text | Interviewer of Xinjing Think Tank Xiao Longping
After the outbreak of the new crown pneumonia epidemic, people need to undergo temperature checks in public places such as train stations and airports. From the initial single detection of the temperature measuring gun to the current pass-through (a "door"-like) infrared temperature measurement. In the short term, body temperature detection has undergone a transition from non-intelligent to intelligent.
"This is all because of the results of artificial intelligence (hereinafter referred to as "AI") processing," Gong Ke, former president of Nankai University and executive director of the Institute of New Generation Artificial Intelligence Development Strategy, told the Beijing Think Tank. A few meters away, identify the person's forehead for detection, and the measurement results will not be affected by someone eating popsicles or holding ice water.
Gong Ke, former president of Nankai University and executive director of the Institute of Development Strategy for the New Generation of Artificial Intelligence. Photography Beijing News Xiao Longping
In July 2017, the State Council issued the "Plan for the Development of New Generation Artificial Intelligence" (hereinafter referred to as the "Plan"), proposing the strategic goal of the development of my country's new generation of artificial intelligence in three steps. Among them, the first step is that by 2020, my country's "artificial intelligence industry competitiveness will enter the first echelon of the international community. It will initially build artificial intelligence technology standards, service systems and industrial ecosystems, cultivate several world-leading artificial intelligence backbone enterprises, the scale of the core artificial intelligence industry exceeds 150 billion yuan, and the scale of related industries exceeds 1 trillion yuan." Gong Ke said that China is undoubtedly in the first echelon. Among the various factors that promote the rapid development of AI applications, policy effectiveness is only one of them, and more importantly, demand traction. For example, the COVID-19 pandemic has unexpectedly created rigid demand for AI applications. Globally, the application of technologies such as AI, big data, and blockchain are accelerating.
When the first step of achieving the goal of my country's AI development plan, Gong Ke said, "Now is the time to summarize and review." So, what changes have occurred in my country's AI applications in recent years and what achievements have been made? What is the current situation? What other challenges are they facing? On June 7, the Beijing Think Tank interviewed Gong Ke.
The new crown epidemic unexpectedly promoted the development of AI
New Beijing Think Tank: The World Intelligence Conference has been held in Tianjin for five consecutive times and has become a window to reflect the development of AI. What changes do you think have happened in the past five years?
Gong Ke: Before talking about "change", I would like to emphasize one of the "unchanged" characteristics of the World Intelligence Conference, which is to focus on application. This World Intelligence Conference is not an academic conference for experts. At the same time, we see that the scope of AI applications is becoming wider and wider, and the fields displayed at the conference are becoming more and more diverse; the topics discussed are becoming more and more abundant, such as AI governance, AI social attributes, and the issues of AI applications helping vulnerable groups, which are introduced to the World Intelligence Conference year by year, which are all issues that need to be paid attention to in actual applications. The fourth session was held in 2020. Due to the epidemic, the conference highlighted related topics such as the application of AI in medical and health care.
data picture. Image source: Unsplash
This year's conference also has some highlights, such as the AI businesses developed by Xiaomi, iFLYTEK, etc., which not only help them understand and familiarize themselves with the world and improve their quality of life, but also open up employment channels for "semantic markers" for visually impaired people. For example, AI promotes poverty alleviation. Last year, the World Intelligence Conference emphasized this topic to help the fight against poverty. This year, our Strategy Institute introduced in its report that a survey of 52 counties that have not yet been lifted out of poverty by the end of 2019 showed that a large number of new jobs were generated in AI and big data application scenarios. Among the 680,000 online employed people, the largest proportion is live streaming salesmen. They not only sell local agricultural and sideline products, but also get employment. These employment skills are also learned from the Internet.
New Beijing Think Tank: Can that be said that the biggest highlight of AI development in recent years is its fruitfulness at the application level?
Gong Ke : It should be said that. I think we have definitely done the "entering the first international echelon" included in the first step of the plan, especially in terms of application.In technical applications, algorithms are optimized, including improvements and innovations, and innovations in underlying algorithms have also emerged. Some improved algorithms are embedded in the hardware independently developed and manufactured. AI companies are widely distributed in 20 application fields, among which the number of enterprises in the two application fields of enterprise technology integration and solutions, smart business and retail accounts for the highest proportion, 17.2% and 10.31% respectively (the data comes from the "China New Generation Artificial Intelligence Technology Industry Development Report 2021" released by the China New Generation Artificial Intelligence Development Strategy Research Institute). The number of companies in the fields of intelligent robots, intelligent hardware, technology finance, smart medical care and intelligent manufacturing accounts for a relatively high proportion.
