|
Post by tkorrovi on Oct 3, 2008 8:56:51 GMT -5
Many may think so about someone, who proposes some seemingly far-fetched solutions such as a system which is aimed to be absolutely dynamic. But this is not true at all, i am not extreme, the problem is extreme. So many sure want to find some "simpler" solutions, to escape the complicated problems related to systems aimed to be unrestricted, but nothing in science can be solved just by escaping problems. One of these miracle solutions is often said to be neural networks, but neural networks by themselves cannot implement whatever functionality, they can be used as parts of a system, designed to implement certain task in the system. In the widest sense, neural network is a set of connections, where the properties of every connection can be changed by some genetic algorithm. Connection is a general term, yes, and so is genetic algorithm, but all the structure of such systems must still be pre-determined, and pre-wired, thus such system is no way made to be unrestricted. Some may think that it is not so big problem when the system is somewhat restricted. But it is the biggest problem, the whole idea of AI, if by intelligence we really mean intelligence, is that such systems should have some advanced capabilities, they must be capable of finding themselves many kind of solutions. But when a system is restricted, then it is never slightly restricted, the number of possibilities is so immense, that every restricted system is severely restricted, what we perceive like a system which doesn't seem to be capable to do almost anything. Sure the creators of such system would not be satidfied afterwards, another wasted effort among a huge number of others in AI, and all for the same reason, and because people are never willing to admit that reason. What are the other alternatives, cellular automata... Something which most already reject, because it is almost impossible to control the development of the patterns there, almost no one seem to have an idea what to do with these, thus they prefer to call them games and not "serious" systems. But even these are likely not a solution. We have a certain capability of developing patterns, we have no idea whether these patterns can never by itself develop into some universal Turing-complete system. For these who cannot understand the arguments, i must again say that i know that a Turing machine has been implemented in a cellular automaton, but this doesn't mean that a Turing machine can emerge there as a result of self-development. But even if such patterns can by themselves develop into some universal system, all the question is still what system it should be then. And this should likely be some interconnected system, such as ADS, all the guns etc in cellular automaton are finally for implementing connections. Thus a cellular automaton, in addition to the fact that there is no known way how to train it if it even is anyhow possible, is not by itself even a solution, we may also have a pile of logical elements on our desktop, and indeed whatever system can be made of these, but this pile of logical elements would not by itself develop into anything, we stil must know the system which we should make from these logical elements. Thus cellular automaton is not a solution of a problem, but rather a mere substitution of one problem, with the same problem in a slightly different form, which may only seem to be a solution because one may not see the problem once the way of finding the solution is somewhat altered.
Sorry again for such raw writing of ideas, which i'm often used to do. This used to be my version of "release often" principle of open source development in all these forums, channels and instant messages, which used to be the form of discussing these ideas for many years. There, you have to write quickly, for any discussion to be possible at all. There was no way to talk about these things at all, if everything had to be written in thoroughly perfected peer-reviewed papers. Well and i should ask this question again, how many thousands of years did it take to develop the Linux kernel, if every change there had to be published in a peer-reviewed paper? No, this doesn't mean at all that i'm against scrutiny, but there are other democratic forms of scrutiny which simply enable a thousands of times faster development. Finally still, the result of the work matters.
Thus no, i am not extreme, the problem is extreme. But there are ways to research even such things, this has been done, and there is no reason why it cannot go ahead, the one and only reason what there may be, is the reluctance of people to deal with such problems.
|
|
|
Post by ano on Oct 26, 2008 0:26:19 GMT -5
who r u trying to convince? any why? r u waiting for someone to come and make something good out of your project? or for some mirracle? well you'll keep waiting cos it aint coming dude.
btw ur texts are not easily understood. try making s h o r t e r sentences!!
about restrictedness: humans developed from primitive very restricted forms into very unrestricted. we can develop AI the same way. by keep improving it to be less and less restricted. slowly step by step man, not all in one magic step like u wanna do it.
|
|
|
Post by tkorrovi on Oct 26, 2008 11:41:07 GMT -5
Well, you may guess that people like you are not these who i would expect to see, or perhaps when you change your attitude or learn to think more. I'm not trying to convince anyone, everything is written, who is interested would come. If no one would start to develop the project, well, it is sad then that no one takes seriously enough something which is the most basic and essential, but in that case there is nothing i can do, even if i did everything which is humanly possible. In that case it is even not important whether i develop the project further or not, and how good results there would be, i cannot create a system which can control a factory or such, and thus the matters would always remain the same.
You should understand that it is not possible to explain complicated things with very short sentences, i'm not writing a children's book. Yes i know that for some "I don't understand" is a conclusive argument, this can really never be an argument for any intelligent person. And if you read even this forum, you see that there are people who understood at least some of the most basic things, thus "No one understands" is by far not an argument. I know that for you it is difficult to believe it, but there are people who at least by now can understand better than you.
You say the same way, but you don't even care to think whether it is the same way or not. Development of life is not exactly the same as adding more and more software to your operating system. A restricted system cannot develop into unrestricted just by adding complexity, not every kind of complexity can produce whatever results. It is only possible when an unrestricted mechanism would emerge, and the development would continue based on this. This unrestricted mechanism can also work in parallel with the restricted system, and this is what seems to happen in brain as well, some neurons there likely implement some very simple functions, and may even work as a fully pre-programmed stateless automaton. This would not make the whole system restricted, when there is also an unrestricted mechanism, but can somewhat influence the development and functionality. Thus the development from primitive to unrestricted is possible, but this involves emerging unrestricted mechanisms. Thus what are important for advanced systems, are still the unrestricted systems, and it is therefore important to research these. When this development happened, we have no idea now, as we don't even know how such mechanisms are implemented in neurons. But they should be based on something which cells always had, and one of my suggestions is the changing system of microtubules within each cell. These are the predictions which a theory have, there are many predictions what concerns unrestricted systems, and once it would be possible to find out whether these predictions are true or not, this, in general, is how science works.
