政府新闻
新一代大型语言模型在上海发布 2024-01-18
Visitors learn about the AI-powered large language model during the 2023 World Computing Conference in Changsha, central China's Hunan Province, Sept. 15, 2023.
A new-generation large language model, capable of accepting and processing about 300,000 Chinese characters at one time, was released in Shanghai on Wednesday.
The limited context length of large models is a significant problem in artificial intelligence.
The new model, InternLM2, was developed by the Shanghai Artificial Intelligence Laboratory, SenseTime, the Chinese University of Hong Kong, and Fudan University.
In an experiment, the researchers fed the model a three-hour public meeting transcript and asked it to extract salient information.
The test results showed that although there were many typos in the unproofread text, InternLM2 can accurately extract key information and summarize the speaker's main points.
The developers aim to advance the language modeling ability of the model base by improving the corpus quality and information density.
The Shanghai Artificial Intelligence Laboratory said that it continues to license InternLM2 for free commercial use.
Source: Xinhua
