The mechanisms of dreams

[Adron]

That's the subject of an entirely different discussion: The one about how things are represented inside the brain. That observation you made sounds like dreams happen at a higher level than the one that parses images into internal representations. Which makes a lot of sense - no need to emulate the actual image when you can just emulate the internal representation of "object #6239123718: Human 'EvilCheese'"

[Raven]

I think the way dreams work, as that when you are in sleep mode, your brain is still trying to parse a sort of non-sleeping consciousness, so it taps into different fragments of things in your mind, such as abstracts from games you were playing, movies you've seen, or thoughts you had, and then sort of compiles it all into a dream where different things (often strange) will happen based on the thoughts that were tapped. So for example, if you're attempting to break out of a prison, and you were casting Fireballs at stuff, maybe it tapped into a recent recollection of the Shawshank Redemption or The Rock, as well as a videogame you may have been playing. If you think hard enough, I believe you can associate most things that happen in your dreams to something in real life that have spent time thinking about.

[hismajesty]

when you are in sleep mode

Oh goodie we're all turning into computers

[iago]

That's the subject of an entirely different discussion: The one about how things are represented inside the brain. That observation you made sounds like dreams happen at a higher level than the one that parses images into internal representations. Which makes a lot of sense - no need to emulate the actual image when you can just emulate the internal representation of "object #6239123718: Human 'EvilCheese'"

I don't think your brain would number the objects.. at least, not a single dimensional number.  

I think that me and you discussed that awhile back when I was writing my final paper for philosophy (which, by the way, I got an A+ on, giving me an A overall), and it's interesting how the brain processes information.  I should start keeping a dreamlog and a diary

[Eibro]

That's the subject of an entirely different discussion: The one about how things are represented inside the brain. That observation you made sounds like dreams happen at a higher level than the one that parses images into internal representations. Which makes a lot of sense - no need to emulate the actual image when you can just emulate the internal representation of "object #6239123718: Human 'EvilCheese'"

I don't think your brain would number the objects.. at least, not a single dimensional number.  

I think that me and you discussed that awhile back when I was writing my final paper for philosophy (which, by the way, I got an A+ on, giving me an A overall), and it's interesting how the brain processes information.  I should start keeping a dreamlog and a diary

Don't forget the perfume. Oh, and ribbons for your hair.

[Adron]

I don't think your brain would number the objects.. at least, not a single dimensional number.  

No, probably not. But it's most probably using the same concept as assigning numbers to tokens in a computer application.

I think that me and you discussed that awhile back when I was writing my final paper for philosophy (which, by the way, I got an A+ on, giving me an A overall), and it's interesting how the brain processes information.  I should start keeping a dreamlog and a diary

Nice to hear that you got a good grade for that one. It was an interesting conversation! You never did show me what you wrote...

[EvilCheese]

Here's my take on things:

I think the concept of the brain numbering or storing information and experiences as objects in some kind of multi-dimensional array would just be a convenient simplification for a process that's much more complex.

The human brain remembers by strengthening certain inter-neural links in response to repeated stimulus. The more often a stimulus is repeated or encountered, the stronger the bias of the neurons to pass on electrical signals along a given path. IIRC this is believed to be the main mechanism in both memory and learning.

What makes it interesting from a programmatic point of view is that the biasing is an essentially analogue process, rather than a digital one.

The brain contains nothing like an if(X){y;} structure, rather it contains a structure which would look more like:

if(x is quite likely)
[
Probably Do {y;}
Sometimes Do {z;}
On remote occasions do {something different;}
]

With the number of possibilities and the biases which affect the outcomes being essentially inumerable and changing each time the structure is executed, depending on results.

Writing code which is analogue, devious, and above all else self-modifying in this manner will (I believe) be the key to the development of true artificially intelligent systems.

The main trouble (there are many) from a coding perspective is that in attempting to implement the biases and the uncertainty in the decision making we are limited to using digital representations of the analgoue processes, making the accuracy and depth of possible results from a given occurence essentially and unavoidably limited.

