[goyabee]

First of all, I would like to explain that this post is mostly going to concern what most people would call " artificial consciousness".  Personally I think the term "consciousness" is probably the most misunderstood and abused term in the fields of psychology, neurology, and AI research so to preface the discussion, I'd like to specify that by "consciousness" I mean the system in the human brain that controls motivation, allows for the construction of perceptual models, and controls executive processing.  I am not a dualist, and as is consistent with evidence, I do not believe that consciousness is anything more than a physical phenomenon in the brain.

Developing an artificial model of these systems in the brain has been a goal of mine for a long time, but as it happened in my lifetime I have failed to get involved with any existing projects and my programming skills have never reached the appropriate level, so I just sort of assumed that someone else (someone with superior programming and systems designing skills) will eventually come up with the same design philosophy within a few years and AI will finally get started.  Well, it's been a few years now that I've been checking up on AI and consciousness research and I found that people still seem to be very much in the dark ages to the point that the dualists are still able to get away with their ridiculous claims and no one has really been able to tell them how wrong they are with any considerable degree of certainty.  So, having recently come across what I consider to be new and important insights, I have decided to seek out a forum of communication, and share my ideas, hoping to jumpstart someone's creativity enough to help get some models built and tested.

It has become obvious to me, and hopefully to anyone else who takes their involvement in AI research seriously, that step one in developing a model is understanding how consciousness works in the human brain.  Here, I'll describe a few principles that I believe to be the central mechanisms of human consciousness.

First of all, we are always hallucinating.  The brain is made up of memory systems and processing systems, one of these memory systems is responsible for holding the information which corresponds to our current perception of reality.  This memory system is fed with information which is generated by a separate system of the brain which has access to sensory information.  All parts of the brain that respond to the world around us do so by reading information out of this memory system which contains a model of reality, not by accessing sensory information directly, hence everything that we "experience" is a hallucination generated from actual sensory data, and everything that we think we are experiencing is actually a prediction of what a particular system in the brain thinks will happen next.  Once the model is read out by other systems of the brain, the accuracy of the prediction is tested against real sensory data and the next prediction is updated to keep it in-line with reality.  This theory of brain function is consistent with every phenomenon in which a human enters a hallucinatory state, namely sensory deprivation and drug induced hallucinations.  E.g. sensory deprivation deprives the brain of data with which to update its predictive perception model, and drugs such as LSD inhibit the system in the brain which is responsible for performing the update.  This causes the algorithms that generate the perception model to run wild and start performing operations on redundant data much like a fractal generation algorithm, hence why fractal hallucinations are so common in these states.

That is what I believe to be the central mechanism behind our brain's perception system, now I will explain how I believe the executive processing system may work, and then I will touch a bit on the importance of implementing dream algorithms in the development of a functional AI consciousness system.  The most important aspect of the brain's executive processing system in my opinion is a continuous assessment of pleasure and pain.  These don't have to be the words that we use but the idea is that the concept of pleasure and pain is intrinsic to the development of motivation and therefore a necessary element for autonomy.  Pleasure and pain are a simple and effective way for the brain to decide what it wants to do and I believe it may be the only effective way to develop true systematic autonomy in machine intelligence as well.  I believe the implementation should be very straight-forward so I won't go into great depth to explain the process.

Okay, so finally, what is so important about dreaming?  In human beings, dreaming is an essential way to accelerate the learning process.  It is a way to press fast forward on experience and develop strategies for how to react to stimuli without having to wait for them to happen in reality and without exposing the body to the dangers of real-world experiences.  Certainly the exact way in which the process of dreaming makes use of the same hardware in the brain that is normally used for waking consciousness is something that will have to get ironed out as we go along but I think the basic concept is fairly straightforward.  It must be a state of mind similar to the sensory deprivation or the drug-trip states that I mentioned earlier, in that the brain is cut off from all sensory data.  The brain shuts off the systems which have access to real-world sensory data and most likely a previously unused system is activated in its place, providing a secondary virtual perception model fed with data from memories which is used to generate world data (mock sensory data) to feed to the primary predictive perception model that is normally used during waking states.

