PDA

View Full Version : Best way to store this data and connections to other data



blackwyvern
01-30-2002, 11:43 PM
To give you an idea of what I am talking about. I am working on the memory processes that humans have and trying to implement a model for computers. It seems kind of tough, but I know there is a way. Humans connect single words with ideas and expiriences they have known. Though computers don't have expiriences in the same sense, they do have learned knowledge, in a working system.

This is how a human may associate one word with many other words or ideas

Starting with the word fly:
When I think of Fly I directly think of bird, milkman and plane. Bird: I think of parrot, eddgs, and a peacock. milkman: a novel, freedom, and the hunt. plane: wright brothers, crashes and security.

so from Fly I can think of security or even a hunt without any thought because my memory has connections between the words. A computer can achieve the same I believe. It could search a brain(a storage area for its learned input) and locate the words that have been input into the program and from there the program can be drawn in several different directions because the computer has flagged associations with other words. I believe this is the true key to A.I. Does anybody have any ideas to add? or any way to implement such designs? This could be a very interesting chat to have here.

kooma
01-31-2002, 05:09 AM
Sounds very interesting... But to assosiate words as you said one must have everyday memories about social interaction, feelings, information, etc. (do remember that you also think a picture, perhaps a smell, feeling or a sound as you think of a concept) To implement all that to a computer is to get a Nobel!

Anyway, you might want to check gzigzag.sourceforge.net for an interesting kind of technique to create a dimensional (somewhere between 3 and 4d) structuces (closest one that I can think of as equivalent to brains).

mithrandir
01-31-2002, 07:31 AM
>>I am working on the memory processes that humans have and trying to implement a model for computers. <<

Are you talking about some kind of neural network or an actualy program?

shtarker
02-01-2002, 03:19 AM
The limitation that you face is that the human brain (as far as researchers can find out) has quite a few sepreate modules (for lack of a better word) all operating independently. Each of these run seperatly, processing a lot of data (again in lack of a better word) in some form we haven't yet cracked and at relativly high speeds.
Some thing tells me that people in hardware will get there first, not us programmers. . .

kooma
02-01-2002, 04:56 AM
> human brain (as far as researchers can find out) has quite a few sepreate modules (for lack of a better word) all operating independently

The image I have according to few psychology courses is that when your brains process something it's a quite holistic event...

shtarker
02-01-2002, 05:06 AM
Yes but each section of the brain is responsible for its own thing, so like one section is responsible for recognising faces, others control different bits of the nervous system. Of course they are all connected and working as a whole, but each bit still only performs its individual task and due to the cellular structure of the human body, it would still continue to function with out the rest of the brain. For a short time, estimated at somewhere between 30 seconds and 2 min 30 seconds.