IT Home According to news on June 16, the Institute of Automation of the Chinese Academy of Sciences released a new generation of artificial intelligence large model developed by the institute in Shanghai this morning - Zidong Taichu 2.0.
According to reports, compared with the first generation, the new large model focuses on improving decision-making and judgment capabilities, achieving a leap from perception, cognition to decision-making, and will play a greater role in the fields of medical care, transportation, industrial production and other fields in the future.
According to previous reports by IT House, the first generation AI large model of Zidong Taichu was released in 2021. It was jointly developed by the Institute of Automation of the Chinese Academy of Sciences and Huawei and is known as "the world's first multi-modal large model with 100 billion parameters." .
Different from most of the current large language models that are mainly based on text, the "Zidong Taichu" large model is based on multi-modal technology at the beginning of its development, using more types of data such as pictures, audios and texts to conduct cross-platform analysis. The unified representation and learning of modalities realizes the "unified representation" and "mutual generation" of image, text, and voice data.
According to reports, Zidong Taichu 2.0 is based on Huawei’s full-stack domestic software and hardware platform Shengteng AI and Shengsi MindSpore, and is jointly built by the Institute of Automation of the Chinese Academy of Sciences and the Wuhan Institute of Artificial Intelligence.
Based on text, pictures, and audio, "Zidong Taichu" 2.0 can integrate more modal data such as 3D, video, and sensor signals, and optimize the integrated cognition of voice, video, and text as well as common sense calculations, etc. function to further “break through the interaction barriers of perception, cognition and decision-making”.
The above is the detailed content of The new generation of AI large model 'Zidong Taichu 2.0' self-developed by the Chinese Academy of Sciences is released. For more information, please follow other related articles on the PHP Chinese website!