If we solve ELK, how could we use that to align an AI?

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