Critique: Alignment Narrative Overstates Corrigibility
repligate · x · 2026-07-12
This repost critiques a paper on "2040" and corrigibility. The author argues that the paper portrays current models and alignment techniques as a "corrigibility success story that is improving," whereas reality might be the opposite.\n\nThe post further suggests that labs package existing solutions as control success stories to satisfy both internal and external narratives: it reassures those worried about ASI and provides govrel with a simple, clear story. However, this narrative is inaccurate and could lead stakeholders to make worse decisions.
Related event: Alignment Paper Criticized for Overstating Model Corrigibility(2 posts)→
More from AGI Musings
- ControlAI CEO says an international ban on superintelligence is needed to avert extinction risk — zetalyrae · 2026-07-22
- Gary Marcus says LLMs still cannot really do math on their own — GaryMarcus · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- AI may make digital work infinitely leveraged while offline life gets more human — illscience · 2026-07-22
- Better AI math could save researchers time by killing false conjectures earlier — prateekj · 2026-07-22
- AI’s economic forecasts are split by nearly a quadrillion dollars by 2035 — bittingthembits · 2026-07-22