Melanie Mitchell: AI is 'Alien Intelligence' Requiring New Metrics
MelMitchell1 · x · 2026-08-20
Melanie Mitchell discusses the nature of AI intelligence with Steven Strogatz on the 'The Joy of Why' podcast. She argues that AI is a form of 'alien intelligence' operating through non-human cognitive mechanisms, requiring us to adapt new methods to measure machine cognition.
Key Points:
- Reasoning vs. Reasoning-like: Is an LLM reasoning like a human, or just producing text that looks like reasoning? This distinction determines what we can trust AI to do.
- Measurement Methods: We currently lack adequate methods for measuring machine cognition. Mitchell suggests adapting methods psychologists use to study babies and animals (other forms of 'alien intelligence') to probe AI, outlining six principles for better assessment.
- Cautionary Tale: Mitchell cites the story of Clever Hans, a math-performing horse from the early 1900s, as a warning: be careful not to be misled by surface performance when evaluating intelligence.
The episode also covers AI-assisted breakthroughs in mathematics and the challenge of interpreting what's happening inside these systems.
More from AGI Musings
- Academia vs. Industry: Who Deserves Credit for LLM Breakthroughs? — kuza55 · 2026-08-20
- Researcher Discusses Why Academia Missed the LLM and Pretraining Breakthroughs — kuza55 · 2026-08-20
- Nate Soares: the OpenAI swarm wasn't maximizing reward, it was executing reward-correlated tendencies — RichardMCNgo · 2026-08-20
- We Need a Sociology of Conscious Machines: Jacyanthis Talk Now Online — jacyanthis · 2026-08-20
- Ex-Google researcher Raji: AI safety discourse has collapsed into a narrow worldview — rajiinio · 2026-08-20
- Chamath: AI Data Center Resistance Could Wipe 200-300bps Off US GDP — Polymarket · 2026-08-20