Authoritative Publication: A New Generation of White Paper on the Development of Artificial Intelligence (2017)

Source: Robot Lecture Hall

Guiding units, expert consultants and writers

Gu asked

Pan Yunhe Academician of Chinese Academy of Engineering


Directed by

Ministry of Industry and Information Technology


steering Committee

Xie Shaofeng Director of the Ministry of Industry and Information Technology

Li Guanyu Deputy Director of the Ministry of Industry and Information Technology

Xu Xiaolan Vice Chairman and Secretary General of Chinese Institute of Electronics

Zhang Hongtu Secretary of the Party Committee and Deputy Secretary-General of China Electronics Society Headquarters

Supermarket Director of Software Department, Ministry of Industry and Information Technology

Fu Yongbao, Investigator, Software Division, Xinsoft Division, MIIT


Expert Committee (in no particular order, sorted by strokes of last name)

Wang Shijin, Deputy Dean of HKUST Xunfei Research Institute

Wei Qing Chief Technology Officer, Microsoft China

Song Bo General Manager of Guo'anrui (Beijing) Technology Co., Ltd.

General Manager of JD Finance

Wu Gansha Co-founder and CEO of Yushi Technology

Ji Xiangyang, Professor, Department of Automation, Tsinghua University

Chen Lijuan Head of Alibaba Artificial Intelligence Lab

Liang Jiaen Chairman and CTO of Yunzhisheng Information Technology Co., Ltd.

Cui Yan Dean, Sino-German Institute of Artificial Intelligence

Cai Xiongshan Deputy Director, Legal Research Center, Tencent Research Institute


Writing unit

Chinese Institute of Electronics


Writer

Li Zheng Zhou Yifeng Ma Liang Ling Xia

Li Yan Zhang Yani Xu Hualei Zhang Yan

Zhang Li Chen Mengmeng Fan Jiangyang Zhu Yi

Li Junping Yan Deli Xie Zhongye Chen Yan



Report link: https://pan.baidu.com/s/1oAn8flo Password: 3gme

Preparation of summary

(I) Compilation background

Since the concept was established in 1956, artificial intelligence has developed for more than 60 years. With the profound changes in the information environment and data foundation, it has begun to enter a new stage of development, showing big data, cross-media, group nature, and autonomy. The new development characteristics of humanization and human-machine integration have rapidly changed from academic-driven development to demand-driven development. It must be closer to human intelligence than at any time in history. It can not only master urban development, ecological protection, Economic management, financial risk and other macro systems provide guidance, and can also provide solutions for design and manufacturing, health care, transportation management, energy conservation and other micro-fields. China is at the tackling stage of industrialization, urbanization, informatization, and agricultural modernization. There is an urgent need to accelerate the innovative application of artificial intelligence in various industries and fields of the national economy and society, to promote industrial quality and efficiency, and to improve people's living standards. Major structural imbalances in economic operations. In view of this, it is necessary to study and develop a new generation of artificial intelligence development white paper, to clarify the technical framework, key links and application prospects of artificial intelligence in the new period and new situation, and to provide measures and recommendations for promoting the advancement of artificial intelligence's key technologies and industrialization applications. , To further promote the continued healthy and rapid development of China's intelligent frontier emerging industries, and strongly support China's deep integration of information technology and industrialization to a new level.

(2) Establishment goals

The first is to clarify the main development direction of the new generation of artificial intelligence. The system summarizes its main driving factors and the most typical characteristics.

The second is to study the technical framework of the new generation of artificial intelligence, sort out the technological evolution track, and propose a basic and general technical system.

The third is to explore the industrial boundaries of next-generation artificial intelligence, divide industry categories and application scenarios, and study related investment and financing characteristics and trends.

The fourth is to propose feasible measures to promote the development of next-generation artificial intelligence and related technologies and industries, provide decision-making references for relevant industry authorities, and provide guidance for the healthy and orderly development of the industry.

(3) Preparation method

The first is to study and study relevant domestic and foreign strategic policy documents, and to fully draw on and refer to major domestic and foreign research trends and achievements. Mainly include: "Preparing for the Future of Artificial Intelligence", "National Artificial Intelligence Research and Development Strategic Plan" issued by the White House of the United States; "Robots and Artificial Intelligence" issued by the Science and Technology Committee of the British House of Commons; released by the British Government Science Office Opportunities and Impacts of Artificial Intelligence on Future Decisions, as well as the Modern Industrial Strategy announced by the British government in January 2017 and the Digital Strategy announced in March; the Roadmap for the Industrialization of Artificial Intelligence developed by the Japanese government ; China's "Guidelines on Actively Promoting the" Internet + "Action of the State Council" and "Thirteenth Five-Year Plan for National Strategic Emerging Industry Development Plan" on artificial intelligence. At the same time, they learned about the EU ’s “Human Brain Project”, Germany ’s “Smart Data Project”, Japan ’s “Super Intelligent Society” and “Advanced Integrated Intelligent Platform Project”.

