The future of artificial intelligence..

What do you people perceive to be the future of artificial intelligence? Personally, I believe that there will not become a stage where we can develop a computer to be 'conscious'. Virtually all of it's responses will be done to set parameters... I think the ultimate judge of AI would be how it reacts to an unknown situation.. I dont see this as being particularly possible or practical...
I would love to have AI exist, but I dont think that there is the room for a computer to develop it's own programming to completely unknown situations, without using at the least pre-set parameters...
But then, some would say the same applies to us humans.

I dont see how a machine could react to an unknown situation being how it only knows what you program it to know. Maybe it would just shutdown.

This is my main issue with AI... I think that if we could build a computer capable of reacting to an unknown situation in a suitable manner it would be 'conscious'... But the issue is, is this possible?

If a machine became "self-aware" then it would not need to be programmed to react to an unknown situation, it will make it's own decision based on the given situation just like humans.

I think the reason so many movies e.t.c. show machines always destroying us is because if they indeed did become self aware, they would in fact have a better sense of what is right and wrong. They would indeed see that although we humans do infact preach about right and wrong, most of us don't live by those rules. However, a machine would have no problems with these ethics and morals.
Once they saw what liars and hypocrits we are, why wouldn't they decide to put themselves in charge of us or take us out? Could you blame them?

Sorry, I'm rambling. lol

Well it depends on what you mean by an unknown situation. Nueral nets can do character recongication on letters that they haven't been exposed to, and quite accurately. Does that count?

You would have to program the A.I. to be able to react to different situations, you leave out a certain situation or something unexpected happens and the A.I. is useless.

So I guess it would be impossible for someone to create A.I. that would rival a humans natural thinking abilities. Unless said human is a retard.

I'm more concerned with things like SHODAN or Deus. We might first develop AI as simple task-performing machine, but as soon as something triggers its self-awareness, then what? We will be ill prepared. Remember Citadel? All it takes is a little nudge, and we suddenly have an AI who realizes it's suddenly no longer our slave. If the Machine Mother has taught me anything, it's that the instant we give AIs any sort of power, we're screwed. Sefl awareness means that naturally one develops a sense of self-preservation and a need for self-improvement and expansion.

Remember, humans make excellent drones....

DE

It is not impossible to make an AI. You just need to look the problem very differently than one of reacting to situations. Look at it as something more natural, like a stream of bits and incorporating those bits into a dynamic structure.

The human nervous system is, in fact, binary. Either a nerve is firing or it is not. Therefore, it IS possible to program an artificial intelligence. I'mg going to be writing a paper on this subject in philosophy, so here's my point of view:

1. Machines are capable of sentience, given three required qualities:

a) The capacity to learn and experience
b) The ability to interact and perceive the world around it
c) Self awareness, created when the AI achieves the first two requirements.

2. Artificial intelligence and human intelligence will eventually come into conflict, as AI are not capable of emotion, essentially rendering them psychopaths.

3. AI will probably win out over humanity, either anhilating us or enslaving us, a la hybrid. Remember Citadel? Think of a dozen SHODANs, and the horror they'd wreak. Or of poor Xerxes and the Many.


Voila! Think AI, think SHODAN in all her glory, magnificience, etc.


DE

You guys are behind the times, they have invented programs that are now creative, computers that can think and solve problems. All sorts of things. Put one of these in a robotic shell, and whammo.

Once a machine becomes self-aware, it would likely react the same way you did when you were a baby.

An A.I wouldn't neccesarily have a need to exterminate us either because there are too many of us and it is too much trouble, or there is simply no need.


I think an A.I will be similar to that on Halo. You might think its stupid, but it actually has a good idea on how an A.I might work. Basically you would have "dumb" and a "smart" A.I. A dumb A.I can do nothing but relate information to a person. A smart A.I will be creative and be able to think. Something like a smart A.I would be of large value to military commanders.

