Help!

[bascule]

Okay, what part of this sentence:

 

Human consciousness results from a hierarchy of temporospatial classifiers which intercommunicate within a global workspace, the bottom of the hierarchy being fed a constant stream of sensory data, and the top of the hierarchy storing and replying unrecognized patterns backwards down the hierarchy"

 

...is the most difficult to comprehend?

 

I'd love to expand that into a layman-oriented description of how consciousness works, but, it's hard.

 

I guess some potential questions would be:

 

What's a temporospatial classifier?

What's a global workspace?

I don't get the hierarchy bit!

 

Where does your comprehension of the statement break down?

[Cap'n Refsmmat]

I believe your three questions are accurate. I get the model, but only after pausing to think about what the heck a "temporospatial classifier" is (it's not too difficult to figure out).

[padren]

temporospatial classifiers is definately a term you don't hear everyday.

 

My best guess is it would be something that classifieds data that is both spatial and temporal in nature - but its hard to find a definition even with google, and the "Merriam-Webster Online Dictionary" wants you to sign up to find out.

 

I am not sure if "Global Workspace" is the best term. Unless my comprehension of the statement already broke down - aren't you trying to say that this occurs within a "confined workspace" more specifically? If you used the term "Global" to convey some nuance of the workplace, then describing that more fully could help.

 

 

To be clear, you are basically saying that consciousness is a side effect of sensory data being passed between nodes of micro-data-handlers, which either forward on the data or pass back additional data-patterns as an automatic response, resulting in a system that has no conscious nodes but causes a sum consciousness-effect?

[bascule]

A temporospatial classifer is essentially a self-training classifier, similar to a Bayesian network except:

 

1) The structure is hierarchical

2) The processes of training and classifying are unified. The temporal component (i.e. what symbols are seen in regular intervals) trains the spatial component (i.e. what data represent symbols?) and vice versa

 

Probably the biggest component of understanding this is understanding how classification algorithms work. Here's Wikipedia to the rescue:

 

http://en.wikipedia.org/wiki/Naive_Bayes_classifier

[alan2here]

Like this?

 

[input from Body] -> [Try and Resolve] -> [Resolved?] -Yes-> [Output to Body]
                            /\               ||
                            ||               ||
                                <--  No  <--

 

I see a problem with this.

The form of pyscoligy NLP covers 'loops' like this one happening.

[JohnB]

Have you ever considered writing Legislation? To be honest it comes across as the most complete psychobabble.

Human consciousness results from a hierarchy of temporospatial classifiers which intercommunicate within a global workspace,

Human consciousness results from the heirarchical interactions of data classifiers within the mind. The word "temporospacial" is superfluous, "communicate" can be used instead of "intercommunicate" and "within a global workspace" sounds like you work for the government.

the bottom of the hierarchy being fed a constant stream of sensory data, and the top of the hierarchy storing and replying unrecognized patterns backwards down the hierarchy

The bottom of the heirachy sending data up the line while the top sends interpretations down.

 

The problem is not that it can't be put in simple english, it's that if it's put in simple english the person saying it doesn't sound as smart. Just because a word has more syllables, doesn't mean it's more accurate or more descriptive.

 

Bascule, I know you're smart and I respect your opinions (Even those I disagree with. ) but that statement sounds like it's been parroted out of the summary page of a book chapter.

[bascule]

The word "temporospacial" is superfluous

 

The word is "temporospatial", and it certainly isn't superfluous. I could substitute "self-training hierarchical classifier", but then the question is how does a self-training classifier actually work?

 

These classifiers are looking for sequences/sets (i.e. groups of symbols over time), hence the temporal component, and symbols must be discovered from underlying input data, hence the spatial component.

 

The two feed off each other, using each other to provide input for their own classification.

 

This is a very important distinction when comparing to other types of classification networks, which typically have to be trained explicitly and provide no sort of feedback mechanism.

 

"communicate" can be used instead of "intercommunicate"

 

That's an apt criticism

 

and "within a global workspace" sounds like you work for the government

 

Well, it's trying to describe, in four words, the operation of the thalamus. Therein lies the entire conscious experience and perception. The words "global workspace" come from global workspace theory, a specific interpretation of the operation of the thalamus.

