What is intelligence?

[Amorphos]

Intelligence.
What is intelligence? Most IQ tests are mathematically based, yet intelligence can be of many forms & a calculator is far better at maths than any human!
A computer linked to a massive library may ‘know’ mare than any human, yet still it is relatively dumb!

[LegionnaireX]

Intelligence is the ability to have a strong grasp of a certain aspect of thought and life. For example you can have academic intelligence, athletic intelligence, musical intelligence, etc.

[Prince Laridus Konivaich]

Intelligence is being able to make connections between things - so that even if you don't know something, or have never seen/heard it before, you can have a very good idea of what it is. A computer, or other 'simulated intelligence' can't make connections! You have to tell it exactly what everything is, and how they will receive it.

[Nihil]

Prince Laridus Konivaich is on the right track here, and has identified a key element of intelligence. Intelligence grows organically, it is a living thing. As primates our opposable thumbs enabled the use of tools which in turn enabled us to grow neural networks in our brains that are more advanced than those of other lifeforms. Therefore our intelligence is something that evolves and grows. It begins as intuition. We sense a connection between things that at first glance seem unrelated, then ideas are refined until they become knowledge. A lot of great scientists got their initial inspirations from dreams - intuitions form in the deeper strata of the mind and gradually take the form of knowledge. In this way, the intelligence of our species broadens and deepens, and our brains evolve.

Therefore, I would say that, in answer to your question "what is intelligence", it is subject to the criteria that we apply when asking "what is life", in particular, it must be adaptible. Some of the biologists here can tell me whether we have yet reached a final definition of life - I think it may still be somewhat loosley defined. But whatever about that, mere data is not intelligence as it is not alive.

However, I believe true artificial intelligence may exist in the future. This may be through the invention of organic computers, or possibly even through the development of a thinking machine that can grow its own data through experience, just like we did with our opposable thumbs, and adapt its behaviour in the light of this new data.


EDIT - woah, Attila, what happened your avatar!?

[Amorphos]

Nihlil: nice answer! I like the organic computers idea, asks questions like ‘are we organic machines’? but would they have the nature of mind [a questionable I know]?
avatar? i am using my laptop - desktop is very ill! :angry

[Erik]

Originally posted by Prince Laridus Konivaich@Apr 30 2005, 09:15 AM
A computer, or other 'simulated intelligence' can't make connections! You have to tell it exactly what everything is, and how they will receive it.

Yes they can.

I have studied Artificial Intelligence at University so I have learned a lot about what intelligence is and what the capabilities of computers are.

There is no real solid defenition for intelligence (yet).
Many different concepts, processes and behaviours are being called "intelligent", but things normally needs to be capable of a vast combination of different "intelligent" concepts, processes and behaviour to be called truely intelligent.

If, for example, I show you a computer that can make connections in exactly the way you thought about as being intelligence you wil likely say: "yeah, your computer can do that, but it can't (..new defenition of intelligence...) so it's not realy intelligent".
Then I wil go back and work on a computyer that also fills your new defenition of intelligence but if I show it you wil come up with yet another defenition of intelligence etc. etc.

All we (AI programmers) know is that humans are intelligent and things like rocks aren't.
So the only way to create "real intelligence" for sure is by turning something that has no intelligence (like rocks) into something that does exactly the same things humans can. (this is also known as the Turing test).

Lucky for us rocks can be turned into computers, and computers are realy good as simulating things.
And a computer with enough RAM can simulate the entire human brain up into the sub-atomic level of detail.
If you couple this simulated brain to a simulated body that includes sensors and motors (robot) you have made real intelligence.

[Amorphos]

Erik: nice answer but… I don’t think a simulated brain is the same as our minds? Who or what is the ‘experiencer’ & is this aspect of the mind crucial to a kind of intelligence that finds answers ‘out of the blue’ - a more human intelligence?

[Nihil]

Originally posted by attila of nazareth@Apr 30 2005, 11:11 AM
Erik: nice answer but… I don’t think a simulated brain is the same as our minds? Who or what is the ‘experiencer’ & is this aspect of the mind crucial to a kind of intelligence that finds answers ‘out of the blue’ - a more human intelligence?

