The launch of the first knowledge map fusion large model platform in China explores a new path for the industrialized application of large language models.

CCTV News:Since the beginning of this year, the big language model and generative artificial intelligence represented by ChatGPT have become global hot spots in science and technology, and have affected human life and production mode. However, users around the world soon found that in the process of interacting with the big language model, they would encounter it "seriously talking nonsense" and output specious or even absurd results. This technical feature called "big model illusion" hindered its application and development in enterprises and industries.　

On September 8th, the research team led by Zheng Weimin, an academician of China Academy of Engineering, launched the first "knowledge map and large model integration application platform" in China. Knowledge mapping is a recognized "brain-like" artificial intelligence technology, which is widely used in knowledge reasoning. The platform combines the fast and accurate characteristics of the two, and explores a new technical path for the industrial application of large language model.

According to Zheng Weimin, the platform effectively overcomes the "illusion" and mistakes that may be caused by the big model technology by combining the whole process of knowledge map with the big model technology.

Zheng Weimin, Academician of China Academy of Engineering and Professor of Computer Department of Tsinghua University:If all models are made, the accuracy of some systems is only 37%, while based on knowledge map, adding knowledge map, the accuracy reaches more than 90%, which is very good.

Since the big language model came out, its cross-domain universality, fast autonomous learning and self-improvement ability are undoubtedly revolutionary.

Zheng Weimin, Academician of China Academy of Engineering and Professor of Computer Department of Tsinghua University:What questions the big model asks it, industrial, agricultural and national economy, it can answer it quickly and will tell you later. This is a very good thing, but we also find that its answer is sometimes inaccurate, and sometimes some people say it is nonsense.

In Academician Zheng Weimin’s view, when a large model moves from personal application to enterprise application, it changes from a toy to a tool, and the accuracy of the tool is very important. Therefore, as a more "brain-like" artificial intelligence tool, the accurate knowledge derivation of knowledge map can complement the large model, and the fast learning ability of the large model accelerates the knowledge generation of knowledge map, thus realizing the use of science and technology.