The Alignment Problem: Machine Learning and Human Values - Brian Christian