The second is to interview well-known domestic experts and scholars to conduct a full discussion on the connotation, extension and characteristics of the new generation of artificial intelligence. The new generation of artificial intelligence is both innovative and inherited. It is both connected and different from the artificial intelligence that has been discussed in the past. It must have both breakthroughs and sufficient absorption and reference in the research content. Through discussions and discussions with well-known academicians, academics and industry experts in the field of artificial intelligence in China, especially the in-depth study and discussion on the article "Artificial Intelligence Towards 2.0" by Academician Pan Yunhe of the Chinese Academy of Engineering, it laid a systematic theory for the compilation of the white paper basis.

The third is to investigate well-known domestic and foreign artificial intelligence companies, collect and analyze typical cases from practical applications. Attach great importance to specific products, services and solutions in the field of artificial intelligence, enter a group of domestic and foreign artificial intelligence companies with advanced levels and distinctive advantages in technology or industry to conduct in-depth research, and invite some companies to be responsible for the technology or strategy People participated in the preparation of the white paper.

(4) Special statement

1. The research theme fully considers interaction and response with national planning

The concept of artificial intelligence has been calculated from the day of its birth. It has been more than 60 years old and is not a new thing born out of nowhere. It is just influenced and driven by the continuous optimization of algorithm models, the massive increase of data and information, and the significant increase in computing power in recent years, showing different levels of development and characteristics. The research theme of this white paper was named "Artificial Intelligence 2.0", and it has been changed to "New Generation of Artificial Intelligence". It is to echo academician research articles, speeches from leaders of ministries and commissions, and the forthcoming national plan, focusing on the new artificial intelligence. Research on trends, new features, and new models is not meant to suggest a completely new research object.

2. Research scope focuses on technology and industrial development

In the field of artificial intelligence, major innovations, changes, and breakthroughs comparable to relativity, quantum theory, computers, and the Internet are being bred. Artificial intelligence has historically stood in the vanguard of the times and will bring a subversive and violent impact on the intelligent wave of human economic and social development. To study artificial intelligence, we must study its detailed status and role in human production and life, involving all aspects, including morals, law, ethics, culture and other fields. This white paper has been prepared to provide decision-making reference for relevant industry authorities and enterprises. It focuses on research at the two major levels of technology and industry, and does not involve other aspects for the time being.

3. The research content still needs to be further enriched

At present, various research and consulting institutions have launched related reports around the subject of artificial intelligence. Their views are consistent and there are some different opinions. The main points and contents of this white paper only represent the current research and judgment on artificial intelligence by the compilation team. Experts, scholars and business representatives from all parties are welcome to put forward their valuable opinions to jointly promote the timely update and correction of the white paper. At the same time, with the advancement of artificial intelligence technology, the development of the industry, and the change of models, the content of the white paper will be further enriched.

Chapter 1 Summary of New Generation Artificial Intelligence Research

The origin of artificial intelligence was the "Artificial Intelligent (AI) Summer Seminar" held at Dartmouth College in the United States in 1956. It reached two peaks in the late 1950s and early 1980s. However, due to limitations such as technical bottlenecks and application costs, they all fell into a trough. Currently, under the guidance of the new generation of information technology, data is rapidly accumulating, computing capacity is greatly improved, algorithm models are continuously evolving, industry applications are rapidly emerging, and the artificial intelligence development environment has undergone profound changes. Cross-media intelligence, swarm intelligence, autonomous intelligent systems, Hybrid intelligence has become a new development direction, and artificial intelligence has stood on the top of the tide of technological development for the third time.

(I) Brief Development History of Artificial Intelligence

Since its birth, artificial intelligence has a 60-year history of development, and has roughly experienced three waves. The first wave was from the late 1950s to the early 1980s; the second wave was from the early 1980s to the end of the 20th century; the third wave was from the early 21st century to the present. In the first two waves of artificial intelligence, due to the failure of technology to achieve breakthrough progress, related applications have always failed to achieve the expected results and cannot support large-scale commercial applications. After two highs and lows, artificial intelligence finally fell Quiet. With the rapid development of information technology and the rapid popularization of the Internet, marked by the introduction of the deep learning model in 2006, artificial intelligence has ushered in the third rapid growth.

