Can simple artificial intelligence be created with Grid Computing?
GeneralThis forum is for non-technical general discussion which can include both Linux and non-Linux topics. Have fun!
Notices
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
If you have any problems with the registration process or your account login, please contact us. If you need to reset your password, click here.
Having a problem logging in? Please visit this page to clear all LQ-related cookies.
Get a virtual cloud desktop with the Linux distro that you want in less than five minutes with Shells! With over 10 pre-installed distros to choose from, the worry-free installation life is here! Whether you are a digital nomad or just looking for flexibility, Shells can put your Linux machine on the device that you want to use.
Exclusive for LQ members, get up to 45% off per month. Click here for more info.
Can simple artificial intelligence be created with Grid Computing?
My friend who introduced me to linux is the professor of computer science at CIT, he shared with me a grant proposal that he wrote that would help a company (like federal express) calculate the shortest route between two points with a continually changing matrix of points.
The thought was this: Utilizing grid computing could you have a series of processors all containing a simple 3 dimensional bionary matrix of randomly generated numbers (or user/host cpu generated) like; 1,0,1 0,1,0 1,1,1 and so on. Each computer having a different number assigned by it's user and by using a cluster of users, having a new user joining the grid select a prime identifying number (or randomly generated number then utilizing the program to trace through the numbers as assigned by the cpu's the shortest point using a sting of like numbers, ie; 0,0,0 or 1,1,1.
Would that constitute a ("Decission") being made by the Grid?
If each o or 1 represented the answer to an argument ie:
Do you want an apple (1) or an orange (0)
Do you prefer black (1) or white (0)
Your spouce cheats but is sorry, you (1) leave, (0) forgive
Then we could have our matrix of answers, have the computer check the shortest points between the answers then answer the following question.
Your spouce is wearing black (1) and eating an apple(0) and informs you they have strayed.
Computer, what do you do?
CPU surveys it's grid of input and determines shortest points for prefering black x,1,x and weighs the percentage of leave forgives, then references the shortest point for eating apples 1,x,x weighs the percentage of leavers to stayers and gives the answer.
If my spouce wears black and eats an apple and informs me they've strayed then I:
0 Forgive.
By using a matrix of answers to questions and inbeding them in files you then enable to have the computer create it's own collective judgement.
Artificial Initelligence.
You would have to create the grid in a pyrimid of matrix comprised of odd numbers so the computer would not receive a 50/50 response in whcih case it would'nt know.
does anyone have a thought of how using shared files created by individual users on a small network that this could be created if I can get my friend at CIT to donate the use of their algorithm for the shortest point calculation?
Is there a more appropriate scientific forum of programmers in which I should post this problem?
I'm no authority in this field, but I have done some reading. I think you are right that the concept of grid computing itself could be classified as artificial intelligence... Perhaps one of the first forms of artificial intelligence ever implemented (packet routing).
However, I think that this approach is not very well equipped to handle the type of decision making that you're talking about. It seems to be fairly inaccurate, even you're example was quite illogical I would say...
The problem I see with this approach is that the computer (or computers, or whatever) would have no way to check to see if a rule/entry was correct, logical, contradictory to another rule elsewhere in the grid, etc. Also, because of the fact that you're just finding the shortest distance b/w two points, the computer is not actually doing any reasoning, it's comparing sets of 0's and 1's, not really understanding what they represent. (Chinese room)
Essentially you would be creating a very complex chatterbot... I don't think that a shortest point calculation or a regular expression matching program would constitute ai...
It is a very interesting idea though... I had never even thought about ai in that way before. Maybe it would work
I have an ai theory of my own ( i think so at least ) .. it seems a bit more, uh, conventional (boring) but I think it could work... http://www.linuxaddict.cjb.net/thinking.txt is where you can find it, I need to redo it sometime. That was the first version where I wrote it down so I wouldn't forget it I'll rewrite it someday though ... and update it with the new Ideas i've had since then
I'll check out the link. Don't know what prolog is but I'll look it up.
Yes I realize my example is a bit of a ("Majic 8 Ball") but human reason is partially based upon opinion and not entirely on logic. The idea being we could create a simple survey of let's say a matrix of 18 questions 6 groups of three, for matters of data storage referenced like a pyrimid.
These survey answers could be stored as a simple shared file.
We would only have to set up an array of users who answer the survey and store their answers in a shared file.
We connect the users through the use of a triangular structured array, that links each like answer in one user file to the corresponding answer in another user file.