My intention in providing what I hope to be valuable insight is that they may be used to develop functional models in such a way that we may test their function and then modify and add systems as needed in order to avoid the need for expensive brain research to confirm every aspect of what I believe to be a basic universal system which simply cannot work any other way.  In other words, I think the way the human brain works isn't just one possible way for it to work, it is perhaps the only way and indeed an effective and efficient way to achieve valuable cognitive phenomena such as learning and creativity.  Evolution is known to create systems with none other than the highest possible degree of efficiency.

[ranch vermin]

LOOKS LIKE THIS THEORY IS GETTIN' AROUND DUDE!!!

Just like EVERYONE tells you on game programming forums,  no-one is gonna write your game for you, same goes for ai.

But im different, (not much different tho, dont get your hopes up.)  I know heaps,  and your self dictation sounds good,  I didnt read it after you said you were doing no programming, or electronics,  but I read the start,  but I think making consciousness is impossible, its like trying to make god,  and its best left to (mad) biological scientists than us programmers and mathematicians.

I think what you have in mind, isnt conscious,  but its a machine,  same as mine.   Dont be foolish and think your brain works this way,  in a way less diabolical, as a tank or helicopter is diabolical,  but in a way more,  because of the fact we have consciousness, something it doesnt have.   and all we think of... is just superfiscialities,  that work as machines, but are not us.


I can lead a beginner to the easiest way to make a game, and tell people fiendish simplifications to ai, that ive prooven to work myself, but theres no HOPE for a teacher with a lazy boy,  I can make it SO EASY, like darth vader,  but its worthless occupation trying to get a lazy boy to do work, no matter how easy you make it for him.


Check out my latest program - automatic 2d to 3d,  guess how easy it is to build a "perceptive model" with this baby!!!

[goyabee]

LOOKS LIKE THIS THEORY IS GETTIN' AROUND DUDE!!!

Just like EVERYONE tells you on game programming forums,  no-one is gonna write your game for you, same goes for ai.

But im different, (not much different tho, dont get your hopes up.)  I know heaps,  and your self dictation sounds good,  I didnt read it after you said you were doing no programming, or electronics,  but I read the start,  but I think making consciousness is impossible, its like trying to make god,  and its best left to (mad) biological scientists than us programmers and mathematicians.

I think what you have in mind, isnt conscious,  but its a machine,  same as mine.   Dont be foolish and think your brain works this way,  in a way less diabolical, as a tank or helicopter is diabolical,  but in a way more,  because of the fact we have consciousness, something it doesnt have.   and all we think of... is just superfiscialities,  that work as machines, but are not us.


I can lead a beginner to the easiest way to make a game, and tell people fiendish simplifications to ai, that ive prooven to work myself, but theres no HOPE for a teacher with a lazy boy,  I can make it SO EASY, like darth vader,  but its worthless occupation trying to get a lazy boy to do work, no matter how easy you make it for him.


Check out my latest program - automatic 2d to 3d,  guess how easy it is to build a "perceptive model" with this baby!!!

So which part do you need help understanding?

There is no evidence whatsoever for the notion that consciousness is something that a machine can't have.  Consciousness is just a poorly defined word, you're right that I don't actually mean that, and you're right that I'm just describing a mechanism.  It just so happens that I think there is nothing to "consciousness" beyond those mechanisms.

I never said I wasn't going to work on this myself or that I don't do programming/electronics, I do in fact do programming and electronics.  I joined this forum to try and learn other people's approaches to this kind of programming problem.  What I meant in the essay is that I haven't reached a skill level that is good enough to implement these ideas, and it's clear to me now that this is not the type of project that a single person could begin to tackle all by himself.  The current state of AI dev is basically developing machine learning but that is only the very beginning.  Machine learning would basically be a component of the memory consolidation systems and the main executive processing system, there's so much more to it, so considering how much difficulty we're having developing just one of the building blocks of an AGI system, I obviously would be wasting my time trying to implement one all by myself.  I made the assumption that others would be coming up with algorithms and systems to model human brain function so I thought sharing some of my ideas about it might be helpful.  Not trying to get someone to do my work for me, just trying to share some food-for-thought with someone who may already be working on something.  If people don't understand how the brain works then they can't make a computer work like one.

