The Alignment Problem: Reward, Values, and Policy
Why I think reward always carries assumptions, and why AI systems should remain uncertain and correctable.
Why I think reward always carries assumptions, and why AI systems should remain uncertain and correctable.