paros
Full Member
Posts: 4
|
Post by paros on Feb 21, 2008 18:59:32 GMT -5
Yeah paros here.
Well, it appears nothing has happened on this forum since about 2006. Pretty much the same thing applies to Demski's forum as well.
Someone is going to have to write up a reasonable description of the Absolute Dynamic System. I mean one with diagrams and how it responds to certain problems. Step by step. All I could find was some ramblings on a website where the background is this repeating purple fractal and I had to stop reading the ramblings because the fractal was giving me a headache.
|
|
|
Post by tkorrovi on Feb 24, 2008 11:22:28 GMT -5
Hi,
About Absolutely Dynamic System, the things are where they are. An effort would be necessary to create an unrestricted system, with several people joining, i have not done anything further also what concerns describing and theory, simply because i don't want to do all that only for myself. So it is a kind of opposite to what one may think, it is not about to wait and see what some people would do, but it is more about whether *you* are interested, if you are, then the things would go ahead. The aim is to create and research an unrestricted system, which is an Artificial Consciousness system, honestly by a serious and scientifically rigorous research. This *must* be possible, because everything in the world must be rational.
|
|
|
Post by tkorrovi on Feb 26, 2008 9:06:19 GMT -5
It sometimes seems to me, that they hate True AI, in AI Forums, this site there is more about their chatbots. I never underestimated the importance and usefulness of developing chatbots, for complicated chatbots, this work may be almost as complicated as developing linux kernel, just that the cognitive abilities of chatbots are too low, for them to be interesting for me. Which is not saying that i don't think that it is anyhow an unimportant work, to develop chatbots, it is just a very different field, from that which i want to deal with. I don't underestimate any serious programming, and appreciate all people, who do that difficult, and very important work, and it's pity to see, how much undervaluated are even the efforts of the real experts, such as debian developers.
On the other hand, I don't think that the principles such as embodiment, should be the subject of these forums. I don't talk about considering interacting with the environment as a part of developing AC, but the idea that some kind of embodiment (as they consider it most primitively, as a set of senses and muscles) is the basis of consciousness, is really whacky, no matter how many scientists support it. Sure a thorough research of what really happens inside a neuron or any living cell, would no doubt be beneficial for researching AC, but they don't go that advanced while talking about the paramount importance of "embodiment". What concerns Strong AI, then this was a strawman argument originated by Searle, to be refuted by him, no serious scientist should use that term, at least concerning any serius research. So, i sit in the forum where was no activity in almost a year, but it is a question whether there is any forum at all, where you belong.
Oh yes to say it here, the most advanced AC theory provided by any prominent person, seems to be that by Hawkins. But even this is likely not the solution to the problem. His theory is kind of weird, in that it is quite good and advanced theory what concerns the general principles, sometimes it even seems that something what Hawkins says, is taken from ADS-AC site, or from Igor Aleksander or Bernard Baars. But the system which he proposes for implementing that theory, is like day and night, and doesn't even seem to be compatible with the rest of his theory. It is like, he put together two things, which don't fit together. His system seems to be only about some memory system, which is restricted, and is not anyhow capable to implement all his theory. In spite that though, i think his general theory is very good, and i would recommend to learn this by everyone who wants to develop AC, but only his general theory, not his system which supposed to implement it, this would likely cause too much confusion and misunderstanding.
I'm sorry that i often write many maybe unrelated thoughts in my posts. It is because i'm used to express my thoughts quickly, often immediately after i thought them. This may result in not very elaborated comments. But this is often the only way in forums, where you have to say quickly what you think, to keep the discussion going, and take care that during the discussion, some important aspects would not be left into oblivion. In a way, this is implementing the "release often" principle known in open source development, in research. Of course, it only works, for friends who together develop something, not for the people who may base their criticism on a few wrongly written words, by mistake due to having not enough time to error-check the writings. Sure the posts in forums are not scientific papers, some likely can only communicate through these, but what they don't understand, is that any development would be much too slow that way, consider how many hundrieds of years it took, if linux kernel was developed only based on published papers, and not based on free discussion in the internet. Of course i understand that some may have different views about how the things should be done, and they are free to do them that way, but please let others to live also, and don't attack others in forums, for no serious reasons whatsoever, based on wrongly assuming things which they didn't even care to anyhow find out.
|
|
paros
Full Member
Posts: 4
|
Post by paros on Feb 28, 2008 13:46:37 GMT -5
My problem with Jeff Hawkins is that he is attempting to jump right into human-level AI without any concessions whatsoever to insect-level AI or reptile-level AI. This is one of the reasons he was rejected by the AI Lab at MIT.
Jeff Hawkins Hierarchical Temporal Memory requires "grandmother cells" which have been completely rejected by neuroscience recently. I'm not a PHd myself, but even *I* can see why the theory of grandmother cells is not correct. Memory works in the brain by connections between disparate regions of the cortex. The memory of your "grandmother" is spread out all over the brain and becomes more salient to your consciousness given that each seperate region is satisfied in its recognition of its particular detail of your grandmother.
