Knowledge and Systems Sciences Symposium 2024

Title: Anomaly detection with graph learning

Abstract: Anomaly detection is crucial because it allows us to proactively identify problems like fraud and equipment failure, but also uncovers hidden patterns that might signal exciting new discoveries or opportunities, making it a powerful tool for managing data across various fields. Traditional anomaly detection methods often struggle with the complexities of interconnected data, where anomalies can manifest as subtle deviations in node properties, edge connections, or even entire sub-graphs. The advent of graph learning techniques has revolutionized anomaly detection, empowering us to uncover subtle deviations and irregularities within complex networks. This talk explores the frontier of anomaly detection through the lens of graph learning. We will provide an overview of graph learning, highlighting its capacity to capture rich relational information inherent in diverse datasets. Drawing upon real-world applications, we will then showcase the versatility of graph learning techniques, particularly graph neural networks, in detecting both node-level and structural anomalies across diverse domains, such as knowledge graphs, social networks, and healthcare. The talk will conclude with a glimpse into the future of anomaly detection with graph learning, highlighting ongoing research and promising directions for this dynamic field.

Bio: Dr. Feng Xia is a Professor of Computer Science at RMIT University, Australia. He is/was Associate/Guest Editor of over 20 journals (e.g., TKDD, TOMM, TAI, TNNLS, TITS, TETC, and TII). He has served as the General Chair, PC Chair, Workshop Chair, or Publicity Chair of over 30 conferences, and PC Member of over 100 conferences (e.g., IJCAI, AAAI, KDD, WWW, WSDM, SDM, CIKM, and NetSci). Dr. Xia has authored/co-authored over 300 scientific papers in journals and conferences (such as IEEE TAI, TKDE, TNNLS, TC, TMC, TBD, TCSS, TNSE, TETCI, TETC, THMS, TVT, TITS, TASE, ACM TKDD, TIST, TWEB, TOMM; IJCAI, AAAI, ICLR, WWW, SIGIR, WSDM, CIKM, JCDL, EMNLP, and INFOCOM). He was recognized as a Highly Cited Researcher (2019). Dr. Xia received a number of prestigious awards, including e.g. IEEE Vehicular Technology Society 2020 Best Land Transportation Paper Award, ACM/IEEE JCDL 2020 The Vannevar Bush Best Paper Honorable Mention, and WWW 2017 Best Demo Award. He has been invited as Keynote Speaker at several international conferences, and delivered a number of Invited Talks at international conferences and many universities worldwide. He is Vice-Chair of IEEE Task Force on Learning for Graphs. His research interests include artificial intelligence, graph learning, brain science, digital health, and robotics. He is a Senior Member of IEEE and ACM, and an ACM Distinguished Speaker.

Feng Xia

Affiliation: RMIT University, Australia

Xiwei (Sherry) Xu

Affiliation: CSIRO Data61, Australia

Title: Responsible Agentic Navigation of Knowledge: A System Perspective

Abstract: Generative AI (GenAI) have revolutionized the creation of new content, such as text, images, and music, by leveraging learned patterns from existing data. But their integration into complex software systems introduces challenges such as hallucination, outdated knowledge, and non-traceable reasoning processes. These issues raise significant concerns for responsible AI, particularly in ensuring the reliability and accountability of GenAI-generated content. Retrieval-Augmented Generation (RAG) has emerged as a promising approach to address these challenges by enabling GenAI systems to dynamically incorporate external knowledge sources.
This keynote will provide a system-level perspective on designing and implementing responsible RAG systems. It will focus on the critical decision points and trade-offs involved in building these systems, highlighting how they can be designed to meet the demands of responsible AI. Additionally, it will cover the latest advancements in agentic RAG, where GenAI agents autonomously orchestrate retrieval processes, integrating both passive knowledge bases and active domain experts. This approach enhances the factual accuracy and adaptability of GenAI systems, ensuring they operate within ethical and responsible frameworks.

Bio: Dr. Xiwei (Sherry) Xu is a principal research scientist at CSIRO Data61. She is currently leading the software systems research group (SSRG). Xiwei is a conjoint senior lecturer with UNSW and is responsible for delivering a subject in Computer Science and Engineering School (CSE). Her current research areas include responsible AI (RAI), software engineering for AI-based systems (SE4AI) and designing blockchain applications. Xiwei is identified by the Bibliometric Assessment of Software Engineering Scholars and Institutions (2013-2020) as a top scholar and ranked 4th in the world as the most impactful SE researchers by JSS (Journal of Systems and Software).

Title: A Socio-Technical Approach to Build Safe and Ethical AI

Abstract: Building Safe and Ethical AI needs complementary approaches from both Sociology and Technology, both Protective thinking and Constructive thinking, both Reactive actions and Proactive actions. From the protective and reactive perspective, I will firstly introduce recent advancement on systematic AI Safety assessment, and how to quantify and control the balance between Safety and Capacity. In addition, I will discuss thinkings on AI safety redlines and the challenges to avoid catastrophic and existential risks. From the constructive and proactive perspective, I will introduce Safe and Moral AI, compared to AI Safety and Ethical AI, I will discuss Safety and Morality as first principles to develop AI, and will introduce preliminary results along this vision.

Bio: Dr. Yi Zeng is a Professor from Chinese Academy of Sciences, serving as the Director at the Brain-Inspired Cognitive Intelligence Lab, and the International Research Center for AI Ethics and Governance. He is the founding director of Center for Long-Term AI, and the Beijing Institute of AI Safety and Governance. He is a member of the United Nations High-Level Advisory Body on AI. He is also an expert in the UNESCO Ad Hoc Expert Group on AI Ethics, an expert in the WHO Expert Group on the Ethics/Governance of Artificial Intelligence for Health, and a board member of the National Governance Committee of Next Generation Artificial Intelligence in China. He has briefed at the UN Security Council (9381st Meeting) on AI for International Peace and Security and been listed as one of the TIME100 Most Influential People in Artificial Intelligence (TIME100/AI) in 2023. His major research interests focus on Brain and Mind inspired Artificial Intelligence, AI Ethics and Governance, AI Safety, AI for international peace and security, and AI for sustainable development.

Yi Zeng

Affiliation: Chinese Academy of Sciences, China