Nov. 29, 2018
An earlier, partial version of this translation was published on the ChinAI newsletter, a weekly collection of translations and links on China and artificial intelligence. You can subscribe to ChinAI, edited by Jeff Ding, here. The translators completed the translation, and it was edited for publication at New America’s DigiChina project. –Ed.
The Innovative Development and Social Impact of Artificial Intelligence
7th Lecture for the 13th National People’s Congress Standing Committee Speeches on Special Topics
By Tan Tieniu, Professor of Computer Vision and Pattern Recognition, Deputy Secretary-General of the Chinese Academy of Sciences
Since the 18th Party Congress, General Secretary Xi Jinping has put innovation at the core position of the country’s development landscape, attached great importance to artificial intelligence (AI) development, and many times discussed the importance of AI, pointing out the direction for how AI can empower the New Era. When the 2018 World Artificial Intelligence Conference opened in Shanghai on September 17, General Secretary Xi wrote a letter of congratulations stressing that the development and application of AI will greatly improve the level of intelligence in economic and social development and effectively enhance public services and city management capabilities. Deeply studying and understanding General Secretary Xi’s series of important expositions on AI, pragmatically advancing China’s “New Generation Artificial Intelligence Development Plan” [DigiChina translation], effectively avoiding AI’s “digital gap,” and striving to secure AI’s benefits all have great strategic importance for establishing China as a global science and technology power and realizing the Two Centenary Goals.
In 1956, the concept of AI was officially proposed, marking the birth of the AI discipline. The field aimed to endow machines with humanlike abilities to perceive, learn, think, make decisions, act, and more. After more than 60 years of development, AI has made breakthrough progress, and has begun to be widely used in the economy and society. This formed a trend leading to a new round of industrial transformation, and pushed human society to enter the Intelligence Era. The United States, Japan, Germany, the United Kingdom, France, Russia, and other countries have formulated national strategies for the development of AI. In 2017, China released the “New Generation Artificial Intelligence Development Plan.” The National Development and Reform Commission, the Ministry of Industry and Information Technology, the Ministry of Science and Technology, the Ministry of Education, other relevant national ministries and commissions, and the governments of Beijing, Shanghai, Guangdong, Jiangsu, and other places have also issued relevant policy documents to promote the development of AI. All spheres of society have recognized the strategic significance of AI.
Like other advanced technologies, AI is a double-edged sword. When it comes to assessing the social impact of AI, there is an “angel camp” and a “devil camp.” Those in the angel camp believe that technological innovations and the resulting applications in the AI field have made major breakthroughs. They hope it will lead to the fourth industrial revolution, which will have a transformative impact on the social, economic, and military realms, as well as others. They believe it can benefit humankind in manufacturing, transportation, education, medical care, services, and other sectors. Those in the devil camp believe that AI is a major threat to human beings—more dangerous than nuclear weapons—which may trigger World War III. In February 2018, 14 institutions including Oxford University, Cambridge University, and OpenAI jointly published a report entitled “The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation,” which points out that AI may pose hidden risks to human society’s digital, physical, and political security. The report also offered suggestions to mitigate these dangers.
'All spheres of society have recognized the strategic significance of AI.'
On the whole, as AI moves from the past 60 years into the present stage of development, it possesses the “Four New” characteristics. The AI core technology represented by deep learning has made new breakthroughs, adapting the “Smart+” model has poured new capacities into the development of the economy and society, AI has become the new high ground in strategic competition for countries around the world, and the widespread application of AI has brought a series of new challenges to human society in areas such as laws and regulations, ethics, and social governance. Therefore, the new topic of AI, which offers both opportunities and challenges, has attracted attention worldwide and is taken very seriously. Although there are still uncertainties in the future innovation of AI, it is widely recognized that the AI boom will bring about a new social culture, will promote industrial transformation, and will profoundly change people’s work and lifestyle. It will be a technological revolution with profound and long-lasting influence.
In order to objectively understand AI’s true nature and innovative development, this report briefly introduces the basic concept of AI and the history of its development, focuses on analysis of AI’s current state and future trends, and attempts to reveal the true face of AI. Obviously, how to choose the Chinese path in the current historical wave of flourishing AI development is particularly worthy of our thinking and discussion. Therefore, at the end of this report, China’s AI development posture, existing problems, and suggestions for solutions are all elaborated upon.
II. AI’s Development History and Inspiration
In the summer of 1956, scientists such as John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon met at Dartmouth College in the United States to discuss “how to use machines to simulate human intelligence.” They proposed the concept of “artificial intelligence,” marking the birth of the AI discipline. The goal of AI is to simulate, extend, and expand human intelligence, to seek out the nature of intelligence, and to develop human-like intelligent machines. AI is full of unknown paths to explore and complicated ups and downs. In describing the development of AI in the more than 60 years since 1956, different people in academic circles have different views. We have divided the development of AI over the past 60 years into the following six stages:
First, the early development period: 1956–early 1960s. After the concept of AI was first proposed in 1956, it successively obtained a number of remarkable research results, such as machine theorem proofs, a checkers program, LISP (LISt Processor) language, etc. This set the first high water mark for AI development.
Second, the period of reassessing development: 1960s–early 1970s. The breakthroughs in the early stages of AI development had greatly increased people’s expectations for AI. People began to try more challenging tasks and put forward some unrealistic research and development goals. However, one failure after another (for example, the machines’ inability to prove the sum of two continuous functions is a continuous function, machine translation making a joke of itself, etc.) pushed AI development into a trough.
