There is a big gap between symbol manipulation like Searl's Chinese room and a conscious understanding of a sentence. The Chinese-room-algorithm is a rule based algorithm and could be done by a computer or a zombie (I be in doubt about that concept). If a human is reading a sentence emerges a representation of the sentence's sense in the mind of the human, because the human is conscious and the representation is experienced.

If Searl's Chinese room did his work well and all sentences are correctly executed the programmer is very clever or did hard work to edit all concepts in Lisp, Prolog or sth. else but the room did not understand (!) what it is doing. The problem is in this case: This program is passing the Turing test even it is totally stupid.

A conscious mind is self organized and experienced. Here is the point of your question.

Dear Dimitrios,

I think this is an interesting proposal. At the same time I think that consciousness is not an hard problem just because of the absence of a good definition, and thus things are never straightforward. One could argue you can be conscious of things you don't recognize/you don't have a concept for. Also, I don't think consciousness is a unitary concept, since the animal in your example could perceive, process and react to the recognized chair or squirrel in a fully automatic manner. As a counter-example I bring up the absence seizures, in which a human can fully work (even if with limitations) without being conscious. In the animals consciousness could be absent, as in computers. Otherwise, your proposal alone could be enough to justify people saying that a computer that correctly processes a stimulus and activates the correct label is conscious.

If we were to assume concepts to be the quanta of consciousness, we would lack explanatory power over situations in which there is processing and recognition but not consciousness. The gain in explanatory power over consciousness per se, on the other hand, may be a huge step in understanding consciousness, since we basically start from zero...

Alan,

I think Dimitrios is trying to avoid the pitfall you outline by using the notion of recognition in the way he does. Searle has famously argued that computers do not have this capacity.

http://w3.uniroma1.it/episteme/file/Dispense20130329_MindsBrainsandPrograms.pdf

As for fully automatic behavior, we may indeed want to reserve a category of consciousness for that. As for whether recognition has taken place in any instance of consciousness behavior, might this not depend on whether the organism in question is itself presented with (some representation of) the object or whether the organism's response is fully explained by reference to nothing other than physical stimuli?

On the lowest level of mental objects are the subsymbols.

Next are the symbols. They are like signals: they are usually

associated with a meaning. Of course we can be conscious

of single symbols. Our mental hardware makes it possible

to store tuples of symbols. How this is realized is still a riddle.

The working memory operates like a shift-register, which makes

it possible to have "dog chases cat" or "dog dog cat cat"

in conscious mind, regardless of grammatical rules.

On a higher level, maybe on the highest, are the concepts.

In AI courses one often encounters Lisp macros like

defconcept, a special syntax to define concepts.

I don't believe that a special syntax for defining concepts

in human brains is necessary, nor do I think concepts

exist as neural hardware. Rather, it seems to be possible

that knowledge about concepts exists as an abstraction,

in the form of a set of related propositions.

One of my favorite examples for an abstract concept

is mounting a harddrive. Obviously not an innate concept.

On Unix systems you have to mount a harddrive.

A person wants to see the content of a device in

a directory, and this device is a harddrive. Someone

with appropriate rights has to execute a special

command to make the drive visible, and this is the

action of mounting. Often the command is executed

after booting by a shell script...

As you can see, a definition of a concept, how it

is used, how it interacts with other objects, other

concepts, can exist as a bundle of propositions.

In this example it wouldn't make sense to search

for a neural circuit for the mount-concept in the brain.

Regards,

Joachim

Ian's question, on whether dogs utilize concepts, made me think on the nature of concepts and how we use them ourselves.

The perception of the sight of a chair may start off as a series of synaptic events that are recognized as colours, shapes and textures but the (preexisting) concept of a chair does not become animated until certain level of processing has been completed and combined. Only then is the person conscious of seeing a chair.

How about a person who has been blind by birth? He too has the concept of the chair that will become animated - only, in the absence of visual cues, auditory, olfactory and tactile cues will converge to achieve this. And he will become aware of the sensing the chair as well, when its concept becomes active.

