-
another possible relational system, lets say that you numbered different types of relevance, 1. color, 2. synonymous, 3. alive, 4. organic, 5.edible , 6. species, 7. <photographic and visual records> (these are just examples)
so for instance, "apple" would contain entries such as this:
apple, (1)red, (2)<links to list of synonyms>, (3) <links to list of things that are alive>,(4)<links to list of things that are organic>, (5)<links to list of other things that are edible> (6) links to list of other plants> (7) <links to pictures and visual records>
so an inquiry "what are apples?" returns red, food, living, organic, edible,plant, <picture>
put through a program designed to construct coherent speech according to the rules of the english language, the application could concievably return "Apples are a red fruit, used as an organic and healthy source of food. Would you like to see a picture of an apple?"
-
Mmm... I would personally simply allow the output to display the top few links.
The front-end filter might be able to clump things, but you still want links. Otherwise, let the AI build its own clusters.
-
perhaps another way to increase the effectiveness of the weighting system could be built using human interactive evolution, for instance. the program give the user a word, such as apple, and a random list of words from its database, and tells the user to choose the word that is the most relevant to the given word (apple) or the option to choose "none" if none of the words are at all relevant. the process repeats over and over, soon incorporating some of the previously chosen words in with the random words from the DB, and each time a word is chosen as the most relevant word in the list, it is given another "weight" point. this way the machine quickly learns how relevant the words are to each other. A similar process could be done to determine HOW the words are relevant and which words are more relevant in certain ways than in others.
-
Lets face it. Artificial intelligence is not actual intelligence.
I added a degree of AI to the server here with VB as in it checks time performs tasks sends me e-mails and is able to formulate actual sentances from what I have "Taught it"
It would take years to map out the human brain. While you could have preprogrammed responses or even a learning curve AI will never be self aware. (See Terminator 1,2,3)
I also note that the average human brain holds more information than any mega terrabyte disk so storage is going to be a bugger (I think I have some old 100 mb hdds is anyone wants to daisy chain em)
-
It's a possibility, and maybe as a 'set up' tool to define a base structure for the AI... but typically you want the AI to teach itself through normal interaction. Also need to be carefrul from becoming too compartmentalised, otherwise you limit the database structure.
Basically, the aim is to generate a system that edits itself continually. The next step would be in allowing the system to adjust the database structure by itself.
-
Another thing, if you are going to teach a program to recognize visual information, we will have to give it more sophisticated visual equipment than the average web-cam. most humans can't even distingush what is going on by those crummy quality pictures. much less a machine that is just learning to interpret visual information for the first time.
-
mark, human intelligence is not intelligence either. We're all a collection of preprogrammed responses and we all have a learning curve. Self-awareness is an illusion in itself.
acid, once we develop a base patterning system, then a visual system will be easy. The trick there is a form of fuzzy logic comparing system, and there will be a specialist subsection of the database to compare and link patterns in the images to patterns from the database.
-
damn, that'll take some serious processing power
-
yeah, that's where it gets scary... that subsection would need to be an entirely different dedicated machine, most likely... or even a separate cluster.
It's actually amazing just how powerful the human mind is... mark's exaggeration aside, it's quite a scary hunk of processing power. Most of the trick is the number of threads it can handle at one time, but that aside, it's still amazingly efficient.
-
I wouldn't say were pre programmed responses as we can "edit" our response and change it to suit. Different volcabulary for different people is an example.
Choice is not an illusion it is a reality. The ability to decide is an ability that computer cannot possess without becoming self aware. IE Do I go out on Friday or not. I decide to go.
The final scene in matrix reloaded is a great example of how many choices we could make at each given point in time.
"mark's exaggeration aside"
So far my brain has recorded a 22 year high resolution mpeg file with 80,000 hz audio.
Find me a disk that can hold it and I'll buy it from you.
