Can computers engage in concept-formation?

[Bowzer]

...for the same reason that the above-mentioned tornado, no matter how close the model gets to approximating the real thing, will never destroy a real trailer. 

And that's exactly it. But cognitive scientists do not use scare quotes when they peddle their models. They are dead serious when they say that this or that particular model is "conscious."

I grant that computer models and simulations can speak for our understanding of a certain fact. I also grant that knowledge can be gained from their use. Neural networks, for instance, have demonstrated principles exhibited by real biological synaptic networks. There is real science to be found in some of these simulations.

The context of cognitive science (call it "consciousness research"), however, is fundamentally corrupt as Mr. Speicher and I have both noted. Just as a simulated tornado cannot blow over a real trailer so a simulated brain cannot produce a real mind. Even taken to the extreme--as some here have suggested, e.g., computers the size of the Milky Way--a simulation is just a mathematical or object-oriented model of a real phenomenon. To actually re-produce a phenomenon, you have to sufficiently replicate its identity in reality.

Let's say that I want to reproduce a sheep. Further, pretend that I can build a super computer the size of the galaxy and I simulate a sheep in every detail. My simulated sheep leads a virtual life (virtually) every bit as complex as a real sheep. My second experiment is just like the example of Dolly (everyone knows Dolly). Which is the experiment that actually reproduces a sheep? Or are they both examples of real sheep? Neither?

[Edit: added the following]

This point is implicit in the above but I want to make it crystal clear. Computers don't intrinsically model anything. It is only in the mind of the scientist that the image on the screen or the readouts from the program model anything.

Edited May 14, 2004 by Bowzer

[amagi]

These analogies are not valid.

Both a sheep and a tornado are physical objects. A mind is not a physical object, though it depends on physical processes to act.

Mental action formed by the processes in some future computer, if indeed such a thing is possible, would be merely the same kind of information processing that occurs during the physical processes in a brain. There would simply be a different physical mechanism underlying the same essential process.

The sheep analogy is exactly equivalent to saying:

no real mental action can be occuring in my brain, because

when I think of a sheep, it's not a real sheep in my brain.

[Bowzer]

These analogies are not valid.

Both a sheep and a tornado are physical objects.  A mind is not a physical object...

It wasn't meant to be a literal analogy. I am trying to point out the difference between simulation and replication.

For something to possess consciousness it must replicate, say, a nervous system to a sufficient degree. A simulation on a computer screen, no matter how complex and how well it models the causal powers of the brain, will never lead to consciousness. SheepSIM1.0 is not really a sheep. Dolly is a sheep. Just a simple point.

[amagi]

As I see it, the source of my disagreement with you on this particular point is that I maintain that the source of the "causal powers of the brain" must be:

the capacity of a system of neurons to process information.

Therefore, a conscious computer would not be "modelling" or "simulating" the information-processing capabilities of the brain.

It would be performing the information processing itself.

Apparently your argument is that we have to assume, until evidence arises to the contrary, that other properties of living brain cells are necessary to support consciousness beyond their information processing abilities, because living brains are the only things we know of that can now create consciousness.

But all I see here is an argument from correlation. It's therefore arbitrary to assume that some other, unknown biological function of brain cells is required simply because brains happen to be biological. What we know is that the brain uses biological means to accomplish the end of information processing. Accomplishing that end is the reason brains evolved. I strongly suspect that a man-made device built to accomplish that same end can lead to the same result--consciousness--acheived by the brain.

Now, another argument of yours seems to be that it is arbitrary to assume that a non-biological substrate can support consciousness until we have evidence to show that it could. But the evidence is already contained in this post!!

Our evidence of the brain leads to the conclusion that, as far as we know, it is ability of the brain to process information that leads to consciousness. For example, the manner in which animals process sensory information is relatively well-understood (compared to, say, the higher order functions of the human mind). We know that there are information-processing pathways--neural circuits--in the brain that lead from raw sensory input to the conscious state of perceiving an image, a sound, etc. Of course, a vast amount remains to be understood about the brain, but at this point there is every reason to believe that similar processes probably underlie the whole of the source of consciousness. (Arguments about human volition and the like aside--only simple animal consciousness is necessary at this point.)

