TranceAddict Forums (www.tranceaddict.com/forums)
- Political Discussion / Debate
-- Artificial Intelligence
Pages (2): [1] 2 »

Artificial Intelligence

Every once in a while you'll hear some story about how it is inevitably that computers will be able to think. Most of these predictions lean towards the sooner (a decade or two) than than later.


But I can't see how it is possible for it ever to happen, because I can't think of a way that a computer could ever *understand* anything. So I'm curious what others think.


It seems to me that the best a computer can achieve is a series of pattern recognition steps followed by a lookup table based on what was recognized, without ever understanding anything that is going on.

For instance say you had an advanced android. And you ask it to do the dishes. It would use its senses to pattern recognize where the kitchen sink was, which dishes were dirty, and where to put them once cleaned.

It could do all of that given an extension of today's technology. But there is no component of understanding in it at all.

Whereas everything we do, the understanding element is an integral component of action. It is what separates us from being biological machines.

I think it will eventually happen. I remember reading that the estimated computational power of the human brain is around 100 teraflops, or 100 trillion operations per second.

Our brain receives inputs through the eyes, ears, skin, etc...and there is no reason to think an advanced computer (capable of more than 100TFLOPS) couldn't interpret the same things about our world with cameras, microphones, sensors, etc.

A true artificial intelligence is actually a long way away off...more like 50-100 years in my opinion, but I think we will definately see people like the guy above in the future...Robots capable of understanding the world around them, communicating with non robots, and making rational decisions about their "lives". Once we develop the power of quantum computing, the growth of AI will be limitless. We will have AI programs which creating a more advanced AI programs, which create more advanced AI programs, all in a matter of MILLISECONDS..

With quantum computing, artificial intelligence will expand exponentially, and it really is impossible to predict just how far it could go. It's exciting to think of the possibilities, but also pretty damn scary.

I've been thinking about it some more...

And I think the problem is that eventually 'concepts/things' leave the realm of language, and become something else.

For instance, I imagine the following conversation...
Me: What is a cat?
Computer: it is a feline animal.
M: What is an animal?
C: It is a living creature.
M: What does living mean?
C: ...

It's an infinite series of lookup tables, but the thing that is labelled a 'cat' or described as 'living' are both concepts that are not encapsulated by the word 'cat' or 'living'. Those are merely words which allow us to access the something else that they are.


What is this something else?

quote:

Originally posted by Capitalizt With quantum computing, artificial intelligence will expand exponentially, and it really is impossible to predict just how far it could go. It's exciting to think of the possibilities, but also pretty damn scary.

I can't see how it would ever leave the realm of lookup tables. More processing power = bigger lookup tables... but I can't see the move from lookup tables to actual thought...

And the idea you are putting forth is the classic one which I disagree with which is, "Increased computation power implies an inevitable switch from lookup tables to actual thought", that conclusion doesn't seem to be based on anything as far as I can see.

quote:

Originally posted by Subey I can't see how it would ever leave the realm of lookup tables. More processing power = bigger lookup tables... but I can't see the move from lookup tables to actual thought... And the idea you are putting forth is the classic one which I disagree with which is, "Increased computation power implies an inevitable switch from lookup tables to actual thought", that conclusion doesn't seem to be based on anything as far as I can see.

Do some reading on quantum computers. I don't pretend to be an expert...but from what I've read, if we develop the technology, a computer will be able to do a near INFINITE set of calculations instantaneously.

That means, when trying to understand something (like a cat) a quantum computer will be able to ask *every* "what is..?" question that can be asked about cats. If it asks every question that can possibly be asked about cats, then the mental "image" it develops as a result should be pretty darn close to what the human brain interprets as a cat..

Hmmm...

I don't really like the term "artificial intelligence", because there really is no such thing.

Like you said, it's a database (memory) and an algorithm to perform the function (cleaning the dishes by moving its parts, putting the dishes away, etc). Ideally this algorithm will be part of a dynamic core that is able to create its own algorithms.

Essentially, exactly the same thing human beings (and every other animal) do. You've got your reptile mind with the basics, then everything else is what you learn.

The advantages to AI will be pretty drastic once they get rolling. I'm going to bet it won't be 50-100 years til its even rudimentary.

