Lingxiang Wu
Associate Researcher, Institute of Automation, Chinese Academy of Sciences
Lingxiang Wu, PhD, Associate Researcher at Institute of Automation, Chinese Academy of Sciences, and Algorithm Director at Wuhan Institute of Artificial Intelligence. He graduated from the University of Technology Sydney with a degree in Computer Vision and Natural Language Processing. He mainly researches on key technologies for AI applications, such as large-scale pre-trained models, multimodal understanding and generation. With academic and industrial background, he has participated in many large-scale projects and completed them online. His team's work on the world's first tri-modal large-scale model "Zidong Taichu" based on Rise AI won the SAIL Award at the World Conference on Artificial Intelligence in 2022. Translated with DeepL.com (free version)
Topic
Innovation and Practical Applications of Zidong Taichu's Multimodal Large Models
The next generation of artificial intelligence is expanding from traditional data intelligence to encompass five major modalities. Researchers in the field of information intelligence, on the path to general artificial intelligence, are focusing on multimodality as the core, breaking through the interactive problems of perception, cognition, and decision-making. Zidong Taichu 2.0 introduces for the first time a fully multimodal multitask unified generative learning framework, transitioning from the three modalities of text, image, and sound to encompassing all modalities, covering images, text, speech, 3D, signals, and videos. Additionally, the team has a series of innovative works, such as multimodal multitask understanding and generative unified models, dialogue-based target localization using large-scale text-image models, and resolver-free virtual outfitting based on diffusion models, among others. In terms of practical applications, the Zidong Taichu large model open service platform provides a complete large model fine-tuning toolchain, allowing for rapid customization of user-specific industry models, leading to applications in smart healthcare, smart governance, and more. Outline: Prospects of General Artificial Intelligence, Zidong Taichu 2.0 Framework, Other Innovative Works of the Team, Open Service Platform for Large Model Landing Applications, Case Studies.
