Christopher Lord

Laurentius, thanks for the reply. I didn't mean to be quite as bitter as I sounded in my reply before last. To offer an exuse, I was under the influence . I realize how many replies you have been getting from various people, so feel free to take your time.


I can relate, I dont typically get into the 'advanced' debate style. I prefer informality.

deterministic is quite loaded here. The universe, due too insurmountable chaos, is only deterministic when things are held to exacting specifications. just because the universe is deterministic, doesnt mean I cant wig-out and go on a killing spree tonight, or go join a cult. It just means you can know everything (past/present/future) about the universe if you the exact rules and the exact state of enough demensions (think pythag. for triangles). And I use the word exact in its literal and exact meaning.

Same here. we do have fairly limiting skills, even compared to some advanced contemporary computers.

but a cat has at least one, and probably all of those things as well, as far as we can tell. If you try and kill it, it will fight or run. this is what you mean by preservation, correct? And will, cats have plenty of that. Almost everything they do fits their agenda of eating, sleeping, and eating. They just cant make dynamic agendas because they lack the ability to speak to themselves in their head with a language. This is perhaps the single critical skill that makes a mindful creature. A non-speaking human is, i expect, much like an ape.


I mostly agree. For example, if you rewrote (c) to say "computer chip" instead of "brain", I would disagree (as you do). The chip is not enough to exibit a software application. Its the substrate from which abstractions are built. I consider the world of software completely seperate from the world of transistors, electrons, and etc. And thats with a few hundred abstraction layers. So of course more than the brain is required for a mind. At the very least it must be trained for 10-20 years. Pre-training, the brain is there, but no mind exists. Its the process of learning combined with a language that creates a mind. I think 'towers of abstraction' is a rather good way of putting this, since the process of learning can be thought of as building a bridge between towers so that mimes and technology can be passed to other towers.


Regard the tower concept again. Each person built their own plumming with the training passed over the bridge. Thus each person seems similar on the surface, but is actually only built to appear that way on the upper levels. If this method existed in computer programming, it would probably be labeled 'high-level logic transfer', and would have the computer writing most of the plumbing for the desired interface, defined in english.

I'm having trouble understanding your example and how it relates. Is it irrationality? This is not a problem for a 'mechanic' mind. Just because the mind is 'software', does not mean it must be rational at all times. In fact, the taller the tower, the more I would expect irrational behavior. Huge software applications are more prone to errors, just as a human is more prone to being irrational compared to a dog.

Out of need and history lifeforms do these things. Machines do not do these because they are not yet complex enough on top of the fact that they dont actually need these things. Modern computers are barely at insect level complexity with rodent-level abstraction capabilities, which is importaint to keep in mind here. When the time comes, machines will be outhinking us, or at least have the complexity for the potential to do so. That depends on if we can also train such a thing to think as we do our children, since that is key.


Quite rightly its an extension of the human. But what happens when it extends further than we can normally extend ourselves? same thing that happened when we extended our eyes (micro/telescopes) and legs (cars); things are gonna change. This is the final extension we will have to reach, though. At what point will an AI be held responsible? Already, actually. If a trained nural network fails, the fault is with the network and not with the manufacturers, since the device learns as it goes, and quickly escapes 'factory' params. When these nural chips are churning out plans for a space shuttle in 30 years, and one puts a screw in the wrong place and the dam thing explodes, it surely wasnt the nural-net manufacturers who will be held responsible! the device will be wiped and re-trained (perhaps even by other nets), with the mistake now in mind. At most, the trainers can be held responsible, as parents are of children. When the complexity is there, any abstract tower can be built.

Good stuff so far, I actually want to work in AI eventually, so its good that I challenge my definition of mind as well. So far I am pleased with my current working model, the abstraction tower. I find I can apply it to most arguments with a good helping of success, although I do sense some of the weaknesses which need to be addressed. For example, the tower fails to help with emotion, although evolution helps with that. And really, emotion is probably one of the more basic systems in a creature. We just have language to express them.

Laurentius

Owleye
Well, this is indeed something that needs to be worked on. No matter how deep the thought of Deep Thought is I suspect that is not deep enough since it lack self relectivity. This is what I was talking about after all. The power of a conscious system to objectively view itself and act according to the inhanced information and analysis thus generated. But I'll certainly come back later. Now I've just poked in during a lunch break.
AVE

owleye

Laurentius...

"This is what I was talking about after all. The power of a conscious system to objectively view itself and act according to the inhanced information and analysis thus generated."

Here's hoping you do. Consciousness, in the form in which most philosophers speak of it, implies self-consciousness. That is, when we are conscious of X, we can be, reflectively, conscious of being conscious of X.