Those advantages includes things such as instantly being able to perform calculations, instant access to the "core" and all the information we've compiled in our history (meaning they don't need to do the whole fetus>baby>child>teenager>adult cycle, all that shit will come preinstalled and probably be connected to a larger collective supernetwork, thus increasing efficiency and speed), plus they will be able to build themselves out of whatever materials they need to accomplish new tasks (IE: exploring the ocean floor, breathing in space, etc).

Personally, I don't think we'll be able to "program" consciousness. I think we'll program a bit, then let it self-program, and then it will eventually collect enough information to actualize itself and attain consciousness. It will have a different set of factors contributing to its continued survival, thus it will also have a different set of emotions than human beings do, and a different consciousness as well. Consciousness is based on our senses and our language, and they'll understand all languages and more, and be able to use more senses than ours (infra-red, sonar, frequency scaling, touch, smell, and who knows what else).

In a way it's the next extension of our evolution, and it should be interesting. One rule is that whatever you expect probably won't be within our range of cognizence. Just as the ancient Romans never could have imaged we'd be talking to each other through the glow of computer monitors.

p.s.: When I was younger I wanted to dedicate my life's work to actualizing the human dream of self-awareness in a computer system.

quote:

Originally posted by Subey I've been thinking about it some more... And I think the problem is that eventually 'concepts/things' leave the realm of language, and become something else. What is this something else?

Information, man.

The universe isn't strings, or atoms, or photons, or electrons... the universe is pure information.

The crazy thing is, we/what becomes of us will eventually make "robots" so small that they actually become a part of our universe.

When I think of "atoms", and viruses, and everything really, etc etc, I can't help but wonder if it is all some advanced entity's "nanobots" in action.

PS: If some AI mechano-organism in the distant future reads this page by accident while gathering data for a school project:

shibby says hi!

lol

quote:

Originally posted by Capitalizt That means, when trying to understand something (like a cat) a quantum computer will be able to ask *every* "what is..?" question that can be asked about cats. If it asks every question that can possibly be asked about cats, then the mental "image" it develops as a result should be pretty darn close to what the human brain interprets as a cat..

To avoid the added complexity of a living thing, I'm going to replace 'cat' with 'Ayers Rock'


It is trivial to imagine being able to store and retrieve a million pieces of data about the rock. Speeding up this process at no point creates an overall cohesive picture in the way a million pixels creates an image.

But as a thinking person you are constantly creating in your mind a cohesive concept of the point I am trying to make in this text, and that message has a place in your cohesive view of reality.

It's all a continuum that you are able to seamlessly manipulate in your mind. At no point do you have to think about creating a new cohesive view.

But to program any kind of cohesive view is a monumental challenge. And even then I think it would be just a data lookup table in cohesive clothing

quote:

Originally posted by DJ Shibby Information, man. The universe isn't strings, or atoms, or photons, or electrons... the universe is pure information. The crazy thing is, we/what becomes of us will eventually make "robots" so small that they actually become a part of our universe. When I think of "atoms", and viruses, and everything really, etc etc, I can't help but wonder if it is all some advanced entity's "nanobots" in action.

I think that's the wrong direction. Information to me implies a step down the road towards reductionism when I think the correct step is towards the 'sum being greater than the whole'

quote:

Originally posted by Subey But as a thinking person you are constantly creating in your mind a cohesive concept of the point I am trying to make in this text, and that message has a place in your cohesive view of reality. It's all a continuum that you are able to seamlessly manipulate in your mind. At no point do you have to think about creating a new cohesive view. But to program any kind of cohesive view is a monumental challenge. And even then I think it would be just a data lookup table in cohesive clothing

Like I said, me creating in my mind a cohesive concept of the point you are trying to make IS my mind metaprogramming new algorithms (segments of code to handle reality) into itself as I go along.

lol, that sounds so mechanical, and it's not to say that I don't appreciate the art and beauty of it all. And some AI will too. =)

9 relies and nobody mentioned the Turing test ?

To begin with, you need to ask what the word "understanding" means. Is understanding just a series of algorithmic steps implemented by our brain ? Human beings have the property of consciousness. What does that mean ? Is understanding related to consciousness ?

