NeurIPS workshop on interpretability for discovery: AI found vowels in sperm whale calls
begusgasper · x · 2026-11-19
Gašper Beguš (UC Berkeley) previews the "Interpretability for Discovery" workshop at NeurIPS 2026 (December 12, Atlanta), exploring how model interpretability can turn learned representations into novel, testable scientific knowledge.
Highlight case: interpretability methods helped find vowel-like structures in sperm whale vocalizations. Invited speakers include Katrin Franke (Stanford), Nicholas Wang (Goodfire), and John F. Wu (Johns Hopkins). The workshop is accepting papers until September 2, AoE.
More from Research
- CoRL 2026 best paper SuperMimic accused of repackaging 2019-era work with slick marketing — animesh_garg · 2026-11-21
- ELLIS Institute Tübingen Brings 48 Papers, 2 Orals to Three-City NeurIPS 2026 — maksym_andr · 2026-11-21
- Agentic science study sparks call for reproduction fields on every agent preprint — tallmetommy · 2026-11-21
- NanoGPT Speedrun Record Falls to 5,962s, Now 49x Faster Than Baseline — Madisonkanna · 2026-11-21
- Harvard chemistry polymath George Whitesides dies, leaving ~1000 papers across chemistry — CatAstro_Piyush · 2026-11-21
- a16z researcher: AI can solve Millennium math problems with Lean proofs but can't explain them — a16z · 2026-11-21