I think I have a big problem with the "chinese room"

[Dark Jaguar]

For those who aren't aware, the "chinese room" is a thought experiment some philosophers like to toss out saying that even something that fully functioned like a concious being, it could still not actually be aware.

It basically gives an example of some massive city smashing robot with a box inside containing just a single man cut off from the world and a big book o' instructions. The man gets a set of chinese characters (he doesn't know chinese) which he then just matches up in the big book and from that he writes down another set of characters, which he still doesn't understand, and feeds them back into the machine. In this way, the robot, to the outside world, appears to fully respond in an intelligent fasion, but it's all just instructions from a book. The man and book can be replaced with a computer program to do the same function.

Anyway, I just wanted to say I think I've finally put into words my biggest problem with this thought experiment. It presupposes that such a book is actually possible, one that so perfectly accounts for set stimulii that it can always give a rational answer. It also supposes that all stimuli can be converted into a set of codes for the man to recieve. I think this critical assumption is false, and I conclude this from my experience with all the "chat bots" I've ever tried out online. It's not enough to just take CURRENT stimulii and respond to them. One can recieve exactly the same stimulus and would be expected to respond a different way depending on context. MEMORY, and the ability to take those past patterns into consideration, is just as important, if not more so. A bacteria can respond for "the moment" but it can't change. A human can. The difference is memory. There is no way ANY static book can respond to a question like "what were we just talking about?". In fact, that is the question I always ask a chat bot to see if someone has finally coded memory of previous topics of conversation in a REAL way into the program. A book wouldn't even be able to do as well as a chat bot that can store previous topics and add them as sentences to a database. It would fail because it would only have ONE specific response to that specific and exact stimulus. Someone can ask you "what were we just talking about?", with every single aspect of that situation exactly the same, but the answer would be different due to what happened in the past. Any attempts by a philosopher to modify this example would require a system that can modify and develop this book as new experiences show up, and the more I think about it, the more it is clear the example would basically BE a concious entity at that point, only with a person tossed in on top of that.

So basically what I'm saying is I think the "chinese room" thought experiment is fundamentally flawed in it's very conception. Score 1 for functional awareness.

[Jekyll]

Any algorithm can be written down (Church-Turing thesis), so any AI could (theoretically) be enclosed in a very large book that takes into account past data. In the cases you are worrying about the instructions would say if criteria X, Y , Z have been satisfied respond with statement J which uses element I of previous conversation.

Basically, the man is allowed to make notes (which he need not understand) according to instructions contained in the book and refer back to them when told to do so.

My objection is that, just like the man in the box, none of my individual brain cells would understand this conversation. Therefore (according to this argument) I am not aware.

[Dancing David]

I am not sure of the point of the argument in the first place. If something behaves as though it is conscious then it is conscious. If there is a self modifying book that contains all the instructions, and can sit down to tea and crumpets with the queens and then go out boot scooting and fall in love with it's partner at the bar, then it is conscious by observation. Now the conscious residing somewhere might be a problem, we are going to have to say that the man in the robot can write some really complex notes and refer to them really quickly to be able to translate all the stuff into the ciphers for the robot.

But functionaly the problem is that you have the human in the middle. That person is conscious and preforms the learning algorith for the computer. How does the robot sense and process information when he has tea and crumpets?

[Ichneumonwasp]

I think I might add one other aspect -- feeling and emotion. When we say that we know something, we seem to have a feeling -- yeah, that's right. We match either the internal and external or new information to memory. The Chinese room also leaves out that aspect of consciousness as though it were magical. It is not.

Essentially, Searle described a much lower level sort of processing than what we call consciousness. In his defense, his argument was really that a computer program couldn't do it, not that a computer couldn't do it if it were linked in a way similar to our brains.

As many seem to do, he leaves out qualia, as though feeling and emotion cannot be explained by mechanical processes. I have always found this ironic, since it used to be reason and rationality that made us human and unique while emotion was bestial. Now it seems feeling is "unexplainable". Except that we are beginning to explain it. The reason that philosphers rush to it and call it by a new name (qualia) is because feeling provides a god of the gaps -- it was always thought beneath us, so few ever investigated emotion/feeling.