We know that we can build sophisticated devices that can themselves both receive sensory data and perform complex information processing. We also know that the science of computing will advance phenomenally in the coming centuries, or decades for that matter.

Therefore I do not think it's arbitrary to hypothesize that computers could be conscious. The hypothesis is based on evidence.

Bowzer, what other kind of evidence would you demand? How could anything, short of being presented with a conscious computer, convince you that consciousness could be supported by a computer? After all, demonstrating any sort of "intermediate stage" to a conscious computer would prove nothing--it's still not conscious, so you could still claim that it's not evidence. What would qualify?

[stephen_speicher]

Are you serious?

Yes, I am quite serious. And, I truly wish that you would treat your own actions with the same degree of seriousness that I treat mine.

An honest approach would have been:

"Speicher said A, B, and C, and from this I conclude D."

Instead, your approach is:

"Speicher said D."

You are putting words in my mouth that I did not say. What you conclude is not the same thing as what I actually say. I am responsible for the things that I say. I am not responsible for the conclusions which you draw from the things that I say. Putting words in my mouth that I did not say is misrepresenting me, and that is dishonest.

I will not discuss issues with you because I refuse to sanction that sort of behavior, and I will not give you any fuel for further misrepresentations of my views.

This is completely ridiculous, and I'm through with you.

I agree with the first part of your sentence, but since I cannot control your behavior I sincerely hope you abide by the second part. I do not enjoy having to post public disclaimers warning others about misrepresentations of what I say.

[Bowzer]

As I see it, the source of my disagreement with you on this particular point is that I maintain that the source of the "causal powers of the brain" must be:

the capacity of a system of neurons to process information.

Yes, I emphatically disagree with this point. And just as an aside, not even all cognitive scientists agree with you. Paul Smolensky, for example, has been arguing that connectionist models show the information processing "paradigm" to be false.

Bowzer, what other kind of evidence would you demand?

My demands are to pretty much scrap cognitive science (do I even need to mention current philosophy of mind???). Consciousness should be researched from within fields like biology and psychology. Most importantly, it should be done under the guidance of a proper philosophy (i.e., Objectivism of course ).

I have studied more than enough research in cognitive science to know that there is no evidence to support the conclusion that anything other than living organisms are/can be conscious. In fact, the more reading I do in the field, the more certain I am. Until my demands are met, there can be no substantive progress made. There are a remarkable few who will work within the field and actually discover knowledge but they will be few and far between. That about ends it for me on this topic at least until someone makes it interesting again.

[amagi]

Bowzer, I wonder if you might answer some questions.

Are you saying that brains do not engage in information processing at all?

Or that they do, but that information processing is not involved in any way in giving rise to consciousness?

What do you mean when you say information processing?

What name would you give to the process by which brains integrate basic sensory data into more coherent forms?

[Bowzer]

Are you saying that brains do not engage in information processing at all?

Yes, that is what I am saying. Further, I don't believe that you can even say that computers process information. "Information" is a concept of consciousness and to apply it to physical things is a form of reification.

What do you mean when you say information processing?

I don't say much at all about "information processing" since it is not a term that I favor using because of the way it is used/abused in fields like cognitive science and artificial intelligence. Its roots are in representationalism which is a terrible modern version of Kant.

What name would you give to the process by which brains integrate basic sensory data into more coherent forms?

I call that "perception."

[stephen_speicher]

Yes, that is what I am saying. Further, I don't believe that you can even say that computers process information. "Information" is a concept of consciousness and to apply it to physical things is a form of reification.

I want to underscore just how correct Bowzer is in this identification. This loosely held attribution of "information processing" reflects those views which have never grapsed the proper relation between consciousness and the brain. Unfortunately, "those views" run rampant in the cognitive fields, and are often swallowed whole by otherwise innocent students. This uncritical approach just begets the next generation of myth perpetuators, and the proper role of consciousness is never understood. Hence the corruptness of the field. One consequence of this mistaken "information" crowd, is that it inevitably implies, as anyone who surveys the current literature can see, the notion of biological "information" applying to arbitrary causal connections, not necessitated in the nature of the genes and the proteins which they code.

