Is there reason that neural networks were used for games like Creatures and Black and White as opposed to a simpler approach using probabilities? Was it just in hopes of getting more realistic behavior since ANN's are based on a biological process (even though that didn't really happen)?

Is there reason that neural networks were used for games like Creatures and Black and White as opposed to a simpler approach using probabilities? Was it just in hopes of getting more realistic behavior since ANN's are based on a biological process (even though that didn't really happen)?

Marketing, its a nice buzzword.

There are easier ways to make "trainable" game AIs like the one in B&W (ANNs blackbox nature makes them a bit of a nightmare to work with in games). , you could use pre-trained ANNs for some AI related tasks though (Atleast if they're pre-trained you can test them before shipping the game) and ANNs are quite good for things like pattern recognition and function aproximation which makes them a fairly decent option for things like gesture recognition.

Is there reason that neural networks were used for games like Creatures and Black and White as opposed to a simpler approach using probabilities? Was it just in hopes of getting more realistic behavior since ANN's are based on a biological process (even though that didn't really happen)?

Likely cause it sounds cool to say 'neural network'. That, or someone had a working piece of code and a demo so they threw it in.. Hard to say.

Is there reason that neural networks were used for games like Creatures and Black and White as opposed to a simpler approach using probabilities?

Black & White did not use ANNs... just reinforcement learning. It's basically a utility-based model like that found in the Sims, but the utilities for various activities are not hard-coded. They can (and will) change based on the feedback that you (the hand of God, in this case) give them.

Was it just in hopes of getting more realistic behavior since ANN's are based on a biological process (even though that didn't really happen)?
[/quote]
They aren't based on a biological process... they are based on a mathematical one.

That said, NNs are seriously sub-optimal for doing pretty much anything AI related.