The Pulse of AI Conversation with Anupam Datta Co-Founder and Chief Scientist at AI Start-up TruEra
Dec 3, 2021
Anupam Datta, co-founder and Chief Scientist at TruEra and a renowned professor at Carnegie Mellon University, dives deep into the pressing issue of AI bias. He discusses the various forms of bias in AI systems and their implications for businesses. Anupam emphasizes the need for fairness and transparency in AI to foster trust. He also highlights the evolving regulatory landscape and the importance of external audits. With practical examples, he sheds light on how corporations can hold vendors accountable, ensuring a fairer AI-driven future.
AI Snips
Chapters
Transcript
Episode notes
Google Ads Bias Sparks TruEra
- Anupam Datta's research uncovered gender bias in Google's job ads targeting men more than women.
- This early study sparked ongoing work in fairness, explainability, and led to founding TruEra.
Statistical Bias Is Double-Edged
- Statistical bias is essential and inevitable in machine learning to differentiate outcomes like credit risk or fraud.
- However, this bias can inadvertently cause unfairness when data is imbalanced or proxies correlate with protected groups.
Fairness Metrics Expose Biases
- Metrics like disparate impact ratio compare group outcomes to reveal fairness issues, like lower credit approvals for minorities.
- Facial recognition errors disproportionately affecting people of color expose serious fairness concerns.