Talk of Chris Russell (AWS)

Thumbnail

Event details

Date 21.02.2023
Hour 11:0012:00
Speaker Chris Russell (AWS)
Location
Category Conferences - Seminars
Event Language English
Title of the talk: The limitations of algorithmic fairness

 Abstract: This talk will take a hard look at what works and doesn’t work in the field algorithmic fairness, and the implications of existing approaches.
We will discuss the challenges of enforcing fairness in the high-capacity models common to computer vision (Leveling down in computer vision: Pareto inefficiencies in fair deep classifiers – CVPR’22)
How methods for enforcing fairness may violate anti-discrimination laws (Are Two Heads the Same as One? Identifying Disparate Treatment in Fair Neural Networks _ Neurips ’22)
As well as a general discussion about how the behavior of fair ML is often undesirable, and  possible alternatives (https://www.wired.com/story/bias-statistics-artificial-intelligence-healthcare/ Wired ’23)

Bio: Chris Russell is an ELLIS fellow and Senior Scientist at AWS. A cofounder of the Governance of Emerging Technology Research Programme at the Oxford Internet Institute, his work on XAI is cited in the guidelines to the GDPR, and has influenced the AI act. His previous work has won best paper award at ICRA; ECCV; and BMVC.
 

Practical information

  • Expert
  • Free

Organizer

  • Professor Volkan Cevher

Contact

  • Gosia Baltaian

Event broadcasted in

Share