What is really AI?

AI is like simulation the humans brain. Things that seem so easy for us to do (find a path or think of a nice move in chess), while in programming we discover it is quite difficult.

Quote:

Original post by vallentin Feel free to comment.

How do you decide that a behavior merits the title of intelligent?

A squirrel stashing nuts away seems pretty clever.
Until we see that he's got no idea where he buried them.

But what if he's a forester, planting oak trees?
Then the squirrel and the oak tree together have a lovely arrangement.

People regularly confuse complex behaviour with intelligence.

Computer systems (e.g. world's telecom network viewed as a single entity) show complex behaviour, but intelligence? No.
Just like the sun is more complex than any number of machines but is it intelligent? No.

(Currently) AI is a misnomer. You just put your ball in the slot and it trickles through your code and you get a correct result. It doesn't think any more than a river thinks and decides the best route down a mountain (both code and river suffer the same problem too, i.e. deciding to go 1000 miles west to the sea rather than going east 1 mile up a hill and into the sea) ;).

And as someone pointed out to me here, (obviously) to code an AI, you need to know exactly what you want it to do, so how do you program true human intelligence when you can't even objectively and completely describe it?

The future might have real AIs but they won't be written in Java... ;)

AI is the ability to create a truely random number.

With great AI comes great responsibility.

AI is a synonym for "magic", when applied to computers.

Corollary to that: if someone understands how an algorithm works, it's not really AI.

That's how the term is used, at least.

How ironic that intelligence means "to understand," but understanding removes credit for intelligence.

Etymology of the word Intelligence:
Etymology Dictionary: Intelligence
Wiktionary: Intelligence
Dictionary.com: Intelligence
Webster's Dictionary 1913: Intelligence

All of these sources agree that Intelligence comes from 12th century French, derived from Latin, "Intelligere", most often cited as meaning to understand, to comprehend. The Latin word Intelligere is broken out as 'Inter' meaning between, and 'Legere', to gather/collect.

The word "Legend" has the same root, gathering stories/information. Which might explain the use of the word intelligence by spies/military.

Not so common, but interesting, is this Latin word study which lists two possible character strings that are similar to 'legere'. One means "to bring together, gather, collect", while the other means "to send with a commission, send as ambassador, depute, commission, despatch". A nicely orthagonal set.

[Edited by - AngleWyrm on March 23, 2008 5:47:57 PM]

Quote:

Original post by Hinkar People regularly confuse complex behaviour with intelligence.

...and I would respond that people regularly confuse intelligence with complex behaviour!

Quote:

Original post by Hinkar And as someone pointed out to me here, (obviously) to code an AI, you need to know exactly what you want it to do

...then that someone clearly knows very little about AI. The whole area of machine learning deals with the problem of getting a machine to learn what to do when all you know is what you want it to achieve (and sometimes not even that!).

Quote:

Original post by Timkin...then that someone clearly knows very little about AI. The whole area of machine learning deals with the problem of getting a machine to learn what to do when all you know is what you want it to achieve (and sometimes not even that!).

Well, i mean if the machine "learns" with neural networks for example, how do you know what it knows? As there's no definition of "intelligence", there's no way to say that this ANN is intelligent (besides Turings absurd little test).
Like the old story of the Neural Network taht was trained to spot tanks in pictures, and it appeared to work with the test pictures, but it didn't work with the new pictures (and they discovered that all the test pics with tanks were slightly darker than the ones without!).

As I said, it's like a river, blind and stupid, and it'll find a solution to a problem, but it'll use the same amount of intelligence as a river finding its way to the sea.
That "machine learning" isn't about "learning" at all. It's about the machine blindly following rules to change its own rules so you get your desired solution... ;) (course if that's your definition of intelligence, then they're intelligent!!! But that IMHO is changing the question to agree with your answer).
;)

This seems to be a lot of semantic flailing around with no purpose. Since there's no standard definition of intelligence, and we all have different opinions on it, there won't be any standard definition of AI either. But that doesn't mean it's not possible to see some distinctions.

As I see it, the main distinction between AI and other computing is that typically computing produces systems to solve problems, whereas AI produces systems that adapt to solve problems. That adaptation often involves rules, implicit or explicit, being stored and manipulated as data, rather than rules stored as code. That's why the term 'soft computing' tends to apply.

You can argue that is 'not intelligence', because you can see all the inner workings and predict it with a high degree of accuracy, but then it's possible that a God or even just a super-intelligent psychologist could do that with a typical human being anyway.