Figure 1 Schematic diagram of the development process of artificial intelligence

Source: Sorted by Chinese Institute of Electronics

1. The first wave: artificial intelligence was born and developed rapidly, but technical bottlenecks are difficult to break through

Symbolism prevails, and artificial intelligence is developing rapidly. From 1956 to 1974 was the first golden age of the development of artificial intelligence. Scientists introduced symbolic methods into statistical methods for semantic processing. Knowledge-based methods emerged, and human-computer interaction became possible. Scientists have invented a number of algorithms with significant impact, such as the prototype Bellman formula for deep learning models. In addition to new advances in algorithms and methodologies, scientists have also made machines with preliminary intelligence. If you can prove the applied machine STUDENT (1964), you can realize the simple human-machine dialogue ELIZA (1966). Artificial intelligence is developing so fast that researchers generally think it is only a matter of time before artificial intelligence replaces humans.

The model has limitations, and artificial intelligence has entered a trough. From 1974 to 1980. Bottlenecks in artificial intelligence are gradually emerging. Logic provers, perceptrons, and reinforcement learning can only complete specified tasks. They cannot cope with tasks beyond the scope. The intelligence level is relatively low-level and the limitations are more prominent. The reasons for this limitation are mainly reflected in two aspects: one is that the mathematical model and mathematical methods based on artificial intelligence are found to have certain defects; the second is that the complexity of many calculations has grown exponentially and cannot be completed according to existing algorithms Computing tasks. The inherent defect is the bottleneck encountered in the early development of artificial intelligence. R & D institutions' enthusiasm for artificial intelligence has gradually cooled down, and funding for artificial intelligence has been reduced or cancelled accordingly. Artificial intelligence has entered the trough for the first time.

2. The second wave: model breakthroughs promote preliminary industrialization, but there are cost barriers to popularization and application

The mathematical model achieved a major breakthrough and the expert system was applied. In the 1980s, artificial intelligence was once again in the public eye. The mathematical models related to artificial intelligence have made a series of major inventions, including the well-known multi-layer neural network (1986) and BP back-propagation algorithm (1986). This has further spawned highly intelligent machines capable of playing chess with humans ( 1989). Other achievements include the realization of a machine that can automatically recognize postal codes on envelopes through artificial intelligence networks, with an accuracy of over 99%, which has exceeded the level of ordinary people. At the same time, Carnegie Mellon University created an expert system for DEC (1980). This expert system can help DEC save about $ 40 million in annual costs, especially in terms of decision-making. Encouraged by this, many countries, including Japan and the United States, once again invested heavily in the development of so-called fifth-generation computers (1982), which were then called artificial intelligence computers.

The cost is high and difficult to maintain, and artificial intelligence has once again entered a trough. To promote the development of artificial intelligence, researchers designed the LISP language and developed the Lisp computer for the language. This model has higher instruction execution efficiency than general-purpose computers, but it is expensive and difficult to maintain, and it is always difficult to popularize it on a large scale. At the same time, from 1987 to 1993, Apple and IBM started to promote the first generation of desktop computers. With the continuous improvement of performance and the continuous decrease in sales prices, these personal computers gradually took advantage in the consumer market, and more and more Many computers entered personal homes. The expensive Lisp computers were gradually eliminated by the market due to their age and obsolescence, and expert systems gradually faded out of people's vision. The artificial intelligence hardware market has shrunk significantly. At the same time, government funding has started to decline, and artificial intelligence has once again entered a trough.

3. The third wave: the information age has spawned a new generation of artificial intelligence, but there are many hidden concerns in future development

Emerging technologies are emerging rapidly, and artificial intelligence development has entered a new stage. With the popularity of the Internet, the ubiquity of sensors, the emergence of big data, the development of e-commerce, and the rise of the information community, data and knowledge are cross-fused and interacted between human society, physical space and information space. The information environment and data foundation of the process have undergone huge and profound changes, and these changes constitute the external driving force driving artificial intelligence to a new stage. At the same time, the goals and concepts of artificial intelligence have undergone important adjustments, new breakthroughs have been made in the scientific foundation and the carrier of implementation. A series of technological buds such as brain-like computing, deep learning, and reinforcement learning also herald the growth of intrinsic motivation. Development has entered a new phase.

The level of artificial intelligence is rapidly increasing, and human beings face potential risks. Thanks to the rapid growth of data volume, the dramatic increase in computing power, and the continuous optimization of machine learning algorithms, the new generation of artificial intelligence has demonstrated the ability to reach or surpass human work in certain given tasks, and gradually moved from dedicated intelligence The transition to general intelligence is expected to develop into abstract intelligence. With the continuous expansion of the scope of application, artificial intelligence is increasingly closely linked to human production and life.On the one hand, it has brought many conveniences to people, and on the other hand, it has also created some potential problems: First, the acceleration of machine substitution, and structural unemployment may More serious; Second, privacy protection has become difficult, and there are difficulties in defining data ownership, privacy, and permissions.

(B) the main driving factors of the new generation of artificial intelligence

At present, with the rapid and iterative evolution of next-generation information technologies such as mobile Internet, big data, and cloud computing, the binary structure of human society and the physical world is advancing to the human society.