Barath Raghavan and Bruce Schneier: Why AI Needs a "Genie Coefficient" Proposing a new metric for whether AI does what you actually want.

"Major benchmarks measure what AI can do. None measure whether it does what you mean: the distance between what you ask an AI to do and the unspoken assumptions about how you want the AI to do it."

I love the name!

Such a measure also captures prompt injection, and things like prompt injection where it just naively doesn't do what you want.