Artificial Intelligence & Android Prospects

[Oxygon]

I wanted to start a discussion on this stuff...

I have been interested in this stuff recently because of the limited release of the movie "Ghost in the Shell 2"... And the android design based off of "Hans Bellmer" dolls, Plus the imagery from "I, Robot - 2004" movie.. Aside from the butchery of asimovs story...

I first wanted to express some of my ideas and concerns...

First I think that A.I. is in a "lull" due to the way in which we are trying to evolve the software, This also ties in with the android concern... I think that it would be better to evolve "thinking machines" in a completely virtual environment, this virtual environment will be randomized for error and deviation from projectable responces... Programs emmersed in this environment will recieve as "virtual participants" a limited amount of data on its environment - comming only from the allowed pathways (virtual eyes, ears, limbs, etc.) These programs can learn at a greater speed and lesser cost than "robots" built in reality, there parts due not fail (unless programmed to do so) and they are cheap to build. Programming a virtual world to "vary" the laws of a "virtual world" would allow the "A.I.-bot" inside to effectively prepare for the actual variances in the real world, I think an A.I.-bot could learn to walk much faster if this method was utilized, eventually this A.I.-Bot could learn how to extrapolate the data recieved thru its "virtual senses" and determine what to do in real life when someone pulls the rug out from under it... or a step fails to hold ground... if you can randomise the environment enough virtually (including the variation of limb response, etc.) the android would in time be able to perform olympic movements of super human proportions... in the real world.

The Mental Aspect of A.I. (Personality and Understanding) I think would benifit greatly from the method of "virtual training" giving an A.I.-Bot a simaler frame of reference in the world to that of humans, then allowing a three demensional ability to determine the shape and location of objects with relation to self... (this too can be perfected in a virtual environment with two virtual eyes, etc.)

With the ability to extrapolate the potential identifications of an objects classification and A.I. could then come that much closer to understanding the world around it in reference to the words we use to communicate our ideas or intent. (Axam. pick up the red ball, or tighten* the second lowest left screw) *=Using tool in reference to the objective.

I think that the "Android Golden Age" is near... if we last that long...

Still a word of concern... I recently watched a "video.wmv" of the Sony "QURO" in action and couldn't help but visualize this pack of androids with knives and not paper fans...!?!?!

they would be tools just the same...

Robots don't kill people, people kill people...?

That phrase is a bit worrysome...

http://www.plyojump.com/qrio.html

Say something if you want... I just needed to express my opinion...

[Jonathan]

I sincerely doubt AI is possible. Maybe after a lot of work there would be Imitation-Intelligence, but machines just won't ever understand. This reminds me of what I was saying to ssmith, about levers. It is well known throughout physics that there has not yet been a device which was greater than the sum of its parts. Because of this, a bunch of non-thinking bits put together just makes one big non-thinking system.

[ovyyus]

...Because of this, a bunch of non-thinking bits put together just makes one big non-thinking system.

A biological 'thinking system' appears to be put together from the same fundimental 'non-thinking' atomic ingredients as everything else. Perhaps a machine might become more than the sum of it's parts beyond a certain level of complex interaction? 1 + 1 = 3

[Jonathan]

That is the general idea, but it seems unsupported by evidence. Consider the size and density of diverse molecules in the brain, and compare that to the earth. Since the earth is bigger, more complex, and has many more kinds of molecules than brains have, you'd think that it'd have some kind of uber-consciousness, but this doesn't appear to be the case.

[ovyyus]

I see what you're saying but I don't think you can easily compare the complexity of the Earth and the brain in that way. Complexity in each case appears to serve a different purpose, with a different result.

Personally, I think the 'sum of the parts' arguement doesn't take into consideration the often unseen processes riding on the mechanics. Without energy continually entering and leaving the complex brain structure I'd guess it would just be another lump of unthinking meat. In this case 1 + 1 + energy = 3

I think this is also the case with our wheel - without real energy entering/leaving the machine, it can never be more than the sum of its parts.

[Oxygon]

You guys seem pretty sure of yourselves...

I believe that if you can make a "bot" in a virtual game understand certain stimulus, I would assume given an ability to determine a real world 3D environment from the experience in a artificial one... and "bot" could then associate sounds and visual data, etc. to objects...

Then all that would be left to do is fill its database, because if it then has the ability to identify and articulate certain object identifications... its not that far from a full blown language itself...

So all we need is to develope a 3D understanding simaler to ours and the object identification, then the line between man and machine would begin to blur... imo.

[Jonathan]

If consciousness were a result of complexity, you'd think it would turn up in any sufficiently complex system, regardless of form, design, or "purpose". To clarify, I consider the energy to be a part of the system, so it would be included just like all other parts without special treatment. I partially agree with your last point, but it should be noted that if energy is coming or going, it must be coming from or going to somewhere, and this "where" should itself be considered part of the system. For our PM purposes, no device need be greater than the sum of the parts, it could just be that one of the parts is infinite (or astronomical and finite) and that'd be good enough. (Consider one of the system's parts to be zpe or whatever).
You are right Oxygon, if a machine could understand, then it wouldn't be too hard to make it capable of living in the real world. But that is a really big "if".

[John Collins]

If consciousness were a result of complexity, you'd think it would turn up in any sufficiently copmplex system,

Animals are complex systems and I think they're conscious but not in the way humans are. You can train a chimp or a horse or a dog to appear to count but it doesn't understand what it is doing. It's a trick but it doesn't know it. A dog may pick the right answer or bark the right number of times but it is doing either, for example, by signals received from its trainer.

You could make a robot which could be "trained" to carry out simple tasks in the same way as a dog, but I doubt it could think for itself. Having said that, it occurs to me that a dog can think for itself, but only in a limited way such as "if I fetch my lead and bring it to the boss, I can get him to take me for a walk". Did he work this out himself? Did he reason it or was it just a learned set of actions and reactions? Is that different ot what we do?

John

[Jonathan]

I agree. My dog clearly has dreams, marked by foot and face twiching and making little woof sounds. :)