• Advertisement
Sign in to follow this  

Need Some Help Here (Where to start AI)

This topic is 3485 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I wanna learn about Artificial Intelligence but I have no idea where to start. I don't have computer science background( I am doing business studies) so I am kind a stuck right now. These days I am learning C# (I hate Prolog and Lisp they are not fun languages). But problem solving, algorithm design and implementation that's the real thing.I have no idea how I learn this stuff. So would you please advise me where should I start? and what skills I need to develop intelligent softwares. Its highly unlikely that I will develop softwares professionally. I have a bunch of crazy ideas and wanna work on them.

Share this post


Link to post
Share on other sites
Advertisement
The place to start is graph search algorithms. An enormous amount of the AI field is effectively the representation of a problem as a graph, and then whacking that graph with a graph search algorithm.

The obvious first choices are 'Breadth first' and 'Depth first' search. Also google for some graph representations of various problems, such as a rubix cube, or maze path finding [both popular and easy to find choices]. Once you got that down, look into A* and heuristics.

This'll get your feet wet, though it's not likely the AI'ish sorts of sources you're looking for. It is a pretty essential starting point though.

Share this post


Link to post
Share on other sites
I suggest you get an introductory textbook on AI used in university courses. Most likely this will cover what Drigovas mentions (find one that does) and will cver other things also. The "cool" AI is stuff like 'neural networks' and 'genetic algorithms', but in reality 'finite state machines' are probably the most widely used thing in games. 'fuzzy logic' might be worth looking at too.

I wouldn't agree most AI boils down to graphs, but it is certainly a core thing to look at. I'd probably say graphs and state machines are the key things to get a good grip on.

Can anyone suggest a good book? Who's done this at university?

Share this post


Link to post
Share on other sites
Learn probability and use few dices. Then try to invent few algorithms. Majority of AI is about self education, so look at basic concepts.

Also look at FSM.

BTW searching isn't AI, and graphs are just mathematical abstractions. If you don't study at mathematical school graphs are useles to you.

Share this post


Link to post
Share on other sites
I guess it depends on what kind of AI you are going for. Board game AI is almost always just a tree(a type of graph.) If you are interested in FPS style AI then I would take a look at the book Programming Game AI by Example. It covers path finding, flocking, and a simple fps and a simple soccer game.

Share this post


Link to post
Share on other sites
Actually my main purpose is to learn neural networks. But there is only problem, I couldn't find an easy book on this topic. I read a lot of articles but couldn't understand a single word because they just put too much math without explaining how to use it. Moreover, too much theory without implementation. I always learn by doing. So majority of books are totally useless.

So can u suggest any books on Neural network (which teach this stuff without getting me bored).



Share this post


Link to post
Share on other sites
Quote:
Original post by Raghar
Majority of AI is about self education
No, it's a recognised field into which huge amounts of research have been done. There are many established algorithms and techniques which can be learnt/used.

Quote:
BTW searching isn't AI, and graphs are just mathematical abstractions. If you don't study at mathematical school graphs are useles to you.
Searching can be AI, or used in AI. Making your creatures walk across a level is AI. Looking at the various moves you can make and picking one based on a heuristic in chess is AI.
And graphs are not horrendously complex. Many algorithms that operate on graphs are horrible, but a graph itself is very simple, taught in introductory CS courses to people with no math skills.

Share this post


Link to post
Share on other sites
Quote:
Original post by ehsenmaster
Actually my main purpose is to learn neural networks. But there is only problem, I couldn't find an easy book on this topic. I read a lot of articles but couldn't understand a single word because they just put too much math without explaining how to use it. Moreover, too much theory without implementation. I always learn by doing. So majority of books are totally useless.

So can u suggest any books on Neural network (which teach this stuff without getting me bored).
NN are complicated. You should be able to understand the basic principles from reading (there are probably some articles on this site) but to write one yourself is more complicated. However, I am sure there are NN programming libraries which can save you a lot of time.