This way of analyzing the problem, of course, takes us down the path of phenomenology. I would recommend pursuing this, and, if possible to become acquainted with Husserl, who has us being able to carry out a transcendental reduction whereby we are able to decouple consciousness, in the form he refers to as a natural attitude toward the world, from that which we are conscious of without such a reference. There is a distinction then between the object as intended, and the intended object. Frege would refer to the former as the sense, the latter as the reference. Intension and Extension as aspects of meaning are also in common use.

Following in the footsteps of Immanuel Kant here, Husserl realizes that in our natural mode of consciousness we take the world as it appears to us (not taking this to mean as it seems to be) to be the real world. In that sense, following Kant, our mind in its capacity to experience the world takes up the position of an empirical realist. But, Husserl goes on to notice that we also have the ability to perceive objects without attributing what Kant would have referred to as substance.

There are in fact examples of this occurring spontaneously, as for example, when we believe the image of the person speaking to us from Afghanistan has been time delayed and at the time the image appears to us the person so speaking has moved on. That is, on this realization, we've decoupled the person from the image. Of course, because of the pressing needs of life, we don't ordinarily occupy this condition for long, slipping back into the natural attitude rather quickly. (Another example, is how the image of the sun can be thought decoupled from the sun itself, which clearly is not in that same position at the time the image is presented.)

In any case, he goes on to describe a vast number of modes of consciousness toward what he believes to be a science of phenomenology.

owleye

Laurentius

Christopher Lord
AVE
Okay; so, everything there is could be described physically were we to be aware of all the dimensions and possess the relevant data. The physical world is pretty deterministic; the incredibly high gravity will prevent light from leaving the black hole or the prolonged cold will badly damage the crops. These are phenomena. But there are actions that are not deterministically triggered and neither do they happen coincidentally due to random encounter of different causal chains - such as those you're just mentioning about. Well, that's the plan of reality that I had in mind when talking about the superiority of the Mind over the Brain. The brain as an organ functions pretty deterministically, like most of the observable universe. But the Mind (that unique aspect of the Brain) is quite free from many causalities in this respect.

I meant more by preservation than just fleeing - all those ability of preserving the same organic form that is developed to its full capacity by each individual, on the one hand, and by the whole species, on the other - and so on. Cats have all this. They have will as well, but lack self-reflectivity - which only the "supreme" Mind can pride with (in fact, the Mind is the self-reflectivity itself.

Yes, but are computers really self-training. I mean, can they teach themselves to aquire skills that they have not programmed that they will ever need? Living things seem to adapt to an extremely wide range of new situations. Can such a program be devised for compters? Maybe, but I don't think so.

Okay, I'm going to present a coherent persepective of the achievements for my theory quite soon.
AVE

[ March 19, 2002: Message edited by: Laurentius ]</p>

excreationist

Laurentius:
I just read a little of your last post....
Well some preprogrammed behaviours are necessary - even humans have this when we are born. This helps them get along in the world and defines their goals to learn to solve (what to seek/repeat and avoid). They learn about how the world works along the way.
Anyway, I think cutting-edge AI could teach itself new skills but it wouldn't be much smarter than a kitten. It would need to have a craving for newness to motivate it to discover and explore things.
Computers on their own aren't self-training though. They just do exactly as they are told, step by step.

Christopher Lord

Oh yes, I've actually built software 10 nuron self-tranable nural nets. These basic little things can do tricks like recognise lines and whatnot. I've also built a simple GA which searches via evolution for the best solution to a problem. GA's express unique solutions that people cant even understand at times, much less program into them. The state of the art would probably surprise you.

My own projects typically aren't too intresting, typically because I dont have time to train a net properly or crunch huge GAs. Usually I write a conventional app to train them, which cant train complex things.

Game AI is typically 2-5 years behind whats at the state of the art, and current game AI has achieved plenty of 'creativity'. Pick up 'black and white', and you have a fairly sophisticated little AI package with which you can train a creature who is driven by an interesting reward/punnishment (slap/pet) system. The game creators never even imagined some of the things this AI is now being reported as achieving. It 'learns' and acts appropriately based on what the operator rewards them for. I've coaxed my own creature to eat only cows, and to heal people. Some people train the opposite. some people train it to always lift weights, some train it to throw a certain person in the ocean.

This is in a comercial game released a year ago, using AI methods developed long before.

<a href="http://ai-depot.com/" target="_blank">ai-depot.com</a> is a site offering help in the practical implementation of AI in current systems.


<a href="http://technetcast.ddj.com/tnc_play_stream.html?stream_id=526" target="_blank">This</a> is a transcript in which Marvin Minsky talks about why current nets cant 'reflect' on output.

And remember, current hardware is only as complex as a small insect, so by that metric our AI software is actually ahead of nature in terms of the complexity-behavior ratio.

[ March 19, 2002: Message edited by: Christopher Lord ]</p>

Synaesthesia

Laurentius,

You are taking up the position that mental events are more than physical structures and manifestations of physical law. What I want to know is what more are they? I don’t think they are any more. I think that the mind transcends merely what you imagine matter to be capable of, not what we observe it to be.