Joachim, even if symbols may evoke a tangential emotional parameter, I am not sure that we become conscious of them until they are placed in a conceptual narrative.

What I found too hard not to comment upon is your hardware analogy. For me the brain-as-a-computer paradigm always brings up the problem that, contrary to computers, in the brain, the hardware (neurons and synapses) is the software (processing) and the software ever modifies the hardware.

> Ian's question, on whether dogs utilize concepts, made me think

> on the nature of concepts and how we use them ourselves.

My notion of concepts is the same as of frame or scripts that

were invented by the early AI pioneers (Marvin Minsky, Patrick

H. Winston, Roger Schank...). Concepts are like object classes

in the sense of OOP. They can have instances which resemble

sentences of natural language.

Concepts can exist in relationship with other concepts,

and these relationships are called slots. For example

a concept that expresses a verbs usually has a "who" slot

that indicates the actor and a "whom" or "what" slot that

indicates the receiver of the action, the object.

For example we can express "John saw Sue this morning" in

a form that can be read by machines easily, in Lisp:

(see (who john) (whom sue) (when this-morning))

Another example. We've got a concept of food and a concept

of (to) eat. A thought might be: if someone eats something

then this is food. The rule could be

(eat (who ?x) (what ?y)) => (food ?y)

With more such rules we could describe the meaning of

the concept "eat". And at the same time we would be able

to perform operations in a mechanical way which would

be considered thinking when done by a human.

(eat (who peter) (what apple)) => (food apple)

In the late seventies there was an interpreter for such

forward chaining rules, OPS5. According to the Wikipedia

article thousands of rules could be handled. Thousands

sounds ridiculous today. Computer technology has developed

in the meantime, also inference machines like that of Cyc.

Maybe there are already thinking machines out there,

in the cellars of the secret services?

Regards,

Joachim

Joachim,

I have referenced Searle's famous "Minds, Brains, and Programs" above. In that paper he explicitly makes use of Shank's work as a paradigm. He also gives, and this is what the paper is famous for, the chinese room example. It purports to SHOW that running programs could never entertain a meaning nor understand a concept (or anything else). What is your response to this argument, which most in western philosophy take seriously?

My opinion on the Chinese room argument is that

one should apply induction over the number of symbols.

Chinese writing, as far as I know, has the nice property

that one written symbol expresses one word.

So discussions around the Chinese room should

always begin about messages of length 1. The question

is a) whether a system is able to represent and

understand the meaning of a symbol, and b)

whether it is possible, that, if a system understands

the meaning of all messages of length n, it could

also be able to understand messages of length n+1.

For me the answer to a) is yes. The easiest way to

achieve understanding of symbols on a machine is

to use a dictionary and map a symbol to an appropriate

subroutine. A different approach is this: use a graph,

where nodes are symbols and edges express similarity.

If you want to define the meaning of a new symbol S1

and you have already defined the meaning of other

symbols S2, S3, S4, and they have a similar meaning,

draw edges S1-S2, S1-S3, S1-S4. A recursive

definition.

The answer to b) seems to me to be obvious:

the opposite would sound strange.

Regards,

Joachim

Joachim,

I can truly say I find this answer astonishing. Thank you for your clarity.

Ian

There is a big gap between symbol manipulation like Searl's Chinese room and a conscious understanding of a sentence. The Chinese-room-algorithm is a rule based algorithm and could be done by a computer or a zombie (I be in doubt about that concept). If a human is reading a sentence emerges a representation of the sentence's sense in the mind of the human, because the human is conscious and the representation is experienced.

If Searl's Chinese room did his work well and all sentences are correctly executed the programmer is very clever or did hard work to edit all concepts in Lisp, Prolog or sth. else but the room did not understand (!) what it is doing. The problem is in this case: This program is passing the Turing test even it is totally stupid.

A conscious mind is self organized and experienced. Here is the point of your question.

In the simplest case the Chinese room reads a symbol

that is either 0 or 1. Is there a difference between reading

a 0 and reading a 0 consciously?

