AI and large models: Are we close to Artificial General Intelligence?

A lecture given by Prof. MA Xingjun, organised on Thursday 13 April 2023 from 11:30 to 13:00.

Online: Webex (the link will be sent after registration)

Free entry

Registration required via confucius@uni.lu ou +352 46 66 44 4964

Dr. MA Xingjun is an associate professor in the School of Computer Science at Fudan University and a member of the Fudan Vision and Learning Lab (https://fvl.fudan.edu.cn/). He received his PhD degree from the University of Melbourne, and previously was a lecturer at Deakin University and a postdoc at the University of Melbourne. His main research area is trustworthy machine learning, aiming to design secure, robust, explainable, privacy-preserving, and fair machine learning models for diverse AI applications. He has published 40+ papers at top-tier international conferences, including NeurIPS, ICML, ICLR, AAAI, IJCAI, and S&P, and have received two best paper awards and one best paper runner-up award. His work has also been reported by MIT Technology Review and Computer Vision News. He also serves as a PC member/reviewer of a number of leading conferences and journals.

Recent advancements in large language/vision models, such as ChatGPT, GPT-4, CLIP, Stable Diffusion, and DALL-E 2, have sparked a growing interest in Artificial General Intelligence (AGI). However, questions remain regarding the definition of "real intelligence" and the differences between human and AI intelligence. Is it possible to achieve AGI by simply creating larger models and providing more data for learning? In this presentation, we will take a retrospective approach by examining the key techniques that led to major AI breakthroughs throughout history. By exploring the new knowledge embedded in models from different periods, we hope to gain insight into the emerging intelligence in current large models. Additionally, we will discuss trustworthiness concerns that arise with large models.