Rethinking interpretability: 'machine interpretability' by AIs steering black-box AIs
burny_tech · x · 2026-11-21
Researcher burnytech challenges the notion of interpretability: if it implies 'interpretable by humans,' it is inherently subjective and shifts as AI systems take on more tasks beyond human comprehension.
He proposes 'machine interpretability' as a distinct concept: knowledge about a black-box AI system that allows other AIs to causally steer the model into specific behaviors, outputs, or representations — making the interpreter another model rather than a human.
More from AGI Musings
- Terence Tao: Riemann Hypothesis needs something beyond current AI, maybe human-AI by 2050 — ChrisGPT · 2026-11-21
- Build for automation that serves humans: staying in control — gdechichi · 2026-11-21
- Multiplayer coding agents: agent attention is unbounded, human attention isn't — manosaie · 2026-11-21
- Konrad Kording doubles down: AI people should hold put options as a hedge — KordingLab · 2026-11-21
- The bitter lesson, interpretability edition: humans steer only the weakest AI in the stack — burny_tech · 2026-11-21
- Turing Award winner David Patterson claims superintelligence is here, all science solved by 2030 — davidpattersonx · 2026-11-21