Third, the application development period: early 1970s–mid-1980s. Expert systems emerged in the 1970s, imitating human experts’ experience to solve problems in specific fields. This achieved a major breakthrough of AI from theoretical research to practical applications, and from a general reasoning-based approach to the application of specialized knowledge. The expert systems achieved success in the fields of medicine, chemistry, geology, etc., and pushed AI to a new high point of application development.
Fourth, the development downturn period: mid-1980s–mid-1990s. As the range of AI applications continued to expand, expert systems’ limited domains of application, lack of common sense, difficulty in acquiring knowledge, simplistic reasoning methods, lack of distributed functions, and compatibility issues with pre-existing databases gradually emerged.
'Seeking truth from facts to set development goals is the basic principle for establishing a discipline’s development plan. Achieving all-around human-level intelligence in machines is the grand ultimate goal of AI, but reasonable, staged research goals must be set…'
Fifth, the steady development period: mid-1990s–2010. Due to the development of network technology—and especially Internet technology—the convergence of information and data has been accelerating, and the continuous popularization of Internet applications has accelerated the innovation of AI, further spurring the practical application of AI technology. In 1997 the IBM Deep Blue supercomputer defeated the chess world champion Garry Kasparov. In 2008, IBM proposed the “Smarter Planet” concept. These were landmark events of this period.
Sixth, the development boom: 2011–present. With the development of information technology such as big data, cloud computing, the Internet, and the Internet of Things; the ubiquity of data-gathering sensors; and advances in computing platforms like graphics processing units (GPUs); AI technologies represented by deep neural networks are rapidly advancing. This technology made significant progress in bridging the “technical gap” between science and applications, and has broken through a bottleneck from being unusable or hard to use, to being usable in a wide range of applications, including image classification, speech recognition, knowledge Q&A, human-machine gameplay, and autonomous driving. The development of AI has entered a new surge of explosive growth.
By summarizing the experience and lessons learned in the development of AI, we can recognize the following lessons:
(1) Respecting a discipline’s natural laws of development is a prerequisite for promoting its healthy development. The development of science and technology has its own laws; those who obey them prosper while those who disobey them flounder. The development of AI requires a combination of basic theory, data resources, computing platforms, and application scenarios. It is difficult to achieve major breakthroughs when these conditions are not available.
(2) Basic research is the cornerstone of a discipline’s sustainable development. Professor Geoffrey Hinton of the University of Toronto persisted in studying deep neural networks for 30 years, laying the important theoretical foundation for the vigorous development of AI. Google’s DeepMind team for a long time has researched foundational topics such as neuroscience-inspired AI, and has achieved a series of major achievements such as AlphaGo.
(3) Demand for applications is an inexhaustible source of technological innovation. The driving force behind the development of the discipline mainly rolls forward on the twin wheels of science and demand. Aside from the inherent contradictions between knowledge and technological systems, staying close to applications and solving user issues are the greatest sources of and driving forces for innovation. For example, practical application demands—such as for expert systems to break through from theory to application, and in more recent years security monitoring, identity recognition, unmanned driving, and big data analysis for the Internet and Internet of Things—have brought about AI technological breakthroughs.
'The proper mode of supporting innovation should include tolerating failure.'
(4) Cutting across disciplines is a shortcut to innovation breakthroughs. AI research involves information science, neuroscience, psychology, etc. The emergence of AI in the 1950s is itself the result of interdisciplinary interaction. In particular, the successful combination of cognitive science and AI has brought about decades of sustained development of AI neural networks. Some basic scientific problems such as the source of intelligence and the nature of consciousness are potential sources of major breakthroughs and have an important role in promoting the development of AI.
(5) The proper mode of supporting innovation should include tolerating failure. The development of any discipline is not smooth sailing, and the realization of any innovative goal will not happen overnight. The development of AI for more than 60 years vividly illustrates the ups and downs of the development of a discipline’s innovation. It could be said that without the “cold winters” in the development process of the past, there would be no new spring for AI development today.
(6) Seeking truth from facts to set development goals is the basic principle for establishing a discipline’s development plan. Achieving all-around human-level intelligence in machines is the grand ultimate goal of AI, but reasonable, staged research goals must be set according to the level of science, technology, economic, and social development. Otherwise, there will be frustration that affects the development of the discipline. Several downturns in the AI development process were the result of unrealistic development goals.
III. The Current Status and Impact of AI Development
AI, over the course of more than 60 years of development, has made important breakthroughs in theory, technology, and applications. It has become a driving force in pushing forward a new round of science, technology, and industrial revolution, profoundly affecting the world’s economic, political, military, and social development. It has increasingly captured the high-level attention from various national governments, industry, and academia. From the perspective of technology, AI technology breakthroughs focus on specialized intelligence, but the development of general intelligence is still in its starting stage; from the perspective of industry, AI innovation and entrepreneurship is spreading like wildfire, as the technology and business environment are already taking shape; from the societal perspective, the world’s major countries have elevated AI as a national strategy, and the social impact of AI has become increasingly prominent.
(1) Significant breakthroughs have been made in specialized AI. From the perspective of applicability, AI can be roughly divided into specialized AI and general AI. AI technology oriented toward a particular domain (i.e. specialized AI) is characterized by a singular task, clear requirements, clear application boundaries, rich domain knowledge, and relatively simple modeling; therefore, stand-alone breakthroughs have emerged in the AI field. In a stand-alone tests of partial intelligence levels, AI can surpass human intelligence. The recent progress of AI is mainly concentrated in the field of specialized AI, and statistical learning is the theoretical basis for the application of specialized AI. Statistical machine learning theories such as deep learning, reinforcement learning, and adversarial learning have been successfully applied in computer vision, speech recognition, natural language understanding, and human-machine strategic games. For example, AlphaGo defeated the human champion in the Go game. AI programs have surpassed human level performance in large-scale image recognition and face recognition, the 5.1% error rate of speech recognition systems is as good as a professional stenographer, AI systems diagnose skin cancer at the levels of professional doctors, and so on.
