Editor's note: China launched a large AI model in Shanghai earlier this month aimed at improving the accuracy and efficiency of carbon accounting. Wei Wei, vice-president of the Shanghai Advanced Research Institute of the Chinese Academy of Sciences; Lai Xiaoming, chairman of the Shanghai Environment and Energy Exchange; and Xu Huaqing, chief scientist of the National Center for Climate Change Strategy and International Cooperation, shared their views on the model in People's Daily. Below are excerpts of their comments. The views don't necessarily represent those of China Daily.

Most carbon accounting systems worldwide focus on the production side while neglecting the consumption side. In 2024, about 2 million metric tons of greenhouse gases were emitted during the production of wind turbines and photovoltaic products exported from China, but their operation helped reduce greenhouse gas emissions by about 350 million tons around the world. Without panoramic accounting that covers both production and consumption, this contribution would go unaccounted for.

The world needs a more scientific and equitable emissions accounting system that not only clarifies where greenhouse gas emissions come from but also for whom the emissions are produced. The Chinese team has developed a large AI model to address this need. The purpose of the large model is not to deny the existing international framework for carbon accounting, but, with due respect to the Intergovernmental Panel on Climate Change's scientific consensus, to make the accounting more comprehensive, dynamic and intelligent.

Carbon accounting faces several challenges, including complex industries and high knowledge barriers. The data used often come from various sources and in different forms, and the data are also updated with respective frequency.

Traditional methods of carbon accounting are time-consuming, involve high labor costs and may lack accuracy. The large model helps resolve these challenges. The research team constructed eight core datasets covering production, consumption, natural sources and other key areas. They also collaborated with government departments, industry institutions and companies for data updates and fusion, forming a multidimensional "carbon knowledge base".

The team developed the model based on ScienceOne, a scientific foundation model developed by the Chinese Academy of Sciences, and created five professional agents with distinct functions.

In addition, the collaboration between an internal, high-performance server cluster and external computing centers enables the optimization and flexible supply of computing power. This way, a complex accounting task that used to take weeks or even months can now be completed in just a few minutes. The model not only marks an upgrade of technology and tools for carbon accounting, but also provides an important scientific basis for global energy conservation and emissions reduction efforts.

It is hoped that the team will strengthen the research and development of high-precision accounting and measurement equipment, and use modern information technologies such as blockchain to ensure data security. At the same time, it should actively benchmark against relevant international rules and standards to build internationally compatible data interfaces.