An interesting historical sidelight: Claude E. Shannon's 1948 "A Mathematical Theory of Communication," The Bell System Technical Jounal, Vol. 27, pp. 379-423, 623-656, July, October, 1948, is usually considered to be the first infiltration of information theory into the field of neurobiology. The fact is, two decades prior, in 1928, that future Nobel Laureate Edgar Douglas Adrian connected "information" with the content of a nerve impulse, in his "The Basis of Sensation: The Action of the Sense Organs," W. W. Norton, 1928. Adrian was awarded the 1932 Nobel Prize in Physiology or Medicine (which he shared with Sir Charles Scott Sherrington), for "their discoveries regarding the functions of neurons." Adrian was the first (along with Yngve Zotterman) to record action potentials in a sensory nerve.

[Bowzer]

Fascinating, Stephen! Thanks for the leads!

[stephen_speicher]

Fascinating, Stephen! Thanks for the leads!

You're welcome.

And, if by "leads" you mean the 1928 Adrian reference connecting "information" with the content of nerve impulse, I learned of that from a paper I read from about six months ago. If you are interested, I can dig up the reference for you.

[jimmay]

Yes, that is what I am saying.

I would infer from this that the *mind* or consciousness is what processes information? otherwise this is a surprising statement.

[amagi]

Bowzer, in other words, you object to describing the kind of computation that occurs in either brains or computers as information processing, because the term ascribes to the "information" being processed a level of reality it does not possess.

I agree that reification of the data processed by computational devices is an error, but I'm not at all convinced that the term "information processing" must be categorically shunned in relation to computers or brains.

It is simply necessary to realize that the concept of information processing is only an abstract model of the computation that occurs. I don't accept that the word information can only refer narrowly to consiously-held knowledge.

I grant that there is a point in wanting to avoid a package-deal with the word, and that the distinction should be identified more explicitly when referring to information processing machines.

But if you reject entirely the use of the word "information" as a model for physical phenomena, you must abandon a large range of useful applications of the word. I must assume that you also reject the idea that a CD or a memory chip is an "information storage device." You must reject the idea that you could ever put information down on paper, or that a book could ever contain information (or that a computer screen could). And you must reject even that spoken words contain information.

I assume that you would say that a book or a word conveys information, but does not "contain" it. And you would be right, if information was defined only as conscious knowledge. But my point is that the word also applies quite validly, and always has, as a conceptual model that can refer to something other than conscious knowledge. To try and erase this aspect of the word is an awkward and artificial manipulation of language.

What would you replace "information" with? You must acknowledge that computers still process something? Is it data? Data is a synonym for information. Computers compute. So what is it they're computing?

Again, I'm not disputing that there should be a clear distinction made between the two meanings of "information," or that philosophical confusion has resulted from the lack of such a distinction. I just don't see any grounds to ban the word information from discussions of computing or of the brain.

The situation is a bit like the use of "infinitesimals" in math. The idea of an infinitesimal length is a useful concept in mathematics, even though reification of such infinitesimal lengths is an error. Much confusion has no doubt resulted from a general failure to identify the fact that the concept is only an abstract model, but I do not think the solution is to abandon and forbid the idea of infinitesimals itself.

[amagi]

At any rate, regardless of what you choose to call the act of computation performed by information processing devices--my point remains:

the "causal powers of the brain" (in giving rise to consciousness) must be:

the capacity of a system of neurons to compute.

Therefore, a conscious computer would not be "modelling" or "simulating" the computational capabilities of the brain.

It would be performing the computation itself.

Also, Bowzer, when I asked you, "What name would you give to the process by which brains integrate basic sensory data into more coherent forms?" you responded:

I call that "perception."

I should not have to point out that this answers little. How does perception occur? Processing of sensory data perhaps? Neural computation of input? What name would you give to this mechanism?

[Bowzer]

I would infer from this that the *mind* or consciousness is what processes information?  otherwise this is a surprising statement.

That is exactly what I am saying.

[Bowzer]

Bowzer, in other words, you object to describing the kind of computation that occurs in either brains or computers as information processing, because the term ascribes to the "information" being processed a level of reality it does not possess.

