I've taken on Intelligent Systems(AI) subject.

Miscellaneous Topics
#1

It’s a 3rd year subject, which I’m doing in my first year because of my previous qualifications.

Anyway, I’ve partially read 2 books on AI. One explains how AI is implemented in computer programs, the other explains concepts on fuzzy logic, expert systems, nuero systems, genetic systems and evolving systems.

Neither book is suited to my taste. The first book goes on about implementation on AI(Which I can do a better job by doing it my own way) and the later goes into way too much detail when a simple sentence would be enough to explain a paragraph.

I’m going to search GameDev for useable information on AI. Honestly, I’d rather be given the concepts and do the rest on my own.
Hopefully the subject will be better(even though the subject uses the latter text book).

Anyone have any advice for me?

#2

Basically what you need (if you haven’t got it already :wink: is a good grasp of high level logics. Beginning with first order logics and moving on to fluent based approaches etc etc.

When/If you’ve got that the importants principles of AI can, probably, be understood without any book with tons of “text-per-concept” - google is your friend. I.e. when you have no problems with mathematical logic, the reason behind ramification problems etc. stop being difficult to understand/solve and you can move on to grasping the motivations behind different techniques used to develop AI systems today.

#3

If you’re reasonably technically aware it may be worth going back to some of the papers that the books reference. Your college library should have most of the journals in question and often the original papers in a given field are among the best references for it.

#4

What books are those ?

I use “AI a Modern Aproach from Russel Norvig”.

Its not very specific in anything but it covers the basics on almost everything very well.

It doesn’t have anything particulary good for 3d games. For that i sugest you go to one of those AI academic sites and AI repositories for some solid info. GameDev and GamaSutra are good for path finding algorithms but they won’t give anything better than what you can find in a good AI book besides path-finding.

The problem i have with 3d games is that NPCs have to decide when is their turn and what is their objective at the moment, since the environment can change radically in a 3d game. Its completely different than a chess game for instance where each player has a turn the other has to respect and the rules and objectives are defined.

#5

[quote=“K.I.L.E.R,post:1,topic:23669”]
You will finally learn a use for calculating Eigen vectors

I took an A.I. course in my final year. “Pattern Recognition”.
The bloody exams killed me, but I enjoyed the course. We didn’t get to far into the Neural Net stuff… the main bits I remember were for extracting interesting feature measurements from sample sets and using them for automatic grouping and classification.

It’s still a pet project of mine to work on computer opponents for simple games that learn to adapt to the human player’s strategy. E.g. a fighting or shooter game that learns move combinations that the human player is least skilled at defending and changes those combinations as the human player gets better at countering them. The idea being that it would do this all by analysing how previous attacks were defended against and how previous defences failed.

If I come up with something close before I die I will let you all know :wink:

#6

That’s a very good book. It is one of the standard references for general AI. However it is not focused on games at all (except maybe some search techniques) and has about 1000 pages, so it is not a quick help. The book has the advantage that it is mostly easy to understand and manages to give you a good overview of most topics without going too much in the details. If you like it, you should go for the second edition, which contains less errors.