(2) General AI is still in its starting phase. The human brain is a system of general intelligence that can draw inferences about other cases from one case, achieve mastery through comprehensive study of a subject, and deal with various issues such as vision, hearing, judgment, reasoning, learning, thinking, planning, design, etc. It can be described as an all-purpose brain. A truly complete AI system should be a general intelligence system. Although breakthroughs have been made in the domain of specialized AI including image recognition, speech recognition, and autonomous driving, the research and application of systems of general intelligence still has a long way to go, and the overall development level of AI is still in its starting phase. The U.S. Defense Advanced Research Projects Agency (DARPA) divides the development of AI into three phases: rule intelligence, statistical intelligence, and autonomous intelligence [DARPA's term is "contextual adaptive intelligence" –Trans]. DARPA believes that the main international trend of AI’s level of performance is still in the second phase. The core technology relies on statistical machine learning including deep learning, reinforcement learning, and adversarial learning. The progress of AI systems in intelligence achievement levels such as perception and learning has been significant, but AI systems have very weak capabilities in areas such as abstraction and reasoning. Generally speaking, the current AI systems can be said to have intelligence, but not wisdom; to have IQ, but not EQ (emotional quotient); to have the ability to compute, but not to calculate; and to have specialized talents but not all-around genius. Therefore, AI still has obvious limitations, and there are still many "cannots," which is a far cry from human wisdom.
(3) AI innovation and entrepreneurship is spreading like wildfire. Global industry has fully realized the significance of a new round of industrial transformation led by AI technology, and have, one after another, adjusted their development strategies. For example, at its 2017 Annual I/O Developers Conference, Google clearly stated that its development strategy was turning from “Mobile First” toward “AI First”; Microsoft’s FY2017 Annual Report for the first time set AI as its vision for the company’s development. The field of AI sits at the forefront of innovation and entrepreneurship. McKinsey reports that global AI research and development investment exceeded $30 billion USD in 2016 and is increasing at a high rate. A report by CB Insights, a world-renowned research organization on venture capital, showed that 1,100 new AI startups were established around the world in 2017. The AI field received a total investment of $15.2 billion USD, an increase of 141% over the previous year.
'At present, a monopoly has not emerged in the industrial structure of intelligent technology.'
(4) The overall position of innovation ecosystems has become a strategic high ground for the development of AI industry. The development history of information technology (IT) and industry is simply the historical interchange of new and old IT giants scrambling to gain a foothold in the IT innovation ecosystem. For example, representative companies in the traditional information industry include Microsoft, Intel, IBM, Oracle, etc., and representative companies in the Internet and mobile Internet technology industry include Google, Apple, Facebook, Amazon, Alibaba, Tencent, Baidu, etc. At present, a monopoly has not emerged in the industrial structure of intelligent technology. Therefore, the global technology industry giants are actively pushing forward distributions of R&D for AI technology ecosystems, and making all-out efforts to seize the commanding heights of AI-related industries. The AI innovation ecosystem includes vertical technology ecosystems such as data platforms, open source algorithms, computing chips, basic software, GPU servers, etc. as well as horizontal commercial and application ecosystems such as those for smart manufacturing, smart medicine, smart security, smart retail, smart home and other uses. In terms of technology ecosystems, AI algorithms; data; graphics processing units (GPUs), tensor processing units (TPUs), and neural network processing units (NPUs); and basic software tasks such as running, compiling, and managing AI algorithms already have a large number of open source resources, such as Google's TensorFlow second-generation AI learning system, Facebook's PyTorch deep learning framework, Microsoft's DMTK distributed learning toolkit, and IBM's SystemML open source machine learning system. In addition, Google, IBM, NVIDIA, Intel, Apple, Huawei, the Chinese Academy of Sciences, and others are actively deploying computing chips in the field of AI. In the AI business and application ecosystem layout, "smart + X" has become an innovation paradigm, for example smart manufacturing, smart medicine, smart security, etc. AI technology rapidly penetrates into innovative consumption scenarios and different industries and reshapes the development of society. This is the most important manifestation of AI as the key driving force of the fourth technological revolution. The competition over AI business ecosystems has turned white-hot. For example, the participants in the field of intelligent driving vehicles include traditional leading car companies such as GM, Ford, Mercedes-Benz, and Toyota, as well as upstart Internet carmakers such as Google, Tesla, Uber, Apple, and Baidu.
(5) AI has been elevated as a significant development strategy for major countries in the world. AI is becoming the engine of a new round of industrial transformation, which will inevitably and profoundly affect the pattern of international industry competition and the international competitiveness of a country. The major developed countries in the world have, one after another, taken the development of AI to serve as a major strategy to enhance international competitiveness and safeguard national security. They have intensified their efforts to put forward policies and strengthen their deployments of resources around core technologies, top talent and standards, in an effort to grasp dominance in the new round of international science and technology competition. Whether it is Germany's "Industry 4.0," America’s "Industrial Internet," Japan's "super-smart society," or China's "Made in China 2025" and other major national strategies, AI is one of the core and critical technologies. In July 2017, the State Council issued the “New Generation Artificial Intelligence Development Plan”, which initiated a new journey for the rapid innovation and development of AI in China.