A human knows that he has read 0: a memory of the

moment of reading is stored. But we can have that in

computers as well: a program could notify the reading

event in a knowledge base or in a log file.

A human, when he consciously reads a 0, activates

a lot of memories of episodes in his life where 0 played

a role. But in a program we can have that likewise,

upon reading a symbol the computer can activate

associations, associated tasks, launch processes

and programs.

Regards,

Joachim

Joachim,

we neither have an idea what consciousness is nor we have a definition of that phenomenon, so we can not compare simple computer tasks with consciousness or something we think it could be.

To store a symbol like 0 in a computer storage could be compared with the notation of that symbol on a sheet of paper. Nobody would argue that this paper is conscious in any way in this symbol. In the same way the computer is not conscious.

The Chinese room (Searls) as an algorithmic program is also a daring attempt. Nobody can prove if the possibility of such an algorithm exists - I am in doubt about that.

I think the "irreducible quantum of consciousness" is qualia. Here emerges metaphorical a light in the mind which represents a stable input data analyse. Simultaneous active neurons were clustered by Hebb's rule to a specific mind symbol. May be a ART ANN is a good model for such a process. With such clusters are further activities possible and with connections and relations between also higher consciousness will occur.

Let me reduce qualia. In the Chinese room thought experiment

the symbols are handed out to the human in the room written

on paper. Maybe the symbols are in kalligraphy on perfumed

rice paper. That would obviously lead to a difference between

human reading and reading by a computer program. It would be

fairer if the input symbols were presented to the human

devoid of anything that could convey extra information like

the one dubbed qualia. On the other hand, why shouldn't we be

able to write a computer program that is able to discover the

beauty in written characters and smell the odor of paper and ink?

Qualia trigger associations. If a person perceives a color,

he feels reminded of experiences of his life. Green reminds

of grass, moss, green fruits, red reminds one of fire, blood...

But why shouldn't computers of the future have associations

triggered upon experiencing a color or an odor or a taste?

Nobody is talking about the qualia of yesterday's wine.

Of the past we only have memories, memories of the

wine's components. What makes the present moment so

unique, why do qualia only act here?

The difference between experiences of the presence and the

past is the number of possible input channels. If we have an

image in front of us, we would be able to focus on every of

millions, if not billions, of pixels. The number of choices of

potential input sources in the present moment is overwhelmingly

large. That is it what makes the presence so rich. I think

that's what is making qualia work. But why not have that

in computers as well?

Regards,

Joachim

One way of characterizing conscious, at least consciousness of an object, is as an act of "taking as." A mind takes an object x as a type F. It is aware of the object by virtue of recognition of it as an F. So, this tradition goes, the x is some manner of input, and the F is some means by which the mind can entertain "That (x) is an F." One way of putting Searle's problem is by asking, can this "taking as" be accomplished digitally? Does a computer recognize the object as an F, or does it merely produce output we recognize as its being/having been in the presence of an F? There is, arguably, a real distinction here. If so, what conditions must be met for a system to accomplish a genuine "taking as" rather than merely the corresponding output?

In many programming languages is an operator present for "taking as".

In C it is the typecast, in C# we have the "as".

Apple a;

Orange o;

...

Orange x = a as Orange;

x would be of type Orange if the application of "as" was successful,

null otherwise. "Taking as" means that an object is supposed to

have the properties, methods, interfaces of another object.

However, the above "as" is not performed consciously. A computer

is supposed to execute instructions without reflecting on them.

But that isn't a necessity, not on all levels. Programs can be written

that don't execute their steps immediately, rather they could

analyze the potential outcome and the value of their actions before

performing them. Interpreters do that to a certain extent. Future

computer programs will discuss with humans: "I know I did do

that, and I did it because...". How deep the introspection in their

inner workings is is for practical purposes mainly bounded by the

skill of the programmers.

We humans can sometimes use "taking as" as a tool. Computer

programs can use tools if necessary and within reach. Another

possibility of "taking as" is in recognition. We recognize 2a + 4c

as 2x. It is pattern matching and we are not guaranteed to

discover such a pattern. It depends on training and knowledge.

