In the view of Guo Zhaohui, vice chairman of the Industrial Big Data Group of the Industrial Internet Industry Alliance, chief scientist of Shanghai Youalso Information Technology Company, and former chief researcher of Baosteel Central Research Institute, the academic community always regards industrial software as a scientific and technological problem, which is completely wrong. Its essence is an economic problem, depending on where the market of industrial software is, the engineering problem is the industrial software problem. The algorithm is secondary. In Guo Zhaohui’s view, the industrial large model is mainly used to assist the research and development service process, and the impact on the production process is not very large, and the production process is the most basic industrial process.

IS200SAMBH1ABA Ge Dongdong, co-founder and chief scientific officer of Sugigu Technology, once said that China’s basic industrial industry is facing the following three challenges in the process of turning to intelligence: The complexity of the operation system has increased significantly. The globalization of the industrial chain and the expansion of production scale have magnified the difficulty of collaboration, and the industrial progress has greatly deepened the degree of cross and integration of various links. Higher response sensitivity is required. In the production and operation of modern industry, in order to achieve the best benefits, the responsiveness of production and decision-making needs to be improved; High technical barriers. Basic industry has the characteristics of technology intensive and capital intensive, technical barriers and research and development costs are high, and China’s technology accumulation and talent reserve in this area are insufficient.

“There are a lot of startups doing optimization and fine-tuning of industry models. Compared with other industries, there is relatively little public data on core processes in manufacturing, making it difficult to pre-train large models.” Said Gu Fanzeng, general manager of strategic business development of Amazon Cloud Technology Greater China. For manufacturing customers, the core of large model application is the need to find a balance between model accuracy and inference cost. If the small model can solve the problem and the cost is manageable, it is not recommended to replace the large model.

IS200SAMBH1ABA According to the industrial software industry research report released by Zheshang Securities, the AI large model is currently in the preliminary exploration stage in the industrial field, and the next four types of core models are expected to enable multiple subdivision scenarios in the industrial field. The AI+ industry is expected to enable a significant increase in productivity in areas such as process control, quality optimization, and robotics applications. As industrial software gradually moves toward standardization, ecology and microservice, and core technologies toward platformization and intelligent development, China’s domestic industrial software shows an obvious development trend in development, technology, market application and service mode, and there is a broad market space for the future combination of industrial software and intelligent technologies such as cloud services, AI, and big data.