At present, the technological and industrial revolution represented by the new generation of artificial intelligence is gestating and emerging. Digital, networked, and intelligent information infrastructure is accelerated. Integrated innovation and cross-field innovation characterized by the intersection of information communication, life, and material science have gradually become mainstream. New industrial applications built around “smart+”, New business formats, and new models continue to emerge, and the “head goose” effect of artificial intelligence can be brought into full play. It is expected that in 2021, artificial intelligence will accelerate to become an important driving force for building a modern digital economy system and promoting high-quality economic and social development. As part of the “new infrastructure”, it is deeply integrated with new technologies such as 5G, cloud computing, big data, and industrial Internet. Integration, forming the core capabilities of a new generation of information infrastructure, and providing underlying support for the development of the digital economy.
Basic judgment on the situation in 2021
(1) Emerging technologies continue to be nurtured, and integrated technological innovation with artificial intelligence as the core will accelerate
Since 2020, the application of artificial intelligence single-point technology in my country has become more mature, but the collaborative scale and industrial application of artificial intelligence and related technologies are still in the early stage, and the efficiency of enabling high-quality economic development needs to be improved. We judge that in the future, a single technology of artificial intelligence will face a ceiling when it functions independently. It is expected that in 2021, new technologies and products such as virtual reality, ultra-high-definition video, and emerging automotive electronics will continue to emerge, and accelerate cross-integration with artificial intelligence to promote an economic form of intelligent transformation of production, lifestyles, and social governance; and this At the same time, artificial intelligence and 5G, cloud computing, big data, industrial Internet, Internet of Things, mixed reality (MR), quantum computing, blockchain, edge computing, and other new-generation information technologies support each other. Through the industrialization of intelligent technology and the intelligentization of traditional industries, artificial intelligence will provide the underlying support for the development of the intelligent economy and the digital transformation of the industry, and promote the deep integration of artificial intelligence and 5G with cloud computing, big data, and the Internet of Things to form a new generation of information Core capabilities of the infrastructure.
In the specific direction, integrated innovation characterized by cross-integration has gradually become the mainstream, and the value of cross-integration of multiple emerging technologies will enable artificial intelligence to exert greater social and economic value. It is expected that in 2021, artificial intelligence will accelerate integration with automotive electronics and other fields to realize special functional modules such as perception, decision-making, and control, promote the formation of autonomous driving, driving assistance, human-vehicle interaction, and service entertainment application systems, and further innovate the traditional automotive industry chain. Accelerate the intelligentization and networking of automobiles; the combination of artificial intelligence and virtual reality technology is expected to provide tools for manufacturing, home improvement, etc., and provide rich scenarios for virtual manufacturing, intelligent driving, simulated medical care, education and training, film and entertainment, etc., Interactive and timely platform environment.
(2) Smart economy is beginning to take shape, and ubiquitous intelligence is developing rapidly
The new crown pneumonia epidemic has become the “new normal” for global development in the next period of time. Both domestic and foreign countries are in a period of economic and social innovation and development, transformation and upgrading. There is an urgent need for the use of artificial intelligence. The accumulation of data resources, the construction of intelligent infrastructure and traditional infrastructure will achieve intelligent upgrades. Artificial intelligence technology is expected to promote the intelligent innovation of all elements of economic development.
Looking forward to 2021, artificial intelligence will further promote the digital economy to enter a new stage of the smart economy. The new economic form of smart economy has begun to take shape. Artificial intelligence will accelerate integration with the real economy and become an important source of energy for industrial transformation and upgrading under the new normal. One of the first, not only to promote the innovation of smart manufacturing, smart logistics, smart agriculture, smart tourism, smart medical, smart city, etc., but also to promote the development of new products such as smart operations, smart software, smart hardware, and smart robots. The development of an intelligent and intelligent economy will take shape. Artificial intelligence will give new connotations to the cyber-physical system (CPS), making it a more universal human-machine collaboration system. In the future, the Internet of Everything will inevitably bring about the ubiquity of networks, the ubiquity of data, and the ubiquity of application requirements. The application scenarios of artificial intelligence will expand to more industries, more fields, more links, and more levels. , Ubiquitous intelligence that can be used by any unit at any time and anywhere will be accelerated, which will further promote the in-depth integration of artificial intelligence technology and various fields of the real economy.
In terms of specific directions, it is expected that in 2021, the manufacturing industry will be the field with the richest and most potential for artificial intelligence application scenarios. Its application requirements will run through the entire life cycle of the manufacturing industry and will become the key field of artificial intelligence integration applications in the future. Artificial intelligence and manufacturing The in-depth integration of AI will be promoted and deepened in more links and more levels in the manufacturing industry. Demand orientation and pain point focus will become one of the keys to the integration of artificial intelligence and manufacturing. Artificial intelligence products and services will fall into specific industrial intelligent products. Or system solutions in specific industries. In addition, since most industry chain companies have not yet obtained large-scale value from artificial intelligence applications, security and the input-output ratio will become an important decision basis for manufacturing companies to apply artificial energy. The key point of increasing its added value will shift from equipment value mining to user value mining bottle by bottle.
