Problems with Artificial Consciousness

[purplefrap]

The meaning of any term is only useful when the idea it is based on can be shared amongst individuals. The term 'consciousness' carries a lot of ambiguity. It can mean very different things to different people. For me, I've tended to see 'consciousness' as just a state of a system. It is the point where a system exists in the flashing that is the present moment. All systems have this nature. Humankind is just a little different in that we are aware of this flashing (at certain times). Realising this fact just means we need to better define 'awareness' but as we do this we'll find ourselves drifting further down tree (or some would say, round in circles) looking for the term that is the key to understanding exactly what it is we are trying to create and the secret behind human nature.

If we are looking to create systems that are 'conscious' then I would tend to believe we're putting the cart before the horse. We are not trying to create a mechanised conscious being - we are looking for something more. In my mind, as I mentioned before, all systems have a conscious state so we should perhaps look at the fundamentals of what we are trying to do in the first place. Are we trying to solve particular problems with our creations? Are we just trying to create something more intelligent than ourselves? Something that can answer our questions for us and help make the world the idealistic paradise we want it to be? Or are we looking to build something that allows us to put our feet up and let it bring us immortality?

What is it we are trying to build?

Cheers,

p-frap

[tkorrovi]

adsproject.sourceforge.net/ads-ac/index.php/Artificial_Consciousness the axiomatic approach, and awareness of processes. And, as already has been said many times, this is about Artificial Consciousness, not about conscious systems, there is a big difference in that Artificial Consciousness is only about modelling these aspects of consciousness which are objective, ie not about implementing the subjective experience, and therefore not about making conscious systems. But why we call it Artificial Consciousness is because intelligence is by most of its definitions too narrow term and therefore too restrictive, we should proceed from awareness in its wider sense. And this forum is about creating unrestricted systems, ie the systems which are derived and aimed to be unrestricted in every way, to solve the problem of generality defined by McCarthy, implement the type 0 grammar defined by Chomsky, and being able to implement all the aspects of consciousness necessary for Artificial Consciousness. Please also see this project adsproject.sourceforge.net/ads-ac/index.php/Main_Page as the wiki there is not only about the systems which aim to be absolutely dynamic, but also about Artificial Consciousness in general. I'm sorry that the documentation is still a work in progress, there simply are no people who would have enough time to do that, time is money as it is said, and we don't have neither. Also, a full documentation of everything is almost a hopeless task, as there likely is no field other than AI where there are so many different solutions and different systems, which is because if the solution has not been found, then the human mind moves to immense number of different directions. Therefore there is necessary a criterion, about what solutions to accept and what not, and this criterion here is that there should be at least some reason to consider that the system supposed to be unrestricted. This is the only way to have one approach, and to avoid going to immense number of possible different directions what concerns the attempts to create True AI. Well, otherwise the task is not so immense, as it is about research and not about creating any very advanced systems.

PS This is the paper which defined Artificial Consciousness as a field of study, BibTex dblp.uni-trier.de/rec/bibtex/conf/iwann/Aleksander95 , Internet Archive web.archive.org/web/19970302014628/http://www.ee.ic.ac.uk/research/neural/publications/iwann.html . This paper also provides several aspects of consciousness which Artificial Consciousness has to model, somewhat similar to these suggested by Bernard Baars, and also includes the ability to predict.

[purplefrap]

I'd just like to point out - "Aspects of consciousness that are objective" - seems like a contradiction in terms. Nothing is objective. Everything we have is based on a form of perception. How can anything be objective if it can only be known when it is perceived? Or to be put another way, if something is not perceived how do we know it exists?

So I take it that your answer to "What is it we are trying to build?" is a really dynamic and flexible system for purpose that will be considered later on or indeed, it will be able to solve all problems - period.

There are obviously a lot of problems in having a system based on solving all problems. From what I gather, ADS would not be well suited to critical systems if it's going to decay nodes fairly randomly. Boot-strapping is also a long process, the results of which must be checked to insured they are correct. I'm guessing not a lot can be carried over from boot-strapping process to boot-strapping process hence you are starting at zero point each time. This is identical to traditional problem solving techniques where one starts with defining the problem and figuring out the best way to interface with the problem to develop a solution. It seem likes ADS is just a convoluted way of performing Test-Driven Development.

