Since the beginning of this year, large-language models and generative artificial intelligence represented by ChatGPT have become global scientific and technological hotspots and affected human life and production methods. However, global users also quickly found that in the process of interacting with the big language model, they would encounter its "serious nonsense" and output plausible or even absurd results, a technical feature called "big model illusion" that hindered its application and development in enterprises and industries. On September 8, the scientific research team led by Zheng Weimin, an academician of the Chinese Academy of Engineering, launched the first "knowledge graph and large model integration application platform" in China. Knowledge graph is a recognized "brain-like" artificial intelligence technology that is widely used in knowledge reasoning. The platform combines the fast and accurate characteristics of the two to explore a new technical path for the industrial application of large language models.
According to Zheng Weimin, the platform uses the whole process of knowledge graph and large model technology to effectively overcome the "illusion" and errors that may be generated by large model technology.
Zheng Weimin, academician of the Chinese Academy of Engineering and professor of Tsinghua University: If all the models are made, the accuracy rate of some systems is only 37%, and based on the knowledge graph, if the knowledge graph is added, the accuracy rate reaches more than 90%, which is very good.
Since the advent of the big language model, its cross-domain versatility, rapid self-learning and self-improvement ability have undoubtedly been revolutionary.
Zheng Weimin, academician of the Chinese Academy of Engineering and professor of Tsinghua University: The big model can answer any questions you ask it, industrial, agricultural, and national economy, and it answers very quickly, and it will tell you in a while. This is a very good thing, but we also find that the answer is sometimes inaccurate, and sometimes some people say it is nonsense.
In the view of Academician Zheng Weimin, when the large model moves from personal application to enterprise application, it changes from a toy to a tool, and the accuracy of the tool is crucial. Therefore, as a more "brain-like" artificial intelligence tool, the knowledge graph can complement the large model with its accurate knowledge derivation. The rapid learning ability of large models accelerates the knowledge generation of knowledge graphs, so as to realize the use of science and technology.
Source: CCTV News Client