Can we create AI ethics before we finish creating AI?