foreword
The construction of new urban infrastructure is the carrier of urbanization development, and it is also the demand of urbanization development. At present, the scale and speed of my country's urbanization process are impressive. In 2019, my country's urbanization rate reached 60.6%. However, there is still a large gap with the urbanization level of 82% to 91% in developed economies such as the United States and Japan. With the advancement of the urban-rural integration strategy and the continuous reduction of the threshold for urban settlement, the urbanization process will also advance rapidly. According to the "City Blue Book: China Urban Development Report No.12" of the Chinese Academy of Social Sciences, my country's urbanization rate will reach 70% by 2030 and about 80% by 2050. The labor force in rural areas will flow to cities at an accelerated rate, and the construction of urban infrastructure should match it. According to the data released by the National Bureau of Statistics, the year-on-year growth rate of urban fixed asset investment was only 3.8% from 2018 to 2019, and this indicator reached 42.16% in 2009. The construction of 5G base stations and UHV will become the basis for a new round of urban development. The construction of intercity high-speed railways, urban rail transit, and new energy vehicle charging piles will further help improve the urban transportation network system; the development of big data centers and artificial intelligence will Accelerate the establishment of a digital city management platform, and greatly enhance the comprehensive planning ability of city management.
1. Development of artificial intelligence technology: Cognitive intelligence represented by natural language processing is the next development stage of artificial intelligence
Investment in the artificial intelligence industry is at a trough, and the dividends brought by deep learning technology based on big data-driven neural networks are close to the ceiling. According to the "Global Artificial Intelligence Industry Data Report" released by the China Academy of Information and Communications Technology in 2019, in the first quarter of 2019, the global artificial intelligence financing scale was 12.6 billion US dollars, a decrease of 7.3% from the previous quarter, and the number of financing transactions was 310, a year-on-year decrease of 44.1%. The amount of financing in the field of artificial intelligence in China was US$3 billion, a year-on-year decrease of 55.8%. At the same time, it is difficult for artificial intelligence companies to make profits. In 2018, nearly 90% of artificial intelligence companies were at a loss.
Artificial intelligence is in the stage of developing from perceptual intelligence to cognitive intelligence , and AI in the post-deep learning era is developing from data-driven to knowledge-driven . It is difficult for deep learning driven by big data to achieve the same effect as cognitive intelligence, and it needs to be driven by rich knowledge (such as knowledge graph). A net-like knowledge structure is formed through the correlation of different knowledge, which is a map for machines. The essence of the process of forming a knowledge map is to enable machines to establish cognition and understand the world. And just as human language is the form of knowledge transfer, knowledge graph is the core of cognition , and NLP is a bridge for machines to build the core of cognition , allowing AI to use natural language to interact with people . Therefore, natural language processing (NLP) and knowledge graph are the key technologies of cognitive intelligence , and NLP is the precursor of knowledge graph. It can be said that NLP is the core of AI technology . Represented by computer vision, the perceptual intelligence technology based on neural network deep learning technology "touches the ceiling". In the current stage of perceptual intelligence, we can see that the application of AI in the industry is actually relatively limited. The main reason is that the development of cognitive intelligence is not mature enough to simulate human interaction well, resulting in poor user experience . Therefore, the development and application of NLP is an important prerequisite for AI to truly understand human language, and it is also a guarantee for the application of AI-related products.
Cognitive intelligence is currently facing some bottlenecks, making the gap between technology and industrial applications, which is the key reason for the lack of related applications and products, but we believe that it will definitely come in the future.
Some problems facing the development of cognitive intelligence at present:
1) It is necessary to explore deep learning combined with knowledge graphs. As the dividends of deep learning for big data are exhausted, the ceiling of the effect of deep learning models is increasingly approaching. On the other hand, although a large number of knowledge graphs are emerging, they have not been effectively utilized by deep learning. How to integrate knowledge graphs and deep learning to improve the effect of deep learning models is an important development direction in the future.