[Paul C. Anagnostopoulos]

I am not sure of the point of the argument in the first place. If something behaves as though it is conscious then it is conscious.
We could debate this assertion until the cows come home, but I'm not sure it has anything to do with the Chinese room, where the AI is only translating Chinese. If you ask it whether it's aware/conscious, it would probably respond with "Input some Chinese, please," and you'd know it wasn't conscious.

~~ Paul

[Ladewig]

It basically gives an example of some massive city smashing robot with a box inside containing just a single man cut off from the world and a big book o' instructions.

I've heard "Chinese Room" descriptions before but they were always rooms and not giant robots. Where did you get that example and what is the purpose of putting it in a massive robot?

[Ladewig]

We could debate this assertion until the cows come home, but I'm not sure it has anything to do with the Chinese room, where the AI is only translating Chinese. If you ask it whether it's aware/conscious, it would probably respond with "Input some Chinese, please," and you'd know it wasn't conscious.

~~ Paul

Would you? If you asked a person that same question and the person responded with "I prefer speaking in Chinese" that would not be evidence that the person was not conscious.

[Correa Neto]

Imagine two robots, one being that from the thought experiment and the second piloted by a human who can operate it with no textbooks. How would you know wich giant robot is driven by a consious being actually taking decisions?

One can even go further, since the chinese characters from the first robot are instructions created by a sentient being... Thus this being is the robot's self, in a certain way...

This experiment has many points in common with the "homunculus fallacy" and the P-zombies, including their limitations.

[Corpse Cruncher]

What is awareness how is that classified? Is a dog aware, am I self aware? Is a tree aware? This is what gets me how do you define and know what is and what is not. Who decides what to toss in or out and on what parameters?

[cyborg]

I hate philosophers. They have to take such a simple proposition and make it into a ludicrous story.

The question is:

If f(x) is equivalent to g(x) and we ascribe the property 'consciousness' to f does g also have this property? Now, as you say, such a book cannot be constructed as the above but we can surely posit a mathematical function that is equivalent to any human's input/output responses. We can even have memory as a function of the input - no problem - we adjust our functions to f(x,m) -> (y,n) and g(x,m) -> (y,n).

Either way if we are talking at a functional level if f and g are equivalent functions any properties we assign to f we must assign to g.

Hence I conclude that we do not consider consciousness a property of function but of computation - but one should also note that the construction of the computation as posited is only possible to define, not compute.

As such anyone talking about consciousness needs to do so in terms of the internal computational process and not the input/output and state.

[Taffer]

We could debate this assertion until the cows come home, but I'm not sure it has anything to do with the Chinese room, where the AI is only translating Chinese. If you ask it whether it's aware/conscious, it would probably respond with "Input some Chinese, please," and you'd know it wasn't conscious.

~~ Paul

I know this is not really the topic of this thread, but if you'll allow a slight derail for a moment:

What is your position here? Are you saying there is a difference between a human and a robot which acts in every way like a human?

[c4ts]

This looks more like something for programmers than philosophers. They might say attempting to create a self-conscious AI is like trying to use the Chinese room to manufacture conversation.

Another problem with the analogy is the man inside. He could pick up this Chinese code language on his own, based on his experience with the box, and start formulating his own responses. He'd fix the book, and the responses would get all screwed up according to his model. Then outsiders could tell the difference between him, and the box with the guy who understands Chinese.

[Beerina]

My objection is that, just like the man in the box, none of my individual brain cells would understand this conversation. Therefore (according to this argument) I am not aware.

Well, I don't think that was quite it -- the point of the thought experiment was that you could accomplish self-awareness purely by non-sentient symbol pushing. That behind the scenes, there was no "understanding" going on except by the system as a whole.

And, of course, you could put in algorithms and memory to compensate -- but you wouldn't have to. Although it would be impossible in practice, you could in theory have 10^^3000 pages of lookup tables to contain every conceivable conversational input you may receive. Humans talk a lot, but there are only a finite number of word combos of any given length. If you plotted out a response to, say, all possible combinations a hundred billion words long or less, that would be huge, but would cover every possible thing someone might say over the course of their entire life.*


* Although descriptions of algorithms would open that can of worms rather easily. But adding simple processing of algorithms as an ability the human in the box pushes around without thought buys nothing in the "bringing in" the "understanding".

