Responsible AI – Principles & Practice
summary
Defines Microsoft’s six Responsible AI principles and three ML pillars and cites a Capgemini finding that 9 of 10 executives report ethical AI issues. Covers concrete tools: InterpretML, Fairlearn and DiCE, explains the explainable boosting machine (EBM) and uses differential privacy (WhiteNoise) and homomorphic encryption for protected predictions. Demonstrates an EBM classifier on US Census data predicting income >$50k where global and local explanations show age, marital status and capital gains drive decisions, and the demo runs end-to-end.
AI-written summary · about
chapters (14)
- 0:00Intro
- 3:20Why Responsible AI
- 5:15Fairness in AI
- 6:38How do I embrace responsible AI
- 8:35Principles of responsible AI
- 12:48What is responsible machine learning
- 13:21Understand your model
- 14:28interpretability
- 18:09interpret text
- 19:30fairness mitigation
- 20:25protect
- 22:36homomorphic encryption
- 24:55control
- 26:09demo
related talks
35mResponsible AI: Ethics in Software Development
03mDeveloping AI responsibly
16mResponsible AI at Scale
48mDo No Harm - A Checklist for Responsible AI
40mResponsible AI Practices: Fairness in ML
12mNavigating the Future: Principles of Responsible AI
27mWriting the Playbook for Fair & Ethical Artificial Intelligence & Machine Learning
25mResponsible AI/Ethics in AI engineering