The hottest industrial intelligence is still in th

2022-07-30
  • Detail

Industrial intelligence is still in the stage of development exploration. Industrialization and intelligence are two-way infiltration or become the core development path. A new round of information and industrial reform is booming. The development of digital, networked and intelligent industrial economy has become the focus and trend of the world, and industrial intelligence has ushered in a new stage of development. However, industrial intelligence is still in the development exploration period, and has not yet formed a clear and large-scale commercial application, and all parties have not yet formed a consensus on the industrial development of industrial intelligence

under this situation, recently, the industrial interconnection industry alliance has jointly prepared and released the white paper on industrial intelligence (2019 discussion draft) (hereinafter referred to as the white paper) with a number of enterprises to study and analyze the development pulse and latest status of industrial intelligence from the aspects of application, technology and Industry, and to some extent predict the direction of future development and reform

in the context of the development of industrial intelligence analyzed in the white paper, we can find that the industrial view of industrial intelligence is mainly manifested in "two horizontal and two vertical", the horizontal is knowledge map and deep learning, and the vertical is general technology and application integration

at the horizontal development level, ICT enterprises, research institutions and relevant industry associations provide general technical capabilities to provide basic support for industrial intelligence in an "integrated" way. Among them, ICT enterprises such as Microsoft, Google, Amazon and Alibaba provide almost all general technical research and engineering support covering knowledge map and in-depth learning; Research institutions such as the University of California, Massachusetts Institute of technology, Tsinghua University and the Institute of automation of the Chinese Academy of Sciences mainly provide theoretical research on algorithms; Industry associations provide relevant standards or general technical support, such as the formulation of enterprise integration standards such as unified modeling language provided by OMG object management organization, which lays the foundation for the industrialization of knowledge atlas; Khronos group has developed a deep learning compiler

at the application level, equipment/automation and software enterprises, manufacturing enterprises and ICT will be listed and start-ups in May. They will take integrated innovation as the main mode, face the actual business field, integrate various industries and technical elements to realize industrial intelligent innovation and application, which is the core of the industrial intelligent industry

among the four types of application subjects, traditional enterprises such as equipment/automation, software enterprises and manufacturing enterprises (such as Siemens, abb, KUKA, Autodesk, Foxconn and Xinsong) face their own business fields or demand pain points, and improve the product performance by introducing artificial intelligence with a pre load of 30KN specified in the experiment; ICT enterprises (such as Cognex, Hikvision, Daheng image, keens, Microsoft, konux, IBM, Alibaba cloud, etc.) rely on the accumulation and advantages of artificial intelligence technology to expand their existing businesses to the industrial field; Start up enterprises (e.g., innovation Qizhi, Kuangshi, element AI, and Tianze Zhiyun) provide solutions for segmented fields with technological advantages; Research institutions (such as the University of Massachusetts and the University of California, Berkeley) rely on their theoretical research advantages to carry out application exploration of cutting-edge technologies

industrial intelligence relies on general technology to realize various innovative industrial intelligence applications. However, general technology often cannot meet the complexity and particularity requirements of industrial scenarios and problems. At this stage, there are still a large number of characteristic problems to be solved

the white paper points out that at present, ICT giants occupy an absolute dominant position in general technology fields such as deep learning framework, compilers and chips. However, at this stage, the end-to-side inference framework is mainly dominated by five ICT giants such as apple, Facebook, Tencent, Google and Baidu. It is preliminarily judged that Baidu is more likely to make efforts in the industrial field. There are portability and adaptability problems in the industrial field, and there is an urgent demand for compilers. However, the market pattern of compilers is not clear, and there is no field oriented development trend. It is predicted that Intel and Amazon may become the choices in the industrial field

in addition, the theoretical research of deep learning tends to be stable, and the application becomes the key. Lifeifei, the top scholar of artificial intelligence, Zhengyu, the top scholar of Microsoft Asia Research Institute and artificial intelligence, and Yukai, the founder of horizon, all believe that the mainstream architecture of deep learning theory research will converge, and it is difficult to make a revolutionary theoretical breakthrough. At present, the bottleneck lies in the connection between technology and traditional industries. At the present stage, there are two main trends in algorithm research: first, algorithm interpretability research. Stanford University has carried out interpretability research based on "tree is normalized according to conch profile announcement", Texas A & M University has carried out migration method to solve the problem of deep learning interpretability, and Nanjing University has proposed RNN interpretability method; Second, research on relevant cutting-edge algorithms. Leading research institutions at home and abroad, such as Massachusetts Institute of technology, Israel Institute of technology, Tsinghua University and Institute of automation of Chinese Academy of Sciences, have carried out research on cutting-edge algorithms related to deep learning such as capsule network, transfer learning, (deep) reinforcement learning and generative countermeasure network

