Some personal & unsuccessful stories in AI safety research

15^th December 2022, 13:00 Ashton Lecture Theatre
Dr. Chih-Hong Cheng
Fraunhofer Institute for Cognitive Systems (IKS), Munich

Abstract

In this (informal) talk, I will reflect on some of my research projects in AI safety. In particular, I will focus on verifying DNN modules for autonomous driving developed by Audi and scenario-based testing for the Baidu Apollo AD stack. I will discuss the key technical challenges and consider the limitations of our proposed solutions. Finally, I summarize some research directions (only personal view) for moving AI from running demos to safe products.

Biography

Chih-Hong Cheng is currently a researcher at Fraunhofer IKS, acting as the head of the research department on safety assurance for AI. He is also an adjunct faculty at the Research Center for Information Technology Innovation (CITI), Academia Sinica, Taiwan. His research interests include software engineering, formal methods, and AI/ML for trustworthy autonomy. He had extensive collaborations with leading industrial automation, railway, and automotive companies. Apart from his daily duties, he also serves as an international expert (sub-team lead) for the ISO 8800 publicly available specification (PAS) in safety and AI for automotive. Chih-Hong obtained his doctoral degree from the TU Munich.