A computer can be made as sophisticated as possible, with a range of sensors reading data and the algorithm giving an appropriate response. But pain, pleasure and all such emotions are real physical experiences of the human brain which cannot be quantified. A look up table is something that has been quantified.

How are emotions quantitatively related in the brain is what we need to understand first, before attempting to answer these questions. Also, consciousness has to be explained by the laws of physics if we are to bring the machine closer to replicating human intelligence.

The computers will replace us. I would consider them human though.

here's an interesting article on:
The Uncanny Valley
http://en.wikipedia.org/wiki/Uncanny_valley

A theory that deals with human's emotional response to non-human entities that emulate human behavior. While the article mostly deals with what the human SEES, I think it needs to start extending to more than just how fluid the animations or facial details are, such as Artificial Intelligence (concepts of morals, emotions, learning abilities, etc).

One of the MOST memorable moments I've had was being 12-13 and having a small 30 second conversation with John Carmack (possibly the greatest computer programmer EVER???? no, i didn't have sports heroes haha). The Quake 2 game has just been released, which sported the "most advanced AI system" for the time. Rather than the characters moving around like zombies, they would sometimes sidestep shots, duck, look for cover or run away to get another monster to double-team.
I asked him "you ever think that maybe AI will get so good that the monsters realize they are being killed and don't want to be killed? So they leave the game and find other things to do, like crash the computer out of anger?"
he says "that's going to be our little secret"
then about a month ago, i found this quote about the next Unreal Tournament:

quote:

quote: When I first got the Darkwalker up and running, I was really surprised to see bots jump in and start really effectively sowing destruction and navigating around with it, even though I hadn't written any custom AI for the vehicle, and it had very different attributes than any other vehicle in the game. There have also been quite a few cases during development of what seemed like bots cheating turning out to be bots effectively exploiting bugs in the game. Hear that? BOTS WERE FINDING EXPLOITS AND... USING THEM!

quote:

Originally posted by Omega_M 9 relies and nobody mentioned the Turing test ?

A bad joke with no context...

http://www.cs.bris.ac.uk/~kovacs/au...ations.old.html

The person linked here has a PHD in Artificial Intelligence. When I last saw him I used the hilarious joke, "But would you pass a turing test?". If you knew him it would be hilarious...

quote:

Originally posted by Omega_M How are emotions quantitatively related in the brain is what we need to understand first, before attempting to answer these questions. Also, consciousness has to be explained by the laws of physics if we are to bring the machine closer to replicating human intelligence.

I see what you are saying but these seemed like later steps to be dealt with. That the first step would be to try and envision something outside of the "pattern match + response lookup" model that was some semblance of true thought.

quote:

Originally posted by SuspicionVandit One of the MOST memorable moments I've had was being 12-13 and having a small 30 second conversation with John Carmack (possibly the greatest computer programmer EVER???? no, i didn't have sports heroes haha).

I saw him on tv a few days ago in reference to civilian spaceflight, I thought it odd that they didn't mention his computer background at all.

Neat video going over some basic bot programming here:
http://blogs.msdn.com/coding4fun/archive/2007/02/19/1719502.aspx

Machines have the advantage of being devoid of all emotion. They can choose the best possible solution based on mathematical algorithms. Logic-wise, computers certainly overtake us. But can a computer be aware of its own existence?

1. We're forgetting that we did not always have these same emotions and consciousness and self-awareness. I think we're operating under the assumption that we've always existed as we do now; it is not so. Everything about being alive is about dynamics and adaptation, from us right down to the very first cellular organism, and perhaps even into the mud and water and whatever else "made" life.

2. We're not even aware of what all of the potential "laws of physics" even are. If you're saying that we'd need to tune it to the best capability we can in the environment we exist in, then yes, I agree...

I think we're still quite young as a species, and we don't quite yet have a good enough grasp of the universe and our philosophies to make any assumptions. The theories will only get stranger, as well.

Like I said, at least 50-100 years until this even becomes a basic, vague beginning. We're thinking inside the box by assuming that we'll always have the same beliefs, theories, and way of problem solving that we do now. As a matter of fact, the exact opposite appears true: major revolutions of thought are happening at exponential rates, and will continue to as the population grows and the foundations of knowledge are expanded upon and passed down.

