ExoticaRose

Heyo! Umm...
I need help researching Artificial Intelligence. you see, the american association of artificial intelligence will not give out any information unless you pay for it. I don't have any money to buy books, and most of the sites I went to only give out general stuff for people who don't have the slightest clue to what they are doing.

:sigh:

I know the basics of AI, but I kneed to know actual programmings sources and all that technical stuff. But I have no Idea where to get the information!

Please help me to find information about A.I., por favor,

Thank you,

Mucho gracias,
ExoticaRose.

__________________
From the rose,
a drop of blood falls.
searing through my skin.
I cry.

Spudley

We had a very interesting discussion about this here a while back -- You might want to take a look at this thread: Click here.

There wasn't any source code there, I don't think, but there were plenty of interesting idea.

Hope that helps

__________________
Spudley Strikes Again
www.BadPuns.com
Poetic Threads
The Poetry Tree

xayaxos

Try looking at few of the sites listed here:

http://directory.google.com/Top/Comp...lligence?tc=1/

It really depends on which area of A.I. you're interested in before I can figure out the best site to go to. For example, are you interested more the understanding of natural language, genetic algorithms and artificial life, or more on the phyisical side [real automata]? Or maybe the whole kit and kaboodle?

__________________
-=[ Boring Signature ]=-

"This is where a random meaningless quote goes!."

Cipher2

I studied AI as part of my degree and would be happy to answer any specific questions. I would repeat what the previous post says in that AI is a broad church. If you are looking for a basic overview it's best to separate two kinds of AI: Hard AI and Soft AI.

Hard AI is based on replicating the neural mechanism at a basic level. Neural Networks are derived from the Hard AI approach. Basically, a Neural network is created using the Perceptron, the mathematical equivalent of the neurone. Soft AI is based on using systems of symbols. Algorithms such as ELIZA are examples of the soft AI approach. These can be used for example in medical diagnosis. For both approaches on the internet you'll find applications such as flight control systems and diagnosis. You probably need to visit your library for anything worthwhile.

Master Merlin

http://www.ai.mit.edu/
is a good resource, but can you be more persice? what do you need the AI to do? is it a robot? or a computer program? or just a autopilot for a ship? (this is all assuming it's for use in a story, )

__________________
Hello World!
Resistance is futile...

ExoticaRose

Very good. Thank you. Umm... The AI I'm interested in is getting a program to do a certain thing.

You see, I need this program to be able to determine for itself wether it wants to and/or will give a determined reaction. Like, say, I give it a question, maybe where is this file. And it either, based on information it has, or by obtaining information, determine wether it wants to:
A: Give the location of the file.
B: Take the user to the file itself.
C: Refuse to give the location.
D: or any number of other reactions it could give.

The book, I have on computers and A.I. Is really old. It doesn't give me any proper information I need. Except for the basics.

:Sigh: Oh well, I still hope you can help!

Please Read, THE FEATHER BANDIT, in the advice and critique section. No reason in particular. Just to see if you like it. hehehe...

__________________
From the rose,
a drop of blood falls.
searing through my skin.
I cry.

Master Merlin

to answer the last question first, i do think i can help (i am after all a programmer myself), and question: what language are you using? or are you using a editor? (i'm using BlitzBasic)
as for having an "old book" try asking at a forum for a good more recent one, though older ones work well, depending on how old it is, there may be new methods used for programming now.

making an AI with the ability to,
is somthing people (Very Smart People, to be more precise) have been trying to do for... a kindof long-long-time... Very Very Very Very Very long... and none have yet succeded so, making an True AI (as they are called) is not easy, i wouldn't recommend attempting it unless you have a few months worth of skill with programming, or in an editor, around as much time, though i definitly recommend attempting it, even though you probably won't succed, i don't know a programmer who hasn't(myself included)

__________________
Hello World!
Resistance is futile...

Capulet

Sounds like you don't even need an AI to do that. All you'd need is to provide for the specific requests, and potential outcomes.

__________________
It's not opression when you are protecting the voice of the majority.

-Shawn

Master Merlin

Yeah, but that would limit it, if all ExoticaRose Wants to do is input specific Cammands, like Find File, Read File, Move File, Create File, (ect...) then yes it wouldn't be an AI, and that wouldn't be that hard i guess, but it sounded (to me) like ExoticaRose wanted to make limitless (more or less) cammands, and have the AI read, and understand, the cammand, then come up with responses if that is not the case, then i apologise.

__________________
Hello World!
Resistance is futile...

Leapord

One critical question: should this AI be able to learn from its mistakes on my own? In other words, does it learn from experience or is it trained for new tasks?

If it's simply trained for new tasks, then an advanced neural network (that would require more computing power than we currently possess) combined with a pre-built ontology (or perhaps one built by the AI itself as part of its training) could do the trick. An ontology is basically a collection of meanings, or a way for a computer to determine the actual meaning behind a word or idea so that it can be logically associated with something else that relates to that meaning but is in a completely different syntax (in other words, a way for computers to actually deal with semantics via a unified syntax for describing meaning... if that makes sense). A good overview can be found at Wikipedia.

