Enterprises use big models to generate production schedules with one click, steel mill converters are equipped with "fire eyes" and big models become "personal assistants" for clothing designers... At present, artificial intelligence big models are aimed at vertical industry applications, deeply integrated with industrial fields such as electronics, raw materials, and consumer goods, and promote cost reduction and efficiency improvement in segmented business scenarios.

  The reporter learned that in the next step, relevant departments will strengthen the research and release of general big models and industry big models, and promote the intelligent upgrade of the entire manufacturing process, key industries and key products. Industry companies are also actively exploring, accelerating technological innovation and product upgrades, and accelerating the refinement of large-scale applications.

  The latest data from the Ministry of Industry and Information Technology shows that my country's artificial intelligence industry has formed a complete industrial system covering the basic layer, framework layer, model layer, and application layer. It has cultivated more than 400 national-level specialized and new "little giant" enterprises, and has built high-quality industry data sets for steel, coal and other industries.

  The big model industry is accelerating innovation and evolution, and its professional fields are constantly being segmented.

  In Zhejiang Jiayi Garment Factory, the designer enters the description words in the chat box, selects good category, style, color and other labels, and can generate a design drawing in a few seconds. In the past, it usually took about two days from designing to determining the upper body effect of a clothing, including placing and modeling.

  "In the clothing industry, big models are quietly changing the process of clothing design, quality inspection and other aspects." A relevant person in charge of China Unicom said that based on China Unicom Yuanjing Clothing Big Model, enterprises can achieve rapid design, one-click plate making, virtual fitting, and convert design originals into accurate industrial-grade electronic versions through AI digital version rooms, shortening the design and plate making cycle by 80%.

  The carbon black industry where Black Cat Group is located provides raw materials for tires, rubber, coatings and other products, with complex processes and high safety production requirements. "Carbon black production is like "frying vegetables" at a high temperature of 3000℃. The quality of the product depends very much on the experience of the craftsmen." Shang Guangyong, deputy general manager of Inspur Yunzhou Industrial Internet, told reporters that the team used the Zhiye big model as the basis to transform the 40 years of operating experience of 20 masters into 36,000 knowledge graph nodes, and trained to build a large model exclusive to the coal chemical industry, which promoted the passing rate of Black Cat Group's carbon black products from 82% to 94%, which is equivalent to reducing 13 tons of waste per 100 tons of products.

  Xie Shaofeng, chief engineer of the Ministry of Industry and Information Technology, recently said that large models are being implemented in electronics, raw materials, consumer goods and other industries, and have been applied in R&D and design, testing and certification, production and manufacturing, operation and management. "We are continuing to carry out empowering 'in-depth tour' activities, build online and offline platforms, promote accurate docking of supply and demand enterprises, and guide the intelligent transformation of the industry."

  In the industry's view, for specific scenarios, not all companies need the "all-round" of general big models, but the accuracy of the model is more needed.

  Gu Weixi, deputy director of the Institute of Intelligence of China Industrial Internet Research Institute, believes that artificial intelligence has been implemented in the industry at an accelerated pace, forming a collaborative application paradigm of "general model + vertical field model". Industry-specific models are booming, and a large number of finely tuned models have emerged in the fields of finance, medical care, law, industry, etc., which have higher accuracy and lighter deployment in professional tasks. “As model capabilities increase and costs decrease, generative AI will be more 'down to earth', accelerating the move from concept to real-life productivity.”

  However, due to the diversity and complexity of vertical industries, the threshold for artificial intelligence to empower applications remains. Taking industrial scenarios as an example, Pu Songtao, director of the Future Industry Research Center of the China Institute of Electronic Information Industry Development, said that core scenarios such as industrial manufacturing have high accuracy and low fault tolerance, and the processes of each sub-industry are highly differentiated, so it is necessary to further explore typical scenarios that can be used as experience replication and promotion.

  Relevant policy measures will be further strengthened. The Ministry of Industry and Information Technology stated that it will focus on supply on the one hand, strengthen the research and release of general big models and industry big models, accelerate the construction of high-quality data sets in the industrial field, consolidate the foundation, and promote "intelligent industrialization"; focus on application on the other hand, promote the intelligent upgrade of the entire manufacturing process, key industries and key products, accelerate the industrialization and commercialization of super smart terminals, and continue to carry out the collection of "in-depth processes" and typical cases, select "ten industries, 100 scenarios, and 1,000 benchmarks" to promote "industrial intelligence".

  At the same time, many companies have recently targeted practical applications to accelerate the launch of new AI models and lower the threshold for implementation. iFLYTEK's newly upgraded iFLYTEK Spark X1 model provides enterprises with a low threshold deployment plan; Conch Group and Huawei jointly launched the "AI+Cement Building Materials Big Model"; based on the Tongyi big model, China FAW has built enterprise intelligence to improve enterprise operation efficiency...

  "With the improvement of base capabilities, technological innovation will promote the combination of 'general' and 'professional' capabilities of the big model, making its deployment easier." Wen Jie, an expert from iFLYTEK, told reporters that the generalization of the Spark model has made progress in the training because it incorporates more complex types of data in scenarios and types of data in training. While improving general capabilities, it will continue to deepen into multiple industries such as education, medical care, and justice, and open up more new scenarios for production and life application.

  For the next step of development, Pu Songtao suggested that it is necessary to encourage industry enterprises, industry experts and research institutions to strengthen cooperation, further open up data flows such as enterprises, upstream and downstream of the industrial chain, supply chain, and partners, create industrial large models in vertical segments, carry out demonstration application actions for large models to empower advantageous industrial clusters and key industries, and build an industry ecological cooperation system. (Reporter Guo Qian)

[Editor in charge: Ran Xiaoning]