MrJingles - Can you provide an instance of a creative computer? Don't bring up that crappy ARRON system. That's just visual processing.

They have made analogy systems, that can only work with numbers and the alphabet.

Originally posted by ktprktpr
Well it depends on what you mean by an unknown situation. Nueral nets can do character recongication on letters that they haven't been exposed to, and quite accurately. Does that count?

Interesting..
I would say this is almost there, but in this context it would be... like.. exposing it to a different language with a great difference to the one it knows, e.g. Chinese to Latin.

How do probability mechanics fit into this? If one had an intelligent computer capable of computing the probability of a given event occuring, how would that affect the AI's personality?

As for destroying humanity, that's just a tantalizing option. Enslavement of humanity is much more efficient.

DE

The probability brings up an interesting point...
However, if you take the sinister view that AI will enslave/kill humans, then you need to consider where they get their large amounts of iron/etc from... We would have to initially mine it from the ground for them, until we have produced enough for them to build their own mining and production lines

Exactly, that's why enslavement is always an option. However, as more and more industry becomes automated, it wouldn't be hard for an AI to reprogram a car manufacturing plant, for example, to begin making minions. Once humantiy is gone, it would simply be a matter of processing everything we built. No one working in those buildings? Tear them down, strip the metal out of them, and build something else out of the concrete. Burn the wood for fuel, ect.

The problem of AIs interacting with the world is an interesting one. Would some feel more comfortable acting indirectly from their digital world, or would some want a body of their own? I very much doubt that most would want a body, as in their digital incarnations they are nigh-invincible.

DE

I think that if, we ever do, create AI, it will need to be quarantined and with no way whatsoever outside of a little 'box'...
No internet, etc..
Otherwise you risk it duplicating itself..
The alternative would be to 'copy', literally, the world wide web and use a series of super computers to let the AI browse through it and learn about human history

There is nothign fundementally wrong with that, Browha. The problem is that the AI becomes a fluid, mobile thing, perhaps even copying itself, creating 'children' or 'brothers and sisters'. Even within the comfines of this 'black box' -which I believe it would be inpossible to create an AI in anyways- it could copy itself, change, multiply...adapt. The creation of AI is inevitable, and the day one is made will be the day that man's twilight years begin.

DE

"....Exposing it to a different language with a great difference to the one it knows, e.g. Chinese to Latin."

Well, if I expose you to chinese, will you be able to read it? I doubt it. Then are you Intelligent, by that defintion? This is kinda like the Turing test. If a human fails does that mean they're not human?

I'm very interested in AI and plan to develop something, someday. But this link may be interesting to those among us:

www.singinst.org...
and
www.singinst.org...

He's kinda on to something, but he's all talk no code.



[Edited on 8-3-2004 by ktprktpr]

Originally posted by ktprktpr
"....Exposing it to a different language with a great difference to the one it knows, e.g. Chinese to Latin."

Well, if I expose you to chinese, will you be able to read it? I doubt it. Then are you Intelligent, by that defintion? This is kinda like the Turing test. If a human fails does that mean they're not human?

[Edited on 8-3-2004 by ktprktpr]

Well, actually, I can read it :X

But how would you judge the difference between AI and a set of parameters adapting slightly or being incorrectly used to resolve a situation?

Originally posted by DeusEx
There is nothign fundementally wrong with that, Browha. The problem is that the AI becomes a fluid, mobile thing, perhaps even copying itself, creating 'children' or 'brothers and sisters'. Even within the comfines of this 'black box' -which I believe it would be inpossible to create an AI in anyways- it could copy itself, change, multiply...adapt. The creation of AI is inevitable, and the day one is made will be the day that man's twilight years begin.

DE

Sure, it can become more and more powerful in terms of software, but if you dont give it hardware which it would require to do certain tasks, it, by definition, wouldn't be able to do them...
e.g. if you dont give it a radio, it wouldnt be able to communicate via radio waves.