Define the Alignment Problem in the context of Artificial General Intelligence. Theoretically analyze the difficulty of encoding complex human values into a utility function, specifically addressing issues of instrumental convergence and the 'orthogonality thesis' regarding intelligence and final goals.