Expert Systems and Logic Games
By ai-depot | June 30, 2002
Overview
A good place to start would be a widely accepted definition of an expert system:
“Expert Systems are computer programs capable of solving problems using a knowledge base generally crafted by human experts.”
Despite their dependence on humans, and their characteristic inflexibility, expert systems are considered as part of Artificial Intelligence. Indeed, they can usually be designed to solve domain specific problems, giving them narrow intelligence. (See our essay on Intelligence and Black Boxes).
As you may expect, an expert system is somewhat limited to the ability of the expert, and the AI programmer. For budget reasons, this is often the same person, but his ability to learn the task of the other will prove crucial for the quality of the final system. Naturally, hiring both a team of experts and a team of programmers will provide the best results, getting them to interact to design the ultimate expert system.
With a bit of time and experience, very good results can be achieved with an expert system. There are a few major problems that need to be taken into account:
- Reliability - When designing an expert system, it is very difficult to take every situation into account. As such, the rules of the system need to be established such that unforeseen circumstances are handled without catastrophic glitches.
- Flexibility - Since it is very tedious to handle every scenario explicitly, expert systems need to have a certain amount of flexibility (more or less limited). This allows them to perform to expectation in their working domain.
Once these problems have been taken care of, an expert system is usually capable of producing very good results. This can also be done in a very efficient fashion, since the programmer has coded a significant part of the system explicitly. This is ideally suited to computer games, where efficiency and predictability are two very important issues in the design.
Tags: none
Category: essay |