New Beijing Think Tank: At present, is there any company in our country that can be called a leading company in the AI field?
Gong Ke : Some, such as Baidu in knowledge graphs, intelligent search, autonomous driving, etc., Alibaba in urban brains, intelligent cloud services, Tencent in intelligent medical imaging, intelligent tour guides, Huawei in intelligent software and hardware, iFlytek and SenseTime, etc., can all be called the world's leading companies. According to our research institute's four consecutive years of surveys on the smart industry, we prefer to use the concept of "platform" enterprises or new platforms to further highlight the "empowerment" role of this group of enterprises in the entire intelligent industry development ecosystem. At the same time, we must also see that we still have a considerable gap. For example, BAT (i.e. Baidu, Alibaba and Tencent) have formed very large platforms, but there is still a big gap compared to Google, Amazon, etc.
New Beijing Think Tank: After the epidemic in 2020, you said that AI applications have made great progress. Can these be attributed to the achievements promoted after the introduction of relevant policies since 2016?
Gong Ke : There is no doubt that the policies of governments at all levels have played an important role in promoting, but policies are only one aspect. After the outbreak of the new crown epidemic in 2020, it created a basic need for AI applications. Under the influence of the epidemic, because of the need to isolate, production, school and market will be resumed, which has created demand for the application of various smart tools on the Internet. From a global perspective, AI applications are accelerating their development.
New Beijing Think Tank: We say that the "SARS" in 2003 unexpectedly promoted the development of the Internet economy. Can it be said that the new crown epidemic unexpectedly promoted the vigorous development of AI?
Gong Ke : Artificial intelligence, big data, including the 5th generation mobile communication (5G), are accelerating the pace of deployment in many places. This is all related to the epidemic, because demand has been created. In 2003, the network bandwidth was narrow, the network speed was relatively slow, and the penetration rate was not high enough. However, due to the occurrence of SARS, the requirements for broadband and penetration rate were increased immediately, promoting the development of the Internet.
After the outbreak of the new crown epidemic, the question of having to attend classes in other places was raised all of a sudden. For example, in universities in Beijing and Tianjin, some of their students are on campus, but many students may be in provinces and regions such as Hubei, big cities and villages, and some are abroad, developed and developing countries, facing many problems such as different distances and delays, different transmission bandwidth and loads, and ensuring two-way transmission of information (video, voice, etc.) to achieve classroom interaction. These require strong enough backend technical support, which brings demand for intelligent technology. For example, online conference systems such as Tencent Conference quickly updated several versions during this period, which are all the results of rigid application traction. Online education and online conferences have been promoted for many years, and the results have not been satisfactory. Now they have been promoted all at once and have gained widespread acceptance from the society.
New Beijing Think Tank: You expressed your hope that AI will take a "pragmatic" development path. Looking back now, do you think your expectations have been realized?
Gong Ke : I think I’m on this road now. The reason why I emphasized "pragmatism" at that time was because I was afraid of bubbles in the development of AI, and worried that "policy favorable policies" would lead to blind launches and overheating investment. If you make up a story, you may get a investment. Now it seems that we can't say that there is no bubble at all, but it's not serious. The development of China's AI is basically to move forward steadily in applications and develop "pragmatically" in the process of solving application problems step by step.
but there are some shortcomings.For example, our country has put forward the eight principles of artificial intelligence governance. Last year, we released the "action suggestions for collaboratively implementing the principles of artificial intelligence governance" at the Shanghai World Artificial Intelligence Conference, but the "implementation" of these principles is not enough. We must use algorithms and corresponding technologies to help these principles "implement" into products and services, help users conduct inspections and make judgments, help users defend against the theft of privacy data, and establish independent, reliable, fair and transparent system security assessment, data quality assessment, algorithm fairness assessment, etc. This aspect is not developed enough.
"We will also have original innovative technology products"
New Beijing Think Tank: The breadth and depth of AI applications in my country are generally deeper and wider than those in developed countries such as Europe and the United States, but disruptive and step-type technologies such as original models of basic theory still lack the leading ability, and even disruptive and step-type technologies are almost oligopoly. How do you think we should break through and catch up on this level?
Gong Ke : Simply put, taking the deep neural network that is most used by machine learning as an example, the application of combining deep learning with specific application scenarios is a specialized process of general technology. In this process, we have many of our own improvements and creations, but the most primitive basic algorithms of deep neural networks were not invented by us. Overall, we have more improvement innovations and less original innovations, which reflects the overall level of scientific and technological innovation in my country and its position in the entire innovation chain.