[Jekyll]

Well, I don't think that was quite it -- the point of the thought experiment was that you could accomplish self-awareness purely by non-sentient symbol pushing. That behind the scenes, there was no "understanding" going on except by the system as a whole.
Traditionally, the Chinese room experiment is used as an argument against AI, even if a computer appears to understand what's going on it doesn't really as neither the book nor the person understand what is happening.

Of course,as I said, the same holds true for splitting the brain up into small enough pieces, eventually none of the bits will understand what's going on. Then again, these arguments tend to be made by dualists and idealist who don't really care for that sort of thing.

And, of course, you could put in algorithms and memory to compensate -- but you wouldn't have to.
True.

[roger]

well, the point of philosophical arguments like this is to reduce the problem without losing the essentials. So, don't worry if the chinese room has memory - assume it does. Same with hormone inputs, knee pain inputs, all that jazz. The idea is to reduce the problem down to something you can talk about in one page, not categorize and explain every single sensory input in the entire human body.

Where Searle falls seems obvious to me; I'm surprised this has ever been treated as a serious argument in philosophy. But perhaps I am missing something. His essential argument is that the person moving the symbols around inside the room does not understand chinese or the conversation, nore do the books (or whatever inanimate objects we have to assume to make the thing work like a human), so there can be no understanding or consciousness in the system.

But of course parts of the system are not conscious. My individual brain cells are not conscious, nor does the materialist position require them to be. The system, or more accurately, the process, the sum total of the machinery, inputs, and states over time has consciousness. This, of course, is not proven yet, merely assumed (with strong evidence), but Searle's argument does not disprove the possibility. Certainly he has proven that a nerve cell is not conscious. Fortunately that is not our (science's) position.

[Paul C. Anagnostopoulos]

Would you? If you asked a person that same question and the person responded with "I prefer speaking in Chinese" that would not be evidence that the person was not conscious.
Sorry, I didn't mean to say that that one response would be enough. But it would make you suspicious and you could start asking enough questions to soon realize that the AI had none of the self-aware responses of a normal human.


What is your position here? Are you saying there is a difference between a human and a robot which acts in every way like a human?
No, but we are talking about a robot that can do nothing other than translate Chinese, aren't we?

Regarding your question, there could be a difference between a human and a robot if consciousness turns out to be an epiphenomenon. However, we couldn't tell that there was a difference. I'm not too concerned about this, since I'm having trouble locating a coherent description of consciousness as epiphenomenon.

~~ Paul

[Paul C. Anagnostopoulos]

Don't anyone dare bring up Mary and the black/white room!

~~ Paul

[Dancing David]

What is awareness how is that classified? Is a dog aware, am I self aware? Is a tree aware? This is what gets me how do you define and know what is and what is not. Who decides what to toss in or out and on what parameters?


Exactly!

That is the whole point of the Chinese room, there is this concept out there that consciousness is some magical and transcendant quality , and that thoughts and qualia are some how not material things.

If the chinese room behaves as though it is conscious, if a p-zombie behaves as though it conscious, it is conscious. So if something behaves as though it is aware then it is aware.

[Dancing David]

I hate philosophers. They have to take such a simple proposition and make it into a ludicrous story.

The question is:

If f(x) is equivalent to g(x) and we ascribe the property 'consciousness' to f does g also have this property? Now, as you say, such a book cannot be constructed as the above but we can surely posit a mathematical function that is equivalent to any human's input/output responses. We can even have memory as a function of the input - no problem - we adjust our functions to f(x,m) -> (y,n) and g(x,m) -> (y,n).

Either way if we are talking at a functional level if f and g are equivalent functions any properties we assign to f we must assign to g.

Hence I conclude that we do not consider consciousness a property of function but of computation - but one should also note that the construction of the computation as posited is only possible to define, not compute.

As such anyone talking about consciousness needs to do so in terms of the internal computational process and not the input/output and state.

That is nice, some processes can be accessed through discussion, but others can only be tested through observation. thgat is how we now know that babies and some animals have a rudimentary mathematical sense.

And while we can speculate on the computational process in a humans head, we can only study it through external behaviors.

Great post!

[Freddy]

I think there is some confusion here about what the Chinese room argument purports to show. The question the thought experiment asks is this: does the man in the room understand Chinese? He chooses all the right outputs to correspond with the inputs. That is, he shows all the outward signs of understanding Chinese. If you give him an input in Chinese, he spits out a response in Chinese, one that makes sense. You could have a conversation with him in Chinese.

