Unlocking societal benefits of AI with privacy protective technologies.
summary
Panel argues concrete tactics: deploy federated learning and differential privacy to do more with less personal data, use Private Join & Compute for joint analytics, and run policy scanners like Claudette to audit long privacy notices; regulators pilot regulatory sandboxes and accountability checklists (the “accountability wheel”) so organizations show traceability, impact assessments and redress. Examples: on-device models and Google-style federated pipelines are live; medical imaging shows higher accuracy but raises bias against minorities and higher carbon costs, so tradeoffs and sector‑specific risk rules are needed.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 15:20Timekeeper
- 22:21Unlocking the potential
- 23:11Next steps
- 26:16Disclaimer
- 26:59Outline
- 27:25What is societal benefit
- 28:42Demystifying AI
- 29:58Public interest
- 31:51Surveillance
- 32:40Data protection
- 36:19Data privacy enhancing technologies
- 37:14Practical approach
- 38:05Apologies
- 38:45A golden age of privacy
- 43:46Protecting users while realizing the benefits
- 47:45federated learning
- 49:00ethical perspective
- 50:09what has changed
- 57:59end
- 58:29Dan Bogdanoff
related talks
43mPanel: Privacy Design Patterns for AI Systems: Threats and Protections
33mPanel: How Privacy Engineers Can Shape the Coming Wave of AI Governance
01h 19mAligning Data Privacy and AI Governance
38mPrivacy: the Last Stand for Fair Algorithms
11mAddressing Privacy, Bias, and Fairness in AI Systems
01h 09mGOVERNANCE AND REGULATION OF AI FROM THE PERSPECTIVE OF AUTONOMY AND PRIVACY
01h 17mPrivacy and New Technologies
31mOperationalizing Responsible AI