Zhu Jun

Keynote Speaker

Department of Computer Science at Tsinghua University

Towards Building General World Models

Abstract:

World models are important to build artifical general intelligence, with key functions of perception, prediction and action in a learnable world. This talk will present some progress on building general world models (GWMs). Similar as the shift from transformers as basic language models to GPTs as general pretrained language models, GWMs are pre-trained large-scale world models, with tens billions or more parameters. I will discuss on the available data, the scalable architecture, and the efficient algorithms to build GWMs, with showcases on high-performance digital content generation as well as embodied intelligence.

Biography:

Zhu Jun is a Bosch AI Professor in the Department of Computer Science at Tsinghua University, an Associate Dean of the Wuqiong (AI) College, and a Fellow of ACM, IEEE, and AAAI. He previously served as an Adjunct Professor at Carnegie Mellon University. His research primarily focuses on machine learning, and he has published over 100 papers in leading journals and conferences in the field. He has served as an Associate Editor-in-Chief of IEEE TPAMI, as well as a Senior Area Chair and Best Paper Committee Members for international conferences such as ICML, NeurIPS, and ICLR. He has received numerous awards, including the China Youth Science and Technology Award, the CAST Outstanding Young Scientist Award, the Tan Kah Kee Youth Science Award, the Science Exploration Award, and the ICLR Outstanding Paper Award. He has also led his team in developing open-source platforms and large models such as "Zhusuan," "Tianshu," and Vidu.