NAII Distinguished Lecture: GLM's Road to AGI - From Chat to Agents
Speaker: Prof Jie Tang, WeBank Chair Professor of Computer Science (Tsinghua University) and Director (Foundation Model Research Center)
Please register for the event here: https://forms.cloud.microsoft/r/vhvgKE8zGa
Abstract:
In just a few years, large language models have moved from answering questions to writing software, operating computers, and finding real-world security vulnerabilities. This talk reflects on that journey through GLM, the open model family we have built since 2020. I will first frame progress toward AGI as a ladder of capabilities: knowledge and reasoning, then coding and long-horizon agents, then autonomous action in the real world. I will then share what we learned climbing it. Parameter count alone matters less than many assume. Scaling post-training and training environments has become the main path to frontier capability, and that path leads toward AI that helps build the next generation of AI. Finally, I will discuss open source and safety. When models become capable enough to attack, how can openness serve as a shield rather than a risk? I will close with the opportunities and open challenges ahead, from China and for the world.
Biography:
Jie Tang is the WeBank Chair Professor of Computer Science at Tsinghua University and Director of its Foundation Model Research Center. He is a Fellow of ACM, AAAI, and IEEE. His research focuses on artificial general intelligence. He proposed the GLM pre-training framework in 2020 and led the development of GLM serial models. Earlier he built AMiner, a researcher-mining system serving over 30 million users across 220 countries and regions. He received the SIGKDD Test-of-Time Award and the SIGKDD Service Award. He is the founder of Z.ai (Zhipu AI), one of the first publicly listed large-model companies.
