The Day AI Solves My Puzzles Is The Day I Worry (Prof. Cristopher Moore)
Machine Learning Street Talk (MLST)
00:00
3 Actionable Recommendations Based on Cristopher Moore’s Core Insights
1) Practice “frog” thinking — work concrete examples first, then abstract
- Action: Pick one real problem you’re facing this week (e.g., a reporting bug, a design decision, a research question). Spend two focused 45-minute sessions making a concrete toy version you can visualize or simulate (paper sketch, small dataset, minimal code).
- Why: Moore emphasizes that deep understanding often grows from tactile, example-driven exploration (be a “frog” rather than only a “bird”), and that moving between concrete instances and abstraction helps you see what really matters.
2) Build and use simple external data-structures/tools to extend your thinking
- Action: When a task demands recursion, long chains of reasoning, or a lot of state (planning, proofs, complex debugging), create an external workspace: a scratch notebook, a digital stack (notes with explicit push/pop steps), or a tiny script that logs and manipulates intermediate states. Use it whenever your mental context window feels overloaded.
- Why: Moore argues humans (and future AIs) gain recursion and universality by extending memory with external tools—this makes hard, multi-step reasoning tractable and trainable.
3) Insist on transparency for consequential systems; test and verify in small experiments
- Action: For any tool or algorithm you rely on for important decisions (hiring filters, legal/medical heuristics, probabilistic genotyping, automated summaries), require at least one independent check: run an alternative (open-source) method, design a small adversarial or edge-case test, or ask for vendor validation/reporting. If unavailable, refuse to treat its output as final.
- Why: Moore stresses that opaque, proprietary systems are unacceptable where rights or high-stakes outcomes are at risk, and that transparency (independent testing, underwriter-like verification) is crucial.
Transcript
Play full episode
