How machine learning, and causal inference work together: cross-pollination and new challenges