Hado van Hasselt, a researcher at DeepMind, mentioned in one of his videos (from 7:20 to 8:20) on Youtube (about policy gradient methods) that there are cases when the policy is very simple compared to the value function - and it makes more sense to learn the policy directly rather than first learning the value function and then doing control. He gives a very simple example at minute 7:20.
What are some other real-life examples (even just one example) of simple policies but complex value functions?
By real-life example I mean an example that is not as simple as a robot in a grid world, but some relatively complex real-world situations (say, autonomous driving).