However, we should not oppose application innovation and original innovation. In the development stage and development environment we are in today, we must make up our minds to strengthen original innovation. Now the relevant documents released by the central government have put basic research first, but the proportion of our investment in basic research is still not high enough, but I believe that this will not always be the case. At this year's "Two Sessions", Premier Li Keqiang proposed that "to build a strong country in science and technology and improve its scientific and technological innovation capabilities, we must lay a solid foundation for basic research and applied basic research." I am not pessimistic about the prospect of China's development of more and more original innovative technologies. However, we must pay attention to preventing and overcoming two wrong tendencies in basic research. First, we are strongly aware that we still have many technologies that rely on others' "bottleneck" problems, which lead to impatience and impetuousness in basic research and original innovation. This is undesirable. The more basic research is, the more we must accumulate strength, the more solid and rigorous it is. Haste will not be achieved; second, we will separate basic research and apply development from the opposite side. In fact, scientific innovation driven by application is increasingly becoming the main driving force of the new scientific and technological revolution. It is not a problem to attach importance to application itself. The problem is whether scientific problems can be extracted from applications to stimulate original innovation. We should explore the path of application-driven originality with Chinese characteristics. Of course, we must not be eager for quick success.
New Beijing Think Tank: Our country has a large market and AI also has successful applications in many scenarios. Is this our advantage?
Gongke : This is undoubtedly our advantage. We have a large market and complete industries, which is difficult for many smaller countries to achieve. Moreover, digital technology (including AI) itself has cross-industry attributes, so our relatively complete industrial system provides very good conditions for the development of AI. The problem with
is that due to the long-term formation of separate industry management and industry barriers, there is still a relatively serious "data island" problem. Data cannot be integrated between different industries and different departments, which is a major obstacle. Therefore, with the development of AI, in order to adapt to the new productivity of AI, our corresponding production relations must further deepen reforms, so that the market can play a decisive role in the allocation of resources including data, algorithms, computing power, etc., and at the same time, better play the role of the government in unifying standards, compliance supervision, and ensuring safety and fairness.
New Beijing Think Tank: This round of AI development is largely related to theoretical breakthroughs in deep learning, but at present, the speed of breakthroughs in deep learning theory has gradually slowed down.Once the dividend release is over, do you think it will lead to AI falling back into a development trough?
Gong Ke : I think the development of any thing will not be linear. There will always be a period of rapid expansion and relatively mature stages along the way. Then, after relatively mature, new bottlenecks will appear and new breakthroughs will be produced. It should be said that the current AI is due to deep learning, and deep learning is due to big data and powerful computing power. However, the current data-driven AI dependence on data has become a bottleneck for many enterprises to achieve intelligence. To this end, on the one hand, we must accelerate the process of digitalization and networking of enterprises to lay the foundation for intelligence, including breaking data silos, standardizing data standards, improving data quality, promoting open flow of data and protecting data security; on the other hand, we must develop AI algorithm innovation driven by joint data and knowledge. I dare not say whether the dividends of deep learning are almost exhausted, but we did see some new algorithm innovations emerging in the research, such as brain-like algorithms and chips made by Shi Luping's team at Tsinghua University, as well as multimodal perception and cognition of speech and images, and have made gratifying progress. It was proposed at the 5th World Intelligence Conference that concluded not long ago that artificial intelligence is forming a trend of multi-point breakthroughs.
data picture. Image source: Unsplash
There is a popular saying now called "involved". This term is quite interesting. I think "involved" is the opposite of innovation. We must also prevent "involved" in AI development. There is an important point at the 5th World Intelligence Conference, called "Application Scenarios Lead Innovation". This may be an important model for the innovation and development of AI. New and extensive application scenarios, such as broadband mobile Internet application scenarios, have driven many innovative technologies, including innovations such as WeChat and Douyin, but also innovations in the Android operating system, breaking Windows' monopoly on the desktop system. Among them, the openness of the Android system is an important factor in its success. The rapid development of machine learning is implemented in open source. It can be said that new and wide application scenarios drive "alternative" innovation, which is also regular. New technologies replace the original technology, not only in the original application, but also in the new application. We look forward to the emergence of more application scenarios, especially the application scenarios of AI+5G/Internet of Things+big data+blockchain, which can drive deeper and broader innovation.
AI development cannot be separated from global collaboration
New Beijing Think Tank: Do you think there are other algorithm theories in my country that can replace, replace or promote the development of AI?