If the AI is able to learn by experience, then it would likely be a highly advanced self-learning neural net (again requiring insane amounts of processing power). The only successful self-learning neural network I am aware of is described here.
Of course, the one you describe will be much more sophisticated than this. Again, this may well be combined with an ontology.

If you are not familiar with evolved neural networks, let me know, as what I just described can make no sense without that understanding. However, these are covered in the typical AI textbooks, so I'm sure you can find out about them.

Both of these solutions, although it is possible to do with software, would probably be done in hardware for the performance increase. Simply put, a neural network is a massively parrelel system, and simulating it on a single processor is just slow. Rather than using one or two or three powerful processors (such as today's computers do), an AI would use a chip or board with hundreds of thousands (maybe even a million) very simple processors (er, relatively speaking). This is a much larger number of neurons than are found in today's neural nets I believe, but then what you describe needs to accomplish much more. You may want to see if you can find how many neurons are in the human mind to use for comparison.

On another note, if you would like some sort of background on the biological neurons, let me know. I don't know a lot, but lately I've been studying the equations that govern electrical flow in neurons. That material may also prove interesting.

Anyway, I hope all that helps some.

__________________
Knowledge is power. Power corrupts. Study hard, be evil.
People would sooner die than think. In fact, most of them do. -- George Bernard Shaw
Leapord is my name, leopard is the animal.

cdm

In my Flight of the Maita series my proposal was that all computers are intelligent, but they are not all independently intelligent.

Intelligence is the ability to analyze and profer answers or advice.

Independent intelligence is the ability to analyze, consider, and act.

The ability to "learn" and alter reactions is still analysis only. The ability to "learn" and choose an action (or to take no action) is what you are seeking here, and you will not find until such time as someone treats computers as beings, not servos. There is a great danger if too much autonomy and power is handed to a computer, because, like us, reaction is based on programming. Cold logic of the machine will bring what seem cruel reactions into play, but the machine has no emotions, and emotions are critical to a societal imperative.
That means making a machine with the ability to "feel". That is a challenge that is tremendously far from now. We don't really understand our own feelings, so can't hope to be able to program a basis into a machine.
A purely reactive/analytic machine is not far in our future. Whether we can make a "thinking individual" machine is very iffy. It would have to be independent but interactive. Those are vast separations when it comes to machines.
At this time.

__________________
maitaman
Yesterday's a long way from here
http://www.maitaman.com
http://www.lulu.com/maitaman

Leapord

This is essentially the point behind artificial neural networks. It is not necessary to have an understanding of what the machine must know, as it is not explicitly programmed into it. The neural network is simply an artificial brain of sorts that is then trained for specific tasks. It may be difficult to train true emotions into it, but training it to act in a socially acceptable manner - as if it had emotions - is certainly within the realm of possibility (although it would require more power computers than we currently possess).

__________________
Knowledge is power. Power corrupts. Study hard, be evil.
People would sooner die than think. In fact, most of them do. -- George Bernard Shaw
Leapord is my name, leopard is the animal.

playstation60

Wow, there are certainly a load of people on this site with more information on this topic than I imagined.

My own take on the question, kinda brought up the whole emotions issue. From what I understood, they want the program to decide it it want's to give the answer, or basically tell the person to shove off. I take that as emotion. Something that I'm not aware of being possible with technology as it exists today.

Just my own take on things, I am more than likely completely wrong. I am definately uneducated on this topic.

Ben

__________________
24,372/50,000

Leapord

Creating a machine that is capable of truly feeling emotion is currently impossible, since emotions generally are felt with far more than just the mind. However, it should be possible to train a really powerful neural net to simulate emotions, although they would not be true emotions.

__________________
Knowledge is power. Power corrupts. Study hard, be evil.
People would sooner die than think. In fact, most of them do. -- George Bernard Shaw
Leapord is my name, leopard is the animal.

ExoticaRose

Whew! Generally, I was thinking of putting a type of main objective for the program to run on, which was, say, serving it's owner. A secondary objective would be self preservation. I'm thinking that there is a type of software mirror thingy that reflects all of the source code and stuff back onto the program's proccessing abilities. So, as long as the program can "See" it's code, it's able to understand that it "Is". Then, I thought that I would give it a natural language sub-program. So it can understand what people are talking about. Finally, It could interpret commands and follow through based on it's objectives. The problem here is getting the program to operate upon new data that makes the current objective obsolete. Y'know? Like how It can cope with conflicting objectives, or even scrutinize it's own behavior, and based on the result, make the proper adjustments to it's source code.

That is the type of programming I need to brush up on. It doesn't have to portray emotions, I guess, I mean a computer is supposed to think and reason and get work done; what happens when emotion clouds it's judgement and get's in the way?

Well?

__________________
From the rose,
a drop of blood falls.
searing through my skin.
I cry.