Cambridge Researcher: Four Unresolved Problems Make AI Safety No Quick Fix

Cambridge's David Krueger argues that AI safety cannot be solved on a deadline, citing unresolved problems in interpretability, testing, alignment, and control. He notes methods exist but are unreliable and cannot yet be fixed.

2026-09-22 ~ 2026-09-22 · 2 related posts

• David Krueger: four unresolved foundational problems stand between us and safe AI — DavidSKrueger · 2026-09-22
• AI safety researcher: interpretability and alignment methods work but not reliably, and we don't know how to fix that — DavidSKrueger · 2026-09-22