Why can't we just solve alignment through trial and error?

We don't have an answer for this question yet. Would you like to write one?