Oversight depends on the supervisor being able to tell good work from bad. That assumption strains as systems become more capable, and scalable oversight asks how to keep human judgment effective even then, through decomposition, verification, and tools that help people evaluate what they could not assess unaided.
Future Systems
Scalable Oversight: Supervising Systems More Capable Than Their Reviewers
As AI grows more capable, a hard question follows: how do you meaningfully supervise a system that can, in places, outperform the people reviewing it? Scalable oversight is the study of that problem.