Gong Ke : Now people use more popular open source communities such as TensorFlow and PyTouch. Many basic algorithms are obtained from domestic applications, and Chinese developers have also made positive innovative contributions to these open source software communities. At the same time, China has also taken the lead in developing an open source innovation community for AI and developing an innovation ecosystem, such as PaddlePaddle, which are also open communities.
has a concern in the current complex and uncertain international environment, whether these open source communities registered abroad will suddenly get stuck in our necks. In this regard, we need to respond calmly, observe calmly, make prudent and rational judgments, and not lose our composure. As for some people who advocate actively "decoupling" from these open source communities from now on, this is obviously a bad self-enclosure. We should indeed support our own open source community and reduce dependence on a single community, but do not proactively decouple. Integrating into the global innovation network is not empty. I am afraid that the major international open source communities, including the open source communities we lead, should be regarded as part of the global innovation network.
We must learn and absorb rather than reject all advanced achievements of world civilization, including AI technology, and actively contribute to the development of world civilization and scientific and technological innovation, and we cannot engage in our "anti-globalization" from the global innovation network outside our own efforts. We sing the "International" and firmly believe that "International must achieve it" and strive for it.
data picture.Image source: Unsplash
So, we oppose active decoupling in passive decoupling. Before they decouple from you, they can’t decouple yourself. For the open source community, we should be committed to maintaining and developing its openness and maintaining the innovative ecosystem of global cooperation. There are also many contributions from us Chinese, and we should make more contributions. This situation of "you have me and I have you" is the inherent meaning of the "community of shared future for mankind", and this is a better development state.
In short, we must grasp the relationship between scientific and technological self-reliance and integration into the global innovation network, and ensure that as General Secretary Xi Jinping said, "The more we face blockade and suppression, the more we cannot engage in self-enclosure and self-isolation, but should implement an international scientific and technological cooperation strategy that is more open, inclusive, and mutually beneficial. "While we focus on self-reliance and self-reliance, we must prevent another misunderstanding, which is to close the door and play with it, which does not conform to the laws of scientific and technological innovation.
New Beijing Think Tank: In a sense, AI is the "nuclear weapon" of the new era, and competition among countries is also fierce, but the international environment is not optimistic. What impact do you think this will have on the development of AI in our country?
Gong Ke : This issue is related to the global innovation ecosystem discussed earlier. It must be seen that after 2017, the global landscape has undergone major changes, and the innovation ecosystem of international collaborative cooperation has been greatly damaged, and there has been a tendency to be split. For example, a new technology alliance was established in a certain country, which not only excluded Chinese companies, but also did not allow Chinese companies to participate in the United States. This trend of decoupling is worrying.
Just now we said that technological innovation requires an open environment, so we hope to do everything possible to integrate into the world innovation system, but first of all, we must have a global innovation system. If this system is fragmented or not at all, then there is no question of integration. Therefore, we must maintain this system. As General Secretary Xi Jinping has repeatedly emphasized, to be a defender of the international order, stand on the right side of history, insist on being a promoter of globalization, and not engage in anti-globalization.
I think that on this issue, we must also prevent and overcome two wrong tendencies. One is to lie on the technical aspects of others and rely on others with peace of mind on the grounds of "integration into the world"; the other is to lie on the grounds of "independence" and engage in self-enclosure. Enhanced independent innovation capabilities to achieve scientific and technological self-reliance and self-improvement is not contradictory to integrating into the world's innovation system. Our country's current independent innovation capabilities have been much stronger than before the reform and opening up, which is thanks to China's reform and opening up. So, we cannot close the door now.
data picture. Image source: Unsplash
AI governance also requires global cooperation. You mentioned nuclear weapons just now. In terms of large-scale nuclear weapons experiments and the deployment of nuclear weapons in outer space, the relevant countries are not technically unable to do it, but are not doing it for the sake of the overall interests of mankind. AI may also face the problem of "weaponization" and must have global control. Another example is whether super AI should develop and how to develop? Where to apply? This requires everyone to sit down and consider for the common interests of mankind, and jointly formulate and abide by the rules.
Regarding AI governance, Europe is now promoting the rules of Europe. The United States promotes the rules of the United States. Relevant groups and enterprises have also proposed their own regulations, but to achieve real and effective control, global consensus is needed. In fact, if you compare the principles and norms proposed by various aspects, including the AI governance rules proposed by European and American countries and China, there is not much difference, and there is a consensus. What is needed now is that all parties have the political will to sit together and introduce common norms to break this situation of "governing each other".
