Since 2023, large models have gradually gone deep into all walks of life, in order to track the application progress of large models in the industrial field, International Data Corporation (IDC) recently released the “Industrial large model Application Progress and Prospects, 2024” (Doc#CHC50964924, June 2024) report. The implementation progress and main application scenarios of the current market industrial large model are sorted out, and the future application direction of the large model in the industry is prospected for market reference.

VI451-10 Industrial Big models are not just industrial ChatGPT

Different from traditional AI, because the large models applied in the industrial field often need to be privatized, the market mostly uses the Transform architecture model with the number of B-parameters. However, in the industrial field, not only large language models like ChatGPT, but also different types of models such as visual models, multimodal models, specialized domain models, and control models are applied.

The industrial large model has been preliminarily applied

After a year of exploration, the large model has landed a small number of business cases on the basis of some POC, and the project amount has been distributed from one hundred thousand to one billion yuan, with the majority of 100,000 and million projects. Power, mining, oil and gas, semiconductors, automobiles, consumer goods and other industries of the head enterprises to explore faster. At present, many industrial enterprise users’ cognition of the large model is also gradually moving toward rationality – from reporting some high expectations to paying more attention to what kind of scene value can actually be played.

The reason why the current application progress can be made is mainly due to two factors. First, all kinds of suppliers in the market have made some breakthroughs in security, computing power, illusion and other issues through continuous efforts in technology; Second, the two technical routes of knowledge retrieval enhancement (RAG) and Agent (Agent) provide more possibilities for specific applications.

VI451-10 Application scenarios are very scattered, and several common scenarios are widely used

The application of industrial large models has penetrated into many links of industry, forming some applications at the functional points of many business processes, which is more fragmented than traditional AI scenarios. Among them, knowledge management/knowledge question answering, data assistant/data question answering, professional content production and visual security monitoring are the four directions of application exploration at present. In the direction of each specific application scenario, the report lists more than 80 cases for market reference.

Suppliers embrace big models

Many excellent suppliers in all directions of the industrial field are actively following up large model technology, launching industrial large models, large model platforms, and some general or professional applications that combine large models.