I can't answer for Erik, but I can express my thoughts on this.

Must there be a ghost in the machine for there to be intelligence? Possibly the aspects of our humanity that Attila believes make us intelligent are something that can grow out of the kind of intelligence Erik is talking about. I'm very intrigued by this concept of a complete model of human intelliegnce - has this actually been done, Erik, or is it just a theory? Do we really have a complete enough understanding of the brain to do this?

I think what you're saying, Attila, is that there must be a "spirit" that is self-aware, the "experiencer" that you talk about, for there to be conscious intelligence. This is why I find the idea of AI so fascinating - by creating self-awareness, can you create a conscious living entity?

In considering this, bear in mind that they say that as new-born babies we are not self-aware. This apparently grows out of our sensory experience and is a process of adaptation to our environment. We develop our sense of a distinction between the self and the not-self through experience. Therefore, is it possible for an artificial intelligence that is capable of learning in the same way as a baby does to develop conscious self awareness?

[Amorphos]

Nihil: I agree in the main. I try to get away from terms like ‘spirit’. I just try to comprehend things on a universal level, thus if there is universal mind/being then all things have these qualities in some way. In its primary form it is undefined, a ‘blank sheet’ if you will. As to creating ‘mindful’ intelligence, well I would say that a computer – no matter how intelligent – is of the ‘primary mindful state’ nature. But if you created artificial intelligence biologically, then you would surely have an intelligence the same as ours imho!

[Erik]

Originally posted by attila of nazareth@Apr 30 2005, 12:11 PM
Erik: nice answer but… I don’t think a simulated brain is the same as our minds? Who or what is the ‘experiencer’ & is this aspect of the mind crucial to a kind of intelligence that finds answers ‘out of the blue’ - a more human intelligence?

Well, this is one of the strongest arguments against AI still alive today.

If you beleive humans are nothing more than physical beings that act purely according to the laws of physics, and I beleive they are, then there isn't much you can say agains my argument.
In this case humans are nothing more than machines, extremely complicated and sophisticated machines, but machines nonetheless.
And therefore it must be possible, in theory at least, to reproduce the entire "human" machine or build a different kind of machine that can do exactly the same tasks.

But if you beleive human intelligence is more than the product of physics then you have a strong argument.
The only problem with this argument is that nobody has ever come up with a single idea about how this "supernatural" intelligence can work, while there are a number of thories about how purely physical intelligence can work.
So it's an argument that is used to disprove other peoples arguments but not to present any kind of alternative.

I'm very intrigued by this concept of a complete model of human intelliegnce - has this actually been done, Erik, or is it just a theory? Do we really have a complete enough understanding of the brain to do this?

It obviously has never been done yet because you would have heared about it for sure.
Only small parts of the human brain have been simulated, present day computers aren't powerfull enough to simulate an entire brain with any usefull speeds.
The workings of a brain isn't realy that complicated on a small level.
It's mainly the size of the thing that makes it able to perform such complicated tasks.
I think we currently know most things about the workings of the human brain that we need to know.
The biggest mysteries have to do with the way the brain interacts with the body.

I think humans generally overestimate how special they are.
They like to beleive they are at the centre of the universe, superior and totally different to all animals, and created by a higher being.
If you tell them we live in a suburb of our galaxy, share 99% of our genes with chimps, and are at the mercy of simple laws of physics you normaly get a lot of resistance.
So I don't expect to convince many people that my colleagues and I are capable of creating real artificial intelligence.... we wil just have to prove it to them by building the damn thing! :happy

[Amorphos]

Is it not possible to have a nature of the universe that is simply ‘mind’? Even if this has no physicality in the normal sense, that does not mean it is any less ‘real’. I think that with the chemical robot approach, there is no answer to the notion of the 'observer' or the 'seer'/self. We experience – that experienced can be mimicked in full detail – yet that is a ‘second hand world’ with no essence or being, if you know what I mean.