I'm also thinking that I can't see how this could be accepted to solve all problems if it up to humans to ensure it has reached a solution. There in itself is one form of an injection of subjectivity. In the case of the game Nim - how does it know the point of the game is to win? If it has been instructed, then subjectivity has come in to play.

I'd be curious to hear your thoughts.

Cheers,

p-frap

Artificial Intelligence

[tkorrovi]

If you say that nothing is objective, then there cannot be no science, science is about theories which are aimed to be objective, and are objective to some degree. The answer is that everything is interconnected, something exists only when it is connected to everything else.

No we can never solve all problems, the aim can only be to create a system which has a potential to solve as many problems as possible. This seems to be merely another version of the argument, that when we want to create consciousness, then we supposed to know everything. No, the aim is to create Artificial Consciousness, not consciousness, and i already explained that there is an essential difference. We don't have an aim to model *all* aspects of consciousness, but only these which are known, are essential, and can be considered objective, and not necessarily even all of these.

No, the information remains, even when the nodes decay, as the system renews itself all the time, the new nodes would contain the information which the old ones did. Look, some areas of your hard disk or flash memory may wear out as well, in whatever locations, and not necessarily any information would be lost. I think it is impossible to create any system which aims to be unrestricted, without some way how the old data can disappear, as this old data would otherwise become an obstacle to the development of the system, its whole structure may once not even correspond any more to the rest of the system. Well, the principle of ouroboros if you like, but the principle of ouroboros is not necessarily wrong only because it is known already from ancient times. No, this is not just a way of performing test-driven development, this is completely wrong, it works very differently and has much more potential.

Well, concerning that decay, if we look at the other self-developing systems, such as cellular automata, then they seem to be perfectly stable, but this is not true if we look what happens in such systems when they really are self-developing. Though it is difficult to imagine a cellular automaton to be fully self-developing, where these different versions of the patterns there supposed to be, in that simple 2d environment? When the patterns in a cellular automaton self-develop, then most certainly certain patterns which cannot survive in their environment, would decay. How can it even be thought that a system can be fully self-developing without such decay?

I said that the aim is not to solve all problems. There can be an objective criterion to determine whether a problem is solved, science works based on objective criteria, at least based on criteria which are as objective, as we can find. Therefore whether the solution has been reached doesn't necessarily depend on humans subjective decision. Science cannot work based on subjective decisions, and therefore also the Turing test, the way many like to interpret it, is a completely wrong approach, we cannot decide about the performance of the system only based on subjective decisions. You are right that the system doesn't know whether it has to win or lose, the systems which aim to be unrestricted have no pre-determined goal, and they ar not built to achieve any particular goal. What ultimately determines whether the system has to win or lose, and when it wins or loses, is the training program, which is usually created by people of course, but the game has objective rules, and the training program plays the game with the system based on these objective well-determined rules, and this is why subjectivity doesn't come in to play.

It is also important, that objective tests of a system must include some information about the system, either known or obtained during the test. I think the biggest problem with Turing test is that Turing didn't consider that something has to be known about the system, one supposed to remotely interact with something which he doesn't know what it is, and make decision based on the system's responses, while no possible questions and responses proposed by Turing were to find out the mechanism how the system works. We cannot decide about the system based on such test, because it is possible to pre-program all the responses and such test would not therefore necessarily indicate any ability to think.

I should finally say that such kind of research needs a very deep philosophy. The biggest problem is that everyone considers that he can think, but when encoutering problems like these, it may occur that one is not able to think, and this always causes wrong decisions concerning any research like this. And it is not only about being able to think all that, but to find the solutions about which one can be reasonably certain that these are the only possible solutions, when the aim is as strict as creating unrestricted systems. This is much more than just being able to think philosophically, and being able to say something, this philosophy must result in systems which work in the computer the way how they are supposed to, this is much more difficult than assuming that everyone can think something about consciousness or whatever, and therefore almost everyone is capable to research such things. This kind of research needs much more thinking and much more responsibility.

