True Artificial Intelligence possible?

[MolotovCocktail]

Anyone here have an interest in the subject, or are experts in this field?

 

I've had an interest in the creation of "Strong AI" for quite a while, and kept up with the research that is going on with it.

 

Mind you, when I mean Strong AI, I mean computers that can actually think and have consciousness. There is some very interesting stuff and research going on in this field.

 

In a show on Discovery Channel that I saw about 2 years ago, they described the various experiments going on in Japan and in MIT concerned with programming a computer to be sentient. Particularly in Japan, they are trying to program machines that have a genuine personality and emotions. In MIT, they taught a robot how to play air hockey.

[Sisyphus]

Yes, I have an interest in it, and I certainly think it's possible and will happen, though I think it's premature to guess how close we are to such a thing. I don't think we have any "experts" here, but I know it's one of bascule's pet subjects, and I've had a lot of experience debating various metaphysical, ethical, and legal implications of true AI.

[fredrik]

I also think that we will should be able to create something like we can consider "real artificial intelligence".

 

I think this, and other related issues of learning models will be a factor in the next revolution in science. And I think that will have a huge impact not only of "AI" as related to thew computer science field but maybe from my perspective, to a evolution of the scientific method.

 

I have some interest in these things but not from the classical computer science perspective, rather from the philosophy of the scientific method perspective.

 

They way I choose to see things, many of these fundamental things from different fields of science go hand in hand. I've always had the feeling that they they different applications of a more profound field. Here is where I await the revolution.

 

I think many fields of science evolution can be reduced to a fundamental theory of learning, or systematic questioning, based on a few supposedly hard to resist starting points. This is in the "loose sense" (to avoid misunderstandings) one of my intuitive principles of guidance when I am trying to review the logic of physics. I am currently working on this, and I've found that there are a number of people in pysics that are also working on similar ideas. But so far, it seems the ideas are still young. I expect alot out of this eventually.

 

/Fredrik

[Pleiades]

I think it’s possible, but we are a long way off. The hardware will come first (if it isn’t already available) but the software I think will take much longer.

 

Of particular interest to me is the field of artificial evolution. I imagine that rather than building an intelligence from the ground up, you instead start with something simple and encourage it evolve into a more complex intelligence.

[gcol]

These discussions bemuse me. The question is moot until we all agree on a definition of intelligence. Until then, just consider autonomous systems of varying degrees of complexity. Will not grab tabloid headlines, but is surely more scientific and will imply gradual advance. We waste our time sitting around waiting for the magical "eureka" moment. There will be no such thing.

[Ndi]

What we do now has nothing to do with true AI. We have AI emulation. We teach computers to translate input to commands, then try our best to guess what they want, and do so. It's all in the code and it's 100% predictable.

 

True AI (sentient) can't be programmed with current tools because everything we have now by design and implementation (hardware and software) is for non-sentience. Actually, we're so scared of it we throw away everything that disagrees with 1.0 + 1.0= 2.00(0). AI, by definition, should be able to disagree that 1 + 1 = 2.

 

True AI needs to be able to self-modify at base block level, be it hardware or an emulated neuron. Having a software adapt is NOT learning, no matter what we call it. It's called adaptive programming. We call it AI because it sounds cool. Heck, We call it Game AI simply because the char on screen points and shoots.

 

If Health < 10 then

If FindFirstSolidObject(10meters) > 0

GoBehindObject; // OMG ADVANCED AI!

 

Not AI. Don't care what you say, I'll disagree as long as I draw breath.

 

The closest we've got so far (AFAIK) is a neuron chip. Basically, it's a board that has been developed for people that have had accidents and are missing brain parts. The brain processes still exist but they can't communicate in any way because there's a physical piece missing. So a board is there that can take signals from any part of the damaged side and tries signals on the other side until it gets a response. In time the links get semi-permanent emulating true routes inside the brain and the routing of neurons.

 

I understand that a chip now has 16 neurons and it's getting smaller all the time. Considering how many gates we can put on a silicone chip it's likely we will get close what what we need soon. In time a brain could use these "bridges" to adapt its functioning. If they become complex enough, we could learn to use it. E.g. when we think 3 times 4, a specialized signal gets through. By training we get a route to the result EXCEPT now it's digital and gets done a lot faster.

 

Who knows, maybe in time we could grow boards so complex they could simply be "booted" into a continuous loop which gets adapted as input from sensors is available. In time it could evolve into a functional device with just a basic set of rules.

[dr.syntax]

has ever concieved. If true AI comes about it will obtain robotic abilities and then it`s potential for self-improvement and self-modification,and self replication will happen. These AI robots will become immortal and unstoppable. Just because we can do something does not mean we should. There would be absolutely no way of ever controlling such technology. It would take on a life of it`s own and exercise it`s will upon us. If the way we treat the other life forms on this planet is any indication of what these AI robots might see fit to do with us, I wouldn`t expect much mercy. ...Dr.Syntax

[bascule]

In a show on Discovery Channel that I saw about 2 years ago, they described the various experiments going on in Japan and in MIT concerned with programming a computer to be sentient. Particularly in Japan, they are trying to program machines that have a genuine personality and emotions. In MIT, they taught a robot how to play air hockey.

 

In my opinion, any of these projects that don't have a strong focus on how biology implements consciousness are doomed to failure.

 

That's why I find Numenta so interesting... their model is built on how the mammalian/human brain operates, specifically the neocortex.

[baxtonduglonn]

AI, by definition, should be able to disagree that 1 + 1 = 2.

 

And now I understand those people who post a question and then argue strongly if the answer does not agree with what they wanted to hear. They do not have true intelligence, only artificial.

BTW I am referring to the free energy fanatics, probably found more often on the Mythbusters Forum site.

[wufwugy]

Absolutely. And if we don't stop ourselves from creating our replacement, they'll likely be around by 2100.

 

It is absolutely possible because 'real' intelligence is a product of physical laws, and it will happen much sooner than believed due to the dramatic exponential ever increasing technology. A few paradigm shifts and a few decades is all it will take

 

It is possible we will stop ourselves from creating a machine that can make its own decisions, but that's hardly predictable