[Erik]

Originally posted by attila of nazareth@Apr 30 2005, 03:49 PM
Is it not possible to have a nature of the universe that is simply ‘mind’? Even if this has no physicality in the normal sense, that does not mean it is any less ‘real’. I think that with the chemical robot approach, there is no answer to the notion of the 'observer' or the 'seer'/self. We experience – that experienced can be mimicked in full detail – yet that is a ‘second hand world’ with no essence or being, if you know what I mean.

Yes, thats why I beleive any true artificial intelligence machine should be able to sense the world and interact with it.
A simple digital videocamera and a robotic arm would to the trick.

The machine should also have natural urges/instincts, for example to find food, to want to stay alive and maybe even to find a mate LOL. Without these urges it would just stand still and have no need to develop itself.

[GodEmperor Nicholas]

yet intelligence can be of many forms & a calculator is far better at maths than any human!

Find me a computer that can find an exact solution to a 2nd order Sturm-Liovile partial differentian equation. (Hint it isnt possible)

I did it last week. (hint: the solution is an ellipic function) Computers arent as good at math as humans.

NM

[Nihil]

Originally posted by attila of nazareth@Apr 30 2005, 02:49 PM
Is it not possible to have a nature of the universe that is simply ‘mind’? Even if this has no physicality in the normal sense, that does not mean it is any less ‘real’. I think that with the chemical robot approach, there is no answer to the notion of the 'observer' or the 'seer'/self. We experience – that experienced can be mimicked in full detail – yet that is a ‘second hand world’ with no essence or being, if you know what I mean.

Are you conceiving of this "mind" that has no physcality as a kind of platonic "form" that exists because we imagine it, or are you saying that something can exist despite not existing physically yet without being something that is only imaginary?

One of the reasons I am so intrigued by the idea of consciousness developing out of perception and experience is that if we can build an artificial intelligence it will presumably help us to answer these questions that you are asking and maybe solve the old mind / body duality once and for all. If we can build a machine that has all the tools to become self-aware and declare "I think therfore I am", what will that tell us?

Are humans the only animals that think this way? It's been shown that chimps are intelligent enough to have a sense of humour. Maybe chimps also ask themselves metaphysical questions about the universal mind...?

[Amorphos]

Eric: a machine or AI would ‘act’ like it has sensed something, yet there is no ‘it’. A digital camera linked to a computer cannot see, it is all a set of ‘mechanical’ actions of information interaction. This is the difference between, the processing of info & the thinking of info!

Nihil: yep, it is real without being either imaginary or physical! I don’t believe it is possible to build a machine that is ‘self aware’ in the context that ‘it’ has no self. But as for solving problems – I am sure we will get our ‘deep thought’ computer in the end! It will probably tell us that we don’t exist!

[Justinian]

Since artificial intelligence needs to be designed by humans, it will never become even more complicated then the human brain, because scientists will need to be able to understand how AI works (and unless I'm twenty years in the past, no one is 100% absolutely sure how every bit and piece of the human brain works).

This would mean that, even if computers can think faster then humans, it wouldn't necessarily think better. How could you design something smarter then you are? You need to have the brains to make it in the first place.

Does this whole post make any sense? :huh Sorry, I'm pretty tired.

[Nihil]

Originally posted by Justinian@Apr 30 2005, 03:50 PM
Since artificial intelligence needs to be designed by humans, it will never become even more complicated then the human brain, because scientists will need to be able to understand how AI works (and unless I'm twenty years in the past, no one is 100% absolutely sure how every bit and piece of the human brain works).

This would mean that, even if computers can think faster then humans, it wouldn't necessarily think better. How could you design something smarter then you are? You need to have the brains to make it in the first place.

Does this whole post make any sense? :huh Sorry, I'm pretty tired.

Yeah, makes sense to me Justinian. There's somebody around here at TWC who has something about that in his sig "we can't understand our brains because if we could understand our brains we'd have to be more intelligent than we acutally are", phrased better than that of course, but it's a good point. To grasp something we have to be able to fit it inside our brains, and we can't do that UNLESS - consciousness is a hologram and contains a microcosm of itself within itself.