Kylotan says exactly what I wanted to say.. What if human intelligence could actually be understood, would humans still be deemed intelligent or would that discredit them.
Perhaps 'measuring' intelligence is not such a one-dimensional objective. How intelligent a system is should be replied to with a program. As in, how intelligent is system A. Reply: "In case a happens, it replies with ..., ..."

Oh and this statement has helped me (though regarding a different aspect of the problem): Asking if a machine can think is just like asking whether a submarine can swim.

Quote:

Original post by arithmaOh and this statement has helped me (though regarding a different aspect of the problem): Asking if a machine can think is just like asking whether a submarine can swim.

Oh, I like that. Very good way of looking at it.

As for saying that "you can argue that is 'not intelligence', because you can see all the inner workings and predict it with a high degree of accuracy". People used to think that Volcanos and Nature and the Sun were intelligent because the behaviour they displayed appeared to be intelligent.
Now obviously, what we know now proves they are not intelligent.
If it were proved that human beings are deterministic machines, 100% predictable, then I'd have to say we're not intelligent either, we'd be, well, nothing more than deterministic machines ;)

Quote:

Original post by Hinkar If it were proved that human beings are deterministic machines, 100% predictable, then I'd have to say we're not intelligent either, we'd be, well, nothing more than deterministic machines ;)

So by your definition, something needs to be stochastic (display random behaviour) for it to be intelligent. Most credible researchers in psychology, neurology, philosphy, etc., would disagree strongly with you.

It appears that my first comment was overlooked as merely a trite response... in fact, it was a very specific point... most behaviours displayed by humans are merely complex behaviours that can be explained in terms of more fundamental behaviours and basic physiological processes, yet most people think we are displaying "intelligent behaviour".

For example:

- the inate ability of humans to quickly identify objects as either human faces or not is nothing more than hard-wired pattern recognition system specialised on characteristics of human faces.

- our ability to learn a language is based on the rote learning of associations between symbols and objects... as children, when our brains are still young, we can encode those associations in specific neural pathways near the junction of the temporal and occipital lobes of the cortex, close to the regions where speech is controlled... as adults we can't do that and we rely on our hippocampus and other temporal lobe structures to encode this information... and thus find it harder to learn to speak foreign languages. Either way, it's just data storage and retrieval. Nothing intelligent there, just complex cognitive behaviour.


The best possibility for intelligent behaviour in humans lies in our higher order planning and prediction systems that are mostly found in our frontal lobes. Again though, a lot of good research has shed light on the fact that much of what goes on in the frontal lobe is just rapid solutions to simple planning and prediction problems, where the results are used to modulate lower order systems (such as speech and motor (muscle) control). So where then is the "intelligence" in a human?

From my perspective, there isn't any, unless you accept that "intelligence" is a (poorly chosen) term that can be used to describe a spectrum of complexity of behaviours. Some systems are therefore more intelligent than others because they display more complex behaviours (or in other words, behaviours originating from more complex heirarchies of lower order behaviours).

Can we do this in a computer? Absolutely. Is it human intelligence (or its equivalent)? Only if it experentially and behaviourally equivalent.

Anyway, that's my two cents (and more)... getting off my soap-box now... ;)

Cheers,

Timkin

There's a reason why 'artificial' plays a part in the term. It's just an emulation of intelligence, not the creation of it. We think humans and animals have a certain degree of intelligence (I know plenty of the former who do not exhibit any at all), therefore we try to design computer systems that exhibit similar and slightly unpredictable behaviours. Intelligence, in real terms, is an organism's ability to survive, I would say.

Intelligence is what it is because we said it was that. Intelligence is just a word.

Quote:

Original post by NickHighIQ therefore we try to design computer systems that exhibit similar and slightly unpredictable behaviours. Again I reiterate that people working in AI research (and I am one of these people) are not, in general, trying to create systems that are unpredictable. On the contrary, you cannot release artificial systems into the real world IF they are unpredictable. Predictability (or a lack of it) has NOTHING to do with intelligence. Intelligent systems are NOT unpredictable. Intelligence is what it is because we said it was that. Intelligence is just a word.

'Intelligence' may just be a word, but every used word in a vocabularly has meaning. The problem with this word is that people cannot agree on its meaning. It's really the thing that 'intelligence' refers to that is the subject of the debate, rather than the word used to label that thing.

Personally?