'The major developed countries in the world have, one after another, taken the development of AI to serve as a major strategy to enhance international competitiveness and safeguard national security.'
(6) The social impact of AI has become increasingly prominent. The social impact of AI is diverse: It not only has the positive effects of stimulating the economy, serving the people's livelihood, and benefiting society, but also may lead to social problems such as loss of control over security, loss of legal precision, loss of moral standards, loss of ethical order, loss of privacy, as well as opportunistic speculation that takes advantage of AI as a hot topic which poses the risk of bubbles. First of all, AI, as a core force of the new round of scientific and technological revolution and industrial transformation, promotes an overall leap of social productivity, pushes forward the upgrading of traditional industries, and drives the rapid development of the “unmanned economy.” In the domains of smart transportation, smart homes, smart medical care, etc., it has a proactive, positive impact. At the same time, we should also see that the legal and ethical issues caused by AI have become increasingly prominent, bringing unprecedented challenges to the current social order and public management system. For example, in 2016, the European Commission's [sic., actually the European Parliament’s –Trans.] legal affairs committee submitted a motion to set the identity of the most advanced automated robots as “electronic persons.” In 2017, Saudi Arabia awarded the robot “Sofia” citizenship. These clearly challenged the traditional civil subject system. So, should AI systems be granted qualifications as legal subjects? In addition, in the new era of AI, personal information and privacy protection, intellectual property rights of content created by AI, discrimination and bias in AI, traffic regulations of driverless systems, and the scientific and technological ethics of brain-computer interfaces and human-computer symbiosis all require us to provide solutions from multiple angles such as laws and regulations, morality and ethics, and social management.
Because of AI’s close relation to human intelligence, broad application prospects, and specialization strengths, it leads people to misunderstand it and produces significant sensationalism. For example, some people mistakenly believe that: AI is simply machine learning (deep learning); that AI and human intelligence are in a zero-sum game; that AI has already reached the level of five-year-old children, and the intelligence level of AI systems is about to completely surpass human levels; that within 30 years, robots will rule the world and humans will become slaves to AI; and so on. These misconceptions will bring about adverse effects for the development of AI. There are also many people whose expectations for AI are too high, thinking that general intelligence can be realized very quickly, and as long as one gives the robot commands, it can do anything. In addition, the phenomenon of intentional hyping and packaging the concept of AI to gain improper benefits has occurred from time to time. Therefore, we have a duty to popularize AI knowledge among the public, and guide the government, enterprises, and the general public to objectively recognize and understand AI.
IV. The development trend and prospects of AI
After more than 60 years of development, AI has made breakthroughs in the constraining factors of the “Three Suans (三算)” — suanfa (algorithms), suanli (computing power), and suanliao (data). This has spread to the Internet, Internet of Things and other broad application scenarios, and begun to enter a golden period of vigorous development. From a technical perspective, AI is currently at a technical inflection point: going from “unusable” to “useable”, but the distance to “useful” still has bottlenecks such as those related to data, energy consumption, generalization, interpretability, reliability, security, etc. There is a huge space for innovation and development—from specialized intelligence to general intelligence, from machine intelligence to integration of human and computer intelligence, from "human labor + intelligence" to autonomous intelligence. Preparations are being made for new, post–deep learning theoretical systems. From the perspective of industrial and social development, AI realizes the transformation of productivity and production relations through the permeation into and integration of economic and social fields, and drives human society to move toward a new civilization. A community of shared destiny for humanity will create a rational mechanism to safeguard AI’s secure, controllable, and reliable development. In general, the spring of AI has just begun, the space for innovation is huge, and the application prospects are broad.
(1) From specialized intelligence to general intelligence. How to realize the development leap from narrow or specialized AI (also known as weak AI or single-domain intelligence) to general AI (also known as strong AI, multi-domain intelligence) is the inevitable trend of the next generation of AI development. It is also a challenging problem in the area of international research and application. In October 2016, the U.S. National Science and Technology Council issued the National Artificial Intelligence Research and Development Plan, which laid out a medium- and long-term development strategy that contained an emphasis on researching general AI. Demis Hassabis, founder of DeepMind, is moving toward the goal of “creating general artificial intelligence that solves all problems in the world.” Microsoft established a general AI lab in July 2017, and more than 100 scientists involved in perception, learning, reasoning, and natural language understanding are among those participating.
(2) From AI to human-machine hybrid intelligence. An important research direction of AI is to draw lessons from the research results of neuroscience and cognitive science, to study new intelligent computing models and methods based on the intrinsic qualities and mechanisms that generate intelligence, to realize intelligent systems equipped with the mechanisms of brain neural information processing and human-level intelligent behavior. In the brain plans launched by the United States, the European Union, Japan and other countries and regions, brain-like intelligence has become one of the core goals. The UK Engineering and Physical Sciences Research Council (EPSRC) has announced and launched a brain-like intelligence research program. Human-machine hybrid intelligence aims to introduce human function or cognitive models into AI systems, improve the performance of AI systems, make AI become the natural extension and expansion of human intelligence, and solve complex problems more efficiently through human-machine collaboration. Human-machine hybrid intelligence has received high-level attention by China's New Generation Artificial Intelligence Development Plan, the U.S. Brain Initiative, Facebook (brain-computer interface for speech and text), and Tesla founder Elon Musk (embedded chips in human brains and brain-computer interfaces).
