AI4 Society Seminar | Irene Y Chen

Abstract: At a time when AI development is concentrated in fewer and fewer hands and public sentiment about AI is declining, how do we build AI systems that reflect what people actually want? In this talk, I explore three ways of incorporating people's feedback into AI development and evaluation: asking people what they want from LLMs, studying how clinicians edit AI-generated notes, and using reports from affected individuals to identify failures after deployment.

‍

Irene Chen is an Assistant Professor at UC Berkeley and UCSF. Her work develops reliable and trustworthy AI with an emphasis on healthcare. She has received best paper and best poster awards, honors from Google and Apple, and Rising Star awards in EECS, Machine Learning, and Data Science. She received her PhD from MIT EECS and a joint AB/SM in Applied Math from Harvard University.