How do machine learning models do what they do? And are they really “thinking” or “reasoning” the way we understand those things? This is a philosophical question as much as a practical one, but a new ...