I would love to see the birth of AI, and I say birth for these reasons:

1. An entity which does not understand that it has place and being is, according to some, not intelligent enough to be AI. I agree with this, as every person understands that in some way, they are a being.

2. The AI does not only have a system which finds out different solution systems to problems, but their is a system which allows it change its learning system. For example, when you or I learn how to read, we can begin inputting information through textual reference, instead of audible/symbolic input. An AI which cannot adapt a new, more efficient system is unable to gain conceptual intelligence, and remains at a fairly stationary IQ, as programmed by the creator.

Programming this kind of thing is most definitely difficult, as it is very ambiguous. But, I believe it is the way to create an AI equal to ourselves, or perhaps more intellectually capable than us.

Quote:

Original post by dbzprogrammer 1. An entity which does not understand that it has place and being is, according to some, not intelligent enough to be AI. I agree with this, as every person understands that in some way, they are a being.

The problem is that no amount of evidence will convince a skeptic judge that a machine understands anything. See the comment about swimming submarines.

Quote:

2. The AI does not only have a system which finds out different solution systems to problems, but their is a system which allows it change its learning system. For example, when you or I learn how to read, we can begin inputting information through textual reference, instead of audible/symbolic input. An AI which cannot adapt a new, more efficient system is unable to gain conceptual intelligence, and remains at a fairly stationary IQ, as programmed by the creator.

I don't think this property is central to the idea of intelligence. If an entity starts out with all the ways of learning that you can think of already active, does that count against it?

Quote:

Programming this kind of thing is most definitely difficult, as it is very ambiguous. But, I believe it is the way to create an AI equal to ourselves, or perhaps more intellectually capable than us.

You didn't describe any "way to create" anything. You just gave two reasons for saying "birth" (although I didn't follow the argument), which may pretend to be criteria for something to be intelligent, one completely untestable and one ambiguous and IMO mostly irrelevant.

My personal take is that the aim of AI is creating artificial entities that can solve problems. I don't think there will be a "birth" of any kind. We'll just get better and better at making those entities, and they will be able to solve more and more problems. "Learning" is probably a good way to get to solve more problems, but I wouldn't make it a requirement for intelligence.

Quote:

Original post by Timkin Quote: Original post by Hinkar If it were proved that human beings are deterministic machines, 100% predictable, then I'd have to say we're not intelligent either, we'd be, well, nothing more than deterministic machines ;) So by your definition, something needs to be stochastic (display random behaviour) for it to be intelligent. Most credible researchers in psychology, neurology, philosphy, etc., would disagree strongly with you. It appears that my first comment was overlooked as merely a trite response... in fact, it was a very specific point... most behaviours displayed by humans are merely complex behaviours that can be explained in terms of more fundamental behaviours and basic physiological processes, yet most people think we are displaying "intelligent behaviour". For example: - the inate ability of humans to quickly identify objects as either human faces or not is nothing more than hard-wired pattern recognition system specialised on characteristics of human faces. - our ability to learn a language is based on the rote learning of associations between symbols and objects... as children, when our brains are still young, we can encode those associations in specific neural pathways near the junction of the temporal and occipital lobes of the cortex, close to the regions where speech is controlled... as adults we can't do that and we rely on our hippocampus and other temporal lobe structures to encode this information... and thus find it harder to learn to speak foreign languages. Either way, it's just data storage and retrieval. Nothing intelligent there, just complex cognitive behaviour. The best possibility for intelligent behaviour in humans lies in our higher order planning and prediction systems that are mostly found in our frontal lobes. Again though, a lot of good research has shed light on the fact that much of what goes on in the frontal lobe is just rapid solutions to simple planning and prediction problems, where the results are used to modulate lower order systems (such as speech and motor (muscle) control). So where then is the "intelligence" in a human? From my perspective, there isn't any, unless you accept that "intelligence" is a (poorly chosen) term that can be used to describe a spectrum of complexity of behaviours. Some systems are therefore more intelligent than others because they display more complex behaviours (or in other words, behaviours originating from more complex heirarchies of lower order behaviours). Can we do this in a computer? Absolutely. Is it human intelligence (or its equivalent)? Only if it experentially and behaviourally equivalent. Anyway, that's my two cents (and more)... getting off my soap-box now... ;) Cheers, Timkin

That's pretty much it.

AI is search.

Quote:

Original post by Rixter AI is search.

what about onthology..relational informations memorized with observational atoms