Recently, IDC released the “AI large model in Industrial Application Outlook, 2023” report, which distinguishes the concepts related to AI large models, and summarizes the two paths for AI large models to be applied to industry, as well as potential application scenarios in three fields of production, manufacturing, research and development design and operation and management. The layout of outstanding technology suppliers such as Alibaba Cloud, Baidu Intelligent Cloud, Innovation Qizhi, Fourth Paradigm, Huawei Cloud, etc. is introduced, and the market outlook and future challenges are analyzed.

2 application directions and 3 major fields of application

The application of large models in industry is mainly divided into two directions: one is to enhance the generalization ability of scene models and improve the applicability of models; the other is to use natural language dialogue and content generation ability to change the application interaction mode and generate documents and reports.

At present, the application of large models in industry has formed some scenes in the fields of manufacturing, research and development design and management, but it is in the initial and exploration stage.

SC560 3BSE008105R1

SC560 3BSE008105R1

Technology service providers and industry solutions in various related fields are also actively laying out the industry application of AI large models. The report focuses on Ali Cloud, Baidu Intelligent Cloud, Innovation Qizhi, Fourth Paradigm, Huawei and other technology service providers that have more accumulation in AI technology and industrial applications, and introduces their relevant business layout.

Three directions and four challenges

The big model is expected to improve the industry in three aspects in the future:

Many enterprise applications will add AI assistants to improve work efficiency;
Enterprises build knowledge base around AI to drive continuous operation;
The generalization ability of AI large models is improved, and the products of AI technology service providers are more adaptable.
At the same time, the industrial application of large models also faces four challenges:

The technical challenge of reliable data content
High cost
Docking and integration with existing enterprise applications
The development and accumulation of models for specific application scenarios are still inevitable