Talk:Absolutely Dynamic System

From ADS-AC

07/02/05 There are no significant results yet, of learning the Nim. tkorrovi

07/02/05 Updated the "It's alive, it's alive!", "Training" and "Learning" sections laurim

06/27/05 Updated the Training and Learning sections. laurim

06/26/05 Tweaked some grammar in the text and managed to mess the pseudo code. Sorry! laurim

06/26/05 Wrote the text. laurim

06/26/05 Neighbour nodes of A -> neighbour nodes of A and Bn. tkorrovi

06/26/05 Filled in the places left for me (initially) -- "Training", "Learning" and "An example implementation". tkorrovi

06/26/05 Removed a section about deleting the old nodes, and added this deleting as an else statement in a pseudo code. That's more correct, though restore if I didn't well understand the necessity of a separate section... tkorrovi

06/26/05 Guess this "else" got lost from the pseudo code. Oh, was it that it deletes all first, and then goes to a loop to create, kind of forgot... But anyway, this deleting has to be somewhere. tkorrovi

Again, what is it all about

In order to try to explain once more, what is it all about, I would like to add a quotation here of what was written to me by someone, after studying the source code of ADS-AC. I'm sorry in that I never write quotations from the letters written to me without a permission of these who wrote them, this is why I didn't say who wrote that, and will take it off if requested. But this time I found it important, to make it clear for all others, who may not understand what is it all about. So, if I could translate it correctly enough from Finnish:

"Your system is not restricted even in that, the knot-link system is completely freely developing, completely dynamic, completely unrestricted. Congratulations, you are a tough type, even because you managed to develop just that so stylishly. :)"

Thank you indeed!!! But why I put this quotation here, is that I want to show, what others find important in ADS-AC, and what exactly is that what I did. Just if I said it myself, these three "completely" did sound nothing more than an exaggeration by myself of what I did. But this is the essence of what I did, and this is all what I did. The quotation above says exactly what there is at present, and what there is not. This is why I found important to add this quotation here, and also to show that some people appreciate that, to show what about the system is important. Oh, not that I am important, I am in no way more important than what is done. I did it because it is a kind of fault of some humans not to stop before the solution is found (recommend everyone to avoid that if it only was possible), and all I want is to see what such system can do, or where the theory goes. Oh yes, to be perfectly honest, I did once also dream about a little money, which would compensate my long efforts and maybe give me a possibility for some time to live my most modest life (I don't have a car, I just want to be alive, not always though), but not recently any more. And to know myself that I did *something*, let it be very little, what others didn't do. This I said to be perfectly honest, as some may want it, there is just nothing more to confess. And, if you anyhow feel that such kind of system is not something you are interested in, better don't become anyhow involved. I don't really know what anyone may think about me, eh, I'm nothing special, it makes me sad to see how much people have to suffer everywhere, I'm not stronger than that, but I can live with all that program and everything related to this, and I'm not sure that you can.

But OK, enough of that, talk about the thing. So, how it is said in the quotation, all there is is a system, which is "completely freely developing, completely dynamic, completely unrestricted". This is all there is at present and all what I did. There is nothing more. And this is all what I can offer. And now it's necessary to see how such system can be used for Artificial Consciousness (or for AI if one prefers). This is what it is all about. The program which implements that system works perfectly now and there is a source code. That's all. It's for you to decide what can be done with that, or can such system be useful at all. I hope this helped to explain what this project is all about, thank you for reading this, and I would be thankful for anyone who would contact me, who have any ideas, even about how to explain all the things better (as I know these are very difficult to understand for many), or just ask questions. My thanks to all these dozens of people who wrote me, and with whom we discussed these issues sometimes very long. Thanks again, you are wonderful people, you helped a lot, and this did give a lot, for me at least. Many, many, many thanks. 07/09/05 tkorrovi

Learning Nim

Yes it is the most important now to train it to play Nim, or then some other such simple game. I guess that some 56% in 100 000 games would be enough, I saw such in the Princeton pear project that 5% more than random is sufficient to show that it is not just a random result, but this must be studied a bit more. 100 000 games seems to be enough for a whole "cycle" for it, where the percentage goes from lowest to highest, and back again. Unfortunately now I even cannot run it, it's too hot, and my computer hangs too often. It's not the best computer for that, I think there must be computers 10 times faster than mine today. And just to write it somewhere, this is how I run it: I have two files, older ends with 0, and newer with 1. Once a day I delete that with 0 and rename that whith 1 in the end to that with 0. Then I wait until it outputs a space, and stop it then, so that a game would be properly ended. Then I run analyzer, and save it as 1 in the end. Then I run it more a bit, and if it outputs a letter immediately, then I usually save it again after another game. Then I see the results from nim.txt, and run it again, load structure file with 1 in the end of its name, and nim training program. It is in fact much simpler than it seems, of course that all could be easily automated, but I don't not yet want to make the user interface more complicated, than it already is. Of course the biggest question is, whether it ever learns to play the game of Nim. Seeing how it develops, and from the experience of training it repeating a word, I would say that it would go somewhere in a few months (on my computer). At least, it certainly would go somewhere, achieving some kind of more regular behaviour. This might be a successful playing of the game of Nim, but at least we would see then where it goes with the current training program, and based on that, we most likely can figure out a way how to train it to play a game of Nim, even if the first attempt was not successful. So I think finally it is anyway possible to train it to play the game of Nim (or some other simple game), in spite the time it takes, this is not just a hopeless trying, and the importance of achieving that is very big. This is the only way I know now, I have no idea yet how to achieve the result with less trying. But I'm the most sure that it is not hopeless, such thing can learn, so we only nead to figure out how we must train it to achieve the behaviour which we want to achieve, it must not be very difficult to figure it out, when we see how it develops. Then, when this step is over, we would have a way how to train it almost anything. Yes it would be terribly slow, but it is important to go it through, then we can figure out all ways how to make it faster, there are certainly many possibilities for that. And then one day, we would have something, which is not so "hopeless", at least enough to study further, how such things learn etc.07/20/05 tkorrovi