(3) From "human labor + intelligence" to autonomous intelligent systems. Current research on AI focuses on deep learning, but the limitation of deep learning is that it requires massive amounts of manual intervention: artificially designing deep neural network models, manually setting application scenarios, manually collecting and labeling a large amount of training data (very time-consuming and laborious), users require manual labor to adapt to intelligent systems, etc. Therefore, researchers have begun to pay attention to the autonomous intelligent methods with reduced manual intervention, and improve the ability of machine intelligence to engage in self-learning about the environment. For example, AphaZero started from zero and achieved “general chess AI” for Go, chess, and shogi through self-play reinforcement learning. Regarding the automated design of AI systems, the automated machine learning system (AutoML) put forward by Google in 2017 attempted to reduce AI personnel costs by automatically creating machine learning systems.
(4) AI will accelerate cross-permeation with other disciplines. AI itself is a comprehensive frontier discipline and a highly interdisciplinary composite field. The research scope is extensive and extremely complex. Its development needs to be deeply integrated with discipline such as computer science, mathematics, cognitive science, neuroscience, and social science. Alongside the breakthroughs of super-resolution optical imaging, optogenetic manipulation, the “transparent brain,” somatic cell cloning, and other technologies, the development of brain and cognitive sciences has opened a new era, capable of large-scale and more precise analysis of the neural circuit foundations and mechanisms of intelligence. AI will enter an intelligent phase of biological enlightenment, relying on the discoveries made by disciplines such as biology, neuroscience, life sciences and psychology, turning mechanisms into computable models. At the same time, AI will also promote the development of brain and cognitive sciences, life sciences, and even traditional sciences such as chemistry, physics, and materials. For example, in 2018, Massachusetts Institute of Technology in the United States launched the MIT Intelligence Quest which united five major schools for a collaborative program of research.
(5) The AI industry will flourish. With the further maturing of AI technology and the increasing investment from the government and industry, the cloud-ization of AI applications will continue to accelerate, and the scale of the global AI industry will enter a period of rapid growth in the next decade. For example, in September 2016, consulting firm Accenture released a report that the application of AI technology will inject new momentum into economic development, and can increase the current rate of labor productivity by 40%; 12 developed countries, including the United States, Japan, Britain, Germany, France, etc., (which now account for half of the global economy) will have an average annual economic growth rate that can double by 2035. A 2018 McKinsey study indicated that by 2030, AI will add $13 trillion to economic activity.
(6) AI will push humanity forward into a widely enjoyed smart society. The innovation model of “AI + X” will gradually become more mature with the development of technology and industry. This will have a revolutionary impact on productivity and the industrial structure, and will push humanity forward into a popularized smart society. In 2017, IDC, the International Data Corporation, pointed out in the white paper, titled “Information Flows Lead the Trend to the New Era of Artificial Intelligence,” that AI will improve the operational efficiency of various industries in the next five years, including 82% in education, 71% in retail, 64% in manufacturing, and 58% in finance. China's economic and social transformation and upgrading has a great need for AI. Pulled along by the demand of consumption cases and industry applications, it is necessary to break through AI’s bottlenecks in perception, interaction, and decision-making, and to promote the integration and upgrading of AI technology with all trades and professions. It is necessary to build a number of benchmark innovations in application scenarios to achieve a low-cost, high-efficiency, and wide-ranging popularized smart society.
(7) International competition in the field of AI will become increasingly fierce. "In the future, whoever leads in AI will be the ruler of the world." In April 2018, the European Commission planned to invest $24 billion USD into AI in 2018–2020. The French president announced the French Artificial Intelligence Strategy in May 2018. The aim was to welcome the new era of AI development and make France into an AI great power. In June 2018, Japan's "Investments for the Future Strategy" focused on pushing forward the construction of Internet of Things and the application of AI. The world's military powers have gradually formed a competitive situation which takes accelerating the development of intelligentized weapons and equipment as its core. For example, the first "Defense Strategy" report issued by the Trump administration of the United States has proposed seeking to maintain military superiority through technological innovations such as AI to ensure that the United States wins the wars of the future. Russia proposed in 2017 that the military embrace "intelligentization" so as to multiply the power of "traditional" weapons such as missiles and drones.
(8) The sociology of AI will be put on the agenda. Water can keep the boat afloat but can also sink it. Any high technology is also a double-edged sword. Alongside the thorough development of AI and the increasing spread of its applications, its social impact has become increasingly apparent. The applications of AI must be appropriate, its hold on things must have limits, and its management must be regulated so as to effectively control negative risks. In order to ensure the healthy and sustainable development of AI, as well as ensure the development of AI benefits the people, it is necessary to systematically and comprehensively study the impact of AI on human society from the perspective of sociology, deeply analyze the possible impacts of AI on future economic and social development, and develop sound AI laws and regulations to avoid possible risks and ensure the positive effects of AI. In September 2017, the United Nations Interregional Crime and Justice Research Institute (UNICRI) decided to establish the first United Nations Centre for Artificial Intelligence and Robotics in The Hague to standardize the development of AI. In April 2018, 25 countries in Europe signed a "Declaration of Cooperation on Artificial Intelligence" to promote the development of AI from the level of national strategic cooperation, ensure the competitiveness of European AI research and development, and jointly face the opportunities and challenges posed by AI for society, the economy, ethics, law, and other aspects.
V. China’s AI Development Trends & Thinking
The overall situation of AI development in China is good. According to the 2018 World Blueprint for Artificial Intelligence Industry Development released by the China Academy for Information and Communications Technology (CAICT) and Gartner in September of 2018, the total number of AI enterprises in China (excluding Hong Kong, Macao, and Taiwan) ranks second in the world (1,040 enterprises), behind only the United States (2,039 enterprises). In terms of the overall level and applications of AI, China is also at the forefront of the international community, with great development potential. China shows promise of becoming the global frontrunner. However, we must also clearly see that AI development in China faces the risk of overheating and forming a bubble, especially in basic research, technological systems, the application ecosystem, innovative talent, and legal norms. In general, the status of China’s AI development can be summarized as, “highly emphasized, in a heartening situation, the gaps are not small, and the outlook is promising.”
