Approaches to alignment stability
I view this as pretty-much a solved problem, solved by value learning. Though there are then issues due to the mutability of human values.
I view this as pretty-much a solved problem, solved by value learning. Though there are then issues due to the mutability of human values.