A New Paradigm of AI: Language-Intelligence-Native and Language-Centric by Dr. Min Zhang

Abstract: 

After thousands of years of evolution, human society has entered the era of artificial intelligence, where the entire society is accelerating its development and AI paradigms are rapidly iterating. Against this broader backdrop, through the relentless efforts of several generations over decades, our team has proposed: 1) a new paradigm of language-intelligence-native artificial intelligence centered on language intelligence; and 2) a new architecture of AI that integrates symbolism, connectionism, and behaviorism. We aim to effectively model the information world, the physical world, and human society, across the spectrum from single agents to multi-agent systems to human-machine integration. This report will introduce the above technical roadmap and the language-native, full-modal mixture-of-experts large model ('Lychee' Lychee-Uni-MoE-Omni, abbreviated as Lychee-UMO) developed by our team.

Biography:  

Dr. Zhang Min is a Distinguished Professor, Distinguished Assistant Principal, and Dean of the Institute of Computing and Intelligence at Harbin Institute of Technology (Shenzhen). He is a Fellow of ACL and AAIS, Chief Scientist of an Innovative Research Group (Category A, NSFC), and a recipient of the National Science Fund for Distinguished Young Scholars. His research interests span natural language processing, language models, artificial intelligence, and AI for software engineering. He has received six Best Paper Awards. According to CSRankings (NLP category), he ranks first worldwide based on number of publications, and Harbin Institute of Technology is also ranked first overall as a team. He has served on the editorial boards of 11 journals. He has proposed a new paradigm of “language-intelligence-native, language-centric artificial intelligence.” Based on this paradigm, his team independently developed the Lychee Uni-MoE-Omni large model, which features state-of-the-art technology and has had significant impact in both academia and industry. It has received positive evaluations from Nobel laureates, Turing Award winners, and leading industry experts. These achievements have been successfully deployed across dozens of enterprises and in his startup company. Over the past three years, they have generated more than USD 2 billion in revenue, with an annual technical invocation volume exceeding 600 billion calls. The systems serve billions of users across more than 200 countries and regions worldwide, with over 600 million monthly active devices.