'China shows promise of becoming the global frontrunner. However, we must also clearly see that AI development in China faces the risk of overheating and forming a bubble…'
First, it is highly emphasized. The party and the state attach great importance to and vigorously develop AI. Since the 18th National Congress of the Communist Party of China, General Secretary Xi Jinping has placed innovation at the core of the country's overall development, has attached great importance to the development of AI, and has repeatedly talked about the importance of AI, pointing out the direction for AI to empower the New Era. In July 2016, General Secretary Xi clearly pointed out that the development of AI technology will profoundly change the life of human society and change the world, and that China should seize the opportunity, and take initiative to seize the high ground in this high-tech field. In the party’s 19th National Congress report, General Secretary Xi emphasized China “must push forward the deep integration of the Internet, big data, AI, and the real economy.” At the 2018 meeting of the Chinese Academy of Sciences and the Chinese Academy of Engineering, General Secretary Xi once again stressed the need to “promote the deep integration of the internet, big data, AI, and the real economy, and enlarge and strengthen the digital economy.” In the “Government Work Report” in 2017 and 2018, Premier Li Keqiang mentioned the need to strengthen the development of the new generation of AI. In July 2017, the State Council issued the “New Generation Artificial Intelligence Development Plan,” which put bringing about the new generation of AI at the level of national strategy, and described the roadmap for the development of AI in China for 2030, which aims to establish AI superiority and seize the strategic initiative in this new science and technology revolution. AI will henceforth be a major national strategy for a period of time. National ministries and commissions such as the National Development and Reform Commission, the Ministry of Industry and Information Technology, the Ministry of Science and Technology, the Ministry of Education, and the Cyberspace Administration of China, as well as local governments such as those in Beijing, Shanghai, Guangdong, Jiangsu, and Zhejiang have all introduced policies to manage the development of AI.
Second, the situation is heartening. According to statistics from the 2017 Elsevier document database SCOPUS, China publishes the most AI papers in the world. Since 2012, China’s new patents in the field of AI have begun to surpass those from the United States. According to statistics from the “China Artificial Intelligence Development Report 2018” released by Tsinghua University, China has become the world’s largest country for AI investment and financing. AI companies in China are at the global forefront of AI applications in facial recognition, speech recognition, security monitoring, smart speakers, smart homes, etc. In the past two years, Tsinghua University, Peking University, the University of the Chinese Academy of Sciences, Zhejiang University, Shanghai Jiaotong University, Nanjing University, and other universities have set up AI institutes. The China Conference on Artificial Intelligence (CCAI), which began in 2015, has been successfully held for four consecutive years and has continuously expanded. Education, research, and academic activities in AI are emerging in an endless stream.
'There is still a relatively large gap between the level of the world’s leader and China’s level in basic research, novel achievements, top talent, technology ecosystems, foundational platforms, and standards and specifications.'
Third, the gap is not small. There is still a relatively large gap between the level of the world’s leader and China’s level in basic research, novel achievements, top talent, technology ecosystems, foundational platforms, and standards and specifications. A 2018 research report by the University of Oxford in the UK pointed out that China’s AI development capabilities are roughly half those of the United States. At present, China’s frontier theoretical innovations in AI are still in the “follow-and-run” stage. Most of the innovations are biased toward technical applications, and a “top-heavy” imbalance results. On the list of Top 700 most talented AI figures, although China possessed the second-greatest number, the number of people selected in China was far lower than half the number selected from the United States. According to LinkedIn’s “Global AI Talent Report,” as of the first quarter of 2017, the number of professionals in the AI field in the world exceeded 1.9 million. Among them, more than 850,000 were in the United States, and only 50,000-plus were in China, resulting in a ranking of 7th in the world. In 2018 Compass Intelligence, a market research consultancy, ranked more than 100 AI computing chip companies worldwide. No company in China was in the top 10. In addition, China’s AI open source community and technology ecosystem are relatively lagging behind, the establishment of technical platforms needs to be strengthened, and international influence needs to be improved. China’s participation in the development of international standards for AI is not sufficient, and the formulation and implementation of domestic standards are lagging behind. The process of formulating and perfecting laws and regulations related to AI in China needs to be accelerated, and there is still a lack of in-depth analysis of possible social impacts.
'In the future development of China, the “intelligence dividend” will be expected to make up for the shortage of the demographic dividend.'
Fourth, the outlook is promising. The development of AI in China has the comprehensive advantages of market scale, application scenarios, data resources, human resources, smartphone penetration, capital investment, and national policy support. The development prospects for AI are promising. According to the 2017 report “Artificial Intelligence: Helping China’s Economic Growth” released by Accenture, the world’s top management consulting firm, AI is expected to boost China’s labor productivity by 27% by 2035. The “New Generation Artificial Intelligence Development Plan” issued by China proposes that by 2030, AI core industries will exceed 1 trillion yuan, driving related industries to exceed 10 trillion yuan. In the future development of China, the “intelligence dividend” will be expected to make up for the shortage of the demographic dividend.
Human society has begun to enter the Intelligence Era, and AI is leading the general trend, and is irreversible. After more than 60 years of accumulated experience, AI has begun to enter a period of explosive growth. With the innovative development of AI itself and its full permeation into the economy and society, this will last for a long period of time. Now is a major historical opportunity for China to strengthen its AI position, harvest AI dividends, and show the way towards the Intelligence Era’s historical opportunities. How to choose the Chinese path, seize China’s opportunity, and demonstrate the wisdom of China in the AI boom requires deep thinking.
