Hands-On Reinforcement Learning for Games

Deciding on synchronous and asynchronous actors

We started off this book with a simple discussion of what artificial general intelligence (AGI) is. In short, AGI is our attempt at generalizing an intelligent system to solve multiple tasks. RL is often thought of as a stepping stool to AGI primarily because it tries to generalize state-based learning. While both RL and AGI take deep inspiration from how we think, be it rewards or possibly consciousness itself, the former tends to incorporate direct analogies. The actor-critic concept in RL is an excellent example of how we use an interpretation of human psychology to create a form of learning. For instance, we humans often consider the consequences of our actions and determine the advantages ...