If we want to solve the alignment problem with AI, I think we need to first solve the alignment problem with Humans.