I believe that in REM sleep, the brain fires off seemingly random neural pulses (I dont think anyone is sure why it does this)... the imagery is likely created by these pulses passing through the brain's regular stimulus-processing system, which is why you're likely to see things that are on "freshly burned" tracks (things on your mind).

The reason we forget dreams so easily on waking is (I expect) due to the fact that the bias alterations caused by the use of said pathways is incongruous with the normal parameters of usage for them, and so the deviations are quickly lost as those pathways are exposed to normal use.

It would be an interesting study to find out if the use of these neural pathways in a dream-state has any permanent bearing on the behaviour, interpretation or cognitive abilities of an individual.

[Note:]

This is just my take on things... it could be (and likely is) totally wrong  

[joykillah]

your all turning into robots.

[Raven]

EC, I'm fairly certain that the human brain functions in a much more complex manner than a computer or a few simple lines of code.

[mavrick_kr]

Human brains are unique.That says it all.

[Adron]

It does make sense that signals aren't digital absolute numbers in the brain, but what do you think about how a new signal is assigned? I.e. when you define something, say a person or so, how do you think the brain assigns that? By dividing an existing space in two? Is that why we can get weird associations between people - the proximity of the signals in the brain's n-dimensional signal space?

[EvilCheese]

EC, I'm fairly certain that the human brain functions in a much more complex manner than a computer or a few simple lines of code.

I thought I'd implied that in what I wrote originally.

Acccurately modelling the behaviour of the human mind in code is beyond any current technology, however, in recent years there has been an explosion of fuzzy logic research and expert systems development.

Weighted probabilistic programming of the kind I described is however used on a much smaller scale for systems such as face, speech and general pattern recognition. (Google for "Neural network programming" or "Fuzzy logic")

Some would argue that it's only a matter of time and technological advance before systems are developed which are approaching the complexity required to carry out tasks or learn in a complex environment.

Whether this would bring computers with personalities, or "true" intelligence is a matter for highly philosophical debate, since many would argue that there is something "aside and seperate" from the brain..... a "soul" perhaps..  which is unreproducable and defines who they really are.

The origin of consciousness would be a debate for a seperate thread though, and might go on for quite a while

[Camel]

I don't think your brain would number the objects.. at least, not a single dimensional number.  

No, probably not. But it's most probably using the same concept as assigning numbers to tokens in a computer application.

All memories are are chemical messages; I would say calling said ID a number is quite a stretch as there are no bits or even digits as said messages exist naturally. I suppose, however, that one could come along and string togeather a list of numbers representing all of the chemicals and their order in the ID, assuming that the brain stores IDs in that mannar.

[EvilCheese]

Adron: I think when you see someone new, someone you havent met before, your brain first attempts to pigeonhole them into a category corresponding to someone you already know.

This is a very poor analogy, but view the brain when recognising people as one of those pyramid shaped boards with pegs in it that you roll balls down.

The balls being rolled down the board are the stimulus representing the person you are seeing, the pegs redirecting the balls are the defining features you use to recognise people, and the bins at the bottom which collect the balls correspond to different people who it could possibly be.

The first time you ever meet someone or see someone, you are likely to produce the response "That person looks like xxxx" where xxxx is someone you can already recognise... since most of the balls will end up in their bin.

However, since the process didnt turn up a 100% match, a new bin will be created for that person, new pegs will be inserted in the board, and the next time you see them you will be able to distinguish them more easily.

The more often you see them, the more refined the configuration of your "pegs" becomes and the more easily you will be able to pick them out.

Apologies for the poor analogy, but it was the best I could come up with in 30 seconds

[Adron]

It's a good analogy. What I'm wondering now is: Do you think there is a simplified representation for a person in the brain, or does the representation contain all the information you know about the person whenever you think about him/her? I.e. are the N inputs from your senses mapped to a set of n (n<N) data points that represent the person to your thinking process?