[purplefrap]

Objectivity is the pipe dream of science. The problem is that every concept in science is based on an abstract concept. Abstracts by their very nature leave out a certain levels of detail. As soon as details are left out - a subject becomes subjective. Even the number 1 is an abstract concept and leaves out a level of detail. For instance 1 is said to be equal to 1.00. They are one and the same. The truth is that 1.00 contains more information than 1. It tells me for instance, that no rounding has taken place and that value is perhaps more pure. As it turns out, the number 1 doesn't exist in nature unless you begin speaking of infinity. The number 1 doesn't exist in nature because no element lives in isolation to be able to be given a value of 1. Everything is connected to something else and as soon as you call something 1 you begin ignoring its true existence. The reason we can say things like "there is one person over there" is because its an approximation that carries an understanding that assists survival.

Science has progressed in an amazing ways by ignoring the infinite (I know it's mathematics but look at the division by zero problem for instance). It sounds like ADS is headed in the same direction. I'm trying to go the other way and I'm not even sure if I can build on anything more than my philosophies but I'm headed in the direction of pure open-mindedness. More thinking is needed before even contemplating coding.

This Universe can be perceived in many different ways. Science has just chosen its perspective with a huge foundation behind it and science is sticking with it. Looking at the history of scientific knowledge one will see a life cycle associated with each discovery. At the time something may appear like an objective truth only until cracks begin to show and it is replaced by something that appears superior. Each discovery will only last until a better approximation takes its place. No scientific discovery will represent the true nature of the universe unless the the theory of everything is found.

... having said all that, one does have to start somewhere.

Cheers,

p-frap


Artificial Intelligence

[tkorrovi]

Well, your logic is in a way right, but has not so much to do with the things like ADS. The paradoxes in mathematics occur because mathematics itself is restricted, the main reason why mathematics is restricted is that it can mostly only model a system of connected objects, where the topology of that system and its connections doesn't change. A system is more general than numbers or other such mathematical objects, which are mostly only artificial. And sure everything is changing in nature, therefore such system is generally changing, including its topology. ADS is such kind of system, and is for these reasons less artificial than mathematical abstractions, and therefore there is no known reason why it should head to paradoxes like division by zero. ADS is also not so absolutist as it may sound, the name is also provisional, it is a system derived and aimed to absolutely dynamic, it is very difficult to prove whether it is such, so it is just a system which is as dynamic as possible. Sure there always would be better approximations, when we would know more, but even the existing ones are right in some way. Concerning the approximations, it is important that they are as right as possible in the most essential, in that sense the approximations like artificial neural networks might be bad approximations, because they leave out some very essential detail, the ability to be aware of processes, as the ANN can only recognize static objects. Such approximations can be questionnable, and lead to totally wrong results. I currently know nothing essential which systems like ADS leave out.

[purplefrap]

It sounds like you know what you are talking about and are happy with the system you've put together. That's great. When you say thing like "Well, your logic is in a way right, but has not so much to do with the things like ADS" shows you're not terribly interested the nature of philosophy because your mind is made up and you seem to already have the perfect solution. You just happen to be at that point of trying to convince everyone of your philosophy and of your system. I understand you are trying to get word out about ADS and perhaps find more people to work on it or contribute to it or simply to recognise it. You sound like a clever guy. My only only advice would be to keep a very open mind, attempt to take on as many perspectives as possible and realise that it is only when you understand the nature of intelligence and the nature of the Universe itself, that the meaning and the essence of a project in AI will become clear.

Good luck and I sincerely wish you all the best.

Cheers,

p-frap

Artificial Intelligence

[tkorrovi]

Well, ADS and this Artificial Consciousness is that which i do, my aim is not to convince people, but to find the solution. If you do something in physics, then this is certainly not what i do. Some people seem to come here just to debunk everything, and they have always failed, this is a useless activity, because my aim is to find the solution for True AI, the same which their aim supposed to be, so we don't supposed to have different interests. And still what i'm used to is endless arguing, which is not at all what i want, i would rather say that i had much more benefit if i didn't ever create any AC systems, all that always appeared to me more negative than positive, and there are very few people who support me (though there are these), in spite that what i want to do supposed to be the same which many others want to do. I'm an unhappy man, not clever at all, i would rather say that trying to create any True AI system can considered to be the most unclever, except perhaps in the circumstances in which i did it. Well, i didn't expect anything better from the moment when you came here, i have seen that so many times before. Maybe one question to you, is there a single thing which you appreciate in all the work which i did, or isn't there anything at all? I wish you good luck too in whatever you do, and when you once would be interested again in Artificial Consciousness and unrestricted systems, then you would be welcome here again.

