Agree with the last one, but we still work on a lot of reinforcement learning, because our world models are not present/perfect. @SahinLale is working at the intersection of learning and control. Check out his work for excellent theoretical background as well as empirical results