That's not the type of AI typically used in games.

I dunno man… there are all of these exercises using ANNs, SVMs, and other regression and classification methods that could be used in games.

While true that machine learning and classification methods are useful, and can be used in some games especially as development tools, they're not what is typically meant by “Game AI”.

Game AI focuses on creating fun gameplay rather than machine learning. Utility functions and weighted ranks are everywhere. Complex decision and behavior trees, statistics functions for providing different probabilities to designers, and using those probability settings and designer-tuned values to run probability-based state machines, those are the norm. More complex methods like monte carlo search trees and minimax solutions are a little less common but still frequently used.

Game AI developers don't use ANNs, supervised learning, relaxation networks, or other machine learning techniques because it is extremely difficult (if not impossible) to tune the game for fun with them. Sometimes techniques can be used as gesture recognizers, as image processing techniques, and in development tools, but not as gameplay AI.

They wrote about blueprint programming, and learning in the Game Development program at SNHU. That's not extensive machine learning. That's game AI.