"There must be a production relationship that is compatible with AI"
New Beijing Think Tank: my country's AI industry is developing rapidly, but the "capital winter" has emerged. Due to the long cycle of AI technology, the speed of industrial development and benefits are not as expected by the capital market. Do you think this will affect the overall development of our country in the future?
Gong Ke : I don’t know much about capital. Judging from our research over the past few years, it seems that it cannot be said to be a “capital winter”.Generally speaking, a large amount of capital is still mainly concentrated in intelligent applications in the consumer field. At the same time, we have also seen that investment in intelligent manufacturing and the real economy is also increasing, so it is hard to say that it is the capital winter for AI. At least our research has not drawn such a conclusion.
Globally, AI development has indeed experienced several ups and downs, but we must see that it was the ups and downs of the scientific research boom. At that time, there was no application in the scale. We must also see that a group of researchers who are willing to sit on the bench have never been interrupted, and their perseverance has finally achieved a breakthrough. In the second decade of this century, AI went out of laboratories and papers, into production and life, which was a historic turning point. Even if there are any further setbacks, it is incomparable to the previous situation.
New Beijing Think Tank: Do you think my country's chip industry may become a factor restricting the development of AI in the future?
Gong Ke : It's already. The recent increase in chip prices has had a great impact on the information industry, including AI. AI needs to use chips in terminals, at the edge, and in the cloud. Chips are the cornerstone of the entire information industry and applications, and the current supply chain situation is not optimistic.
New Beijing Think Tank: How do you think we should break this situation?
Gong Ke : Our country has been deploying the chip industry for many years. To break this situation, we must continue to work hard to strengthen chip development and production capacity, and at the same time maintain a global market. But this cannot be achieved overnight, but requires long-term technical accumulation and the overall industrial foundation needs to be improved.
If we look at design capabilities, we have made great progress, and some aspects have reached the world's advanced level. For example, our Chinese company has designed several AI chips, but it still has a lot of dependence on design tools and software. In terms of chip manufacturing capabilities, it has also improved a lot compared to before, and is close to the world's advanced level. However, there are large gaps and dependencies in basic materials, key preparations, tool software, etc.
requires unremitting efforts to achieve breakthroughs on the chip. On this issue, we must seize the day and not be too hasty, but do it steadily. This is not a fundamental innovation that can be made through slogans, but ambitions.
New Beijing Think Tank: What new characteristics have my country shown in the cultivation and reserve of AI education talents in recent years? What other investments should be paid attention to?
Gong Ke: After the State Council issued the "Plan" in 2017, the Ministry of Education issued the "Action Plan for Artificial Intelligence Innovation in Colleges and Universities" in April 2018. Now more and more universities have opened artificial intelligence majors. I believe that after a period of time, the effect will be revealed and the talent team will gradually grow.
It is worth paying attention to that AI is a general technology that empowers all walks of life. The significance of developing AI lies in promoting intelligence in all walks of life. Therefore, the cultivation of AI talents should focus on cultivating a large number of talents who can adapt to and promote economic and social intelligence, including the intelligence of mechanical manufacturing, the intelligence of chemical materials engineering, the intelligence of civil construction engineering, the intelligence of medical and health care, the intelligence of agriculture and food production, the intelligence of energy production, transportation and distribution, transportation, modern services such as finance, and the intelligence of social management, etc., and intelligent talents are needed in all aspects. I feel that our education community does not pay enough attention to this issue, but is still limited by the traditional thinking of industrialization stage, and cultivating AI talents as a subject and professional talents. This is dislocation from the world-recognized "cross-border integration" characteristic of artificial intelligence. Education of artificial intelligence requires innovation, not only to break out of the narrow "discipline" and "professional" stubbornness, but also to explore new methods and new models. There is a gratifying phenomenon that Chongqing, Tianjin and Shanghai have begun to offer AI courses in primary and secondary schools. AI has entered basic education textbooks and basic education classrooms, but it still faces the problem of lack of teachers and needs to explore innovative education methods.
New Beijing Think Tank: What are your expectations for the future development of my country’s AI?
Gong Ke : Why do we attach so much importance to AI? This is not because of the popularity of capital, nor because of the trend, but because of seeing AI as an advanced productivity. In this sense, my expectation is that we need to have production relations that are compatible with AI and a policy environment that can reflect the requirements of the development of advanced productivity, so that AI can develop healthily and rapidly in practical applications.
The so-called healthy development means that AI applications are more efficient and universal, and are safer and more reliable. To this end, we must not only strengthen innovation, but also deepen reform. Reform is to liberate and develop productivity.
Interviewer: Xiao Longping Editor: Zhang Xiaoyuan Proofreader: Liu Jun
