Staredit Network

Anybody see the movie "Along Came Polly?"

Well the guy has a program that calculates the chances of stuff. Which one has a high risk etc.He has to input the stuff though. Like negatives and positive.
If you could make this automatically happen without a person could that be some form of A.I.?

I think that AI is possible... we just need a lot more decades. but i do not thing that AI will ever become more advanced then the human brain.

why do you people say "we need a little more time", or "in a couple of years"
if something can be done, it doesnt take as long as you think to do it, and instead of sitting here saying that it should be done, why not go out there and do it yourself, cause if you dont do it, and other people dont do it, who's gonna do it? ............

um... scientists? im not into programming stuff

Well, if you didnt scroll up for the link here it is

http://www.physorg.com/news7746.html

That is heuristics (or so called if-then in programs) used to create 'expert systems' that are specialised in certain areas. However, AI will go far beyond that. AI will have to have the ability to recognise, make connections, save patterns, analyse them and then learn from it. (make new if-then statements). Because something like simple starcraft triggers will eventually cause AI to confuse over things not stored in the database when it encounters it. They will also have to have the ability to calculate the probability of something, and choose certain if-then paths to follow if the probablity is high/low enough. Only after a process of elimination to a few possibilities of choices, does the AI start to go through and analyse the possible outcomes/consequences of each choice, so that it doesn't take forever checking all the possibilities (that's how checkers and chess AI works). Also, e.g. A future AI that will be sent to another planet, or automatic cars' AI will also have to have very fast processing speeds, which means only one BUS between a proccessor and memory will cause major delay for the AI to respond to its surroundings. Supposedly they have to analyse 250,000 pixels and make patterns based on colour, shading, distance (calculated from analysing the pinpointed location of the same object from two cameras), other factors to form an outline of objects in its surroudings, and finally go through eliminations, analysation, learning, and decision making, even a fast computer will take several seconds by sending information all through the same BUS. This means that the the probe would have missed something that could've been critical to its purpose, or a truck would've hit the automatic car. Therefore, something had been suggested - which is to have thousands have micro 'cells' that contain a independent memory and processor, with a BUS connected to each other. This means each cell can calculate one pixel and send it to a central major Processor, Memory, and Hard Drive for further analysation and information storage. This idea was inspired by the structure of our brain, with brain cells interconnected to each other. This neural network inspired AI will probably be the basis of future AI.

the new oblivion game uses somewhat AI. They tell them a schedule and its the NPC's job to figure out what to do. They can read books, go to sleep, steal, buy stuff, eat, play with the dog, practice archery, drink potions, and loads of other shiz that they aren't being told to do. They even talk to eachother.

Yeah, but they need to recalibrate the system.

Person 1 is given objective 'Rake Leaves'
Person 2 is given objective 'Sweep'
To 'Rake Leaves' Person 1 requires a rake, which Person 2 has
To 'Sweep', Person 2 requires a broom, which Person 1 has
Person 1 kills Person 2 and takes rake

They are way too violent.

Woman is given objective 'Sleep'
Cannot sleep because dog is barking
Sets dog on fire
Sleeps

Those were both actually things they did to test the AI.

Thats because it was all pre programmed to do those actions. Same sort of thing goes for the sims. Based upon input, they perform an output. But the question remains what would happen if they were givin a value they did not comprehend, they would not do anything.

The first artificial intelligence program was an email blocker that was developed about 3 (correct me if I'm wrong) years ago. It could use specific sayings and quotes on mail that contained viruses or people were tagging as spam, and could analyze new mail adhering to these guidelines.

I suppose what the biggest problem with a topic like this would be the definition of learning.
If you consider that most of our learning is based upon simply linking two things together. These things can be a sound with a visual picture, or a word with a meaning, among many others. If you could take away all visualizations of those things and make them into variables of an equation, it would look like a = b, b≠c, c=d. And by knowing those values, the computer would know, a≠c or d. (Remember that any one variable, if it were used in a real life situation, is not an object in itself, it would be peices of an object, like the sound, or touch, or shape, or color, ect. any sense you can think of)

If this is learning than a simple version of artificial intelligence has already been made, and has exsisted for a long time. Simply attach words and images instead of the variables and you have a program that can match up many things.

But this is not very redily considered the definition of learning. The above paragraph would more closely define knowing something, learning it would be taking a new value and comparing it to older ones to find what it is close to. You would need a certain number of methods for measuring how close it relates to other variables. But in the end it would simply be like running a few differing tests, and discovering what those tests have in common with other known outcomes from other things.

.. Greater efficiency and a much bigger capacity to hold information.

In binary there are 256 possible binary states per byte. Trinary has 6,561 states.

Only problem is binary uses electricity. Positive and negitive charges indicate a 0 or a 1. We would need a third state of electricity. Theories to get around this are neutral energy, or quantum mechanics. Quadratic code (65,536 states) may even be possible using quantum mechanics.

Only a set number of states? Can't you just indefinately increase the place value?