The cooling of capital boom, the shortage of artificial intelligence (AI) talents, the bankruptcy of start-up companies and the difficulty of industrial landing… The development of artificial intelligence is facing many difficulties. Therefore, some observers sound that “artificial intelligence is beginning to decline”.
Is that true? At the 2020 Beijing Zhiyuan conference held recently, many experts expressed optimistic about the prospect of artificial intelligence: “Ai spring is just the time.”
Zhang cymbal, an academician of the Chinese Academy of Sciences, said: “the development of AI is in the prelude stage. The basic problem of symbol has not been solved, and the fundamental problem of ‘intelligence’ has not been involved.”
It’s still a long way from “intelligence”
In recent years, AI has developed rapidly, from “experimental” to “practical”, from “ashamed of expression” to “proud” of AI. However, it is difficult and long to really give human wisdom to machines and let machines stand on people’s shoulders to learn.
Zhang cymbal said: “the existing AI system is unsafe, unreliable, unexplainable and difficult to expand. In this regard, we must strengthen the basic research related to AI, and the road is still very long.”
Gao Wen, an academician of the Chinese Academy of engineering, also pointed out that “compared with the human brain, the efficiency of the existing AI system is not high enough. In this regard, researchers should do something that others have not done and try to make machines inherit knowledge.”
How can AI break through the siege? John Hopcroft, winner of Turing Award, said: “AI research is still based on the mode of asking questions to find solutions. In the next decade, the breakthrough of AI technology will not be in the computer science community, but depends on breakthroughs in engineering, biology, linguistics and other fields. These breakthroughs can enable AI to solve more problems.”
As for the characteristics of the next generation AI technology, Gao Wen said: “the first is interpretable AI, and the second is efficient AI, that is, whether we need to pay huge computing power while pursuing beyond human accuracy is a necessary factor to be considered by the next generation AI.”
In the view of experts attending the meeting, people’s excessive expectation makes the AI wave seem to be declining, but in fact, the AI wave is “higher and higher than each other”.
“I am very optimistic about the future of AI,” said Li Kaifu, founder of Innovation workshop, “In the future, the whole industry can connect with AI, and change from the traditional form of ‘AI +’ to the form of ‘+ AI’. In the past, ‘AI +’ mainly expanded from one entry point to more applications, and continuously rolled into a potential platform. In the future, the traditional industry will be driven by the industry platform, AI as an important auxiliary technology, and ‘+ AI’ will have greater opportunities.”
“The spring of AI has just begun. Especially from the perspective of the Internet of things, there are many application scenarios that do not give play to the ability of AI. With the innovation of products and intelligent scenarios, more technological innovation will be brought.” Cui Baoqiu, vice president of Xiaomi group, said, “how long the spring of AI can last depends on the care of all our employees.”
Where will AI expand its territory
Experts attending the meeting said that the future AI research in China should not be limited to existing problems or areas that have been widely studied. AI research still needs to expand its territory in terms of data open source, knowledge mining, utilization and sharing, self-regulated learning and so on.
Among them, deep learning, as an important breakthrough in the development of AI in China in recent years, has many technical dividends. Bart Selman, a professor at Cornell University, said: “China’s AI research field should not be too narrow, but should explore new fields. There may be a breakthrough.”
Wang Haifeng, CTO of Baidu group, said: “Ai’s mining and utilization of knowledge will be an important breakthrough direction. In addition, AI will also make a breakthrough in the learning field of small samples and low energy consumption, the direction of AI’s’ soft and hard integration ‘and in-depth integration with other application scenarios.”
Cui Baoqiu agreed. He added that open source is also a major variable of AI in the future. It is not only open source of code, but also open source of data and knowledge sharing.
According to Sun Jian, chief scientist of Kuangshi technology company, the current popular “self supervised learning” will be a good research direction. You can learn the same good features without labeling data, which will have a very positive impact on practical application.
Sun Jian also pointed out that the data required for machine learning training are often scattered in all walks of life and difficult to obtain. If we can make progress in data security training and provide AI with a highly secure and reliable machine learning environment, it will help to give full play to the value and function of data.
“Ai development mainly depends on technology and demand. Good technology should be pushed to the product market by good entrepreneurs, and then driven by good demand,” said Lu Qi, founder of Qiji Chuangtan, “In the next decade, we need to further open up data islands. The exchange of data will drive the development of AI industrialization on a large scale. In addition, we should gradually build our own technology ecology and industrial ecology.”
How to fill the talent gap
Continuously accumulating “talent bonus” is not only the key for AI to maintain its competitive advantage, but also the dilemma that needs to be broken in recent years.
John Hopcroft said: “if AI wants to make a breakthrough, education and talent training in various fields are the most important.”
According to the talent development report of artificial intelligence industry (2019-2020 Edition) released by the talent exchange center of the Ministry of industry and information technology on June 24, the current gap of effective talents in China’s AI industry is 300000, especially the imbalance between supply and demand in specific technical directions and posts.
In addition, China’s AI talents are also faced with the dilemma of scarcity of high-end engineers and top scholars and weak aggregation effect, which also leads to the weak innovation ability in China’s AI field.
“There is a shortage of AI talents in China. We should create more opportunities for top overseas scientists and engineers to come back,” Li Kaifu said. “Domestic colleges and universities focus more on the cultivation of research talents and lack the cultivation of application talents.”
Lu Qi agrees with this, “at present, the actual operation ability of talents cultivated by domestic colleges and universities is relatively weak. We need to introduce more innovative culture into colleges and universities to make students become talents with strong operation ability, professional ability and organic combination of technology, products and customer needs.”
In addition to strengthening the cultivation of talents in Colleges and universities, Lu Qi also pointed out that it is equally important to study in the “skill University” of enterprises. Students should devote themselves more to large companies and projects and exercise their practical ability in real scenes.
Sun Jian said: “students must lay a good foundation and cultivate scientific research quality in all aspects. They are not superstitious about authority and dare to challenge cutting-edge wisdom. For newly graduated students, one of the best ways to make rapid progress is to find a good R & D environment and walk with experienced peers and good enterprises.”
Xia Huaxia, chief scientist of meituan, also pointed out: “college teachers should cooperate more with the industry. The industry has a lot of real scenes and a lot of real data. It’s time to apply the accumulation of AI theory in the school to the real scene.”
“In the future, AI will have very broad prospects in theoretical research, technological development and industrial development. As long as we recognize the direction and do it firmly, we will gain something,” Wang Haifeng said.
Editor in charge: Tzh