The same is or will be true for AI programs.

Regards,

Joachim

It seems as if we are reducing consciousness to some exercise of the intellect and, by so doing, have missed the essence of consciousness that we all know when we are not manipulating symbols. The sense of self, the sum of all my wishes, desires, aspirations, urges, shames, joys, and love -- I would have to include all of these in my "definition", and none of them can be codified in the way we are approaching the issue, unless we can be satisfied with a cardboard facsimile of beingness. The research task I have assigned myself is to find some holistic way of measuring the essentials of each state of consciousness (including dreaming, meditational, and mystical states), none of which can be properly reduced to symbols or exercises of logic. Any suggestions along those lines?

Searl's Chinese room is a special Turing test and in Searl's case and Joachim's hypothetical program the test is passed very well. Turing would have said there is a human not a computer acting. Nevertheless both examples are not conscious but only zombies, very clever AI programs.

There is not only a quantitative difference between today computer programs and such Joachim proposed, the difference is qualitative. We do not only need more input channels or more associations, we need a feed back loop between the input and the input analyse which results in a resonant state of this perception. In the input analyse there are all associations and relations of the potential perceived object. There also is a lack of physiological reactions of the body and the perception of such reactions like feelings which influences this perception. In such processes, I think, we could find quale.

Imagine we had a time machine and we went back in the Dark Ages and nuked a city (and I agree this is a psychopathic premise, but keep reading). Imagine further that you are a person from the Dark Ages who escaped his/her city being nuked that day (because (s)he was selling a pig some 100 km away and took him/her 5 days to come back home) and that person , with the scientific tools (s)he, as a member of the the people of Dark Ages has, tries to understand what happened to his/her city.

Well, if you can picture that, if you can put yourself in that person's shoes, you will have a better idea of how preposterous it is that we try to understand qualia (if such thing exists) with our poor (and this is a huge understatement) understanding of the brain and of how it entails cognition.

That much about qualia.

Now, turning to the proposal of Dimitrios, it seems to me that replacing "qualia" with "concept" does not help a lot. Concept is an equally vague thing as qualia. It may seem more tractable but it is not. That's my 2 cents.

Cheers,

SCM

Thank you for your contribution, Serban.

Allow me to clarify: qualias as subjective, too numerous and resistant to an exact definition. Hence, their limited use in corralling consciousness.

Using concepts instead, allows us to move a few steps up yet keep at the basic level of conscious thoughts. What you and me and a million other people have in mind when we hear the word "chair" may be different. However, we will use the concept of the chair to recognize one when shown a relevant picture. 

Thank you Dimitrios.

I still do not agree with what you say, in the sense that I believe that when you are presented with the picture of a chair and you recognize it to be a chair (or name it as such) there is no *concept* of chair being involved. The concept is in the eye of the beholder. 

And it is not that I have something against the concepts per se, speaking of concepts is useful maybe but I just think that i) the notion of concept is not necessary to explain cognition and ii) the concept does not explain cognition better that say, being a Greek god Zeus explains why there are thunders.

Cheers,

SCM

Dimitrios, 

I agree with your "concept" of a chair if somebody say this word but there is a difference between the concept of  chair and the quale of the representation of a chair. 

An AI program is able to find an association of the word "chair" and an image of a chair. But there is nothing else, no quale, no emotion, no other association - nothing! 

You should not exchange these different expressions hoping things become clearer. 

Joachim, 

it is not the question how we note the Chinese words. Type it with a Chinese key board. We do not need the "beauty" of ink on paper to discuss Qualia. 

Quale does not trigger associations, it is the perceived sum of all individual associations produced by the conscious mind enriched by emotional evaluation. 

What makes the present so rich? You are right, it is the number of input informations. If we remember an object we activate a concept and some individual properties but not the total picture. So memories are smoothed and less details. 

Serban, 

Do you believe in the idea of representation? Can something represent something else? And this is really to ask, do you believe in the idea of generality?