2) Cognitive intelligence often needs to extract inherent implicit knowledge, or obtain cognitive results based on background-related knowledge. Based on the same data, the cognitive results may be quite different under different business requirements and different background knowledge. Cognitive intelligence effects are scene-sensitive and relevant, and general cognitive intelligence effects are not satisfactory in specific scenarios. For dialogue systems such as small languages, certain professional fields, and customer service, the relevant training sets are still in a small and fragmented state, and it is still necessary to manually label relevant language information data as a preparation for NLP training.
Figure 1: Need to explore deep learning combined with knowledge graph
Judging from the distribution of AI technology, the United States has the strongest overall AI strength, and China is developing at an astonishing speed. China and the United States are in the leading positions in the world. According to the "2018 Global AI Report" released by Stanford University, 70% of the papers of the International Association for Artificial Intelligence in 2018 came from China and the United States. The number of citations in the United States is the first, 83% higher than the global average, and the number of citations in China in 2016 increased by 44% compared with 2000. China has entered the international frontier group of developing countries in terms of technological development and market application in the field of artificial intelligence, presenting a situation where both China and the United States are leading the way. Overall, in the key field of AI, China has already occupied a relatively high starting point.
2. Development of artificial intelligence industry: multi-scenario empowerment is the core direction of the development of artificial intelligence industry
The artificial intelligence industry chain can be divided into three layers: the basic layer (computing infrastructure), the technical layer (software algorithms and platforms), and the application layer (industry applications and products). The basic layer mainly collects data and provides computing power, mainly including AI chips, sensors, operating systems, cloud computing services, data service platforms, etc. The technology layer develops application technologies for different fields, including computer vision, intelligent voice, natural language processing, machine learning, etc. The application layer is based on the basic layer and technology layer, combining scenarios and industry knowledge to develop applications or products to empower different scenarios, such as the application of AI in security, home, medical, transportation and other fields.
Empowerment is the essence of AI. AI applications are entering an era where scenarios are king. The development of knowledge-driven cognitive intelligence should be combined with scenarios. The success of AI application models requires a deep combination of technology + closed-loop data + scenarios (industry knowledge ) . Cognitive intelligence effects are scene-sensitive and relevant. Cognitive intelligence often needs to extract inherent implicit knowledge, or obtain cognitive results based on background-related knowledge, and recognize based on the same data under different business requirements and different background knowledge. Results may vary widely. In the post-deep learning era, in addition to computing power, algorithms, and data, scenarios and industry experts (knowledge) are particularly critical.
Figure 2: Multi-scenario empowerment is the core direction of the development of the artificial intelligence industry
From a strategic point of view, the layout of US technology giants in artificial intelligence focuses on the technically difficult AI basic layer, while China's layout focuses on the application layer. In the situation of competition with the United States, China is relatively strong in the application layer, while the basic layer and technology layer are relatively weak. Chinese technology giants invest the most in the AI application layer, and have more experience in AI scene empowerment. There are a large number of AI start-ups that are commercialized based on core AI technologies in a certain direction.
3. A-share related companies
In the field of artificial intelligence, we recommend A-share companies TRS and NavInfo, and we recommend paying attention to HKUST Xunfei.
TRS: As one of the domestic companies that have long insisted on independent core technology research and development, the company has accumulated more than 20 years of deep cultivation in the field of semantic intelligence and is a technology-driven company. The company has a leading independent core AI technology. Semantic intelligence includes natural language processing, knowledge graph, image and audio and video understanding, which is a rare artificial intelligence target in the A-share market. At present, the company has empowered AI technology vertically in multiple industries such as finance, security, media, and government, and has strong market competitiveness. In the future where artificial intelligence is better developed, we are optimistic about the possibility of the company's AI technology extending in infinite scenarios.
NavInfo: The company is a leader and a rare target in the field of Internet of Vehicles and autonomous driving. High-precision map technology has advantages and high barriers. The company has long focused on research and development, and research and development expenses have accounted for more than 50% of operating income for many years. Intelligent driving has entered the L2~L3 stage, and the company's ADAS and automotive electronic chip business will also become a highlight.