Verification Dilemmas, Law, and the Promise of Zero-Knowledge Proofs

on YouTube ↗54m

" summary

Defines a formal notion called PAC verification and gives constructive protocols and lower bounds showing verification can be strictly easier than learning (quantitative separation d versus sqrt(d) and a qualitative separation where a class is verifiable from random samples but not learnable without membership queries). Gives sample bounds: realizable verification needs O(1/ε) samples; agnostic verification is harder. Proposes concrete zero-knowledge and MPC uses—zero-knowledge proofs for eligibility, membership, range, equality—to let courts, buyers, or regulators verify claims without exposing raw data, and flags randomness provenance and ground-truth linkage as practical pitfalls.

AI-written summary · about

" chapters (26)

1. 0:00Start
2. 3:44A Journey into "Law For Algorithm"[Y]
3. 4:43California Proposition 25: Replace Cash Bail with Risk Assessment Tool, Referendum
4. 5:54Risk Assessment System Machine Learning Algorithm
5. 13:28Many Verification Tools for Algorithms
6. 15:13PAC Learning Valiant 1984 A theory of the learnable. CACM, 27(11).
7. 16:28PAC Verification GSRY21
8. 21:18Example Application: Delegation of Machine Learning (ML)
9. 21:36Realizable Case (opt=0) is Trivial
10. 22:10Agnostic Case is Interesting
11. 23:11Lower Bounds: Not Always
12. 23:33Sometimes: Quantitative Separation between Learner and Verifier
13. 23:49Sometimes: Qualitative Separation between Learner and Verifier in Type of Data Access
14. 25:02Can PAC-verification be accepted as a Legal Norms?
15. 28:12Open Source vs. Verifiability: Zero Knowledge
16. 28:56Verification Lemmas in Legal and Policy Settings
17. 30:36Each of these settings points to a Potential use of ZKP
18. 34:25Rest of the talk: Verification Dilemmas Across Legal Landscape
19. 35:55Privacy Law: Verification without Identification . Currently the process of verifying eligibility to access digital systemslase, possessing financial capacity forces access- seekers to expose identifying information, risking the creation of largescale aggregation of permanent "digital person"
20. 37:59Privacy Law: Verification using ZK Use
21. 39:16Deal Making: Avoiding Arrow Disclosure Paradox by Using ZK
22. 42:47Deal Making: The Price of Disclosure of Information
23. 43:02Deal Making with ZK: Limit Disclosure Beyond Company Boundaries Using ZK
24. 44:01Deal Making: Limiting disclosures beyond company boundaries (11) • Mergers and Collaborations: Use MPC+ZK.
25. 45:09Litigation of Trade Secret Disputes
26. 48:05Government oversight: Balancing transparency and secrecy

" related talks

• 33mAn Introduction to Zero Knowledge ProofsShubham Patil, Keertana Ganesh · DevOpsDays Geneva 2023
• 51mWhy you shouldn't trust meKeerthana Ganesh, Shubham Patil · NDC Oslo 2022
• 35mEfficient Zero Knowledge Protocols: The Modular ApproachYuval Ishai · 2nd ZKProof Workshop 2019
• 41mMPC-in-the-Head based Zero Knowledge ProofsAmit Sahai · 2nd ZKProof Workshop 2019
• 41mGKR based Zero-Knowledge ProofsYael Tauman Kalai · 2nd ZKProof Workshop 2019
• 26mApplications of Zero Knowledge Proofs in the Banking IndustryEduardo Moraes · 2nd ZKProof Workshop 2019
• 33mEducational Challenges in the Ever-Evolving Space of Zero-KnowledgeMary Maller, Luis Brandao · Zcon4 2023
• 44mZero-Knowledge Proofs: Simultaneously ensuring integrity and privacyTjerand Silde · NDC Security 2026