How Can Artificial Intelligence Learn Human Values?

The Alignment Problem