Inserting privacy controls and due process into semi-autonomous algorithms
summary
Argues that machine learning models are opaque in their hidden layers and that Silicon Valley’s open-source surge (TensorFlow, "mtk") and massive investment spike from 2014–2016 fuel data collection — Google and Facebook control about 56% of mobile media. Proposes a runnable concept: a metadata administration server issuing provenance tokens to track source datasets and allowed analytics; tokens can be revoked so consumers opt out. Concept exists as prototype/design and blog posts, not a deployed global system.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:19The world is changing
- 0:53Innovation and invasion
- 1:18Open source frameworks
- 2:09Investment
- 2:38Microsoft Ireland case
- 2:54Machine learning
- 3:40AI is global
- 3:52AI is arbitrary
- 4:08New metadata management model
- 4:39Metadata tokens
- 5:05Conclusion
related talks
01h 10mDIGITAL PLATFORMS, THE NEW PRIVACY CHAMPIONS? BETWEEN MYTHS AND REALITIES
33mMachine Learning in the age of increasing Data Privacy Consciousness
20mProvenance Without Surveillance: Privacy Engineering for AI Content Transparency
01h 20mUnlocking societal benefits of AI with privacy protective technologies.
28mPrivacy Engineering Networking Break AMA
29mAnti-Privacy Anti-Patterns
44mPrivacy by Design Machine Learning
01h 16mLifting the Hood on Big AI - The Future of Transparency and Accountability in AI