AI has a long way to go
Artificial intelligence is an important driving force for a new round of scientific and technological revolution and industrial transformation, and it is having a significant and far-reaching impact on economic development, social progress, and the international political and economic landscape. The development of artificial intelligence in our country started relatively late. We must strive to achieve transformative breakthroughs in theories, methods, tools, and systems, focusing on key core technologies, strengthening application development, and promoting the formation of technological innovation and industrial applications to promote each other.
When a new round of industrial revolution sweeps the globe, artificial intelligence (AI) has become the core direction of industrial revolution. Is artificial intelligence as amazing and omnipotent as people think? What barriers does China's artificial intelligence development need to break through?
Recently, at the 2019 China Artificial Intelligence Industry Annual Conference sponsored by the China Artificial Intelligence Society, participating experts and scholars had a lively discussion on these topics.
Artificial intelligence is in a boom in recent years. Iconic events such as the Alpha Dog defeating the Go World Championship have made the public eagerly look forward to artificial intelligence. The development of deep learning and big data has brought about a great leap forward in the artificial intelligence industry. Intelligent security, intelligent voice input, intelligent navigation, and intelligent customer service using artificial intelligence technology have entered the lives of ordinary people.
Will the future of artificial intelligence be omnipotent? Experts attending the 2019 China Artificial Intelligence Industry Annual Conference are generally cautious about this. In their view, there are too many bottlenecks in artificial intelligence to be broken.
Computing power, algorithms, and big data are the three core elements of artificial intelligence. Yang Xueshan, chairman of the Academic Committee of the China Information Technology Hundred People's Association, and former vice minister of the Ministry of Industry and Information Technology, said that many of the advances we have made in the field of artificial intelligence are mainly made by the advancement of computing power.
"The artificial intelligence technology represented by deep learning is close to the ceiling." Zhang Huan, Dean of the Institute of Artificial Intelligence at Tsinghua University and an academician of the Chinese Academy of Sciences, won the 2019 Wu Wenjun Artificial Intelligence Achievement Award at this conference. He believes that artificial intelligence technology Actually experienced two generations, the first generation is symbolic reasoning, and the second generation is current deep learning.
"The tool of deep learning has the weaknesses of untrustworthiness, insecurity, unreliability, and poor promotion ability. In order to distinguish it from deep learning, we propose to develop third-generation artificial intelligence." Zhang Ye said that humans have a stronger ability to process knowledge. Computers have a stronger ability to process data, and the development of third-generation artificial intelligence must achieve the combination of knowledge and data.
"There are currently three levels of artificial intelligence: perceptual intelligence, computational intelligence, and cognitive intelligence. There are many bottlenecks at each level." Song Aiguo, dean of the School of Instrument Science and Engineering at Southeast University, believes that at this level, many elements Device technology needs to be broken; artificial intelligence still cannot reach the level of human intelligence in many aspects of computational intelligence and cognitive intelligence, which shows that there are many problems to be broken in the basic theory of mathematics and the intelligent cognitive mechanism of biophysics.
Artificial intelligence is an area where basic research and applications are closely integrated, and the bottleneck of basic research is also transmitted to practical applications.
"When we provided computing power, we found that the main bottleneck lies in the framework. I hope more people will invest in research and development, guide the development of algorithms, and then promote chip design." Said Wang Xiaolei, director of computing chips at Huawei Hisilicon.
Where are the "pain points"
Artificial intelligence is at the core of a new round of scientific and technological revolution. For our country, it is both an opportunity and a challenge.
The development of artificial intelligence in China started relatively late. Compared with developed countries, there is still a certain gap. Many experts believe that the shortcoming of the current development of artificial intelligence in China lies in the fact that the development of the industry relies too much on open source code and existing mathematical models. There are not many things that really belong to itself. Chinese manufacturing is moving from a "hardware assembly plant" to a "software assembly plant". spread.