[purplefrap]

If you didn't expect anything of me when I came here I suggest you retrace your steps...

"First, Strong AI is a misnomer, this term was coined by John Searle only to refute it, and no matter in what sense one may want to use it, what defines the term is its original meaning, and by that this term is a strawman. It is better to use True AI instead or even better to talk about Artificial Consciousness..."

If that's not trolling I don't know what is.

In answer to your question...

"Maybe one question to you, is there a single thing which you appreciate in all the work which i did, or isn't there anything at all?"

I have respect for you and the all work you have done. I've have not coded one line with AI in mind so you've bested me by a mile. It's almost too hard to know where to start. It is true that I could join a project and help out in one that already exists but as you know - time is so valuable. I already spend way too much time on my laptop and I going to be d**n careful before I tie myself up in another project.

I think ADS has the components for true intelligence. However, ADS is not the only one. AI is a weird field in that the desired outcome for a project (especially from a bottom-up approach) becomes so broad that it becomes hard to measure. Flexibility isn't appreciated in a system because in finding any solution we already start with a huge amount of flexibility (e.g. which language to use, etc) and all we are after is the path of least resistance. It is the time to delivery that runs this world and that's why the top-down approach suits the world we live in. Everything was needed yesterday. I am a believer that bottom-up approach will produce a more natural form of intelligence - unfortunately this world is moving away from natural traits. But that's not to say the game is over.


Cheers,

p-frap

Artificial Intelligence

[tkorrovi]

"First, Strong AI is a misnomer, this term was coined by John Searle only to refute it, and no matter in what sense one may want to use it, what defines the term is its original meaning, and by that this term is a strawman. It is better to use True AI instead or even better to talk about Artificial Consciousness..."

Why is that trolling? It seems that you misinterpret something, this is true and this is what i meant, Strong AI is a misleading term to use, as it can be understood in its original meaning. Read the John Searle's Chinese Room, where the term Strong AI was first defined, this is a completely pre-programmed system. John Searle himself refuted that there would ever be possible to make such system, and this is true. Thus this is not what i mean by True AI, and i also think that this is not what others mean, so it is not good to use the term Strong AI, if one wants to avoid any misunderstanding.

"Flexibility isn't appreciated in a system because in finding any solution we already start with a huge amount of flexibility (e.g. which language to use, etc) and all we are after is the path of least resistance. It is the time to delivery that runs this world and that's why the top-down approach suits the world we live in."

Well, this is the approach when we want to create an AI system for some particular purpose, capable of implementing only some restricted task. Trying to research True AI i don't deny the usefulness of that approach, this is simply a completely different thing. And the usefulness of such practical approach no way proves the uselessness of the research of True AI or unrestricted systems. It seems though that people are mostly not satisfied with creating AI only for some restricted task, and not being able to find out what to do, has been exactly the reason why many of the AI projects have stooled and failed. We need both, the same as we need both engineering and fundamental research like physics. I see Artificial Consciousness only useful in the foreseeable future as a theoretical research, and i don't think that this can be overcome with any wisdom, as this is the very nature of such research. But such research would be very useful, the way how a basic scientific research is, giving us more knowledge about the implemented aspects of consciousness and intelligence, thus also helping to improve these practical AI applications. I think that one reason why AI has not moved forward enough during many decades, is exactly that this is what we should expect in a field where there is almost no fundamental research. All the computer science is considered to be an "applied science", and Artificial Intelligence is considered to be a part of it, thus there has never been and would never be while this approach remains, any funding for a fundamental research related to AI. And naturally then, there would be almost no fundamental research, and naturally the result would be stumbling in the darkness, what the AI in many ways is now. When there is not much general understanding, then human mind goes to immense number of different directions, to find the solution, and indeed it is difficult to find a field other than AI, where there are so many different approaches, and an immense number of very different systems made. For True AI research i have a simple criterion about all systems which pretend to be True AI, and it is whether there is any reason to consider that they supposed to be unrestricted, and almost all of them fail. True AI is what i deal with, so i cannot decide about the restricted systems, but sure there also can be better criteria. Well, this is about the necessity and importance of a fundemental research in True AI, for which there is no way to get any funding whatsoever today, and also not in the foreseeable future.

[purplefrap]

I know you meant what you said - I don't doubt it.

