[i_wolf]

Hi guys,
I want to know if any of you have any experience with developing AI Scripts (like frames if that helps jog some memories).
Im looking fora good book to start learning the subject. If you could recommend something that would be great
Kind Rgds
i_wolf

[djohnson]

I have been looking for some things like this as well. I have taken a grad AI class and an undergrad AI class when I was getting my CS degree. Both did not cover AI in a game sense. One was an AI robotics class and the other was an AI theory class. I was going to return and get my Masters in CS and do my work in AI but I have no money... I wish I could have taken some gaming classes for the AI scripts and OpenGL work. Oh well, now I get to learn it for fun! I will look for a couple of AI scripts but am having problems finding them, so any information will be helpful to me too!

[i_wolf]

it was mainly for hobby work that i was interested. When i did my degree in comp sc. we never covered AI scripting much, we tended to stick more to the concept for neural nets.
Thing is now i want to look at scripting AI for game engine just out of interest.

[Arkham_c]

Originally posted by i_wolf:
it was mainly for hobby work that i was interested. When i did my degree in comp sc. we never covered AI scripting much, we tended to stick more to the concept for neural nets.
Thing is now i want to look at scripting AI for game engine just out of interest.

If your goal is less theory and more practical, consider downloading some of the "bot"s for Quake 3. They include the source code and some are quite efficient and effective.

[i_wolf]

excellent suggestion. thanks a lot. i will do just that!

[smeger]

I've done quite a bit of this stuff - mainly genetic algorithms, artificial neural networks, and expert systems. I might be able to answer specific questions, even though it's been awhile since I've played with this stuff.

[TampaDeveloper]

I've got a buddy thats developing a system for military use called Sky-Net. Says its going to be huge.

[jewelskull]

Asking for "AI stuff" covers a pretty wide range. Here's some publicly available stuff I know of that might help you get started in whatever areas you need.

WEKA
http://www.cs.waikato.ac.nz/~ml/weka/
A general statical AI package for Java. Has good stuff for classification, clustering and categorization with the ability to use more methods such as SVM, neural nets, and bayes. Designed for production use and actually has tools.

jBNC
http://jbnc.sourceforge.net/
Bayesian classifier toolkit that also has plugins for WEKA. Once again, for statical reasoning, and they're all the rage today in spam filtering. However, this isn't for Bayesian belief networks.

Bayesian Network Tools in Java (BNJ)
http://bndev.sourceforge.net/
Similar to the above, and in Java.

mySVM
http://www-ai.cs.uni-dortmund.de/SOFTWARE/MYSVM/
Package for doing only SVM (Support Vector Machines) in C/C++. For problems of classification, clustering and categorization the current trend is they're more useful and controllable than neural nets.

CLIPS
http://www.ghg.net/clips/CLIPS.html
Toolkit in C/C++ for FOL (First Order Logic) reasoning and knowledge representation. However, the raw FOL logic is hidden behind APIs that hopefully are easier to use, and data is put into framelike representations. It's an expert system builder toolkit.

ThoughtTreasure
http://www.signiform.com/tt/htm/tt.htm
A knowledge base that hopes to provide a number of assertions to cover a variety of common sense knowledge. It has models for ontology and doing FOL based reasoning on that data. In addition, it has some NLP (Natural Language Processing) to go from NL to FOL. I think it does chart parsing on the NLP side. Implementation is in C/C++ but there is a Java API. Seems to no longer be in active development.

OpenCyc
http://www.opencyc.com/
Probably the largest and longest ongoing attempt to provide an ontology of common sense knowledge and use FOL to reason about it. Has it's own ontology models and FOL reasoning. OpenCyc is only a small portion of the overall knowledge in Cyc's knowledge base, but it's free and has a quite a bit to start with anyway.

SUMO
http://suo.ieee.org/
http://ontology.teknowledge.com/
Attempt to build a standard upper ontology using a version of KIF (SUO-KIF) but with versions also in DAML, LOOM and Protege. It also contains assertions which can be used to reason in FOL about information in the ontology.

WordNet
http://www.cogsci.princeton.edu/~wn/
An attempt to build an ontology directly by classifying natural language words into classes and then relating those classes. Words are grouped into "synsets" which are words that are similar in meaning and therefore related. They provide a simple DB in C/C++ but there are also Java APIs out there to manipulate the data. Honestly, the formats of the files are simple enough you could easily build your own code to deal with it.

GATE
http://www.gate.ac.uk/
General package for NLP and information extraction in Java Like WEKA, it's a framework and a set of tools designed for production use. However there seems to be more emphasis on statistical NLP.

Classic
http://www.bell-labs.com/project/classic/
The original description logic reasoning system in C/C++. Description logic simply puts restrictions on reasoning in FOL to permit representing knowledge in generalized taxonomies, and also make the reasoning faster by forbidding certain things you can do in FOL.