(3) Scene empowerment becomes the main theme, and typical scenes will become the focus of financing
With the gradual maturity of my country’s artificial intelligence technology and the formation of application models and business models, the artificial intelligence market and industrial development will continue to improve. As of the end of June 2020, the scale of my country’s artificial intelligence core industry will reach 77 billion yuan, and artificial intelligence companies will exceed 2,600 companies have become one of the main centers of global unicorn companies. The artificial intelligence investment and financing logic of “scene determine the application, the application determines market, and the market determines enterprise development prospects” has been further recognized by all walks of life. It is expected that in 2021, the degree of segmentation and specialization in the field of artificial intelligence will be further improved, and the commercialization stage of the widespread application of artificial intelligence is coming. The government and the market will be more aware of applications that are closely integrated with specific application scenarios, especially the real economy application needs. attention.
Specifically, it is expected that in 2021, the local government’s enthusiasm for the development of the artificial intelligence industry will continue. Local support policies and measures will also become more pragmatic and operable. The application will become an important content of government attention and focus. More cities (clusters) will focus on smart chips, smart drones, smart connected cars, smart robots, and other advantageous industries, and actively build for key application fields such as healthcare, finance, supply chain transportation, manufacturing, home furnishing, and rail transit. The in-depth application scenarios of artificial intelligence in line with local advantages and development characteristics are expected to attract more capital attention in the coming year, such as new retail, driverless, medical and education, and other easy-to-follow artificial intelligence application scenarios. At the same time, since China still lags behind the United States in terms of the underlying technology of artificial intelligence, with the further development of artificial intelligence in China, the investment in the underlying technology will continue to grow. Those underlying technology startup companies with top-level scientists and strong technology genes will With continuous capital injection from the capital market, the transformation of the capital market will promote artificial intelligence to emphasize rationality. Major companies will take root in the scene and dig deep into the application to make artificial intelligence products truly “useful”.
(4) “New infrastructure” empowers all walks of life, and the underlying support of the artificial intelligence industry continues to improve
The Central Economic Work Conference first proposed the concept of “new infrastructure” in 2018 and pointed out that it is necessary to play the key role of investment, increase technological transformation and equipment update in manufacturing, accelerate the pace of 5G commercial use, and strengthen new types of artificial intelligence, industrial Internet, and Internet of Things. Since then, there have been 7 central-level meetings or documents clearly expressing the strengthening of “new infrastructure.” On March 4, 2020, the Standing Committee of the Political Bureau of the CPC Central Committee held a meeting and proposed to speed up the construction of new infrastructures such as 5G networks and data centers, which aroused greater attention. “New infrastructure” has the rich connotation of the new era. It not only conforms to the future economic and social development trend but also meets China’s current social and economic development stage and transformation needs. It will become a new engine for social and economic development while making up for shortcomings. Artificial intelligence “new “Infrastructure” is of great significance to the development of the artificial intelligence industry. It is expected that in 2021, around the “troika” of new artificial intelligence infrastructure such as algorithms, data, and computing power, the construction of the artificial intelligence industry chain will continue to increase.
Specifically, in terms of computing power, China’s 5G communication network deployment will accelerate in 2021, and the number of devices connected to the Internet of Things will increase to 50 billion. The growth rate of data is getting faster and faster, and the amount of calculation required for artificial intelligence training will be further increased. With exponential growth, the demand for computing power in related industries will be even greater. Leading Internet companies will have large data volumes reaching thousands of petabytes. Leading companies in traditional industries will reach petabytes of data, and personal data will reach terabytes. GPU, ASIC ､FPGA, and other computing units will become the underlying hardware capabilities that support the development of artificial intelligence technology in China, and the industrial chain construction around the troika will continue to be strengthened. In terms of algorithms, the Cafe framework, CNTK framework, etc. are collected and integrated for different emerging artificial intelligence algorithm models, which can greatly improve the applicability of algorithm development scenarios. Artificial intelligence algorithms transition from RNN, LSTM to CNN to GAN and BERT, and GPT. -3, etc., emerging learning algorithms will be implemented more efficiently in the mainstream machine learning algorithm model library.
Several issues that need attention
(1) The basic computing power of artificial intelligence is limited
Diversified artificial intelligence industry application data and more complex deep learning algorithms require powerful computing capabilities like support for implementation. It is expected that the amount of data will continue to increase explosively in 2021, and artificial intelligence algorithm models will become more complex and require higher levels of computing However, domestic companies that can provide large-scale artificial intelligence computing power support are still very limited, and my country as a whole is not well prepared for artificial intelligence computing power infrastructure. According to professional organizations, the popularization of new-generation information technologies such as artificial intelligence and 5G communications will rapidly increase the amount of newly created data worldwide from 33ZB in 2018 to 175ZB in 2025, which requires continuous upgrading of computer computing capabilities; 2010 Since the beginning of the year, with the popularity of GPU chips, FPGA and ASIC chips have accelerated their development and been applied to the field of artificial intelligence. In 2020, the computing power of supercomputers will reach the level of tens of billions of times per second. However, with the continuous iterative and escalation of demand for computing power in the development of artificial intelligence, a large number of domestic artificial intelligence chip companies still rely heavily on Qualcomm, Nvidia, AMD, Xilinx, Meiman Electronics, EMC, Avago, MediaTek, and other international giants to provide compliance In the field of commercial server, international giants such as IBM, HPE, Dell and others rank among the top three in the global server market. Inspur, Lenovo, New H3C, Domestic companies such as Huawei have limited market share.
(2) Lack of open source and open artificial intelligence algorithm platform and framework
This round of artificial intelligence industry development uses deep learning technology as the main engine. The open-source and open deep learning environment provides a basic guarantee for the evolution and innovation of technology. my country urgently needs to expand its technological influence and promote technological innovation through open source and open methods. Focus on the development of industrial ecology and provide new solutions for product traceability and system credibility assessment of artificial intelligence technology. However, China’s open-source ecosystem construction started relatively late, and insufficient participation in the core AI open source platforms and frameworks. The global mainstream AI algorithm frameworks and platforms are dominated by American companies such as Google, Facebook, Amazon, and Microsoft. Baidu, the fourth paradigm The algorithm frameworks and platforms of domestic companies such as Megvii Technology, SenseTime Technology, and Yitu Technology have not been widely recognized and applied in the industry. China’s lack of support in the core technology field of deep learning frameworks is mainly reflected in core technology and related technological innovation Limited capacity, insufficient training performance and cross-platform support capabilities for neural network models; insufficient advanced design and development capabilities for deep learning frameworks, and lagging research on modular development and cross-platform support, which is not conducive to the formation of a complete artificial intelligence industry in China Ecological and potentially negative impacts on China’s information infrastructure security, industrial security, and data security. Chips have given many Chinese companies and developers a ring of overturning the boat, but the deep learning framework has just attracted attention. The lack of core technology will directly affect the development of chips, systems, software, and hardware platforms related to the artificial intelligence industry in the future.
(3) The level of industrial data standardization and interconnection is seriously insufficient
Data is the core element of artificial intelligence iterative innovation. The development of new-generation information technologies such as big data, cloud, Internet of Things, and 5G communications has produced unprecedented amounts of data, and the growth rate of data is getting faster and faster. Although my country’s artificial intelligence technology has been piloted in the fields of manufacturing, transportation, e-commerce, finance, and medical care, the application of industrial data by upstream and downstream enterprises in the industry presents their own arrays, repeated efforts, sporadic scale, different standards, and scenarios. Different characteristics, the successful experience of a single industry or enterprise is difficult to transfer, in fact, has delayed the majority of small and medium-sized enterprises using artificial intelligence technology to improve productivity and achieve high-quality development. Data sources between different industries are more complex, with uneven data quality, different labeling levels, lack of data standards, and integration and sharing channels, resulting in the effective interconnection and organic integration of data between industries and within a single industry. Greatly reduce the availability and portability of data.
(4) A customized AI infrastructure construction evaluation framework embedded in industry scenarios has not yet been formed
Typical application scenarios are important technical “testing grounds” and “accelerators”, and their evaluation, selection, and creation will determine whether all walks of life can effectively use artificial intelligence infrastructure to improve the level of intelligence and realize intelligent transformation. At present, my country has not yet effectively explored the development potential of rich data and diversified scenarios, and has not been able to grasp the requirements and characteristics of artificial intelligence “new infrastructure” embedded in industry scenarios; although it has a huge data scale and richer application scenarios, especially in Finance, healthcare, education, manufacturing, retail, smart cities, government services, and other fields have a huge accumulation of basic data and new-generation infrastructure needs, but there is generally a lack of adequate assessment of the demand for artificial intelligence computing power and lack of deep learning in combination with their own industries. The ability to grasp, understand and apply algorithms, lack the awareness of collecting, coordinating, organizing, and cleaning industry data.
In fact, in the process of preventing and controlling the new crown pneumonia epidemic in 2020, the effectiveness of artificial intelligence as a “new infrastructure” has been fully demonstrated, and it has played an important role in alleviating the bottlenecks in the flow of people, logistics, information, and capital in various industries. It plays an indispensable role in the prevention and governance of major public safety risks, promotes the resumption of production and production of manufacturing enterprises, and maintains teaching and education in universities, primary and secondary schools, and summarizes successful experiences in 2020 in time, and sorts out customized artificial intelligence infrastructure embedded in industry scenarios It is imperative to build an evaluation framework in 2021.
