The world's first!Image voice and text are generated, and "Zidong Taichu" won the highest award

Wuhan Evening News (Reporter Chen Jie Li Jia correspondent Tu Xingpe) Study for more than two years, won the world's highest award, and many technologies were the world's first ... Recently, at the opening ceremony of the 2022 World Artificial Intelligence Conference, the Chinese Academy of Sciences Automation, Huawei and Wuhan The "Zidong Taichu" three -mode model project led by the Artificial Intelligence Research Institute won the highest award of the World Artificial Intelligence Conference -SUPERIL AI Leader, excellent artificial intelligence leader), and became one of the five winners.

On September 1st, the 2022 World Artificial Intelligence Conference was awarded at the awards. Correspondence

At present, the AI ​​field is mostly independent evolution single -modal technology, with low learning efficiency and poor modeling of models. The "Zidong Taichu" project breaks through the current AI technology limitations, effectively solve the pain points of the current AI technology "one dedicated and one energy", lack of small sample learning ability, and cross -modal semantic grooves, realizing the three -modal data of images, text, and voice The "unified representation" and "mutual generation" of the intercourse, the ability to understand and generate is closer to humans, and provides an innovative foundation for creating the application of the multi -modal artificial intelligence industry. It has taken an important step towards the general artificial intelligence.

What is the unified expression of the three -mode state of pictures, texts, and sound? Imagine that when the European Cup was broadcast, according to the players of the two teams, the robot could automatically explain; when shooting a TV series, the robot "watched" the script can automatically generate the picture and scene for the director to process ...

Wuhan Artificial Intelligence Research Institute introduced that the automation of the Chinese Academy of Sciences is based on the basic software and hardware of the full -stack of localized software and hardware. Relying on the computing power of the Wuhan Artificial Intelligence Computing Center, for more than two years, it has been developed for more than two years. The three basic models of maps, texts, and sounds are added to the three basic models. Cross -modal coding and decoding networks are added. Based on the MindSpore AI framework, the three -mode large model "Zidong Taichu" was successfully created, and it was officially released in July 2021.

"Zidong Taichu" is the world's first three -mode model. Since its release, it has won the winner of many international events and awards. It has independent understanding of medical surgery scenarios, multimedia information retrieval, industrial defect inspection, virtual digital tour guide, etc. The scene realizes the demonstration application, showing the broad industrial development prospects.

Wuhan Artificial Intelligence Research Institute stated that in the future, it will continue to aim at the problem of fragmentation and difficulty in applying fragmentation and landing in the artificial intelligence industry. Based on the "Zidong Taichu" model, it will build a general artificial intelligence platform and comprehensively promote the construction of first -class domestic domestic first -class domestic intelligence as the core The technical pioneer area, important growth pole growth pole and application demonstration zone, promoting artificial intelligence empowerment of thousands of industries, and becoming a new momentum for driving the transformation and upgrading of the industry and empowering the development of the digital economy.

This year is the fifth year of the SAIL award selection. Internationally renowned universities and scientific research institutions, the world's leading enterprises have enthusiastically declared, the Yuan universe, smart chips, and AI models are assembled. More than 800 projects participated in competition. As the core representative of the big model, "Zidong Taichu" was recommended as one of the treasures of the eight major towns, and made a hard core appearance at the conference exhibition site.

[Edit: Yu Lina]

- END -