Panel: The shape of AI governance
summary
Panel argues governance rests on concrete practices not slogans and recommends starting at data: track origin of data, version models, add explainability and real‑time monitoring with canary tests to detect data drift. Practical tools cited include Captum and InterpretML for feature attribution, plus bespoke fairness metrics and a fairness and bias quality score; synthetic data is expensive and imperfect, so prefer simple, explainable models for most use cases and reuse existing MLOps rather than rebuild.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:22What is fairness
- 7:25Data quality
- 13:25Synthetic data generation
- 15:08Misconceptions
- 18:24Life cycle
- 20:15Measuring fairness
- 23:13Practical tips
- 23:36Explainability
- 32:39Recommendations
- 35:09Culture
- 36:41Government oversight
- 38:16Two types of rules
- 39:56Federal oversight
- 41:39A neutral body
- 43:32Conclusion
related talks
39mResponsible AI
39mThe Search for Transparency and Accountability in the Age of...
44mFuture of AI
55mPanel: Gender bias in AI - In collaboration with Women in Data Science Belgrade- 39mSponsored Session: The Search for Transparency and Accountability in the Age of AI: RAI & XAI as Essential Tools
22mBuilding Governance To Maximise AI Business Impact
37mBuilding a case for Fair and bias-free AI
44mStreamlining Your Approach to Data