Interpretability essay argues neural networks are "already clean" — a paradigm shift
teortaxesTex · x · 2026-10-16
Commenting on a viral interpretability essay, teortaxesTex calls it "a paradigmatic shift." The quoted text argues: neural networks are "already clean" in a deeper sense than monosemanticity — their parts are built to be read by downstream readers, making the net "a fractal of explanations." The update is a change of kind, not degree: like moving from intelligent design to natural selection, they're 85% sure nearly all mechanisms come from a different kind of process. This lowers the stakes on monosemanticity scores — the cause of polysemanticity matters more than its count — and suggests interpretability's goal is to "become another reader."
More from AGI Musings
- Kroger cuts 3,000 corporate jobs (37%) for AI; humanoid robots next in the 2030s — csuwildcat · 2026-10-16
- Ben Thompson: generating human desire will be the post-AI economy — generativist · 2026-10-16
- Interpretability won't save us: critics call superintelligence project reckless regardless — zetalyrae · 2026-10-16
- John Nash's 1998 Hierarchical Introspective Logics paper recirculates in AI circles — fkasummer · 2026-10-16
- Ex-OpenAI interp lead: most neurons in trillion-param models are now legible — CatAstro_Piyush · 2026-10-16
- Researcher disavows successionism: watch the button-presser who wants to end the universe — jd_pressman · 2026-10-16