Leaning towards artificial intelligence?

[Gear300]

Lets say that a large number of computers, such as 100000000, were networked with as few barriers as possible. Will this system show signs leaning towards artificial intelligence?

 

[mgb_phys]

Not unless someone gets much better at software.

 

[Gear300]

oh...oh well...

 

[CylonMath]

We need new algorithms and theorems to make AI a reality. I want to quote from Modern Approach

A student in physics
might reasonably feel that all the good ideas have already been taken by Galileo, Newton,
Einstein, and the rest. AI, on the other hand, still has openings for several full-time Einsteins

 

[technutz]

It is not the number of computer that makes them "smart", it is the software that drives them. Here is a link to the open source project that links computers together like you want. Now all you need is the AI software to go with it. :P

http://www.open-mpi.org/

 

[TheMadman]

Artificial Intelligence has much critiques. Some seriously doubt whether it is even possible. I read somewhere that some big projects on AI has been closed and some 'Professors' have lost their hope in this field. But, it would be nice if we could simply do one thing. That is, if we can definitely state what do we mean by developing intelligence?
Human being is now able to fly? Is it true? Not wholly. We have been able to fly only after we realized that we have to leave mimicking birds. If we can do something on AI in the same way that might do something.
I got really frustrated when I took this course this trimester [with great anthusiasm]. We were taught a famous algorithm that's called 'Multilayer Feed Forward Neural Network'. That is kinda making 'Neurons' called 'Perceptrons'. Each neuron takes given input and then produce an output. This output then given to the next layer. After some some layer there is one output. Now this output is compared with the expected output and then the errors are calculated and the discarded with backward propagation. I personally didn't like it, since it didn't seem that simple. As I see All GREAT IDEAS are simple. TOO SIMPLE TO NOTICE.
Anyway, We still don't know how neurons work. I mean internally. I read somewhere that some day not the computer scientist rather the biomedical engineers will develop AI.
Though, I vehemently reject this idea :]

 

[michinobu]

Artificial Intelligence has much critiques. Some seriously doubt whether it is even possible. I read somewhere that some big projects on AI has been closed and some 'Professors' have lost their hope in this field. But, it would be nice if we could simply do one thing. That is, if we can definitely state what do we mean by developing intelligence?
Human being is now able to fly? Is it true? Not wholly. We have been able to fly only after we realized that we have to leave mimicking birds. If we can do something on AI in the same way that might do something.
I got really frustrated when I took this course this trimester [with great anthusiasm]. We were taught a famous algorithm that's called 'Multilayer Feed Forward Neural Network'. That is kinda making 'Neurons' called 'Perceptrons'. Each neuron takes given input and then produce an output. This output then given to the next layer. After some some layer there is one output. Now this output is compared with the expected output and then the errors are calculated and the discarded with backward propagation. I personally didn't like it, since it didn't seem that simple. As I see All GREAT IDEAS are simple. TOO SIMPLE TO NOTICE.
Anyway, We still don't know how neurons work. I mean internally.

It'd be nice if they could get MS Word to know the entire English language's grammar. However, it's not possible, as Noam Chomsky's work has shown. Yet, somehow people can understand the English grammar.

Some how, we're able to do specific tasks, which are computationally unsolvable. If there's no conceivable way to explain how to do something algorithmically, then I don't see how anyone could ever get machine to do it.

I don't know a great deal about the field of AI myself. I guess the idea is that if you can get data structures which mimic the behavior of brain cells, you might get software which can actually "think". But, somehow I doubt that will ever work. It seems that you'd have to make some major changes in mathematics field, and I guess philosophy (logic) too.

I read somewhere that some day not the computer scientist rather the biomedical engineers will develop AI.
Though, I vehemently reject this idea :]

Maybe if they started cloning brains! But, wouldn't that then be actual intelligence, rather than artificial intelligence?

 

[michinobu]

Not unless someone gets much better at software.

No, it's not an issue of software. It's more of an issue of mathematics. Again, as I said earlier, I haven't actually taken a course on AI or know a great deal about it. But, I from what I gather from computational theory, it seems that if its mathematically impossible for a computer to do something then no one will ever be able to even think of way in which one could write software to implement it.

But, maybe if someone can get data structures which mimic brain-cells well-enough to where a computer could actually think, then maybe it wouldn't be so impossible. Yet, I don't know any of the problems associated with this idea.

 

[tony134340]

You can network a thousand dogs, and they still wouldn't be Einstein. When you think about it, developing a true AI would be scary. It would have to have comparatively decent intelligence mixed with a human's curiosity and ambition to overcome obstacles, a want to learn. If we truly ever did get to the point where it's a possibility, we would have to be really careful because ambition can cause lots of hazards if something gets in the way of it.