Maybe we are just intelligent enough to build an artificial intelligence that can figure our brains out for us.

Nihil: yep, it is real without being either imaginary or physical! I don’t believe it is possible to build a machine that is ‘self aware’ in the context that ‘it’ has no self. But as for solving problems – I am sure we will get our ‘deep thought’ computer in the end! It will probably tell us that we don’t exist!

So this thing that you don't want to call a soul, which isn't imaginary and isn't physical - how can it be said to exist? Is it a symbol, an idea? Or what?

[smack]

There is no reason why one cannot program a learning system with an Ego. A system that would take every damn thing personally. Even the earliest attempts to pass the Turing Test mimicked this quite well (Alice comes to mind). But you are right Atilla, there is a difference between mimicking self and having self. Still, there is no reason why this cannot be given to a learning machine if you accept Erik's point: Either you accept that we are biological machines, and in being machines, are constructable, OR you must accept that claiming otherwise leads to currently unprovable notions such as soul.

The main reason that AI is not programmed with Ego is that most AI research has been focused on the utility of learning, rather than mimicing human learning. Programming simple learning is tough enough without bogging it down with the inefficiencies of self. That history of AI development is of course changing. 'Kismet' at MIT is a popular one that demonstrates learning and response to human facial expressions. But all of these things are one or two vector explorations into a hugely complex system of actual human learning. So far nobody has been audacious enough to suggest they can start putting all these pieces of Humpty Dumpty together again. But it'll happen eventually.

It's also quite interesting how some simple behaviors will convince a viewer that some machine is 'alive'. Flocking behavior in birds is very simple to reproduce with 5 or so rules of behavior. Yet if you show that graphically it looks 'alive' http://www.aridolan.com/ad/Alife.html

It's more difficult to convince someone of more than animal intelligence, but it's definately been attempted and succeeded to some degree. And when it comes down to it, we call things intelligent as much from impression as from some list of features. Like the definition of life, intelligence is somewhat difficult to nail down with absolutes.

[Vercingetorix]

Originally posted by Nihil@Apr 30 2005, 03:00 PM

Yeah, makes sense to me Justinian. There's somebody around here at TWC who has something about that in his sig "we can't understand our brains because if we could understand our brains we'd have to be more intelligent than we acutally are", phrased better than that of course, but it's a good point. To grasp something we have to be able to fit it inside our brains, and we can't do that UNLESS - consciousness is a hologram and contains a microcosm of itself within itself.

Maybe we are just intelligent enough to build an artificial intelligence that can figure our brains out for us.



So this thing that you don't want to call a soul, which isn't imaginary and isn't physical - how can it be said to exist? Is it a symbol, an idea? Or what?

This is an interesting point. However at the same time we have created things that no one human has the capacity to completely understand. Look at the design of things like computer chips or very complex software, no one person has complete knowledge of everything in a project. It's simply not possible; you need a whole team of people each working on their own part of the project to make it all come together. One human brain may not be able to understand the human brain fully but many brains and many computers may. Just a thought.

[smack]

Originally posted by Justinian@Apr 30 2005, 03:50 PM
Since artificial intelligence needs to be designed by humans, it will never become even more complicated then the human brain, because scientists will need to be able to understand how AI works (and unless I'm twenty years in the past, no one is 100% absolutely sure how every bit and piece of the human brain works).

Frighteningly perhaps, this is false.

I am working on artificial life projects in which intelligence designs itself in an evolutionary framework at high speed, for instance. There is no need for us to understand everything in an intelligent system in order to design it, considering that learning is part of intelligence. Yes, it's a little like Skynet to worry about, but that is absolutely possible. If you unleash a learning and evolving system it will surpass you if it has the proper resources. Its just damn difficult to get over the baby steps and set such an intelligence off out of the nest. Hasn't been done yet, but it will be done.

Like you said, we still don't understand the brain, nor the 'mind', which is a large part of why AI has moved away from trying to model that and just got down to the basics: Learn, Infer, Remember, Ask, etc..

-S