This year's work report proposes to carry out the "artificial intelligence +" action to build a digital industrial cluster with international competitiveness. The reporter learned that Zhou Hongyi, a member of the National Committee of the Chinese People's Political Consultative Conference and founder of 360 Group, prepared three proposals for general large-scale model security, building a security cloud, and deepening the multi-scenario application of artificial intelligence. He believes that 2024 is the first year of large-scale model application scenarios, and China can fully embark on a large-scale model development path with Chinese characteristics, and should support the development of large-scale models in the direction of verticalization and industrialization, and help accelerate the formation of new quality productivity.
The general large model cannot be directly applied in the enterprise-level scenario, and the development of vertical and enterprise large models can better meet the personalized needs of enterprises, improve production efficiency and service quality, and promote industrial upgrading. Zhou Hongyi believes that the general model is a "generalist", but it lacks industry depth, and it is unrealistic to try to solve all the problems of enterprises with a large model. Compared with general large models, vertical large models have lower debugging and deployment costs, less training time, and more cost-effective, and at the same time, vertical large models require less computing power, which helps to deal with the current problem of "stuck neck" computing power.
How to create a large model with Chinese characteristics? Zhou Hongyi put forward three suggestions:
First, the scene is very important, and the large model has a great future in the vertical field, so it is recommended that the central state-owned enterprises take the lead in providing more application scenarios, focusing on "small incision, large depth", and promoting the verticalization and industrialization of large models. Enterprises should not use large models to advance aggressively, but to use AI to gradually transform their business, step by step, and accumulate small wins into big wins. In practice, it is necessary to split the specific analysis of the scenario, find the right entry point in the business process, and select the business link that matches the mature ability of the large model. It is suggested that the central state-owned enterprises continue to provide more application scenarios, open up more opportunities for the development of vertical, miniaturized, and low-cost large models, and help the digital transformation of hundreds of industries.
Second, knowledge is very important, and a vertical model based on "dark knowledge" can better solve enterprise problems. It is recommended that enterprises should be encouraged to do a good job in knowledge management before customizing AI, and upgrade the enterprise big data platform to the enterprise knowledge platform. The data and knowledge of large models are only the tip of the iceberg of human knowledge, and enterprises also have a large number of "dark knowledge", such as strategic planning, product design drawings and other unique knowledge of enterprises, which only exist in specific enterprises. It is recommended to encourage enterprises to build knowledge platforms, aggregate "dark knowledge", build enterprise-specific knowledge bases, do a good job in management, and on this basis, through vertical training, go deep into enterprise-level scenarios to meet the needs of enterprises.
Third, business integration is very important, and it is recommended to encourage and guide enterprises to deeply integrate large models with digital business systems, combine them with business processes, and give full play to the value of large models. Large models are like engines, which should be combined with business, especially traditional manufacturing. The key to the integration of large models and business scenarios is to connect the agent application with the enterprise digital system. At the enterprise level, by connecting the organization, personnel, business, and process, a business collaboration platform is built to promote the full integration with the large model. At the national level, enterprises are encouraged to come up with one or two business scenarios to integrate with large models, create a large number of large models that can be implemented and promoted, and promote the integration of these large models with digital systems, which will have a huge impact on China's industrial digitalization and new industrialization, and essentially become an important part of new quality productivity.
*: Beijing**client reporter Wu Honghong.
Process edit: u028