"Computation" presupposes information so I wouldn't describe the brain (or computers for that matter) as doing computation. Brains don't count and they don't do algebra.

A computer used to be a human being who performed these tasks and it was an innocent metaphorical application at first to apply the word "computer" to a big box of vacuum tubes. That metaphor has now become a literal belief for almost everyone doing work in the cognitive sciences these days (thanks to Stephen I have even learned that "hard" sciences like neurobiology use the term literally).

I'm not saying that we should completely shun the word all together. What I am saying is that in philosophy we need to insist on strict definitions. "Information" is a loaded term in this respect and it is the source of an immense number of errors in the cognitive sciences.

I have a lot to say about "models" and their practicality but I want to leave the discussion to this point which a lot of people get wrong.

[Bowzer]

I should not have to point out that this answers little.  How does perception occur?  Processing of sensory data perhaps?  Neural computation of input?  What name would you give to this mechanism?

Those are not philosophical questions but they are a fascinating topic of study.

[amagi]

"Computation" presupposes information so I wouldn't describe the brain (or computers for that matter) as doing computation. Brains don't count and they don't do algebra.

This is the same as saying that a clock doesn't really tell time--all it really does is move around gears and springs.

Sure, that's true, technically. But it doesn't mean that it's somehow philosophically corrupt for us to say that clocks tell time, any more than it is to say that computers compute.

It is simply necessary, as you said, to strictly define the words and concepts involved.

Are there any words in the English language left for whatever it is a computer actually does, or are they all invalid? You have challenged my terminology but you have never addressed the point I'm trying to make, which I'm sure is quite clear despite my use of the word "compute."

The point is that a conscious computer would not be "modelling" computation--it would be computing.

Replace "computation" with whatever word you deem appropriate for the processing of sensory data. (sensory impulses? sensory input? are these ok?)

[Spearmint]

Can a book contain information? If I write down information onto a piece of paper, can we not say that the paper has information written on it? Likewise if I enter information into a word processor, can we not say that the word processor document contains information? When I close the word processor down, does the information remain in the computer? If not, where does it go? If we can still call it information, then surely the computer is 'processing information' when it performs operations on it.

Surely we can sensibly talk about non-conscious entities storing a representation of consciously interpreted/derived information without suggesting that the non-conscious entity is somehow 'self-aware' while manipulating it.

[Bowzer]

This is the same as saying that a clock doesn't really tell time--all it really does is move around gears and springs.

Sure, that's true, technically.  But it doesn't mean that it's somehow philosophically corrupt for us to say that clocks tell time, any more than it is to say that computers compute.

It isn't invalid to say that a clock keeps track of time but to hold that time is "out there in the clock" is invalid. To say that sense organs process information is OK as long as you don't hold that the information is intrinsic to the process of transduction itself. The information processing view of consciousness holds that information is intrinsic to the brain, that information is "out there in the physical world."

Are there any words in the English language left for whatever it is a computer actually does, or are they all invalid?

There are plenty of valid terms left to describe what computers do: execute programs, store data, transfer data, etc. I wouldn't even have a problem saying that computers process information as long as you don't mean that there are little 1's and 0's in a computer's memory and that the computer manipulates that information (i.e., symbols). But the information processing view holds exactly that; it claims that information and symbols exist apart from consciousness.

Replace "computation" with whatever word you deem appropriate for the processing of sensory data.  (sensory impulses?  sensory input?  are these ok?)

Look, the metaphysics of consciousness is a very tricky subject. If there is any part of philosophy that must demand precise meanings in the terms that it uses, then it is here. Call me a stickler but I'm sticking to precise meanings of words.

[Bowzer]

Can a book contain information? If I write down information onto a piece of paper, can we not say that the paper has information written on it? Likewise if I enter information into a word processor, can we not say that the word processor document contains information? When I close the word processor down, does the information remain in the computer? If not, where does it go? If we can still call it information, then surely the computer is 'processing information' when it performs operations on it.

Strictly speaking, No, No, No, No, Information is not spatial so it can "go" nowhere, and finally, no we cannot still call it information.

Surely we can sensibly talk about non-conscious entities storing a representation of consciously interpreted/derived information without suggesting that the non-conscious entity is somehow 'self-aware' while manipulating it.

