Stephen McAleese comments on How might we safely pass the buck to AI?

[Stephen McAleese]

I agree that this seems like a core alignment problem. The problem you are describing seems like a rephrasing of the ELK problem.