The shortest point algorithm would just manage finding the answers of similar minded respondents. Each Users answers become part of the computers collective unconcious.
You can then enter your array or partial array and ask the computer how it would answer a missing question.
They system then queries the userfiles for x, by aligning all similar respondents on X and Y and giving the majority reponse for X.
Primitive to be sure. Majic 8 Ball, absolutely, but it could be a very easy project to build.
I'll look into the reference material provided and I hope the three of us could have some continued conversation on the topic.
PS, using the right structure for the array, you would be unlimited in the amount of contributors that could link their computer to it. Users could also contribute questions to other users on the array and the array's knowlege complex would increase with every contributing user.
Sure, it wouldn't be Hal.......
but it is possible.
The kernal for Hal?
I heard a little operating system started out the same way.
If you'd like to see some of my computer work please check out my comic movie at gurfrip.com
I think at the very least, a simple and organically growing AI Kernal for Linux files would make a hell of a game.
I'd like to chat with you about it some more, let's keep the dialog open, maybe we can get a team together.
Best regards,
gurfrip
Answer File System 1:
0
1 0
1 1 1
Answer File System 1,2,3 array:
0 0 1 0
1 1 1 0
1 1 0 0
0 0 1
1 1
0
Now consider the array in 3 dimensions as questions and users are added, by using the most simple three dimensional structure the knowlege of questions can grow as rapidly as users/experiment participants can add layers of questions, let's call them ("Knowlege Increments") to the system. As a User joins the grid, they can add questions, those questions are distributed throughout the grid and the users then add their answers to their individual shared file. Thus helping the grid develop a shared opinion through searching the response of participants
Mount points being the same question. This is where the shortest point algorithm comes in.
Have to manage the processing of the information somehow, and shortest point seems to be the simplest way.
Each user sharefile would be like a neuron or an electrical component, you send it a pulse and measure what comes out the other side the pulse being the question and the asnwer being the shortest point through compared participant responses.
I would not classify that as AI. Heck i would not classify anything a computer does as AI since it is just carrying out instructions that where given to it. It's a simple fact no matter how complex the problem, your still doing the programmers bidding. I mean if you think making decisions and being correct is intelligence then your computer is very intelligent it can decide if one register is equal to the other quite easily...
Anyways about the algorithm it would not happen to have any relation with dijkstra's?
as far as the name associated with the algorithm, it very well may be the one you've referenced. My friend who has used it in shipping algorithms is Laurent Michell who is a professor of Computer Science at CIT. He's going to make it available to me and says it is not his propritary invention, so it very well may be who you mentioned, I'll inquire. I'll do a search on the name you mentioned.
What I am thinking about is nothing new, If you do a google search under ("bayesian network") they are using this type of computing for various applications. I guess a better description would not be ("Artificial Intelligence") but a human response simulation process.
The thought being that if we develop or rather manipulate existing processes to accept multiple user input through creation of questions and corresponding answers that ultimately we could develp a simple machine that simulates simple human responses to simple questions.
The key is to keep it simple enough to where an average computer user can develop questions with corresponding answers and immediately implement them for response by other users of the grid.
The more users that are inputing questions and answers into the basic system, the greater the data population that the system can draw on for it's answers.
A great big toy that eveyone could input to from simple code and network.
Do you know of any resources that we could obtain existing code?
Check out the bayesian network page, there have been many scientific and industry applications for such an application and quite possibly there is one that could be modified easily to suit the purpose to just test and see if it is possible.
Academician Prof. B. Sendov (Bulgarian Academy of Science) spoke on "Information and Knowledge"; his main point was that as all intelligence is based on knowledge, artificial intelligence had to be based on artificial knowledge.
(this is what I'm speaking of in general terms, though dijkstra was less than complimentary on the topic)
great reading though. If guys like dijkstra were charged with the advancement of all science.....The world would still be flat. I found most of his letters contemptuous of ideas of his colleagues, but I've only spent 15 minutes reading him and what he had for lunch in innsbruck.
I'm not talking about building some supper machine that can tell you the circumference of sub atomic orbitals and their relationship to pigs farting in france.
I'm talking about a TOY that simulates a human character in simple responses input and created by a bunch of other humans who want to play with the toy.
Not looking to get to complicated here, I leave that to the guys at MIT.
LinuxQuestions.org is looking for people interested in writing
Editorials, Articles, Reviews, and more. If you'd like to contribute
content, let us know.