计算机学院系列讲座名家讲坛第 23期——Reinforcement Learning and AI Safety

报告题目(Title)：Reinforcement Learning and AI Safety

时间(Date & Time)：2025.4.11; 3:00pm (Friday)

地点(Location)：理科一号楼1131（燕园校区）Room 1131, Science Building #1 (Yanyuan)

主讲人(Speaker)：Prof. Luke Ong, NTU Singapore

邀请人(Host)：胡振江

报告摘要(Abstract)：

Linear temporal logic (LTL) and, more generally, ω-regular objectives are alternatives to the traditional discount sum and average reward objectives in reinforcement learning (RL), offering the advantage of greater comprehensibility and hence explainability. In this talk, I will discuss the relationship between these objectives. Our main result is that each RL problem for ω-regular objectives can be reduced to a limit-average reward problem in an optimality-preserving fashion, via (finite-memory) reward machines. Furthermore, we demonstrate the efficacy of this approach by showing that optimal policies for limit-average problems can be found asymptotically by solving a sequence of discount-sum problems approximately. Consequently, we resolve an open problem: optimal policies for reinforcement learning with LTL and ω-regular objectives can be learned asymptotically. I will relate these results to safe reinforcement learning and end with some general remarks about AI Safety.

主讲人简介(Bio)：

Professor Ong holds a B.A. in Mathematics (1984, Triple First), a Postgraduate Diploma in Computer Science (1985, Distinction) from Trinity College, University of Cambridge; and a PhD in Computer Science (1988) from Imperial College, University of London. He joined NTU as a Distinguished University Professor in August 2022 and was Vice President (Research) from January 2023 to April 2024. In May 2024, Professor Ong was appointed Vice President (AI and Digital Economy) and Founding Dean of the College of Computing and Data Science. Professor Ong is the Chief Scientist of AI Singapore and is responsible for AI Singapore’s fundamental research into Artificial Intelligence. Prior to joining NTU, he was Lecturer then Professor of Computer Science at the University of Oxford (1994-2022); Fellow of Merton College, Oxford (1994-2022); and Prize Research Fellow, Trinity College, Cambridge (1988-1994).

Professor Ong's research in computing is broad, ranging across semantics of computation, programming languages, verification, logic and algorithms. He is known for his solution to the Full Abstraction Problem in the semantics of programming languages, and pioneering contributions to higher-order model checking in algorithmic verification. His current research interests include security, Bayesian statistical probabilistic programming and AI Safety.

Professor Ong was General Chair (2013-2015) of the ACM / IEEE Logic in Computer Science (LiCS). He is the joint winner of the ACM / EATCS Alonzo Church Award 2017 for Outstanding Contributions to Logic and Computation and an ACM Distinguished Paper Award. He is a recipient of the President of the Republic of Singapore Scholarship (1981).

欢迎关注计算机学院微信公众号，了解更多讲座信息！

北京大学计算机学院