Ian, 

representation is a coding rule where a symbol represents sth else. Where is your problem? 

Joachim, 

it is not the number of input channels because we do not realize the input but the result of a process which offers a concept. We do not see lines, edges, colours and surfaces but we see a chair. The very best is we know what a chair is. It is not only a symbol in a dictionary, it is much more. We have all associations how it feels if you can sit down and relax. We remember different chairs and we anticipate the feeling when we are sitting in the viewed chair. 

Qualia and consciousness is much more than symbol manipulation and pattern matching. It is related with understanding. This means an individual knowledge of experience, bodily feelings, associations, concepts and sensual informations. There is neither a way to formulate rules in any program language nore to fill slots of a concept. 

If this would not be enough, there is something else: we know that we realize these things, the chair and everything else. 

Wilfried,

You say, "representation is a coding rule where a symbol represents sth else," by which I presume you mean representation is a coding rule by which a symbol represents something else. (You wouldn't want to identify the rule with the symbol.) No problem there. But it seems to be just such rules which Serban's ontology seems to deny.

Ian,

naturally you are right.

@Ian: I do not "believe in the idea of representation" because (cf. below) what you may mean bu that is that representations are effective and necessary causes in cognition. But believing or not in something does not make any difference. Science is not about what we believe but about what we can prove of. What more, "Do you believe in the idea of representation?" is awfully vague a question.

I do not deny that something can stand for something else, i.e., that something represents something else (e.g. the letter M standing in alphabetical languages for the sound /m/). I also doubt that there are many people out there who deny representations in this sense.

What I do deny is that representations play a role in cognition. In other words I deny that a representations be an effective cognitive cause. I'm comfortable with this view because I claim representations are not necessary, but this is maybe outside the scope of your questions. And I am pretty sure there are quite some people who hold this view (though I am well aware that we are not a majority).

"And this is really to ask, do you believe in the idea of generality?" I don't know what to do with that. I really have no clue on how generality would be tied to representations as  effective cognitive causes in cognition. Please elaborate if you wish.

Cheers,

SCM

Serban,

You are certainly right that many people hold your view. And many others should. In saying others should, I mean that others are dedicated to positions that imply this view even if they don't see the price of that commitment. 

When Strawson or someone like that talks about individuals, or Sellars talks about particulars, he is contrasting them with what some call 'universals.' My impression is that you don't believe in universals. For you, what there is to meaningfully talk about are particulars (and their particular causal relations), and this would be so for physics, biology, psychology, etc. I don't hold this view because I don't think it can make sense of notions that are indispensable to holding views, including about particulars. E.g. the notion of a fact, the notion of a set, and most fundamentally the notion of something's being the same.

Ian,

I do not know anything of the literature you cite, and thus cannot make much sense of what you saying about individuals/particulars vs. universals. Could you please suggest me some readings on this topics? Many thanks in advance.

Cheers,

SCM

Sure Serban,

Here is a nice summary on the topic by Michael Lacewing. (Note that he uses as his example of a relation, the relation of resemblance. I think he might just have well used the example of the relation of causality. I say this because you have mentioned your belief in causality specifically.)

http://cw.routledge.com/textbooks/alevelphilosophy/data/A2/Epistemologyandmetaphysics/Universals.pdf.

(I will just add, Serban, that I think there is a circularity in Lacewing's final account of realism: on what grounds can the realist say it is this 1 kg. weight that was weighed? This would be recognition, i.e. a use of a "resemblance" universal, not a "causation" universal.)

Serban,

Allow me also to be courteous enough to mention the texts to which I originally alluded. They are considered by some to be classics of 20th century analytical philosophy. Peter Strawson's book 'Individuals: An Essay in Descriptive Metaphysics' and Wilfrid Sellars' paper 'Empiricism and the Philosophy of Mind' collected in his book 'Science, Perception and Reality.' The latter, in what is now no doubt a poignant metaphor, has been described as one of the "twin towers" of 20th century philosophy, the other being Quine's paper 'Two Dogmas of Empiricism.'