Self-awareness has never been a part of my argument...it's quite obvious that a non-conscious entity could not be self aware. What is apparently hard to grasp is that non-conscious entities do not contain, manipulate, create, understand, or destroy information. There are hard drives with magnetic charges, pieces of paper with ink on them, etc. but without a conceptual consciousness there to make use of these things, there is no information there to speak of.

[Spearmint]

Strictly speaking, No, No, No, No, Information is not spatial so it can "go" nowhere, and finally, no we cannot still call it information.

Well yes, information would be any collection of data which is meaningful to an intelligence. However, if I go over to my bookshelf and pick up a random book, I know in advance that what I find in it will be information (as in my consciousness will process the data contained within in an informative way). Similarly, I know that when I view the internet, I am going to receive information from it. If a non-conscious entity contains data which we know a consciousness will interpret as information, I dont really understand the problem with saying that said entity contains information. I do understand where youre coming from, but I'm not sure theres a need for such a strong dichotomy.

edit: Just to make things doubly clear, I do agree with you that "information isnt out there in the world". What exists in the world is data - information is a result of the processing of data by an intelligence. My point is that when we _know_ a certain collection of data is going to yield information to a consciousness, it seems acceptable to say that the data contains information (you could also say it 'is' information, but I dont like that for several reasons). In any case, I think that we definitely need a word to identify data which we know/think is meaningful (or likely to be meaningful), as opposed to data in which we have so far found nothing of meaning nor have reason to believe that we will. The two concepts are very dfferent, and hence deserve seperate names. Since the word 'information' has been adopted by many to describe the first, it seems easier to just stick with it.

[Bowzer]

O, it wouldn't really be a problem if it weren't for the complete corruption of words like "information" and "computation" in the cognitive sciences. I have been arguing against the information processing view of consciousness which is by far the predominant view in fields like artificial intelligence and cognitive science and it is even permeating fields like psychology and biology. It is for this reason that one shouldn't use these words in this context or at least to qualify their meaning.

I don't jump on my friends when they say that my computer crashed because it has some "bad information" on it. If my friend were, say, Daniel Dennett (yeah right....) then I would jump on him.

[stephen_speicher]

The information processing view of consciousness holds that information is intrinsic to the brain, that information is "out there in the physical world."

Yes, and that illustrates the worst danger of that way of thinking. We had a debate surrounding these subjects a number of years ago on HBL, and for me it was one of those nice times where someone convinced me differently from my previous thinking, or, at least, helped to sharpen my understanding. There are a very nice few words by Ayn Rand in Letters of Ayn Rand, p. 529, in the "Letters to a Philosopher" section.

"Clocks do not measure time. Clocks are a mechanism that produces a certain kind of motion (a uniform motion, of unvarying speed, gauged in a certain manner to the motion of the sun); by taking the clock's motion as the unit or standard of measurement, we measure time."

This ties in well with Ayn Rand's remarks on measurement in ITOE, p.8:

"Measurement is the identification of a relationship—a quantitative relationship established by means of a standard that serves as a unit."

Note the use of the word "identification," which underscores that measurement is a cognitive process, not a process to be attributed to a clock. Bowzer's comment cautioning about those who hold "information" as being intrinsic, is right on.

[Godless Capitalist]

I am arguing for something even stronger than the relationship of consciousness to the brain. I am arguing for its inextricable link to life as such. (I am not arguing that this is part of Objectivism but I definitely consider it to be compatible with Objectivism.)

Consciousness is a sub process of life (like digestion or respiration). It exists due to its evolutionary value for those organisms possessing it. Its sole purpose is for value-satisfaction. I still fail to see how you could ever tie the faculty of consciousness to something that is not alive.

Your argument here seem to be circular. You are assuming that a robot cannot be alive and thus cannot be conscious. But that rests on the idea that only biological entities can be alive, which has not been proven. Just because all of the examples of life we have are biological does not mean that non-biological life is impossible. If it were possible to create a conscious robot that had to act to preserve its existence, it would be "alive" according to the Objectivist definition. Now as discussed above such a robot is an arbitrary speculation, but it is certainly not impossible by definition as you seem to be claiming.