Not only is the comparison relevant, it directly pertains to the topic of this thread. Our conception of a glider is a heuristic, a shortcut in our understanding that neglects direct reference to the squares that make up the shape. We can develop a meaningful understanding of the glider without direct reference to the Game’s laws.

My position is that like gliders, our minds are amenable to being understood without direct reference to physical laws. Physical objects can indeed have things like intentionality(will) and can indeed interact with the world via a representation of it.

If we were to count up the squares, the glider totally obeys the rules, does it not? So in that sense, nothing at all is or can be added. If you like, you are perfectly free to describe “laws” that govern gliders, but it is absolutely imperative that you not forget when you speak of the glider’s behavior, you are implicitly referring to our perception of the emergent properties of ONLY the G-laws.

This is false. Cats have to devote an enormous amount of representational power to getting a sense for it’s own state. Of course it’s nowhere near the degree of sophistication of man’s reflexivity, but that misses the point: self-awareness is not a single ingredient. It is a combination of thousands of propensities and skills.

One can easily make a self-modifying computer program. What is far from easy, and what took evolution billions of years to do, is make a program modify itself in useful and flexible ways soas to cope with a diversity of situations and tasks.

Christopher Lord,
I suggest that when we say “brain”, we automatically suggest the structure within that brain By this semantic formulation, nothing but a brain is required for the mind; The fact that this brain has a developmental history should go without saying.

excreationist,
So do our neurons but we are indeed smarter than kittens.

Regards,
Synaesthesia

Synaesthesia

owleye
Ah, please do not misrepresent Daniel Dennett so. That the mind/brain are complex is a fact that Dennett takes for granted. It is not the crux of his conception of how it works, it’s simply a consequence of how it works. I should remind you that Dennett and his ilk are very outspoken critics of expert AI systems, whatever their complexity and processing power.

This is called the fallacy of division(http://www.infidels.org/news/atheism/logic.html#division) That is, the assumption that since a system has some property each composite element must have it.

1.This post is 170 words long.
2.The words in it are not 170 words long.
Therefore
3.This post is not made of words.

Regards,
Synaesthesia

excreationist

This also means that even if you scanned the contents of someone's brain, it would be very difficult to decode what all the information means... at the moment they're just correlating which brain areas have activity in what situations. BTW it's nEural and nEuron.

Synaesthesia:
What I mean is that we need to explicitly program computers how to solve problems. With neural networks, we can just show them some examples and answers and they work out the pattern for themselves.
Actually it looks like the <a href="http://www.genobyte.com/robokoneko.html" target="_blank">robotic kitten</a> hasn't been finished yet. And I'm not sure how good it would be at exploring its environment and teaching itself new behaviours.
I think an even more advanced robot would be <a href="http://www.cyberlife-research.com/contents.htm" target="_blank">Lucy</a> - which is being created by the creator of the "Creatures" games, Steve Grand. This robot is also controlled by a neural net, but it will begin extremely simply - like a baby - and it learns new behaviours.
And obviously, we're much smarter than kittens... but AI research keeps on moving forward...

Laurentius

Owleye
Year 3333. Astronaut Joe Amstrong and his highly computerized shuttle are hovering Jupiter's troubled gaseous surface in a mission during which the shuttle has learned how to deal with many of Jupiter's gravitational irregularities. Suddenly, while Joe Amstrong is busy abserving the recent unusual behavior of the guineea pigs in the lab area, the computer displays on its screen an enormous gravitational storm in the eye of the planet and announces bluntly: "Shuttle destruction is imminent in 10 seconds." The guineea pigs are frantically running all over the shuttle now, and their crazed squeaks show how they cannot take their little senses away from the danger. In a moment astronaut Joe Amstrong mentally views his highly sophisticated but minuscule shuttle caught in the gravitational turbulence, with him and the guineea pigs trapped inside, and says to himself: "We're doomed."

No matter how complex the AI of the shuttle is, in the case of the computer we can only speak about INTRANZITIVE behavior, the automatic execution of implanted or learned operations. By contrast, the guineea pigs show higly TRANZITIVE behavior toward the environment with regards to which they engage an emotional rapport. Detaching himself to his emotions and mentally picturing the situation in which they are all caught, the astronaut's behavior is REFLECTIVE.

I don't know how convincing this hypothetical situation is, but that's about the most rigorous "philosophy on the run" I am able to provide under these informal conditions. Besides, the great difficulty is to persuade computer-oriented people of the fallacy considering high-intelligence bearing structures as being relatively the same as the human brain.

Since my major is language, I've inevitably studied various theories, from Saussure to Chomsky and beyond. I'm not working in this field though, and I've come to detach myself from all those abstractizations. I think I'm more attracted by a popular philosophy that specialists may scorn, but which can be accepted and assimilated by the common man.
AVE