AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is Objective Maximization a Bad Frame?
Human goals are not necesarily very well modelled as just, you know, objective maximization. In some sense, they trivially have to be. I don't feel like it's a very good specification language for these agents. And so the idea is that by having the agent preserve its ability to do a wide range of different objectives, it'll also,. perhaps accidentally, preserve its right objective - though we can't begin to specify what that right objective is.