Emergence of an unrestricted system is not exactly a "magic step", but this is a qualitative change. Who knows something about philosophy, likely knows that the system don't develop always gradually, but sometimes the quality changes, and as a rule this happens faster that the gradual development, something like a revolution. And the systems don't develop only by gradually adding the complexity, in time the whole system change, including the very bottom of it. Remember the ouroboros principle -- everything disappears once, enabling the development of the new, when the old basic structures would not once disappear, the system would not be able to develop into a higher level, because the old structures would hinder the development, by causing too many restrictions to new formations, due to their outdated format, etc. This is likely different than what you understand that the development is, and this is what they often forget in software development, as they often don't provide a mechanism to delete the old data structures, and they would cause restrictions in further development due to their outdated structure, the system kind of cannot rebuild itself. We see this hapening in evolution, the more advanced species would replace the more primitive species, and this often happens rapidly. Humans didn't develop as a result of gradual development of insects, fish or reptiles, more advanced species at the qualitatively higher level emerged one after another. Emergence of unrestricted systems is also some of such qualitative changes, which are quite normal in development. This likely didn't happen as "one magic step", but by such systems appearing first maybe as minor parts, and then gradually taking over the functions of the more primitive, pre-determined systems. But in that process, the very basic mechanism of most of the system changed, which is also quite normal in development.
Well, as one can see, even the stupid questions may sometimes help to explain the things, though i certainly prefer smart questions, and people who are willing to understand.
|
|
|
Post by John B on Feb 1, 2010 17:54:05 GMT -5
Ok. I did not read the whole thing, because I think not many people will want to read paragraphs that are 30 lines long. I agree with the other guy that said that your posts are too long.
But let's get back to the main subject. The human is extremely restricted by its DNA. The human is intelligent and it is a VERY VERY restricted system. So the key to being intelligent IS NOT to be unrestricted. If we build a random system, the probability that it will develop into a human or something similar is close to zero. For humans, the evolution has taken millions of years and billions of humans to reach the current "human". Moreover, the human is extremely complex. The brain includes many hundreds different types of neurons, and each one can have hundreds of connections. And just modeling one neuron realistically with a computer is far from being possible yet.
So don't expect to reproduce the human evolution from zero to create an artificial intelligence. That would be very naive.
Besides, I had a look at your "conscious" system and it is pretty basic. Sorry to sound cold, but I would call that extremely useless. I don't think anybody will be ever interested in pursuing this idea. Good luck.
|
|
|
Post by tkorrovi on Feb 2, 2010 3:31:19 GMT -5
Yeah, thanks for saying what you think, even if there is unfair criticism, everyone is interested what is it about. First i think we should not start to talk about the length of the paragraphs, as this would inevitably lead us out of topic. You think that the human DNA restricts the human intelligence. This is a quite questionable argument which yet has to be proved. DNA is not exactly like a code in the computer, and often it determines only the direction of development, i hope you understand this, but how DNA exactly works is not even relevant here. One may notice that the most unrestricted self-developing systems there are, such as cellular automata, have mostly a very simple structure or basic mechanism which can be easily determined with a very small pre-determined code. And the ADS which is aimed to be extremely unrestricted, has a simple basic mechanism as well. A not pre-determined self-developing system is not necessarily random, and it is not random exactly because it is self-developing. Like the regular patterns which emerge as a result of development of the structures in a cellular automaton, show quite clearly that this is true. The main reason why the development is not random, is that such structures would survive, which fit better to the environment which they are in. Yes the human evolution has taken millions of years, but this has happened differently in quite inflexible physical conditions, very different from these which can be made present in an artificial device such as computer, so that again it is not a proof that something intelligent cannot evolve from close to zero in a reasonably short time. No one has talked about developing a human, very far from that, but even my system shows that some simple behavior can emerge starting from almost nothing, and this happens in time which can be measured in hours, not to talk about years. How fast can it be, why not experiment, this is not so very difficult, and not exactly a rocket science. Yes, no one has yet modelled a single neuron realistically in a computer, i agree with you in this, and i'm glad that you understand it, different from these why today try to model a brain, without modelling properly even a single neuron. But the aim of the Artificial Consciousness is not "realistically modelling" a neuron, or anything else, it is rather about finding the principles how something works, and then making something which works by the same principles. Like watching how the birds fly, and then making an airplane which works by the same aerodynamic principles, but which is not a "realistic model" of a bird. Like the ADS system can also be considered as such kind of model of a neuron, very unlikely that it can be anything more advanced. That my system is pretty simple, doesn't yet mean that it is "useless", there is no reason for such claims to always be true. I would say that it is useful exactly for that reason, if we research something new, it is the most reasonable to start from the simple. Yes there are people who have been interested in that idea, and as that field is almost not researched, then there is no doubt that some day someone would start to research it, starting from the same simple systems, and some people may do that right now. I have made no "conscious" system, and i have never claimed anything such. The Artificial Consciousness is not about making conscious systems, if you are able to understand the difference between Artificial Consciousness and consciousness. As a conclusion, none of your arguments really holds water. I don't know how naive it is, but there is nothing naive in starting to research a new area, if that research is rigorous and the arguments are much better substantiated than your arguments ever were. Maybe you should rather try to put up and not miss the train
|
|