It's interesting to see that you even hinted that you know Strong AI is a synonym. And as is the case with most words, one can look at the context in which it is being used and determine its meaning. Every time I've used it, it has not been in reference to the Chinese room. I'm not even sure why you'd think that's the 'Strong AI' I'd be referring to. I don't know why this term is troubling you so much.

You seem to hold the term AC very close to heart. I get the impression you'd like everyone to use this term. You feel that it has less ambiguity than other terms. In some way if you can convince the world of this term it will put everyone on the right track and we can all focus on the 'objective' points of consciousness. That my friend, is a massive task and this industry is always going to have a massive amount of ambiguity. I'm starting to think the ambiguity is there for a reason. There is nothing concrete in this field and it's all very nebulous. Fundamental research will only open up a small number of paths to True AI. The truth is, paths to the solution are limitless hence so many approaches and projects going in different directions. As soon as someone tries to lock down a fact about intelligence an example can be found that contradicts it and leads them back to square one.

No one approach to AI is correct. All have their value. You are holding on the unrestricted nature of you system in the belief it is True AI and only path to it. It could very well be a case of True AI. However, even a restricted system could be considered True AI the problem is - it is based on the eye of the beholder.

I can also say the your system is only as unrestricted the same way a blank piece of paper is unrestricted. One can draw on it and rub parts out a draw on it again and again but both systems have their limits. It's not the piece of paper that should be the focus - it is the usefulness of what is drawn on the page that is important - even if it's just a measure of neatness so I can hang it on my wall.

A factor I can see you don't like is the subjectivity associated with AI and intelligence in general. You mentioned the Turing Test is flawed because something needs to be known about the system to show that it is intelligent. That's rubbish. For instance I would say that the Chinese Room is intelligent judging on what I experience with it. I don't need to dissect it because in the end it makes no difference. I don't care what is behind the door - it may fascinate me but I don't care. My calculator is intelligent as it give me a meaningful result when I use it. Most people would say a calculator is not intelligent because it has a limited capacity and can't 'think'. If I created a Turing Test that just had questions based just on arithmetic and performed it on both a calculator and a person, I could easily get the same answers from both parties. The only way I could tell which is the human would be if I got back a incorrect answer as a response. It's funny to think that the intelligent being would be the one that got an answer wrong. Why should I say the human is more intelligent? In the end all that matters is what is useful and what is perceived.

The major problem with the Turing Test is that it has such a limited scope and animals that aren't human would have no chance to show their intelligence.

Perhaps you've realised that intelligence is a fuzzy term and have moved on to consciousness instead. As I mentioned at the top of this thread, consciousness is a majorly ambiguous term and it doesn't change the situation at hand. In the end we all looking for solutions. We just need to figure out what it is we are looking for.

Cheers,

p-frap


Artificial Intelligence

[tkorrovi]

Well, that's your opinion, it seems to me that you didn't understand several things which i said, but of course everyone can understand things in his own way. What concerns Strong AI, then usually every term is used in its original meaning, how it was initially defined, and this gives everyone a full right to interpret Strong AI as it was defined by John Searle, and as you admitted yourself, this is not what you mean. Well to change some understanding etc is always a big task, i never wanted to take such grandeous task, in fact i was terrified after i encountered these problems of understanding after i created my system, and no i'm in that position just almost against my will. But there is nothing which i can do, because if i did agree that whatever kind of complexity leads to consciousness or whatever, and that all what is necessary is only programming and even more programming, and the basic mechanism of a system has no importance whatsoever, then it would be equivalent to admitting that everything which i did is worth nothing and is utterly stupid. Well, i have no reason to see it that way, maybe in the present world i was happy if i did, and i had to be against myself if i did agree with all these musings, which would be the most unnatural. I don't think it is necessary, and i see no reason why all such arguments supposed to be correct.

