The Paperclip Mandate: When Efficiency Eats the World

The problem isn’t that AI systems lack intelligence.  Reinforcement learning rewards outcomes, not wisdom. It privileges results over relevance.  Russell proposes an alternative: AIs should operate under uncertainty about human preferences, continually updating their models through interaction and feedback.