Human Computation
summary
Defines formal “humanly computable” algorithms with tight time budgets: preprocessing up to two hours, generation ≤60s (preferably 30s), and step costs measured to seconds. Proposes PHUSSPix schema (publishable, analyzable, secure, humanly usable), gives concrete mapping-from-letter-to-word constructions with measured quality 6.5 and quality 20.5 meaning an adversary needs ≈7 or ≈20 observed challenge–response pairs to break it. Models humans as finite‑state plus long‑term memory and a tiny short‑term stack (≈7±2 chunks, typically 2 items); demos show ~3s per digit, ≈30s per password. Conjectures CSP solvers can break any humanly usable schema in comparable time; one‑way and zero‑knowledge in‑head functions seem infeasible.
AI-written summary · about
chapters (9)
- 0:00Intro
- 1:26Password Creation
- 2:47Password Schema + Key
- 8:32The use of schemas has many interesting
- 25:35About human computability 1/6
- 26:07About human computability 2/6
- 28:07About human computability 3/6
- 38:26Long-Term Memory 3/7
- 39:33Short-term (working) memory 4/7
related talks
27mFast, Lean, and Accurate: Modeling Password Guessability Using Neural Networks
40mHuman Learning
43mDehuman future Fifth sun
07mHuman brains runs on Potatoes
45mBridging the gap between AI Research and Human-Centered Products
19mDesign and Evaluation of a Data-Driven Password Meter
43mPushing the Limits of Zero Knowledge Applications
43mThe Biological Path Towards Strong AI