The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) How External Auditing is Changing the Facial Recognition Landscape with Deb Raji - #388
Jul 2, 2020
Deb Raji, a Technology Fellow at NYU's AI Now Institute, tackles the pressing issues surrounding facial recognition technology. She shares insights from her work on the Gender Shades project, revealing biases against darker-skinned females. The discussion touches on recent moratoriums from tech giants like IBM and Amazon, highlighting the urgent need for ethical standards and independent audits. Raji also critiques practices like those of Clearview AI, emphasizing the risks of digital surveillance, particularly for marginalized communities.
AI Snips
Chapters
Transcript
Episode notes
The White Mask
- Joy Buolamwini's TED Talk inspired Deb Raji's work.
- Buolamwini's facial recognition software didn't recognize her face until she wore a white mask.
Gender Shades' Impact
- Gender Shades audits revealed a 30% accuracy disparity between darker female and lighter male subgroups.
- Targeted public pressure is crucial for prompting companies to address bias in facial recognition.
Auditing Limitations
- Companies improve specific metrics when audited, but broader bias persists.
- Microsoft improved gender classification after Gender Shades but still showed significant age classification disparities.