Reconciling The AI-Human Conflict With The Centaur Model