Protective Optimization Technologies
summary
Argument that optimization systems (cloud, feedback loops, objective functions) create externalities that privacy fixes like differential privacy don’t stop, and that defensive tools should reoptimize inputs to shift outputs; formal proposal called Protective Optimization Technologies models world states, agents, observations and actions to either internalize costs or sabotage harmful objectives. Cited examples: Waze ghosting and routed surface‑road congestion, GPS spoofing and fake police reports, drivers toggling apps and credit‑score manipulation experiments; some grassroots tactics succeed but corporate unresponsiveness and governance gaps remain.
AI-written summary · about
chapters (19)
- 0:00Intro
- 1:07Prelude
- 2:03Data
- 5:00Box
- 8:36Systems
- 11:00Risks Harms
- 12:51Location Services
- 13:36Waze
- 23:04Optimization Systems
- 27:14Centralized Planning
- 28:27No Optimization
- 30:51Fairness
- 32:30Proposition
- 39:49Conclusion
- 42:00Comments
- 46:46Optimization vs Equilibrium
- 50:20Economics vs Politics
- 52:21Why are these questions fundamentally different
- 55:26Centralization still makes a difference
related talks
21mEthics in the age of AI: Strategies for mitigation and their historical context
01h 08mDo-it yourself privacy protection. Empowerment or burden?
40mAgainst the Tyranny of Optimization: On the Stability of Automated Republics
20mPrivacy Paradigms for Law Enforcement Response
22mPrivacy for Safety - OPSEC When the Threat is in the Home
26mSecurity, Politics, Neutrality, and Protecting Users
50mShifting Privacy In: How Privacy and Security Can Strengthen Each Other
12mPredictive Response Optimization: Using Reinforcement Learning to...