In 100 years, it will be a different ballgame. In 300-500, or whenever self-awareness occurs (which could be a natural immune response of sorts from the universe) technology and development will be very different. We're just going to see the tip of the iceberg in our lifetimes.

Don't get me wrong, I understand why it would trigger fear and apprehension; being evolved over and knowing that its happening and being the cause of it? We think of ourselves as the tragic hero, when in fact we're passing on the torch of what it is to be a cognizent entity in this bizarre universe. That's quite a gift.

It's something only humans will ever be able to experience. But a lot of our mentalities in general right now are still deeply rooted in "us vs them", "us vs nature", "survival". We just now have a semi-solid footing on the planet and space in order to be able to relax. I have a feeling in the future there will be more of an understanding of the shackling nature of many of our most basic instinctive thoughts and reactions; and we'll slowly shed them off as we will, just like Galileo illuminated us on God, and King illuminated us on racism.

It is my belief that philosophy/religion is a by-product of human self-contemplation. If machines were able to consiously question their own existance, would a new philosophy/religion arise from AI? Imagine a religion of machines based on mathematical logic!! Perhaps a philosophical model devised by machines could provide insight to all philosophies currently floating around today.

quote:

Originally posted by Subey I see what you are saying but these seemed like later steps to be dealt with. That the first step would be to try and envision something outside of the "pattern match + response lookup" model that was some semblance of true thought.

"Pattern match + response lookup" models will depend on the mathematical equations that describe the dynamics of our brain. It is necessary to first derive these equations. Measuring a physical variable quantitatively is the first step towards developing a mathematical model.Therefore, the fundamental challenge is to develop the ability to quantify abstract emotions and thoughts.

To develop an intelligent model it is necessary to answer these questions:

What are the inputs to the brain ? How do you measure them ?
What are the outputs ?
What are the variables that define the states of the brain ? Are they measurable ? Are they controllable ?
What are the mathematical relations between the inputs and the outputs ?
What type of a control system needs to be implemented to generate the appropriate response to a given input ?

Once you have this model, it may

1) Either be simulated before hand for all possible input/output combinations and can be used in the lookup table form
2) Run in real time to choose an appropriate response to a given stimulus ( like our brain) --> I think this would be a better approach.

In my opinion all such studies are incomplete without a complete understanding of consciousness. If we simulate the system according to the first method mentioned above, then it would seem that the lookup table has consciousness ! The lookup table may be printed into a book, and the book will have consciousness ! That sounds absurd. Hence, the need to understand this property before even attempting to design an AI system to replicate human brain.

All AI robots have extremely complex dynamic models and control systems in place. This is not just a question of developing an algorithm. An algorithm is meaningless without the underlying mathematical description of the behavior and a mechanism to generate a response based on the math model and inputs.

Yes. I agree with many of the posts above. In my opinion, Artificial Intelligence is a very touchy subject. The art of computers is easy to just program from top to bottom, and tell it what to do based on algorithms, responses, and dictionaries. But what makes us humans different from computers is the art of life and emotion. Each one of us is different in every way, shape and form. We act, think, and move in a particular way unique to us. We can get into a discussion about how our actions, thoughts, and movements are based on our exterior environment, and how it influences us, but that's another topic for another day.

With Artificial Intelligence, yes it does take processing power to compute the actions what a robot can do based on the stored dictionary. However, it also involves the logics of languages, parsers, semantics, and pattern recognition. In the past, if you tell the robot "Pick up a rock", it would go into the dictionary, and attempt to look up the term "Pick up a rock", which we all know is very futile and unproductive.

Developers and scientists these days have broadened this terminology in 2 ways: languages and recognition. Now, if a robot is told "Pick up a rock", it will attempt to figure the primary subject/noun of the sentence, which is "rock". Then, it will attempt to figure the action of the sentence, which is "pick up". Through recognition, it puts together the logic "rock" and "pick up", matches it with it's dictionary through hashing, algorithms, and tables (i know sounds complicated for this sentence, but for longer logic, it's necessary. lol ), and picks up the rock.

Re: Artificial Intelligence

quote:

Originally posted by Subey But I can't see how it is possible for it ever to happen, because I can't think of a way that a computer could ever *understand* anything. So I'm curious what others think.

