CHAIN Seminar #44 T. Speith "Making AI Understandable: Goals, Methods, and Open Challenges of XAI"
Автор: 北海道大学 人間知・脳・AI研究教育センター(CHAIN)
Загружено: 2024-08-06
Просмотров: 89
The speaker of the 44th CHAIN academic seminar is Dr. Timo Speith from University of Buyreuth, Germany. He is also a visiting researcher at CHAIN until the end of September, 2024.
Abstract: As AI systems become increasingly integrated into high-stakes decision-making processes, understanding their operations and outcomes is essential for satisfying societal desiderata such as trust, accountability, and fairness. This talk explores the goals, methods, and open challenges of an increasingly popular field of research dedicated to understanding AI systems: explainable AI (XAI). I will begin by motivating XAI, starting from the above-mentioned societal desiderata. Next, I will discuss the various stakeholders involved in XAI and their respective needs to understand AI systems. Furthermore, I will introduce a variety of methods for achieving explainability, focusing on saliency maps. Finally, I will address the open challenges that remain in the field. By highlighting these aspects, the talk aims to provide a comprehensive overview of XAI, emphasizing its importance in fostering societally desirable AI development and deployment.
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке:
