#DS4SocietySeminar 2022 <> From Human to Social-Centric Tech & AI: Interdisciplinarity in Massively Multilingual Machine Translation.
Skyler Wang
Talk Details
Abstract
The discourse around human-centricity in technology and AI has grown in recent times. Within this landscape, I argue that a more productive way to propel tech development forward requires us to think more ’socially.’ More specifically, going from human-centricity to social-centricity forces technologists to better consider communities, group relations, culture, and institutions. To contextualize this argument, I share my experiences doing sociological research for the recently launched “No Language Left Behind” project, an open-sourced project that pushed the total number of languages offered by commercial machine translation past 200. I share how and why we prioritized low-resource languages, the strengths and limitations of our community-based approach, and lessons for future research. In the end, I will demonstrate how we can better integrate social sciences and humanities into tech and AI, an imperative far from realized, to create a more equitable technological future.
Speaker Bio
Skyler Wang is a Visiting Sociologist at Meta AI and a PhD Candidate in Sociology at UC Berkeley. An economic and cultural sociologist by training, he uses interviews, international fieldwork, and digital ethnographies to investigate whether technologies designed to improve our day-to-day lives meet their moral commitments by honing in on user experiences. More specifically, his dissertation explores how monetary and gifting structures in the sharing economy shape ethical and social exchange between users. At Meta AI, Skyler works on the “No Language Left Behind” team to make machine translation more socially and ethically oriented. Skyler’s research and ideas have been featured in both academic journals and public forums such as National Public Radio, San Francisco Chronicle, Quartz, Vox, HuffPost, and USA Today.
Video, Slides and Notes
Video 1 - Talk + Q&A
Slides
To get future information about Seminars and events, subscribe to the newsletter https://tinyletter.com/datascience-up/