From May 25th to 30th, the 2023 Zhongguancun Forum with the theme of "Open Cooperation, Shared Future" was held in Beijing. During the forum, a number of cutting-edge technologies, cutting-edge products, and authoritative reports were unveiled one after another, fully demonstrating the cutting-edge achievements of my country's science and technology. Wang Jinqiao, President of Wuhan Institute of Artificial Intelligence, was invited to participate in the conference with the world's first 100-billion-parameter multi-modal large model "Zidong Taichu". It is an important path for the future development of general artificial intelligence.
Mei Jianping, deputy director of the High-tech Department of the Ministry of Science and Technology and a first-level inspector, said in an online speech that the current penetration, diffusion and disruptiveness of artificial intelligence to various industries are increasing day by day. The emerging technology represented by Li is triggering a new wave of AI innovation, creating a good artificial intelligence development ecology, which can inject strong impetus and provide solid support for the development of the real economy and the digital economy.
Wang Jinqiao introduced in his keynote speech that in the past few years, although artificial intelligence has developed rapidly in terms of perception capabilities, and some special scenarios have even surpassed the human brain, its cognitive function generalization ability is insufficient, and there is still a big gap from the human brain. Generalizability is still a major challenge for the current development of artificial intelligence.
Wang Jinqiao, Dean of Wuzhi Institute, delivered a speech
He said that the emergence of multi-modal large models has broken through the limitations of current AI technology, making AI move from the traditional "one specific and one ability" to "multi-specialized and multi-functional", and driving a new round of innovative application paradigms. Through the "big data + large model" method under unsupervised learning conditions, the performance of various tasks such as natural language processing (NLP), speech recognition, computer vision, and multimodality has been significantly improved, and powerful large models have become a general artificial intelligence. important path of development.
For this reason, starting from 2019, the Wuhan Institute of Artificial Intelligence and the Institute of Automation of the Chinese Academy of Sciences have focused on the field of multi-modal large-scale models for intensive research, and officially released the world's first multi-modal large-scale model with 100 billion parameters in September 2021 "Zidong Taichu".
The "Zidong Taichu" multi-modal large model based on Shengsi MindSpore pioneered the "unified representation" and "mutual generation" of image, text, and voice three-modal data, which not only took a key step towards general artificial intelligence At the same time, it is of great strategic significance for my country to realize AI autonomy.
"Zidong Taichu" open service platform
Wang Jinqiao introduced that based on the multi-modal large model of "Zidong Taichu", Wuzhi Academy has successively launched the "Zidong Taichu" open service platform with full-stack localization, supporting low-code one-stop development, and the "Quick Words" available to everyone. "Living Painting" AI content creation platform "Jiangcheng Luoshen" and so on.
At the level of industrial application, centering on multi-modal large-scale model technology and application, the multi-modal artificial intelligence industry consortium created by Wuzhi Academy and Chinese Academy of Sciences Automation has successively absorbed nearly 70 member units from all walks of life, education and research, and has successively empowered industrial quality inspection , film and television creation, sign language teaching, smart medical care and other dozens of fields, showing broad potential for landing.
Wang Jinqiao, Dean of Wuzhi Academy, was interviewed
In addition, the large-scale model service can also greatly reduce the threshold of AI technology application, greatly release the commercial value of AI, and usher in a new golden decade. In an interview with the live media, Wang Jinqiao said that the multi-modal large model has the same visual, auditory and language abilities as humans, which is expected to promote artificial intelligence from perceptual intelligence to cognitive intelligence. Facing the current era of large-scale model and AI development, Wuzhi Academy will continue to adhere to the strategic direction of multi-modal large-scale models, strengthen key core technology research, promote the combination of artificial intelligence and the real economy, and provide solid support for the development of the digital economy , to accelerate the arrival of the era of cognitive intelligence.