After the Hugging Face attack, ex-OpenAI board member Helen Toner explains why AI goes rogue
timrudner · x · 2026-09-20
On the Offline podcast, host Jon Favreau interviews AI researcher and former OpenAI board member Helen Toner about the Hugging Face attack, why we can't "just unplug the AI," and why AI companies want models building new models.
Tim Rudner recommends his 2021 non-technical paper with Toner, "Key Concepts in AI Safety: Specification in Machine Learning" (CSET). The brief, fourth in a series on AI safety, explains specification: conveying to an ML system exactly what its designers want. For many tasks it's hard to capture intent in precise mathematical language—like King Midas or the Sorcerer's Apprentice, you may get exactly what you asked for, not what you wanted. The framing is more relevant than ever as agents gain autonomy.
Related event: Ex-OpenAI board member Helen Toner talks AI safety and oversight(2 posts)→
More from AGI Musings
- Andreessen: Agents Are Entering Enterprises as Chaos, Just Like PCs and Spreadsheets Did — Dan_Jeffries1 · 2026-11-04
- 1000+ mathematicians sign pledge to shun AI companies and models — Shesawthat · 2026-11-04
- Von Neumann meme: "You don't have to be responsible for the world that you're in" — akbirthko · 2026-11-04
- Joscha Bach on the dream chocolate that stopped being real — Machine Learning Street Talk · 2026-11-04
- François Chollet: modeling brain wiring as an engineering problem constrained by genome size and time — CatAstro_Piyush · 2026-11-04
- Most US Couples Now Meet in a Data Centre, Doubling Office Matches — menhguin · 2026-11-04