Value Alignment

Why encoding human values into AI is difficult.

Why is aligning AI with 'human values' considered challenging?