 

The bottom of the heirachy sending data up the line while the top sends interpretations down.

 

It's not so much interpretations as predictions. Since different levels of the hierarchy act as temporospatial classifiers, identifying symbols and learning sequences/sets of them which occur in a given temporal sequence, once part of a sequence/set has been identified the upcoming members can be predicted and fed back down the hierarchy.

 

The problem is not that it can't be put in simple english, it's that if it's put in simple english the person saying it doesn't sound as smart.

 

Actually, it's that it's hard to describe in simple english. That's why I'm asking for "Help!"

 

Just because a word has more syllables, doesn't mean it's more accurate or more descriptive.

 

More specific, descriptive language allows you to explain a complex idea concisely.

 

Bascule, I know you're smart and I respect your opinions (Even those I disagree with. ) but that statement sounds like it's been parroted out of the summary page of a book chapter.

 

It's the most concisely I can explain the operation of the human brain, at least from a hippocampal-thalamocortical perspective.

[JohnB]

I thought the idea was to put the concept in simple terms?

 

The word is "temporospatial", and it certainly isn't superfluous. I could substitute "self-training hierarchical classifier", but then the question is how does a self-training classifier actually work?

Oops, my bad, but yes it is, and who cares? Do you drive a car with a "Quadruple Synchromesh, bronze gated, manually activated gearbox" or is it a "Manual"?

Well, it's trying to describe, in four words, the operation of the thalamus. Therein lies the entire conscious experience and perception. The words "global workspace" come from global workspace theory, a specific interpretation of the operation of the thalamus.

Then just say it happens in the Thalamus. Deriving the term "Global Workspace" from Global Workspace Theory means absolutely nothing if the reader isn't familiar with Global Workspace Theory in the first place.

It's not so much interpretations as predictions. Since different levels of the hierarchy act as temporospatial classifiers, identifying symbols and learning sequences/sets of them which occur in a given temporal sequence, once part of a sequence/set has been identified the upcoming members can be predicted and fed back down the hierarchy.

So we make the sentence "The bottom of the heirachy sending data up the line while the top sends interpretations and predictions down."

More specific, descriptive language allows you to explain a complex idea concisely.

Only up to a point, after that it becomes gobbledegook. Compare the two statements.

 

Human consciousness results from a hierarchy of temporospatial classifiers which intercommunicate within a global workspace, the bottom of the hierarchy being fed a constant stream of sensory data, and the top of the hierarchy storing and replying unrecognized patterns backwards down the hierarchy"

 

and

 

Human consciousness results from the heirarchical interactions of data classifiers within the Thalamus. The bottom of the heirachy sending data up the line while the top sends interpretations and predictions down.

 

What essential information is contained in the first that is not in the second?

 

If you intend to get a concept across then use simple english, only later is it neccessary to get into the nuts and bolts of the internal workings of the idea. Similarly, you could say that a car is a Manual or an automatic, once that is established, then you would go into the internal workings of the gearbox itself.

 

Put it another way, one could talk about "overcoming the bonding forces of certain atomic nuclei within fissile materials" or you could just "split the atom".

[jck]

I have this as a basic structure.

 

Information feeds are moved from the original source to be stored and recognised in order for the information to become a concious thought that is understood.

 

Anything that is not recognised is fed back and awaits further input.

 

john

jck

[bascule]

More specific' date=' descriptive language allows you to explain a complex idea concisely.[/quote']

 

Only up to a point, after that it becomes gobbledegook. Compare the two statements.

 

There's no way you're going to keep the statement as concise and correct. What I'm really trying to do is expand the statement into several paragraphs and eliminate all jargon.

 

Human consciousness results from the heirarchical interactions of data classifiers within the Thalamus.

 

The classifiers aren't in the thalamus, they're in the neocortex. They interconnect with each other through the thalamus as an additional, mesh-like mode of persistent storage and communication, as opposed to the rigidly hierarchical structure of the neocortex itself.

