it is better to stay away from us, we are ivory towers. This kind of problem was encountered 70 years ago, but this ivory tower is actually blocking science. Because it really needs an attitude to popularize science to the public and to carry out such work in practice, it is possible to maintain the public's enthusiasm for science, it can not only promote the continuous metabolism of scientists, but also enable science to extend to various industries and the chain of specific details. We are often reluctant to tell you about these things behind it, for an ordinary person, the mobile phone he gets in his hand is the real innovation he can contact, but behind the mobile phone there is a chain of basic scientific research, this chain involves the progress of many fields of electronic science, he has stopped paying attention, what we need to do is to tell everyone the chain in the middle.
We are fortunate to be able to discuss the development of science today, the shift in the center of gravity of science, and the path of scientific development, and we have not had the opportunity to discuss them for 20 years. Today we have the opportunity to do scientific research in China, to try to pass on the advantages of the United States. Just as the so-called ten-year wall-facing map has broken the wall, the opportunity of history has come to us. Only by cherishing and struggling can we live up to this great light.
Title: Science: The Endless Frontier
Price: 59.00 RMB
ISBN：9787521729597
By Van Neva Bush; D. Holt
Publication date: May 2021
Publisher: CITIC Publishing Group
Introduction:
"Science: Endless Frontiers" is the "pioneering work" of American science policy, which makes the United States quickly get rid of its dependence on Basic Research and Scientific Research Talents in Europe, which has achieved the status of the United States as a scientific and technological power today, and has also permanently changed the pattern of human scientific development. It is an important work for us to understand the origin of American science and technology policy, the path of scientific development, and its future direction.
"Science: Endless Frontiers" systematically analyzes the importance of science to the country's economy and security, social well-being and personal development, emphasizing the need to attach importance to basic scientific research, give researchers a high degree of research freedom, the government should allocate funds to fund the smooth progress of scientific research projects, and establish a national research foundation. These ideas and suggestions are still shining to this day and are often read and updated.
The situation in China today is similar to that of the United States at that time. Although our science has developed rapidly, it still relies heavily on foreign basic research results and is stuck in the neck of core technologies in many key areas. Through "Science: Endless Frontiers" to study the reasons for the great development of science in the United States, cultivate the scientific thinking and scientific spirit of the whole people, increase investment in basic science and frontier science, and enhance the scientific and technological strength and global competitiveness of enterprises, it is of incomparable practical significance.