NeurIPS to Host "Interpretability for Discovery" Workshop in Atlanta in December
NeurIPS will host the "Interpretability for Discovery" workshop on December 12 at its Atlanta venue, focused on mining hidden scientific discoveries inside AI models and turning interpretability research into novel, verifiable disciplinary findings.
Confirmed
- The workshop takes place December 12 at the NeurIPS Atlanta venue, under the name "Interpretability for Discovery" (m2, m3, m4).
- Speakers include Gašper Beguš of UC Berkeley (m4, m5), plus scholars from physics, neuroscience, and other fields; the list mentions kfran… (post truncated, full name unknown) and begusgasper, j…, among others (m2, m4).
- Topic directions include interpretability in astronomy and astronomical foundation models (m2), plus cases of interpretability driving scientific discovery across disciplines.
- The organizers cite, as an example, interpretability research that helped uncover "vowel" structure in sperm whale vocalizations (m3, m5) — a representative case of the workshop's mission.
Unconfirmed
- The edition is in conflict: most posts (m1, m2, m4) say the workshop is at NeurIPS 2025, while one original poster, Gašper Beguš (m5), wrote NeurIPS 2026; the exact year awaits official confirmation.
- m1 mentions the event includes a "spotli…" (likely a spotlight session); the post is truncated and incomplete.
Why It Matters
- The workshop represents interpretability research extending from an AI safety topic toward a tool for scientific discovery: if a model's internal structures can map onto real-world phenomena (like "vowels" in sperm whale calls), interpretability could become a new method for cross-disciplinary research — worth following for researchers in relevant fields.
2026-11-19 ~ 2026-11-20 · 5 related posts
Primary sources
- NeurIPS workshop on interpretability for discovery: AI found vowels in sperm whale calls — begusgasper ·
- NeurIPS workshop on Interpretability for Discovery set for Dec 12 in Atlanta — ChenhaoTan ·
- Interp4discovery workshop on interpretability for scientific discovery at NeurIPS, Dec 12 — begusgasper ·
- [source] NeurIPS workshop on interpretability for discovery: AI found vowels in sperm whale calls — begusgasper · 2026-11-19
- [source] NeurIPS workshop on Interpretability for Discovery set for Dec 12 in Atlanta — ChenhaoTan · 2026-11-20
- [source] Interp4discovery workshop on interpretability for scientific discovery at NeurIPS, Dec 12 — begusgasper · 2026-11-20
- NeurIPS workshop on interpretability for scientific discovery, Dec 12 in Atlanta — ChenhaoTan · 2026-11-20
1 near-duplicate retellings: niloofar_mire