Safety researcher: 'Both capabilities and safety' ignores how risk assessments actually work
ambaonadventure · x · 2026-09-25
Safety researcher Heidy Khlaaf pushes back on the 'we can do both capabilities and safety' argument: risk assessments require prioritization by impact and likelihood, which deprioritizes unfalsifiable claims.
If everything is urgent, nothing is urgent — a substantive critique of Pause/Stop rhetoric that lacks risk methodology.
More from AGI Musings
- Stanford Lecture: AI Slashes Build Costs, Entry Barriers and Switching Costs at Once — le_james94 · 2026-09-26
- Stanford Lecture Maps Every AI Era's Bottleneck; Today's Is High-Quality Tasks — le_james94 · 2026-09-26
- As LLMs Grow Context Windows, Redditors Debate the Future of Mathematics — Lolsebca · 2026-09-26
- AI Now at House Hearing: AI Companies Aren't Just Grading Their Own Homework, They're Writing the Curriculum — AINowInstitute · 2026-09-26
- With 60k ICLR abstracts, a researcher mourns LLM-generated figures homogenizing papers — mariyaivasileva · 2026-09-26
- AI-generated French novel shortlisted for literary prizes, professor weighs in to WSJ — TuhinChakr · 2026-09-26