Human oversight is broken: monitoring systems are themselves misaligned in correlated ways
davidmanheim · x · 2026-11-16
AI safety researcher David Manheim argues that even if alignment proves tractable, the current trajectory ends in disaster: AI systems are far faster than humans and deliver most value precisely when unmonitored, while the systems we rely on for human oversight are themselves misaligned in correlated ways — making meaningful supervision fundamentally limited.
Related event: AI Safety Researcher Warns Human Oversight Is a Dead Loop(2 posts)→
More from AGI Musings
- Marcus snaps at critic: 'Bregman or people like him have contaminated your brain' — GaryMarcus · 2026-11-16
- Tim Lee: A society of cognitively impaired adults would collapse long before AGI arrives — binarybits · 2026-11-16
- Demis Hassabis says AI demands flipping education; satirist quips schools will teach 'artificial scarcity' — moultano · 2026-11-16
- Gary Marcus challenges dev: show symbols in your agent harness or you're confused — GaryMarcus · 2026-11-16
- Debate: Was Tool Calling an Unpredictable Leap, or Predictable Since 2022? — OwariDa · 2026-11-16
- Dean Ball: top AI models mode-collapse into 2020-era governance discourse — deanwball · 2026-11-16