I appreciate your interest, but my advice would be that when you'd like to criticize someone's work, it helps to read the entire paper first so you don't miss anything.  It helps when you know everything that was written because then you can discuss the ideas that are in the paper and explain whether or not you disagree and for what reasons and also add alternative explanations or elaborations to ideas in the paper.

Machine learning is great, and the techniques and skills that people are developing in that area are going to be essential components of complete AI systems.   Of course, machine learning algorithms are the most basic element and will be the building block several different sections of a complete AI architecture.

Keep up the good programming work, your enthusiasm would be an excellent contribution to an AI development team, I'm sure.

[ranch vermin]

Note, I read your post already after I wrote what I did.

So which part do you need help understanding?

Nothing, I understand.  Excuse my confidence.

There is no evidence whatsoever for the notion that consciousness is something that a machine can't have.  Consciousness is just a poorly defined word, you're right that I don't actually mean that, and you're right that I'm just describing a mechanism.  It just so happens that I think there is nothing to "consciousness" beyond those mechanisms.

Yes true, but I dont think your machine will have it, (because it sounds very similar to mine.) or anyone elses.

I never said I wasn't going to work on this myself or that I don't do programming/electronics, I do in fact do programming and electronics. 

Ah good.  All the speculation in the world isnt as good as seeing even the dumbest idea actually prooven not to work.  You only further develop after youve done things for real,  and noone is going to do it for you.
Theres no such thing as a guy in a-i that is good that isnt either a practicing!! programmer, electronics engineer or a mechanic (doing the body work.),  unless your a phony entrepreneur going nowhere.

Do you have any examples of your work -  I have a youtube account with 300 videos of my shit.

 it's clear to me now that this is not the type of project that a single person could begin to tackle all by himself.

It seems strange,  but I believe in the movie "jurassic park" where the security programmer guy wrote the entire system himself in his pile of junk food trash around his computer.    He feated the entire thing himself.   You only know that once youve done it,  anyone that doesnt believe in that,  I think isnt experienced.
I dont see how people believe in "one man wrote a whole book??!?? thats unbelievable!!!"   why,  cause they dont,  all authors work alone and a programmer is just another kind of author, dude.



So, thats enough death quoting.   Thanks posting this,  its opening my eyes wider every day,    I told you in my post, your idea is going to work,  but if you want one of these *mega toys* you have to make it yourself, you cant just go to the hardware shop and go lay-by a terminator. (or what I did in that video, for example.)  But programming is a tricky thing that takes many years of practice, doesnt matter how simple things are in theory.

Computers, therefore A.I.   only run on very explicit commands  (and in electronics the most of all! electricity!!!) and its mindbending to go so lowlevel with things,  to explain things to very absolute skeric of what needs to be carried out, for the busty old hunk of junk to "compute" its way around.

So good luck to you,  that video I gave you, already helped you heaps.  So I hope your glad you came here.

[goyabee]

Thanks mate. Cheers.

[Korrelan]

Hi Goyabee… and welcome to the site.

I’ve read the thread so far and…

It’s customary when joining a forum to say hello in the ‘New Users’ thread too give site admin and users a chance to say welcome.  It also helps new users understand that this is a friendly, co-operative and informative place to post.  You are in the company of many like minded individuals and there is a host of information on the site relating to your understanding of how AGI should work. It generally helps the introduction to the site and stops new users being jaded by their first interactions with current users. (How diplomatic was that eh?) lol.

Anyways… I personally agree with your theories and look forward to you posting any progress you make in your research.

Search the threads on this site, you will find lots of information and discussions pertaining to your ideas/ concepts.

What approach are you considering, Heuristics, Connectionism or some custom solution?

Again… Welcome 

[Don Patrick]

So, sensors for information and basic reward system for motivation. Both are forms of input signals, but I don't believe this tells me anything about how to process the information after and inbetween. Neural nets or evolutionary algorithms to figure it out on their own? Go ahead and try, I doubt others haven't.

It is a common theory, and indeed not many successes have been made. However, there are projects that do attempt such an approach, like iCub and http://www.mindconstruct.com

[ivan.moony]

If people don't understand how the brain works then they can't make a computer work like one.

