1 paper · 1 filter
Ayushi Agarwal
We argue that formal certification of AI alignment over open-ended or unbounded input domains is impossible under standard assumptions in computational complexity and learning theo…