NDIF is hiring AI interpretability researchers and opens a Steerability Challenge to suppress AI deception
davidbau · x · 2026-09-21
davidbau announces that NDIF (@ndifteam), the deep-inference interpretability infrastructure effort, is hiring for an AI interpretability/safety role, with more openings to come.
Alongside the hiring, IBM's Erik Miehling and Tara Research, built on NDIF infrastructure, have launched the Steerability Challenge: instead of detecting AI lies, competitors must suppress AI deception. Registration is open and cash prizes are offered; a prior whitebox lie-detection contest preceded it.
Related event: IBM Launches Steerability Challenge to Curb LLM Dishonesty(2 posts)→
More from Companies & People
- ETH AI Center opens PhD/Postdoc fellowship applications with AI-driven biomolecular design tracks — victorgreiff · 2026-09-21
- Investor: Meta's latest acquihire could be the most consequential in its history — sarahdrinkwater · 2026-09-21
- V7 Labs partners with OpenAI to build AI operating system for financial work — nathanbenaich · 2026-09-21
- Shopify CEO Tobi Lütke: AI's New Failure Mode Is Over-Output, 'Slop Grenades' — WalterReade · 2026-09-21
- Gurman on new Apple CEO Ternus: first AI point of view after two years of muddle — The Verge AI · 2026-09-21
- Georgia Chalandidou Elected VP of the German Robotics Society (DGR) — GeorgiaChal · 2026-09-21