in terms of development paths, the white paper believes that the two-way penetration of industrialization and intelligence will become two types of core paths

from the specific analysis of the four application subjects, equipment automation, software and manufacturing enterprises are facing the needs of equipment and product performance improvement or the pain points of their own business development. They are constantly seeking ways to integrate artificial intelligence around the main supply line of artificial intelligence technology. At present, these enterprises mainly develop industrial intelligence in two ways:

first, some of them are in urgent demand Powerful giants in the field have achieved intelligent upgrading through cooperation and merger and acquisition of artificial intelligence technology companies. For example, FANUC cooperates with artificial intelligence start-up preferred networks to enhance the intelligent level of robots; Ge acquired bit stew systems, an AI start-up, and to build Ai strength; Easton acquired 30% equity of Barrett tech, an American high-tech company, to expand the field of AI robots and micro servo systems. Second, through the introduction of talents and the establishment of corresponding research institutions, enhance the comprehensive competitiveness of enterprises. For example, Siemens established a central research institute and promoted the "Vision 2020" plan, developed artificial intelligence and robotics, and built an industrial knowledge map platform for its own financing management. Foxconn, Xinsong, etc. set up artificial intelligence research institutes to accelerate the implementation of artificial intelligence research and industrialization of achievements

on the other hand, information technology enterprises and research institutions, relying on the basis of artificial intelligence technology, continue to enrich their application service capabilities for industrial scenarios, and strengthen cooperation with manufacturing enterprises to export their capabilities to the industrial field through the introduction of industrial intelligence solutions or the industrialization of cutting-edge technologies. For example, the Alibaba cloud industrial brain platform will open up the knowledge maps of the three major industries of chemical industry, photovoltaic and power, so that developers can respond quickly and realize the demands of artificial intelligence in specific business scenarios. Huawei has built an industrial knowledge map for supply chain and parts management. Hikvision's main business is the video monitoring industry. In 2014, Hikvision entered the industrial field and developed in-depth learning quality detection products to be applied in 3C manufacturing, metal processing and other fields

unlike the former two, research institutions pay more attention to technological innovation, and they are also the incubators of the industrialization of cutting-edge technologies. For example, the Massachusetts Institute of technology has conducted research on mind controlled machines, and the accuracy of radio wave recognition has reached 90%, which has a significant impact on the future human-computer cooperation technology. Berkeley robot dexnet2.0 is equipped with a deep learning system. By learning 10000 three-dimensional objects with different characteristics in the virtual database, it can quickly predict objects and select appropriate schemes to grasp objects of various irregular shapes. A German enterprise has made efforts to apply industrial intelligence industrialization

it is particularly noteworthy that under the above general trends, start-ups have become important solution providers in subdivided fields by virtue of their technological and financial advantages. On the one hand, big data technology start-ups provide knowledge mapping solutions for small and medium-sized vertical enterprises. Rely on the advantages of data processing and artificial intelligence technology to help small and medium-sized enterprises with high labor and time costs release the value of enterprise data. For example, minglue data release Intelligent System 2.0 provides complete solutions for vertical fields such as industry. On the other hand, the equipment field has become the main entry field for in-depth learning and application of start-ups, attracting a large amount of investment. The equipment field is a technology and capital intensive industry. The start-up enterprises have inherent advantages, and their products come into the market quickly with a high return on investment. For example, Kuangshi wholly purchased iris robot to develop the manufacturing industry and build an intelligent warehouse; Qizhi innovation focuses on providing services for "artificial intelligence +b2b" enterprises from the perspective of consumers, applying artificial intelligence technology to create solutions such as intelligent quality inspection, and raising billions of dollars; Element AI provided artificial intelligence solutions for global enterprises in manufacturing, logistics, robotics and other fields, and obtained US $102million in financing

Copyright © 2011 JIN SHI