CM Seminar: Dr. Jofish Kaye – “Understanding Humans and AI”
Welcome to the third Computational Media seminar of the fall quarter. Open to all, class enrollment is not required. Food will be provided.
Please join us in-person at SVC-3212 or via Zoom to learn more about the works of Dr. Jofish Kaye:
Title: Understanding Humans and AI
Presenter: Dr. Jofish Kaye, Director of Research for New Product Development, Inflection AI
Description: Millions of people now talk to AI daily, but not in the same ways or with the same trust. Drawing on surveys of chatbot users in the US, India and South Africa, I’ll show how experiences of AI vary across cultures: as just one example, Americans are increasingly less likely to trust the companies building it. Even within a country, people fall into distinct styles of engaging with chatbots.
I’ll then ask whether AI can stand in for these users as synthetic respondents. Our experiments show that models imitate people well when they’re well documented online and poorly otherwise. They also erase dissent: people who abandon chatbots on principle all but disappear from the synthetic data.
I’ll close by considering what it takes to build AIs that take humanity seriously, including using AI collaboratively and creatively. The challenge is to create experiences that recognize the possibilities inherent in working together to accomplish what none of us could alone.
Bio: Jofish Kaye is Director of Research for New Product Development at Inflection AI, where he studies how people use, trust and talk to AI systems. Over more than 20 years in industry, he has led research functions at Wells Fargo, Anthem.AI, Mozilla, Yahoo Labs/Flickr and Nokia Research. He has published over a hundred academic articles, holds a Ph.D. in Information Science from Cornell and an M.S. from the MIT Media Lab, has taught HCI at Stanford, and co-chaired CHI 2016.
Hosted by: Associate Professor of Computational Media, Christina Chung
Zoom Link: https://ucsc.zoom.us/j/94103955103?pwd=od4ueavRt06ECpwAwqvgofwn5Suyal.1