There are two aspects of "understanding how the brain works". One is out-of-the-box experience and this should be more or less equivalent to the real brain functioning. The other is in-the-box implementation and here we have complete freedom, it doesn't need to be implemented exactly the way the real brain works. Like you can calculate the same thing in numerous ways in math, I believe that there is more than only one way to make a brain alike machine. We already took a sneak peak into neural networks and that gave us a decent recognition algorithm, but other recognition algorithms function too.

So there are two different ways in programming AI. One is to scientifically reverse engineer the whole brain machine. The other is to start from an imagination point, and, by using common sense, to try to construct out-of-the-box activities of the brain machine. Somehow I like this other way of complete freedom.

[ranch vermin]

So, sensors for information and basic reward system for motivation. Both are forms of input signals, but I don't believe this tells me anything about how to process the information after and inbetween. Neural nets or evolutionary algorithms to figure it out on their own? Go ahead and try, I doubt others haven't.

It is a common theory, and indeed not many successes have been made. However, there are projects that do attempt such an approach, like iCub and http://www.mindconstruct.com

I guess you could call the reward system as a form of internal input.   About the processing,  all the machine does is need to store and recall it,  the reward system is in place for the "internal pattern development"   and the rest comes in from the sensor,  possibly with a bit more pattern development happening off the "sensor assimilation."

This robot I havent seen it yet,  but i seem to hear about it all the time.  

its a piece of science fiction you can concoct for yourself at home, if you are serious enough to go for the full 120 yards alone.

The other is to start from an imagination point, and, by using common sense, to try to construct out-of-the-box activities of the brain machine. Somehow I like this other way of complete freedom.

That is the way - just have a think about it, as in, dont do so many experiments at first, or observations, even tho thats all good too - but u may have what you need already to solve your problem. and THINKING!!! thats what man has been doing for a long time.

[ranch vermin]

This "fastforward development" is it having a feedback model of itself in its environment.
and we are gonna have some problems!
that its not a free lunch in that situation that it has hampered motor development over a purely virtual GA.

getting to the six million dollar man, in gta 5, BUT IN REAL LIFE!!! is not quite solved, unless youve done at a least a crap one first.

it can actually feedback speaking better than walking - but it hasnt much use for speaking -  even with a fully classed language model off the sensor. (it knows what other people say, but it doesnt know what it should say.)

Your going to then get to the situation of MOTIVATION.  and you have to look up "negentropy" in other words its patterns being found in entropy, slowly over time, and then maybe the stupid thing would end up wanting to talk.

And thats the end of this style of robot.  so far.


If it all works hopefully well have iran in flames real fancy 1 time via our flying giant robotic dragons.  so its well worth the effort.

[madmax]

The consciousness is much more complex than simple neural network pattern optimization, it is more about a sustainability of organic entity balance from which is derived as a result of life evolution.More complex kind of consciousness forms intellect or intelligent being.That is why i wrote previously that we can't wholly understand what is intelligence because we don't know how life began.

It is very wrong theory that consciousness is just vast number of many possible different information.The point is actually in very the way how and for what is this information system, that we call consciousness, constructed.

On the other hand, there is a life form that maintains organic balance and doesn't have consciousness, like plants bacterias, viruses, and other primitives animals, but they also have some input-output information.Another intriguingly question: why then consciousness?


If we see the evolution of life as a complexity of life forms, we could perceive that complex life organism needs some sort of operating system that holds all parts together, and before that differentiation of the parts.So there is now different part for the input and a different part for the output.Next level of the complexity is the breaking point,in my opinion,and that is differentiation of the part for the some sort of voluntary movement.Why is that breaking point,simply because interaction with surrounding starts to become much more complex which emerges development of new parts for interaction with surrounding.While this development is going  the central system for connection of the parts is developing too,but there is not consciousness jet.Next puzzling question: when consciousness emerges?


