Every time we consider a utility function, its constituent components will carve out large swaths of non-existence space which it cannot apply to. Maximizing happiness excludes those which have no concept of happiness/reward (and offends the non-existent ones for whom reward is pain!).
The space of non-existence is simply too vast and self contradictory to merit consideration. Any utility function will cancel out to 0 across the entire space (because one can conceivably contrive just as many universes where it harms as where it helps).
Though, given this challenge, someone will probably munchkin a self referential, turing-complete utility function that can have a uniform effect across the entire space of existent and non-existent entities.
Every time we consider a utility function, its constituent components will carve out large swaths of non-existence space which it cannot apply to. Maximizing happiness excludes those which have no concept of happiness/reward (and offends the non-existent ones for whom reward is pain!).
The space of non-existence is simply too vast and self contradictory to merit consideration. Any utility function will cancel out to 0 across the entire space (because one can conceivably contrive just as many universes where it harms as where it helps).
Though, given this challenge, someone will probably munchkin a self referential, turing-complete utility function that can have a uniform effect across the entire space of existent and non-existent entities.