 

"Classifier" is a bit of unexplained jargon there. Furthermore, the neocortical algorithm is substantially different from "classifier" as typically used in computer science. A Naive Bayes Classifier, perhaps the most common in general usage, operates in a substantially different manner from the neocortical column. Naive Bayes Classifiers must be trained with several examples of a particular pattern and how that pattern should be classified. The closest equivalent to the NCC classification algorithm would be Numenta's Hierarchical Temporal Memory.

 

The bottom of the heirachy sending data up the line while the top sends interpretations and predictions down.

 

The bottom of the hierarchy sends unclassifiable patterns up and essentially asks for help classifying the pattern from farther up the hierarchy.

 

This is why, for example, you can drive a route every day and consciously lose track of where you are without having any trouble driving. The actual act of driving is being handled by the lower levels in the hierarchy, and since they're not running into unrecognized patterns, your higher levels can be concerned with something else entirely. Yesterday I was driving home and had to ask myself "Did I just go through a tunnel?" only to look in my rearview mirror to discover "Yes I did". The lower levels never informed the higher levels because they were anticipating the tunnel.

 

What essential information is contained in the first that is not in the second?

 

How do we learn? How do we use what we learn to make predictions? How do we use predictions to make decisions? Where does perception/conscious experience take place? How are long term memories stored?

 

I think you're also missing I'm trying to go for a functional, not anatomical description. Substituting "thalamus" for "global workspace" essentially introduces a black box into the system. What does the thalamus do?

 

Judging from the above, you don't know what the thalamus does. That's why I'm trying to avoid using specific anatomical terms. A layman has no idea what a neocortex, thalamus, and hippocampus are. I think if you were to give your average Joe a pop quiz and ask what part of the body each of those are a part of, you'd likely not get "brain" as an answer to more than one.

[GutZ]

This makes sense to me, and thats rare so....that's probably as laymen as you can get.

 

The word is "temporospatial", and it certainly isn't superfluous. I could substitute "self-training hierarchical classifier", but then the question is how does a self-training classifier actually work?

 

These classifiers are looking for sequences/sets (i.e. groups of symbols over time), hence the temporal component, and symbols must be discovered from underlying input data, hence the spatial component.

 

The two feed off each other, using each other to provide input for their own classification.

 

This is a very important distinction when comparing to other types of classification networks, which typically have to be trained explicitly and provide no sort of feedback mechanism.

 

Once you get that, the rest seems easy enough to get.

[JohnB]

Firstly Bascule, my apologies. I've reread the OP and had missed the word "expand" in it. I thought you were trying to simplify the phrase, not expand it's scope.

 

I would think that, to a degree, what you are trying to is not hard, it's bloody near impossible. The difficulty you are having is simple. You are attempting to give a specific description. To give a specific description you need to use words from a specialised vocabulary, whereas a layman, almost by definition, does not have access to this vocabulary.

 

BTW, do you mean a layman here on this forum, or just laymen in general? That would have to effect the terms available to you, a layman here would be able to follow your explanation and terms much more easily than John and Mary Public would.

I think you're also missing I'm trying to go for a functional, not anatomical description. Substituting "thalamus" for "global workspace" essentially introduces a black box into the system.

To the layman, exactly how is "Global Workspace" any less of a black box than "Thalamus"? At least they may have heard of the Thalamus before today.

What I'm really trying to do is expand the statement into several paragraphs and eliminate all jargon.

"Global Workspace"

"Naive Bayes Classifier"

"temporospatial"

For the layman, these are all jargon.

The bottom of the hierarchy sends unclassifiable patterns up and essentially asks for help classifying the pattern from farther up the hierarchy.

But the classification and interpretation occur in the higher levels of the heirachy, right? (Just clarifiying for my own interest.)

Judging from the above, you don't know what the thalamus does.

Before you get narky, I've done nothing but logically combine and simplify your comments.

within a global workspace

Well, it's trying to describe, in four words, the operation of the thalamus. Therein lies the entire conscious experience and perception. The words "global workspace" come from global workspace theory, a specific interpretation of the operation of the thalamus.

You're the one who brought up the thalamus, not me.

 

I'm familiar with your "tunnel" phenomenon. I would expect that it is also the reason that the majority of car accidents occur within 10 minutes drive of the persons home. They know the area so they're driving on autopilot.