The machine learning capabilities of today's AI are extremely exciting, but they are only one part of the architecture of today's systems. For example, self-driving car technology uses machine learning to interpret the visual field around a car, but it's good old fashioned rule-based AI that contains the plans on what to do with that information.Similarly, the phenomenal program that plays Go at a master level uses machine learning to evaluate board positions (very difficult for a machine to do) but uses conventional AI for deciding what to do with that information.Written a number of years ago, Building Expert Systems in Prolog is still a classic text on knowledge engineering. The name might be a bit misleading, as it's not so much about building expert systems, but rather on building the knowledge representation and reasoning engine tools used for various types of expert system applications. (Including, of course, examples of each kind of system.)As such, it provides details and working code for identification systems, systems that need to cope with uncertainty, forward-chaining planning and configuration systems, frames for representing complex layers of information, and ideas for prototyping/implementing any way of representing knowledge and algorithms for reasoning over that knowledge.The text is a bit more general than just a Prolog text.

Yes, all the code for implementing these systems is written in Prolog, but Prolog is more than just an implementation language. Because it is logic programming, a Prolog program is really a logical specification of an application. Yes it runs, but it can also be used for rapid prototyping, and the Prolog code itself can be used to precisely specify how to implement the system in any other language.In other words, the architectures of these systems will be the same, no matter what language they are implemented in, and the use of Prolog for rapid prototyping is covered in the book as well.