AI used in the original black and white
I was wondering if anybody knew what AI techniques were used in the production of black and white? I heard it was neural nets but after some research found that it may use reinforcement learning instead. Does anyone know what is implimented and any other information would be greatly appreciated.
Quote: Original post by reidzy
I was wondering if anybody knew what AI techniques were used in the production of black and white? I heard it was neural nets but after some research found that it may use reinforcement learning instead. Does anyone know what is implimented and any other information would be greatly appreciated.
Reinforcement learning can be used in conjunction with backprop nerual nets, both as input for supervised learning or with the ANN providing data to the reinforcement algorithms.
From what I have seen and read, they had a series of AIs that ranged from an individual, a group, and a large area.
I also know that when they first started out, the avitar just sat there and they couldn't figure out why. In debugging they found that it was hungry and was looking for something to eat. It was looking for the nearest edible thing ... it was trying to eat itself. Emergent behavior is great. [grin]
I seem to remember that there was a post-mortem somewhere (on gamasutra?) that discussed it, but I don't see it offhand.
frob.
Actually, they're using the simplest of NNets: perceptrons. These are easily run time trainable, very simple NNets. You could visualize them like this:
(Input Node)------Link with some threshold value-------(Output Node)
So, as you can see, it's essentially the most simple NNet you can make. But, they can be "stacked" (meaning that a perceptron can be attached to another perceptron) so that more complex learning can occur.
(Input Node)------Link with some threshold value-------(Output Node)
So, as you can see, it's essentially the most simple NNet you can make. But, they can be "stacked" (meaning that a perceptron can be attached to another perceptron) so that more complex learning can occur.
I think it was the first AI Game Programming Wisdom book that had about three chapters on B&W's AI.
Check out my new game Smash and Dash at:
December 02, 2005 03:44 PM
http://www.itee.uq.edu.au/~penny/_commercial_AI.htm
So, Decision Trees and Neural Network(perceptrons) as LeatherShorts told.
Thats kinda hard to judge.
So, Decision Trees and Neural Network(perceptrons) as LeatherShorts told.
Quote: I heard that B&W has the most advanced AI to date... is it true?
Thats kinda hard to judge.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement