Educability (Virtual Talk)
Talkat Understanding Higher-Level Intelligence from AI, Psychology, and Neuroscience Perspectives · 2024-06-24..06-28·
summary
Proposes "educability" as a concrete, computational capability composed of three parts: learning from examples (standard PAC-style generalization with error falling algebraically as n grows), robust logic chaining (compose many learned classifiers with sound probabilistic guarantees—e.g., chaining two 99% rules yields predictable downstream accuracy), and receiving programs (explicit instructions/recipes transmitted symbolically); builds a mind’s-eye model for multi-object relational scenes, notes vulnerability to unverified beliefs and suggests testable measures of educability.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:19Subtle Science
- 2:58Understanding Human Capabilities
- 5:26Cognitive Capabilities
- 7:52Civilization Enable
- 11:41Definition of Educability
- 17:14Learning by example
- 22:42Robust logic
- 24:47Robot logic
- 31:52Quantitative things
- 32:55Learning from others
- 33:44Upside of Educability
- 35:54Downside of Educability
- 37:35How to become more educable
- 39:20Parameters
- 42:01Singularity
- 43:30Eugenics
- 45:41Questions
related talks
52mCredal Models for Uncertainty and Logic Treatment
01h 05mAdaptive Generalization: The Role of Dynamic Evaluation in Low-Resource Settings
41mTheory of Modern AI: Learning Theoretic, Game Theoretic, and Algorithmic Perspectives
42mReasonable AI
39mComputational Thinking on Learning Models
47mUnprovability and Other Impossibility Results in Machine Learning
11mThe Unbearable Lightness of (Agentic) Evaluations
50mCyber-Informed Machine Learning: End-User Value through Explainability