Simple,but not so simple,when the organism starts to develop a complex visual sensory system through which could make the mental picture of outside world, and that mental picture with rest of organism balanced system together of course through central system construct first consciousness.But this doesn't mean all organisms with eyes have consciousness.But from this point, further explanation is much more complex.And i think it is more than enough for this post

[keghn]

 I have my AGI theory that might achieve machine consciousness as stated in this post:

https://groups.google.com/forum/#!topic/artificial-general-intelligence/UVUZ93Zep6Y

[madmax]

From my post as i tried to explain, was shown that for the construction of consciousness is important a mental picture of outside world.So from this, we could achieve simulation of the consciousness by making a valuing algorithm for the mental picture of outside world.Of course, that is just a simulation of the consciousness it is not real consciousness.

First, we must determine the outside world in known fundamental laws of geometry, and as complex, we determined those laws through mathematical functions that complex will simulation of consciousness perform.This determination will be achieved through valuing stages of the algorithm for the information input as a common language.

To go back to the purpose of real life consciousness.As i tried to showing the live organism develop through evolution their parts for the interaction with outside world and developed eyes.But simple developing eyes doesn't emerge consciousness.When the visual sensory system becomes complex more than simple registration of movements or light changes on the level as perform e.g. sensors for touch, then is possible to occur mental picture which is not divided from others sensory systems.

The visual sensory system is for optimisation movement in space, so determinating space is the most important.But, to go back and to try to explain why i think that visual sensory system is the key that develops consciousness.First because of movement of the organism as they become more and more complex visual sensory system requires more and more space in the central organisation system for coordination of the parts,so become somehow dominant.And as the eyes further develop the amount of information that eyes collect becomes bigger and bigger and more and more important so they being collect as visual pattern memory which enables the construction of wholeness of the mental picture of the outside world which include the information from other sensory system too.That wholeness emerges the consciousness.But this wholeness is not all about sensory information from outside of the body but from inside too,in the way that this inside and outside information interact with each other and together involve consciousness reaction.

As central nervous system evolve the consciousness becomes bigger and bigger or more comprehensive and make intellect that is able to choose from many possible outcomes.So consciousness as the evolutionary tool is on the simplest dealing with the splace with movement in space or any other interaction in space, with comprehending that space with all the movements in it.But we must not forget that from the wholeness of a mental picture through making consciousness we get a sense of self, as opposite part of the consciousness of outside world, from which will in evolution emerge self-consciousness.The sense of self is inner life maintain information that we take unconsciously,but we get the sense of self in the same time we get consciousness even the inner life maintain information was there before.So the sense of self and consciousness are the different side of the same medal.


And there is no the sense of self without the consciousness and opposite way too.But the sense of self is not self-consciousness.Self-consciousness is evolutionary develop of consciousness through the interaction of the sense of self and consciousness.This interaction happened in the way that consciousness was tried to control unconsciousness reaction on inner maintain life information, which are parts of the sense of self.


So possible is just the simulation of consciousness through the complex algorithm with use of combined scientific knowledge from many science fields.

[keghn]

 If a simulation of consciousness can deal with the out side world dynamically as real consciousness, then they are
equal.


 However the real world is encoded into a real mind or machine mind. The encoding are just shadows and illusions in
the mind. But  what ever illusions that are in a mind, then that will be it's reality.

[madmax]

Simulation of the consciousness is not real consciousness because don't have the sense of self.To figure out how is work the sense of self we must figure out what is the life, and that is a very difficult question.Simulation of consciousness is just some sort of complex optimisation of interface for more efficient interaction with humans and as we could see too complex and very fast processes inside of machines.


But i repeat again that is not real consciousness and dynamic of the processes is not what make consciousness is consciousness.What makes the life to live that is making consciousness is consciousness.That is huge motivation which simulation of consciousness is missing.And that is not just some efficient fuel consumption, it is the balance of life, of the organic form of molecules, again one big question why is life on earth?


Consciousness is the very very complex notion.But misunderstanding and misuse of the technology of consciousness simulation or technology of AI, which is wrongly named because there is no autonomous intelligence but simulation of intelligence, is possible.Intelligence is just level in the evolvement of processes inside of consciousness.And without consciousness, there is no intelligence.But the simulation of intelligence is possible and could be the extension of human intelligence which makes human intelligent but not simulation itself.