Can you think of a way that an assemblage of neurons, synapses, and chemical interactions could allow an organism to "understand" anything?

In any case, since those biological and chemical components apparently are sufficient to, when properly arranged, allow for "understanding," it seems as if we could probably create a "computer" capable of "understanding" out of those materials.

Of course, the question of what exact criteria define understanding, or what can still be called a "computer" for that matter, cloud the matter significantly.

That said, it seems to me the issue is more the presently incomprehensible nature of consciousness than any functional limitation inherent to computer systems.

quote:

Originally posted by Omega_M To develop an intelligent model it is necessary to answer these questions: What are the inputs to the brain ? How do you measure them ? What are the outputs ? What are the variables that define the states of the brain ? Are they measurable ? Are they controllable ? What are the mathematical relations between the inputs and the outputs ? What type of a control system needs to be implemented to generate the appropriate response to a given input ?

Yes, patterning and control are important... you've got to make it able to discern itself.

But.. to think:

We all come from a single cell, embroided with nothing but a set of commands that turns into every organ of us. And from there, well, we all know the rest.

Yet I can't remember my first year of life, nor the womb; why?

That is when we piece together, slowly but surely, the "algorithms" with which we exist. Don't touch the hot stove. Learn your letters and your name. Color within the lines.

Simplicity explodes into infinity.

As arbiter, wisely as always, said.. if it's already been done, we can do it just the same or better.

Furthermore:

If we can't, that will be a damn shame.. because it will mean the end of the line for us.

Re: Re: Artificial Intelligence

quote:

Originally posted by Arbiter Can you think of a way that an assemblage of neurons, synapses, and chemical interactions could allow an organism to "understand" anything? In any case, since those biological and chemical components apparently are sufficient to, when properly arranged, allow for "understanding," it seems as if we could probably create a "computer" capable of "understanding" out of those materials. Of course, the question of what exact criteria define understanding, or what can still be called a "computer" for that matter, cloud the matter significantly. That said, it seems to me the issue is more the presently incomprehensible nature of consciousness than any functional limitation inherent to computer systems.

I agree with Arbiter here. The functionality and 'understanding' of concepts in our brain is not much more a cascade of interactions. Nothing that computers could not mimic. One of the things biopsychologists and neurologists haven't been able to figure out is how all these interactions seem to be controlled by an independent component. Theres plenty of information about emotions and other matters of brain functioning and their relationship to behavior. The missing link is consciousness as Arbiter says.

Re: Artificial Intelligence

quote:

Originally posted by Subey Every once in a while you'll hear some story about how it is inevitably that computers will be able to think. Most of these predictions lean towards the sooner (a decade or two) than than later. But I can't see how it is possible for it ever to happen, because I can't think of a way that a computer could ever *understand* anything. So I'm curious what others think. It seems to me that the best a computer can achieve is a series of pattern recognition steps followed by a lookup table based on what was recognized, without ever understanding anything that is going on. For instance say you had an advanced android. And you ask it to do the dishes. It would use its senses to pattern recognize where the kitchen sink was, which dishes were dirty, and where to put them once cleaned. It could do all of that given an extension of today's technology. But there is no component of understanding in it at all. Whereas everything we do, the understanding element is an integral component of action. It is what separates us from being biological machines.

I strongly suggest you take a look at some prominent philosophers of mind if you're interested in answering your question. Along with the question of what "understanding" is, the problem of what "consciousness" is are the two main issues being grappled with in the field.

Briefly, the main proponent of the strong AI position is Daniel Dennett (click here). The main proponent against the strong AI position is John Searle (click here). You might be partial to his Chinese room argument (click here), although make sure you read the replies to his argument.

You might also be interested in other philosophers of mind that look at these two problems from a different perspective, perhaps someone like Patricia Churchland (click here) or David Chalmers (click here). The latter does a good job of defining the so-called easy and hard problems of consciousness. He also maintains one of the best philosophy of mind web-sites I've found with a wealth of (pre-print) papers that you might be interested in reading (click here).

Powered by: vBulletin
Copyright © 2000-2021, Jelsoft Enterprises Ltd.