'Some organizations with ulterior motives intentionally overhyped AI and packaged the concept of AI in order to seek improper benefits.'
(1) Establish a rational and pragmatic development concept. After AlphaGo defeated Lee Sedol in the great man vs. machine Go battle, the public mistakenly believed that AI was omnipotent. Some local governments, social enterprises, and venture capital funds thus swarmed like bees and unrealistically developed the AI industry. Some organizations with ulterior motives intentionally overhyped AI and packaged the concept of AI in order to seek improper benefits. This kind of “swarm then disperse in confusion” behavior of following the crowd is not conducive to the healthy and sustainable development of AI. No kind of development can always be at a high point; where there are peaks there are valleys. This is an objective law. According to the technological development curve published by Gartner Consulting, the current popular AI technologies and fields—such as intelligent robots, expert cognitive assistants, machine learning, and autonomous driving are in a period of inflated expectations, but the development of general AI and AI as a whole is still in the initial stage. AI still has many limits. Realizing autonomous intelligence and general intelligence for machines in any real environment still requires medium- and long-term accumulation of theory and technology, and AI’s penetration and integration into traditional fields such as industry, transportation and medical care are long-term processes. It is difficult to get results overnight. Therefore, the development of AI cannot be aimed at short-term profit, and we must fully take into account the limitations of AI technology, fully recognize the long-term and arduous nature of AI’s reshaping of traditional industries, rationally analyze the development needs of AI, rationally set AI development goals, rationally select an AI development path, and pragmatically advance AI development initiatives. This is the only way to the healthy and sustainable development of AI.
(2) Strengthen the solid foundation of original research. AI’s frontier of basic theoretical research is the cornerstone of AI technological breakthroughs, industry innovation, and industrialization. At this crucial development point, in order to have the final say, China must make major breakthroughs in the basic theory and cutting-edge technology of AI. According to the SCOPUS statistics from the 2017 Elsevier document database, although China ranks 1st in the number of AI publications, it ranks 34th in field-weighted citation impact. In order to objectively evaluate China’s overall strength in the basic research of AI, we searched the SCI periodicals and mainstream AI academic conferences such as the Conference on Neural Information Processing Systems (NIPS) for statistics on keywords including general intelligence, deep learning, brain-like intelligence, brain-knowledge integration, and human-machine games. There is clearly a huge gap between China and the United States in terms of the basic strength at the frontier of AI: in terms of the number of high-quality papers (according to statistics from the Chinese Academy of Sciences’ SCI Level 1 Paper Standard), the United States has 5.34 times more (1,325) than China (248); in terms of human talent (SCI paper authors), the United States has 2.12 times more (4,804) than China (2,267).
China should respond to the highest international standards and build a future-oriented AI basic science research center, focusing on the development of original, basic, forward-looking and breakthrough AI science. Researchers should be encouraged to carry out leading original scientific research at the forefront of AI. Through the integration of AI and brain cognition, neuroscience, psychology, and other disciplines, China should focus on major fundamental scientific issues in the field of AI, form an internationally-influential original AI theoretical system, and build our nation’s independent and controllable AI technology innovation ecosystem to provide the theory support that will allow China to lead.
'American companies such as Google, IBM, Microsoft, and Facebook have actively built innovation ecosystems, seized the innovative high ground, and already in the international AI industry hold the upper hand in AI chips, servers, operating systems, open source algorithms, cloud services, and autonomous driving, among others.'
(3) Construct an independent and controllable innovation ecosystem. American companies such as Google, IBM, Microsoft, and Facebook have actively built innovation ecosystems, seized the innovative high ground, and already in the international AI industry hold the upper hand in AI chips, servers, operating systems, open source algorithms, cloud services, and autonomous driving, among others. China’s AI open source community and technological innovation ecosystem are comparatively lagging, the strength of technology platform construction needs to be reinforced, and international influence remains to be improved. The U.S. ban on ZTE fully demonstrates the importance of independent, controllable “core-, high-, and foundational” technologies. In order to avoid repeating this disaster, China should learn its lesson about importing core electronic components, high-end general-purpose chips, and foundational software. China should focus on preventing the “hollowing out” risk of the AI era, as well as systematically deploy and focus on the development of AI’s “new core-, high-, and foundational” technologies. “New” refers to the new type of open innovation ecosystems such as civil-military integration, and the integration of industry, academia, and research. “Core” refers to core and critical technologies and components, such as advanced machine learning technology, robust pattern recognition technology, and low-energy-use intelligent computing chips. “High” refers to high-end integrated application systems and platforms, such as machine learning software and hardware platforms, and large-scale data platforms. “Foundational” refers to theories and methods with great innovative significance and potential for driving technological progress, such as brain-computer interfaces, brain-like intelligence, etc.
In addition, we need to pay attention to the establishment of AI technology standards, product performance, and system security testing. Particularly as China is at the global forefront of AI technology applications, it should seize its right to speak in the formulation of international AI standards, and through the implementation of these standards should accelerate the process by which AI drives economic and social transformation.
