January 24, 2020
Many problems related to fairness, accountability, transparency, and ethics in machine learning are rooted in decisions surrounding data collection and annotation—problems akin to those seen in other types of archival record-keeping. Archives are the oldest communal effort of human information management, created for the purpose of recording personal and government events and documenting histories of people and nations.
By drawing on interdisciplinary expertise in archival and library sciences, we can gain insights into issues related to the collection of sociocultural data in machine learning. Among those challenges are consent, power, inclusivity, transparency, as well as ethics and privacy. Focusing on methodologies for data collection can shed light on theoretical bias and potential discriminatory outcomes from the use of datasets.
Friday, January 24, 2020
Robertson Hall, 8th Floor
Jon M. Huntsman Hall
2:00 PM – 3:30 PM (Doors open at 1:30 PM)
Timnit Gebru, PhD is a research scientist at Google in the ethical Artificial Intelligence (AI) team. An alumna of Stanford University where she earned her bachelor’s and master’s degrees in electrical engineering, she also earned her PhD from Stanford’s Artificial Intelligence Laboratory, studying computer vision under Professor Fei-Fei Li. Her thesis examines data mining large scale publicly available images for sociological insights as well as resulting computer vision problems. As a post-doc at Microsoft Research New York City in the FATE (Fairness, Accountability, Transparency, and Ethics) in AI group, she studied algorithmic bias and the ethical implications underlying data mining projects.
Dr. Gebru is the co-founder of Black in AI, an organization which works to increase the presence and visibility of Black researchers in AI.