Speaker: Dr. Anh Nguyen
From: Auburn University
Abstract
Artificial Intelligence (AI) is becoming ubiquitous and a key that enables humankind to reach unprecedented horizons (e.g. in protein folding). However, in high-stake applications (e.g. cancer diagnosis, automated target detection, or face identification), a single incorrect decision of AIs could change lives entirely. My three burning research questions are: (1) How to build AIs that are accurate in new edge cases in the evolving world? (2) How to build AIs that maximize the human-AI team accuracy? (3) How to build AIs whose decision-making can be understood and debugged by humans? In this talk, I will share our recent research in both Computer Vision and Natural Language Processing in these three thrusts. I am also excited to discuss Bottleneck XAI, my proposed architecture for future Trustworthy and Explainable Artificial Intelligence systems. Bottleneck XAI would enable users to (a) understand and learn from AIs; (b) audit; and (c) edit AIs.
For more info, please follow this link.
Read More