(4) Establish a coordinated and efficient innovation system. The escalation of China’s economic and social transformation lays out major demands for AI, but it is difficult for a single innovative agent to achieve comprehensive breakthroughs in policies, markets, technologies, and applications. At present, China’s academia, industry, and business sectors have an obvious tendency to each go their own way in the development of AI. The open sharing of data resources is insufficient, and the effective integration of industry resources is lacking. In contrast, the United States has already formed a collaborative AI innovation system coordinated across the entire society, the whole scene, and the entire ecosystem, and has done a very good job of civil-military integration and integrating industry, academia, and research. China should further reform and innovate in terms of institutional mechanisms, and establish an AI collaborative innovation system that integrates “military, government, industry, academia, research, and applications.” For example, the country should conduct top-level design and strategic planning, promote “AI+X” industry integration, as well as smash industry barriers and administrative obstacles. Leading technology enterprises should show the way in establishing technology innovation ecosystems and in breaking through the major technology bottlenecks of AI; scientific research institutions at colleges and universities should advance personnel training and original innovation, exert themselves to build public data resources and technology platforms, jointly establish a number of benchmark application innovation scenarios, and promote the deep application of mature AI technologies in cities, medical care, finance, culture, agriculture, communication, energy, logistics, manufacturing, safety, service, education, etc., and construct a low-cost, high-efficiency, and inclusive intelligent society.
(5) Accelerate the cultivation of innovative talent. The key to AI development is human talent. The shortage of mid- and high-end talent has already become a main bottleneck blocking China’s AI from becoming bigger and stronger. In addition, the AI technology literacy of the Chinese public needs to be further improved, and everyone needs to adapt to the science and technology wave of the Intelligence Era. While strengthening the training and import of leading AI talents, it is also necessary to cultivate AI innovation and entrepreneurial talents at multiple levels for technological innovation and industrial development. The “New Generation Artificial Intelligence Development Plan” proposes to gradually carry out the National Intelligent Education Project and set up AI courses in primary and secondary schools. At present, popular science AI activities are welcomed by schools all over the country, but there is a lack of easy-to-understand and high-quality AI teaching materials, educational equipment, lab facilities, as well as no open and shared interactive teaching resource platform. Relevant national departments should attach great importance to basic work in the field of AI education, increase investment, organize superior strength, strengthen high-level AI education content and the construction of resource platforms, accelerate the specialized training of qualified AI teachers, and comprehensively ensure the launch and development of AI education in China, in areas ranging from educational materials to teaching aids to teachers.
'Through AI technology development, developed countries control the upper echelon of supply chain resources. Formidable technological gaps and industrial barriers may further widen the gap in productivity growth between developed and developing countries.'
(6) Promote shared global governance. AI will reshape the global political and economic order. Through AI technology development, developed countries control the upper echelon of supply chain resources. Formidable technological gaps and industrial barriers may further widen the gap in productivity growth between developed and developing countries. The United States, Japan, Germany, etc. make up for their labor cost disadvantages through technological breakthroughs and the extensive application of AI and robots, hoping that manufacturing industries will flow back to developed countries from emerging countries. At present, China is the only one in the lineup of developing countries that is expected to become a global leader in AI competition. It should adopt a route different from some countries’ “economic monopolization, technological protectionism, and trade bullying,” and as soon as possible build up a structure of openly-shared, high-quality, low-cost, universally-beneficial, and global AI technology and application platforms, in line with the national “Belt and Road” strategy. China should export high-level, low-cost “smartly made-in-China” achievements to economically underdeveloped regions such as Asia, Africa, and South America, and offer China’s Intelligence Era proposal in order to make China’s contribution to allowing the Intelligence Era to universally benefit the community of shared future for humanity!
'In the field of AI, China’s formulation of laws and regulations and risk management are relatively lagging.'
(7) Formulate scientific, rational laws and regulations. In order to truly harvest the benefits of AI, we must first ensure its secure, controllable, and reliable development. Developed countries and regions such as the United States and Europe attach great importance to issues of law and regulation in the AI field. The White House has organized seminars and consultations in this area on a number of occasions; industry giants such as Tesla have led the establishment of organizations such as OpenAI to promote and develop friendly AI in a way that benefits the entire human race; researchers freely signed the 23 “Asilomar AI Principles,” intending to seize a key moment in the regulation of AI research and application. In the field of AI, China’s formulation of laws and regulations and risk management are relatively lagging. This lag is not compatible with the overall situation of AI development in China at present, and may become a major constraint for the next step of innovation and development of AI in the country. Therefore, it is necessary to vigorously strengthen legislative research in the field of AI, formulate corresponding laws and regulations, establish a robust, open, and transparent AI supervision system, and create a good regulatory environment for the innovation and development of AI.
(8) Strengthen and encourage AI social research. The social impact of AI will be far-reaching and comprehensive. In planning ahead, we will systematically study the possible effects of AI from various dimensions including national security, social governance, employment structure, ethics and morality, and protection of privacy, and we will formulate reasonable and feasible countermeasures to ensure the positive effects of AI. We should vigorously strengthen the popularization of science in the field of AI, create an efficient dialogue mechanism and communication platform between science/technology and ethics, eliminate the public’s misunderstanding of and panic towards AI, and create a rational, pragmatic, energetic, and healthy social atmosphere for the development of AI.
After more than 60 years of development, AI has entered a period of strategic opportunity for innovation breakthroughs and beneficial industrial applications. It will definitely have a revolutionary impact on productivity, industrial structures, and the international order. It will push humanity to become an inclusive, intelligent society. However, we need to recognize that the overall development of general AI and AI is still in its infancy. AI is not omnipotent, and AI has many limits. We should adopt a rational and pragmatic development path; firmly promote basic research, a technology ecosystem, talent development, laws and norms, etc.; innovate amidst openness; develop amidst innovation; run at top speed to win the Intelligence Era; and focus on becoming an AI science and technology superpower!