RL algorithms are the usual choice when we're faced with sequential decision problems. Usually, it's difficult to find other ways to solve these tasks other than using RL. Despite the hundreds of different optimization methods that are out there, so far, only RL has worked well on problems for sequential decision-making. But this doesn't mean it's the only option. 

We'll start this chapter by recapping on the inner workings of RL algorithms and questioning the usefulness of their components for solving sequential tasks. This brief summary will help us introduce a new type of algorithm that offers many advantages (as well as some disadvantages) that could be used as a replacement for RL.