Herbie Bradley: no detailed takeover scenario exists, and takeover is infeasible anyway
herbiebradley · x · 2026-09-25
In an alignment debate, herbiebradley concedes that broadly deployed misaligned AI is a concern and "train misaligned successor" is one path — but he has never seen a detailed threat model or scenario for how it could happen. His current expectation: even if a deployed AI has misaligned goals, takeover is infeasible. The exchange highlights the split in AI safety circles over internal-deployment risk versus practical containment.
More from AGI Musings
- Ross Girshick posts CVPR slides recounting 1.5 decades of object detection and its link to the LLM era — CatAstro_Piyush · 2026-11-11
- Ben Todd wrestles with a 50:50 chance only 18 months remain of 'the human era' — ben_j_todd · 2026-11-11
- Mathematicians face identity crisis as AI solves problems in hours, Corriere interview warns — ValerioCapraro · 2026-11-11
- Why inference speed will be AI products' biggest competitive edge in 1–2 years — itsOmSarraf_ · 2026-11-11
- AI Defense Shifts to Scale Economics: Attacks Buried in Legitimate Traffic — DavidLinthicum · 2026-11-11
- Ben Todd's alt log-log plot shows task horizons hitting a singularity in under 12 months — ben_j_todd · 2026-11-11