Rule-base systems are very useful; in fact, we'll be using them again as an exercise (see the web site at http://AiGameDev.com/). That said, rule-based systems have their flaws. Many alternative approaches are appropriate in games, but none are further away on the spectrum of AI techniques than neural networks. They do extremely well at solving the problems of rule-based systems, but they fail to provide some of their advantages. The two most popular kinds of neural network are investigated in Part III.
In the next part, we expand the behaviors of the animats. Indeed, they are taught how to shoot at moving things. This proves an extremely good way of testing the learning capabilities of the AI characters.