"The problem of" neck neck "in core technology is still quite serious, especially the gap between basic theory, underlying framework and core algorithms." Said Sheng Chaoxun, deputy director of the Institute of Industrial Economics and Technology of the National Development and Reform Commission.
Yan Junjie, vice president of Shangtang Group and chief technology officer of the smart city business group, believes that the gap is not terrible. With the future development, China can catch up, but it takes a long time.
Participating experts said that China has unique advantages in the development of artificial intelligence, such as a stable development environment, sufficient talent reserves, and rich application scenarios.
"From the perspective of the three major elements of artificial intelligence, the domestic environment has advantages in terms of data; in terms of algorithm, the United States is ahead, but from the application point of view, the country is not behind; in terms of computing power, we must do well in domestic chips and build The software ecology of domestic chips still has a long way to go. "Li Xiaohan, co-founder of Yunzhisheng, said that a large part of the upstream of artificial intelligence is still in short supply, which is the key point of China's" cracking neck ".
At the conference, Tsinghua University-Chinese Academy of Engineering Knowledge Intelligence Joint Research Center and China Artificial Intelligence Society Wu Wenjun Artificial Intelligence Science and Technology Award Selection Base jointly issued the "2019 Artificial Intelligence Development Report". According to the report, for China, the number of talents leads the second echelon in most areas. Compared with the United States, China has insufficient numbers of high-impact mechanics, and there is a large space for catching up.
Needs deep integration
How should the future of artificial intelligence develop? "Deep integration" is a keyword mentioned by many participating experts.
The development of basic research in artificial intelligence technology requires deep integration. Speaking of the development direction of the Institute of Artificial Intelligence at Tsinghua University, Academician Zhang Jian said: "The purpose is one core, two fusions. That is to say, taking the basic theoretical research of artificial intelligence as the core, strengthening the relationship with mathematics, neuroscience, psychology The cross-disciplinary integration of science, brain science and humanities is committed to the development of the third generation of artificial intelligence theories and methods; it is also necessary to strengthen the deep integration of technology and industry, schools and enterprises to promote the development and progress of artificial intelligence. "
The practical application of artificial intelligence technology also requires deep integration. At the conference venue, a reporter from the Economic Daily saw that Cloud Track's service robot "Runner" was used for hotel customer service. It would independently plan routes, get on and off the elevator, and deliver slippers, mineral water, takeaway and other items to the guests. Yunfu Technology's founding partner Ying Fuchen said that this product is the result of the company's deep integration of its many years of technology in indoor positioning and navigation, robot movement, and big data with the actual needs of hotel guests, truly solving the "last 100 meters" problem. "We have served more than 1,000 hotels, covering more than 130 cities nationwide and more than 10 countries around the world."
Xiao Jing, chief scientist of Ping An Insurance (Group) Co., Ltd. believes that "the greatest value of artificial intelligence lies in empowering entities. The development of artificial intelligence requires many conditions, some of which are unique to enterprises, such as scenarios, data, and fields Knowledge, etc. Therefore, artificial intelligence requires deep cooperation between scientific research institutions and enterprises in order to really do a good job. "
Participating experts also believe that artificial intelligence has not yet achieved true convergence in many industries.
"Today's industrial intelligence, I personally think that the thunder is still loud and the rain is small." Baidu Venture Capital CEO Liu Wei said whether the new artificial intelligence technology can truly bring new capabilities to traditional enterprises, whether it can improve their efficiency and help them. Rapidly defeating traditional competitors is the hard standard for testing whether new technologies are really effective.
The future development of artificial intelligence requires not only cooperation between disciplines, cooperation between academia and industry, but also cooperation between countries.
Li Xiuquan, a researcher at the New Generation Artificial Intelligence Development Research Center of the Ministry of Science and Technology said, "Artificial intelligence is still in its initial stage of development. No country has mastered the overall algorithm, and no country has all the elements to achieve this theoretical breakthrough. Scientists from various countries are required. Strengthen cooperation and collaboration to jointly tackle these scientific problems in the field of artificial intelligence. "