One thing, many people hear about NNs and think "that's really cool, I should use them", but most games don't use anything so clever. Unless you just want to use it for fun, you need to make sure a NN is the best option. If you describe what you want to do in the AI forum, you can ask those guys what they think the best tools might be. Let us know if you do that, and I'll come read your thread there.

Share this post


Link to post
Share on other sites
Instead of FSMs Check out behaviour trees (do a search there is only one site which explains them in game context) they are far more flexible and their construction does not eat at your sanity.

Quote:
Original post by d000hg
Quote:
Original post by ehsenmaster
Actually my main purpose is to learn neural networks. But there is only problem, I couldn't find an easy book on this topic. I read a lot of articles but couldn't understand a single word because they just put too much math without explaining how to use it. Moreover, too much theory without implementation. I always learn by doing. So majority of books are totally useless.

So can u suggest any books on Neural network (which teach this stuff without getting me bored).
NN are complicated. You should be able to understand the basic principles from reading (there are probably some articles on this site) but to write one yourself is more complicated. However, I am sure there are NN programming libraries which can save you a lot of time.

One thing, many people hear about NNs and think "that's really cool, I should use them", but most games don't use anything so clever. Unless you just want to use it for fun, you need to make sure a NN is the best option. If you describe what you want to do in the AI forum, you can ask those guys what they think the best tools might be. Let us know if you do that, and I'll come read your thread there.

Actually from reading these forums it seems the case is not that nnetworks are clever for game code but that they are too stupid to be worth all their effort. :D

Share this post


Link to post
Share on other sites
I am doing NN for fun guys. And I 'll be more interested in hitting the gym or work on my martial arts training rather then playing a game (Its too feminine).

Share this post


Link to post
Share on other sites
Wow, what a real man you must be. You're right, playing video games is very feminine. That's why nearly everyone that plays them is a guy.

You expect to come to a VIDEO GAME development site and ask for help, and at the same time suggest that playing VIDEO GAMES is somehow feminine?

Share this post


Link to post
Share on other sites
Quote:
Original post by ehsenmaster
These days I am learning C# (I hate Prolog and Lisp they are not fun languages).


Ha, you and I have totally different opinions on that. I consider C# to be a worker language, nice enough for me to use at work and for my xna hobbies. But when I want to program something for project euler or another fun lil project I always reach for my haskell (which is lisp-like)

Share this post


Link to post
Share on other sites
Quote:
Original post by ehsenmaster
Actually my main purpose is to learn neural networks. But there is only problem, I couldn't find an easy book on this topic.

The state of these books accurately reflects usability of neural networks.

Thought you might look at http://www.generation5.org/ they like NN and other rather mathematical algorithms.
Quote:
I read a lot of articles but couldn't understand a single word because they just put too much math without explaining how to use it. Moreover, too much theory without implementation. I always learn by doing. So majority of books are totally useless.

You are doing it correctly. Allways try to implement these simple concepts, never just read about them. (or you might end as A.C. When he implemented more complex FSM, he did critical mistake nobody warned him about, because everyone expected people knows what to do with FSM, and what to avoid when not needed by experience. ~_^)

Quote:
Original post by d000hg
Quote:
Original post by Raghar
Majority of AI is about self education
No, it's a recognised field into which huge amounts of research have been done. There are many established algorithms and techniques which can be learnt/used.
And? When AI developer is doing his work he is self learning a new algorithms by research. No university could afford to teach a student 10/5 new algorithms, force him to invent implementations, and do other stuf to enforce creativity for about 11 years.

It's often easier to invent a new algorithm in few minutes, than look at Internet for a scientific work about that algorithm.

Quote:
Original post by ehsenmaster
I am doing NN for fun guys. And I 'll be more interested in hitting the gym or work on my martial arts training rather then playing a game (Its too feminine).
From when is the corpse eating feminine?

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement