Preview of Biologically Inspired Artificial Intelligence for Computer Games (Book)

A new book recently came out explaining how to apply biologically inspired AI to games. It covers the classics in computational intelligence such as genetic algorithms, neural networks, artificial immune systems and particle swarm optimizations — not forgetting underrated techniques like reinforcement learning, independent component analysis, and radial basis functions.

I found out about the book via the Learning Games blog by Daniel Livingstone, who co-authored the book. As a disclaimer, I was asked to write the foreword for this book over 2 years ago (posted below), but the content has changed significantly since then. It’s great to see this book finally hitting the shelves!

This book’s edition is currently a hardcover, and is available from Amazon.com resellers at round $82 USD, and Amazon.co.uk resellers for around £42 GBP. These things seem to be selling out fast; there are few in stock so move quickly! (I wonder if the cover image has anything to do with it…)

Description

Until I get my hands on a full copy to review, here’s the official description of the book by the publisher:

“Computer games are often played by a human player against an artificial intelligence software entity. In order to truly respond in a human-like manner, the artificial intelligence in games must be adaptive, or respond as a human player would as he/she learns to play a game.

Biologically Inspired Artificial Intelligence for Computer Games reviews several strands of modern artificial intelligence, including supervised and unsupervised artificial neural networks; evolutionary algorithms; artificial immune systems, swarms, and shows—using case studies for each to display how they may be applied to computer games. This book spans the divide which currently exists between the academic research community working with advanced artificial intelligence techniques and the games programming community which must create and release new, robust, and interesting games on strict deadlines, thereby creating an invaluable collection supporting both technological research and the gaming industry.”

Foreword

As game developers a few years from now, we may look back at current generation AI with astonishment and possibly with a hint of nostalgia. We will notice the extreme simplicity of the systems and behaviours we created, asking ourselves whether it was in fact by design or more by necessity. More importantly, we will be surprised by the amount of time taken to prototype such AI. These classical game AI techniques (such as finite-state machines or scripts) appeal to us game developers because they can be easily controlled, but they suffer from this very same property; the content creators must craft each behaviour manually. Any progress in game AI necessarily involves finding a solution to this bottleneck without our designers loosing control over their creations.

This book provides an overview of modern AI techniques that can alleviate the burden. Biologically inspired AI techniques—and machine learning in general—provide us with methods and tools for creating complex in-game behaviours in a manageable way. Not only does this allow designers to prototype simple tasks quickly, but it also means that additional resources can be directed to improving the believability of our characters. While such AI techniques enable the creation of behavioural content in a much more scalable fashion, they are also broadly applicable across the many disciplines of game programming, particularly in off-line tools, as they are already used in parts of the games industry. In general, the next generation of games is relying increasingly on third-party software, rather than home grown systems. These solutions provide strong toolsets and standardised algorithm that empower the game developer to create game content with indirect support from the programmer. Biologically inspired AI techniques fit into this paradigm since they are available as a set of algorithms that process data. However, these techniques are no longer just loading the designers’ data directly from disk, but they are in essence assisting the creation process. For example, the AI can provide statistics about our gameplay prototypes, and find patterns in playtests to detect possible anomalies. Such powerful computational tools help produce higher quality behaviours and increase the scalability of the development process.

The role of us AI programmers in this environment is to provide a framework for the content creators. The designers decide what they want out of the machine learning, and we use our knowledge and experience to get the best results. Of course, there will always be a need for the classical AI techniques, but developers will increasingly default to modern approaches. This book is a great opportunity for game programmers to extend their skill set with modern AI techniques. It may take a bit of time for biologically inspired AI techniques to become commonplace in the industry, but there is certainly a need for them. Most players dream of playing against highly believable adaptive opponents, and with developers that can leverage modern technology to solve any problem at hand, this goal is certainly attainable.

Table of Content

1. Contemporary Video Game AI

2. An Introduction to Artificial Neural Networks

3. Supervised Learning with Artificial Neural Networks

4. Case Study: Supervised Neural Networks in Digital Games

5. Unsupervised Learning in Artificial Neural Networks

6. Fast Learning in Neural Networks

7. Genetic Algorithms

8. Beyond the GA: Extensions and Alternatives

9. Evolving Solutions for Multiobjective Problems and Hierarchical AI

10. Artificial Immune Systems

11. Ant Colony Optimisation

12. Reinforcement Learning

13. Adaptivity within Games

14. Turing’s Test and Believable AI

Get the full details about the book from the publisher’s website, and order it from Amazon US and UK.