How Improving AI for Blind People Sparks Innovations that Benefit the Broader Community of AI Users

Dr. Danna Gurari

Abstract: A natural grand challenge for the AI community is to create computer vision models that assist blind people with learning about visual content. In this talk, I will discuss my team's work on building the first public AI challenges using data originating from blind photographers who were authentically trying to learn about their visual surroundings. I will address questions including: What difficulties arise when creating large-scale datasets from authentic use cases? How do AI models perform on these challenges? And what key obstacles remain for developing AI solutions that effectively support blind users? I will conclude by highlighting how innovations driven by accessibility research not only advance solutions for blind users, but also spark breakthroughs that benefit the broader community of AI users.

Bio: Dr. Danna Gurari is an Assistant Professor in the Computer Science Department at University of Colorado Boulder and Founding Director of the Image and Video Computing group. Her expertise is in computer vision, human computer interaction, accessibility, and (bio)medical data analysis, with an emphasis on interdisciplinary approaches. Gurari's research has resulted in over 60 peer-reviewed articles at top academic journals and conferences, with several papers earning awards at venues such as WACV, CHI, CSCW, ASIS&T, HCOMP, MICCAI, and AAPM. Over 30 mentees, spanning undergraduate students to postdoctoral fellows, contributed to this research before moving on to successful careers in industry and academia. Gurari's work is supported by diverse funding sources, including the National Science Foundation, Silicon Valley Community Foundation’s Chan Zuckerberg Initiative, Microsoft, Adobe, and Amazon.