Dr. Olga Russakovsky 
Associate Professor 
Department of Computer Science
Princeton University 

Building trustworthy Artificial Intelligence (AI) systems has been a long-standing challenge and goal for AI researchers. We need these systems to be capable and robust, performing gracefully even in unfamiliar circumstances. As AI systems are approaching their potential for improving the daily lives of people around the world, we also need to teach end users to trust them -- through exposing their limitations, explaining their decision-making processes, and, importantly, being transparent in our own decision-making processes as their creators. In this talk, I will describe my lab's recent work on building trustworthy and trusted computer vision. I'll focus on the challenges of designing thorough and transparent evaluation protocols, and demonstrate that careful evaluation design enables both technological innovation that leads to more trustworthy systems as well as sociotechnical insight that helps cultivate human trust. I'll describe our research work in the domains of visual recognition and algorithmic fairness, as well as briefly mention our outreach work with AI4ALL for educating a diverse next generation of responsible AI leaders. 

Green Hall, Room 0120