PS You added one of my comments to your Artificial Intelligence blog, and moderated out another. No i have no complaints, this is your blog and you have full right to write there whatever or delete there whatever. But look how much more i appreciate the posts of the others and the opinions of the others. In addition to some spam sent by bots i have deleted only one post in this forum, and this was a post where the one who attacked me asked whether i'm already suicidal, thus clearly threatening me, this incident was clearly connected to banning me from AI Forums. So at least you should see that i'm more open-minded than you, now already more than ten years i have listened all possible and impossible arguments against my system, and against all the rest which i have ever thought, with none of them succeeding to refute anything, though of course they were trying. From these arguments i have learned everything from AI which ever may cause to think that my approach may be wrong etc, yet i have not found any convincing argument against. I would be the most interested if there was such, it would be possible to find out how the things are then, and this would bring us one more step closer to truth. But yet again, this forum is about unrestricted systems, and is therefore for these who think that this is the way. I welcome any criticism but still this forum is intended to be a place for people who support each other in finding ways how to create unrestricted systems, and not only for arguments against. Any good arguments are welcome though.

[purplefrap]

I haven't meant for my words to be harsh or to put you down in any way. I don't think what you are working on is futile. I feel a bottom-up approach will produce or more natural form of intelligence and something we will be able to identify with more strongly. From what I can gather of your project the key is in the bootstrapping, as it looks like you already have the unrestricted foundation. I hope that a lot can be taken from your current training processes and used to create more training processes for more complicated environments and tasks. This will enable you to show others the power of your system. It's when people see results upon application that they take notice. I certainly don't think you should give up.

I'm sorry to hear your about your ban on AI Forums. Surely you'll find a way around it.

The reason one of your two comments on my articles was moderated out was because it didn't relate to the article. The only part that did was your slamming of the term Strong AI and even then, the article was about approaches to AI - not what constitutes the term Strong AI. I would have let it slide if it was just critising the term but the rest of the comment was talking about synergies which was referring to my last post in ai-forums about intelligence in ant colonies. I'm more than happy for people to critise my work as it helps keep me in check. You are more than welcome to comment on my work again, favorably or otherwise, as I do value your opinion. I will only moderate out spam and stuff that's off topic that refers to my activities elsewhere.

Anyway, it seemed like more of a personal comment to me saying to join you here if I was interested or intrigued in what you had said and if wanted to discuss it further.

Good luck!

Cheers,

p-frap

Artificial Intelligence

[tkorrovi]

"It's when people see results upon application that they take notice."

Well look, the result which one can get from systems like ADS, and other similar systems, are some rows of numbers and statistics calculated from these, which certainly would not fascinate most of the people. In AI there has already been too many grandeous claims, so any true claim would not make anyone any more wonder, and just a words from an unknown person when they don't understand the evidence, would certainly not have much effect. No such project can become widely known to many people because of popularity. Any such projects are basic scientific research by their very nature, and expecting that they would become popular because people would become excited from using them, the same way as they become excited while playing video games, is almost the same like for example to expect that people would start to like the theory of relativity only because the equations look so very nice for them, that they would prefer to look at these equations instead of watching movies, playing video games or whatever. One first has to know quite a lot to understand any importance, thus only the very few who understand can appreciate such things, and most of them have no means whatsoever to make such things anyhow widely known. So popularity is certainly not something which can help such things to go ahead.

[tkorrovi]

Well this "problems with artificial consciousness" reminds me one conversation. Someone who again was against me, for the reasons i have no idea about, once said that if this what i say would appear to be right, then we have lost. So i understand that this would break someone's dreams, and this may be why many don't want to hear anything about unrestricted systems and such, or want to desperately argue that true ai doesn't have to be or cannot be unrestricted etc. And the dreams likely are to create some very advanced universal AI system by conventional coding. All such attempts have failed in the past 50 years, but the dreams once created, likely don't end so easily. One should realize that there can be other dreams, one can use all the programming skills for an unrestricted system and research as well, and perhaps even more, as i have realized that the programming necessary for unrestricted systems needs the most advanced methods of debugging and testing, such as thoroughly analyzing the whole system in every step, something which is almost never necessary for conventional programs. Also most generally, any systems with changing structure are difficult to code, as one programmer said that he cannot do that because he cannot imagine it. So such things need some advanced programming, and sure there is a lot to do. Not a worse challenge at all i think and very far from "losing", but it seems that many people see the things that way. Well, when talking about hal, many are interested, many of their systems are called hal. But when i sometimes have mentioned vanamonde, no one seems to be interested, in spite that this is about the writings of the same writer. So there are no these kind of dreams. I dreamed about creating something like vanamonde, not any kind of physical features at all, but the way it works. Thus, it is also me who wants to know, what are the problems with artificial consciousness.