 

[John Creighto]

I got really frustrated when I took this course this trimester [with great anthusiasm]. We were taught a famous algorithm that's called 'Multilayer Feed Forward Neural Network'. That is kinda making 'Neurons' called 'Perceptrons'. Each neuron takes given input and then produce an output. This output then given to the next layer. After some some layer there is one output. Now this output is compared with the expected output and then the errors are calculated and the discarded with backward propagation. I personally didn't like it, since it didn't seem that simple. As I see All GREAT IDEAS are simple. TOO

Don't expect to learn everything about AI in one trimester. There are many many more topics. For instance, genetic algorithms, fuzzy logic, pattern recognition, expert systems, predicate logical, functional languages, logical languages, symbolic algebra systems.

On project that I think is quite interesting is cyc:
http://en.wikipedia.org/wiki/Cyc

 

[TheMadman]

Thanks for the link. Can you please differentiate between functional languages and Structured Languages?

 

[TheMadman]

It'd be nice if they could get MS Word to know the entire English language's grammar. However, it's not possible, as Noam Chomsky's work has shown. Yet, somehow people can understand the English grammar.

Are you talking about Context Free Grammar?

You know when I read The Dragon Book for my compiler course it was really difficult to differentiate between our grammar and grammar for Computers. But, whenever you get the idea that here Grammar means for computers, all the cryptic definitions become so clear! Otherwise who would believe that you can define a Grammar say, English with Regular Expressions & CFGs?!

Maybe if they started cloning brains! But, wouldn't that then be actual intelligence, rather than artificial intelligence?

Even if they were able to clone a brain, I would be dubious about it's functionality. Because, it's not just the 'Brain' that works, it's the whole body.

 

[AUMathTutor]

AI already exists.

If you're talking about AI that mimics human behavior... that's stupid, and even if it never happens, it's no loss. That's as pointless as making a toaster that builds a fire and cooks bread on a stick.

 

[neslte]

No, it's not an issue of software. It's more of an issue of mathematics. Again, as I said earlier, I haven't actually taken a course on AI or know a great deal about it. But, I from what I gather from computational theory, it seems that if its mathematically impossible for a computer to do something then no one will ever be able to even think of way in which one could write software to implement it.

But, maybe if someone can get data structures which mimic brain-cells well-enough to where a computer could actually think, then maybe it wouldn't be so impossible. Yet, I don't know any of the problems associated with this idea.

Human brains also follow math. Everything that exists follows mathematical rules (physics). Unless you believe in souls and stuff...

Check this article out though:
http://news.bbc.co.uk/2/hi/science/nature/8012496.stm"

 

[AUMathTutor]

"Everything that exists follows mathematical rules (physics)."

As far as we know... it would seem...

 

[TheMadman]

Check this article out though:
http://news.bbc.co.uk/2/hi/science/nature/8012496.stm"

Thanks for the url and did you notice it(?):

"Wolfgang Wahlster of the German Research Center for Artificial Intelligence, and a chief German government scientific adviser on ICT, thinks that the reductionist strategy of the project is flawed - that it won't see the forest for the trees.
This is very interesting research and I'm not criticizing it, but it doesn't help us in computer science in having the intelligent behaviour of humans replicated."

 

[AUMathTutor]

That's reassuring. Apparently I'm not the only one who feels that way. See: post #13.

 

[Cyrus R.]

Yeah, AUMathTutor, you are on the right track.

A.I. has suffered a lot throughout the years because scientists have been trying to directly emulate human intelligence. The problem is that we don't know nearly enough about the brain, neurons, or psychology to program these principals into a computer.

However, once computer scientists started taking a logical and statistical reasoning approach to A.I., great advances have been made. This alternate approach still relies on psychological principles, but it focuses on areas which we already know how to program into a computer. Then, from this foundation, mathematical and algorithmic techniques are used to create a new, computer specific, notion of intelligence.

To understand this, think about how you read the sentence, "Time flies." You probably interpreted this to mean, "Time moves fast." If you look closely, however, you could also read it as "Time (with a stopwatch) small flying insects called flies." The reason you thought of the first meaning was because this phrase is more common. That is to say that it has a higher probability of occurring. This notion of probability allows computers to make educated decisions. We can tell it what time is and what flies are and now it can figure out what the sentence is supposed to mean.

Unfortunately, the ambiguity present in "Time flies" poses a real challenge for the future of AI. Humans can determine the meaning almost instantly, but computers have to go through both interpretations and determine which the speaker intended. For an even more complicated example, consider all the different meanings of "Time flies like an arrow."

As for the philosophy of intelligent machines, Alan Turing wrote a fantastic article about this subject in 1950. The article is written for a general audience and very understandable, but it's about thirty pages long, so read it if you want to. =)
http://www.loebner.net/Prizef/TuringArticle.html"