Writing the Playbook for Fair & Ethical Artificial Intelligence & Machine Learning
summary
Company adopts seven AI principles and four explicit red lines (no weapons, no human‑rights violations), builds a central non‑engineering review team with senior exec sign‑off, and requires training plus rewards to slow releases when needed. Human‑impact work uses social equity impact assessments and census‑tract maps (food‑access example) to drive product interventions. Technical tools include adversarial multi‑task learning to remove protected signals, transfer learning to boost underrepresented groups, and transparency via model cards and data cards; prototypes improve tasks like smile detection using consented, diverse data.
AI-written summary · about
chapters (4)
related talks
03mDeveloping AI responsibly
36mKeynote: Ethics in AI: Building Fair and Just Intelligence Systems
35mResponsible AI: Ethics in Software Development
34mResponsible AI: The Key to Ethical, Safe and Inclusive Software Development
42mBecause Having an AI Ethics Code Is Essential for Every Person Within a Company
40mSocially responsible AI: Practical tools for robust design
50mSecurity as Part of Responsible AI: At Home or At Odds?
34mEthical framework for AI : Creating AI powered products that empower people