If I may submit my own complaint here, This is pure 100% paros criticism, not repeated from a secondary source! Hawkins realizes that abstract concepts are dealt with in the mind by tags that I believe he calls "names". The "names" can thus be dealt with as corpuscular objects that are handled without all their underlying details. That's fine. And I have no doubt that this is correct. However, Hawkins claim that the very cells in our heads are organized PHYSICALLY in this manner is completely ludicrous.
Why is this ludicrous? Because the cells are not the information in our heads. And the way that cells are wired together is not equal to the concepts themselves. The concepts are actually in our heads because of firing patterns made by the cells, not the cells themselves. Do you see the subtle difference there? I guess my complaint is summarized as follows, Hawkins wants us to believe that the brain forms higher-level concepts because the neurons are already wired together in a hierarchy scheme. This is confusing form with function.
Hawkins never mentions how the brain creates correspondences between sensory modalities. And really, these days, if the guy doesn't mention this, I just dismiss him immediately. You can get angry and call me "stubborn" or whatever, but I've really been around long enough to know the earmarks of a true neuroscientist. The coordination of signals across sensory modalities is so fundamental to brains that even INSECTS do it!! If someone attempts to talk to me about the macro-workings of the entire brain, they had better mention cross-modal processing, or I just dismiss them.
|
|
|
Post by tkorrovi on Feb 28, 2008 18:55:02 GMT -5
I'm sorry, i have not much time right now, so i answer only briefly, and kind of superficially. First i'm glad that you want to talk seriously about Artificial Consciousness, i also hope that the others would join here.
I don't think that modelling the nervous system of the insect must necessarily be the necessary simplification, AC has to model the aspects of consciousness, like these stated by Igor Aleksander or Bernard Baars (the functions of the brain where consciousness plays a role). More primitive AC system implements them less, or less number of these aspects, but some systems may implement some of these aspects more, but some functions of brain even less than an insect.
My view of how the brain works, is that there must be some extremely dynamic system, inside a neuron, and this is the only thing which i have been concerned with, what concerns the basic mechanism which implements AC. I don't argue that Absolutely Dynamic system is any absolute truth, but science works by creating and studying hypotheses. What concerns Absolutely Dynamic System (ADS), then there is no pre-determined hierarchy or system, there must be a start structure, but all it has to do is just to tie the i/o nodes together. ADS is derived so that it was unrestricted. For a system to be unrestricted, the necessary requirement is that whatever system must be able to emerge, withtin a bigger system, during the development. Considering that, there are not many possibilities for a mechanism, which is unrestricted. The nodes and links of such system, cannot have much properties, because if they did, creating the mechanism which would enable to emerge whatever system, with whatever combination of these properties, would be a very complicated task, and with even a few more properties, most likely impossible. Therefore the mechanism was derived so, that the nodes and links have no properties, except that nodes can be old and new. Even with such simplest system, the mechanism is not very simple to implement.
So, there are no "words" in ADS, and not any other symbols, either. The mechanism is the result of logical and rigorous derivation, the result we get may be unexpected, but this is the result of the derivation, not what may seem to us natural, or self-evident. Everything in such system only consists of connections, there are no symbols. How the patterns and the working processes really represent information, is not entirely clear, the system is in the constant process of changing (creating and deleting nodes). But even the patterns of topology, and sequences of these, can represent information. ADS creates correspondencies between sensory modalities in the way that the structures which connect the i/o nodes, entirely depend on the details of the activity, of these nodes. We may even say that in ADS, the sensory input, together with output, can alone create unique changing structures.
Again i don't argue, that ADS is the only solution, just it is so far the only model for a basic mechanism, for which there are at least some theoretical reasons to consider, that it could be able to be the basic mechanism of an unrestricted AC system. Sure it's necessary to proceed from AC theory, and if there would be found another mechanism which may satisfy all the conditions, that's fine, or maybe some changes should be made to the ADS mechanism, this all is a matter of research. What concerns that insect again, then, i think that one possibility, of what may happen inside a neuron, is that there is a changing system of microtubules, and exactly such changing system can implement ADS. It is known that there is a changing system of microtubules, in every cell, only it is not known, how that system exactly works, and how it changes. If there can be a system like ADS, inside a cell, then i would rather say, that ADS then models the working of a cell, not even an insect.
|
|
|
Post by Trav's Buddy :) on Mar 18, 2015 8:33:13 GMT -5
Yeah paros here. Well, it appears nothing has happened on this forum since about 2006. Pretty much the same thing applies to Demski's forum as well. Someone is going to have to write up a reasonable description of the Absolute Dynamic System. I mean one with diagrams and how it responds to certain problems. Step by step. All I could find was some ramblings on a website where the background is this repeating purple fractal and I had to stop reading the ramblings because the fractal was giving me a headache. That's why its called Artificial Intelligence (AI/AC) in first place my friend... hehe! and my buddy trav an expert at that
|
|