(5) There is a large gap of professional talents in subdivided application fields
my country is still facing the challenge of a shortage of deep learning talents in advancing the further development of artificial intelligence. According to statistics from think tanks under the Paulson Foundation, China is the largest source of top AI researchers in the United States. As of the end of 2019, nearly 60% of the world’s top AI talents have settled in the United States, of which top AI talents receiving undergraduate education in China accounted for The highest ratio is 29% (20% in the United States, 18% in Europe, and 8% in India). China is the largest source of top artificial intelligence talents in the United States. According to LinkedIn’s big data, the overall global supply of AI talents is about 3.4 million people, of which only 95,000 are deep learning talents, and they are highly mobile, which further increases the gap. Among them, China’s AI talents The total is only 50,000. In 2020, the domestic artificial intelligence talent gap will reach more than 5 million, and the supply-demand ratio will be seriously unbalanced; the penetration rate of children’s programming education in the United States will reach 44.8%, while in China only 0.96%; China’s top artificial intelligence talents will only rank sixth, the former The five are the United States, Britain, Germany, France, and Italy. In 2021, it is an urgent task to continuously strengthen the training of artificial intelligence talents in my country and make up for the shortcomings of talent introduction.
(1) Promote the establishment of dedicated AI computing facilities to lay a solid foundation for computing power
Promote the establishment of an AI supercomputing center to undertake large-scale AI algorithm computing, machine learning, image processing, scientific computing, and engineering computing tasks, accelerate the industrialization of artificial intelligence technology in vertical industries, and promote the development of the local artificial intelligence industry. Promote the application of technologies such as elastic computing and massive data storage to improve the efficiency of computing resources. Accelerate the green and efficient development of AI computing infrastructure and build a green and efficient computing center. Strengthen the preliminary planning and design of the computing power center, based on application requirements, taking into account factors and conditions such as energy, climate, natural cold sources, network facilities, and energy consumption indicators, and rationally deploy and construct computing power infrastructure.
(2) Building an intelligent ecosystem to build software and hardware collaboration capabilities
Promote the realization of a high degree of coupling between software and custom AI chips to achieve optimal performance. Build industry collaboration capabilities, promote efficient docking of artificial intelligence companies with vertical industry platforms and general platforms, and ensure the real-time nature of calling the required platform functions. Promote the effective docking of AI dedicated computing facilities with existing business systems in the industry, and build an intelligent application ecological environment based on computing power. Support industry enterprises to provide intelligent computing power infrastructure and general software services, gather and incubate artificial intelligence enterprises, promote the development of artificial intelligence industry, and create an intelligent ecosystem system of “technology research and development, industrial incubation, venture capital, education and training, and supporting policy environment”.
(3) Continue to support the construction of artificial intelligence open source and public service platforms
Create a carrier for artificial intelligence technology innovation, support leading companies to take the lead, unite upstream and downstream enterprises, colleges and universities, professional institutions, etc. to jointly build a technological innovation platform in key areas of artificial intelligence, and support universities and enterprises to apply for national laboratories and state key laboratories, National Technology Innovation Center, Key Engineering Laboratory and other national scientific research platforms. Identify a number of district-level artificial intelligence technology innovation platforms, and provide support based on innovation results. Guide and support the establishment of a batch of artificial intelligence open platforms, open-source projects, and large-scale common-sense databases, the establishment of a batch of platform-based artificial intelligence application test entities such as artificial intelligence technology public service platforms, multi-scenario training, and test verification key laboratories, and support Cloud training and terminal execution development framework, algorithm library, toolset, etc., and open the underlying technology interface and database call interface for colleges and universities, innovative enterprises, and promote artificial intelligence original innovation and independent innovation from the source.
(4) Building a supportive AI policy toolbox
Improve the artificial intelligence data standards, evaluation, intellectual property, and other service systems, focus on creating standardized data sets, and establish metadata sets for artificial intelligence system training, verification, and testing, focusing on industrial terminology, reference frameworks, algorithm models, basic theories, and key technology, products and services, industry applications, safety, and ethics, etc., provide application standards, deployment guidelines, and practical cases for the application of artificial intelligence technology in subdivisions. Introduce quantitative artificial intelligence technology measurement indicators, establish a standardized evaluation method system for artificial intelligence technology performance, and form an accountability system and review tools for artificial intelligence intellectual property and ethical risks. Actively attract overseas scientific researchers, gather global talents, and introduce a package of policies for the introduction of high-end overseas talents in a series of fields such as research funding, personal taxation, visas, household registration, and children’s education, so as to effectively solve the worries of scientific researchers and provide them with scientific research and entrepreneurship. Greater support.