The AI safety problem is wanting

The article argues that the fundamental problem in AI safety is about wanting - the challenge of aligning AI systems' goals and desires with human values. It suggests that the core issue isn't just technical capability but ensuring AI systems want what we want them to want.