But the man in the room chooses his responses by blindly following an algorithm. The Chinese characters are just symbols to him -- he has no idea what, if anything, they are symbols for. The man in the room is just like a computer: he is performing formal operations according to an algorithm, with no knowledge of the semantic content of the symbols he is manipulating (if they even have any semantic content). That is just what a computer does.

If to understand Chinese means something over and above what the man in the room is doing, then no computer running a program could ever be said to understand Chinese. And since it makes no difference to the argument whether the symbols in question are Chinese characters or variables or numbers, the argument can be generalized to conclude that a computer can "understand" if and only if "understanding" is nothing more than the ability to perform formal manipulations of symbols according to the proper set of rules. If this is what it means to understand, then the man in the Chinese room understands Chinese. If understanding requires knowledge of what the symbols stand for, then no computer can be said to understand anything, because computers perform only formal operations. A computer does not know the meaning of any of the symbols it manipulates, or even whether those symbols have any meaning. It just blindly follows an algorithm that matches outputs to the inputs it is given according to a set of rules.

[Dark Jaguar]

Thanks for the comments. The whole "it needs to be thought of in terms of computation" comment pretty much nails it on the head. What possible method using nothing but input, output, and a big static book could actually produce full conciousness. If the human is allowed to take notes and learn from the past, the example is voided right there and you have an adaptive rather than prescriptive AI, which I'd call self aware even if someone jammed a human into the process.

As far as I can tell, these sorts of flawed analogies seem to be invented in some attempt to support a presupposed conclusion (robots can't be aware in and of themselves). Well that's precious. Also, I'm using "robot" in the stereotypical big metal man walking around doing chores and plotting to overthrow humanity sense, so don't bother pointing out the unthinking car assembly bot.

[Dark Jaguar]

I think there is some confusion here about what the Chinese room argument purports to show. The question the thought experiment asks is this: does the man in the room understand Chinese? He chooses all the right outputs to correspond with the inputs. That is, he shows all the outward signs of understanding Chinese. If you give him an input in Chinese, he spits out a response in Chinese, one that makes sense. You could have a conversation with him in Chinese.

But the man in the room chooses his responses by blindly following an algorithm. The Chinese characters are just symbols to him -- he has no idea what, if anything, they are symbols for. The man in the room is just like a computer: he is performing formal operations according to an algorithm, with no knowledge of the semantic content of the symbols he is manipulating (if they even have any semantic content). That is just what a computer does.

If to understand Chinese means something over and above what the man in the room is doing, then no computer running a program could ever be said to understand Chinese. And since it makes no difference to the argument whether the symbols in question are Chinese characters or variables or numbers, the argument can be generalized to conclude that a computer can "understand" if and only if "understanding" is nothing more than the ability to perform formal manipulations of symbols according to the proper set of rules. If this is what it means to understand, then the man in the Chinese room understands Chinese. If understanding requires knowledge of what the symbols stand for, then no computer can be said to understand anything, because computers perform only formal operations.

Thanks for the clarification, I did misunderstand, but you also failed to understand my counter argument. That being, that in this case, the man in the room will NOT appear to fully understand Chinese and CAN'T, because if you asked him "what were we just talking about?" the magic book won't have a satisfactory answer. It won't take context and memory into consideration. He'd HAVE to actually understand Chinese to provide a proper conversation that we couldn't tell from "the real thing".

[Taffer]

Regarding your question, there could be a difference between a human and a robot if consciousness turns out to be an epiphenomenon. However, we couldn't tell that there was a difference. I'm not too concerned about this, since I'm having trouble locating a coherent description of consciousness as epiphenomenon.

~~ Paul

Could a chinese-translating machine be considered some form of turing machine?　　It doesn't really matter how the robot works, as long as it behaves like a turing machine, I see no reason why it wouldn't be able to behave consciously.

I, myself, consider consciousness to be a phenomenon of the brain, and any simulation or replication of the brain will exhibit the exact same features. However, I realise this isn't really the point of this thread, so the properties of consciousness I shall leave alone.

I just don't quite understand your point, is all Paul. Why could this system not be considered conscious? Or do you just feel that translation is not enough to completely replicate consciousness?

[Jekyll]

If understanding requires knowledge of what the symbols stand for, then no computer can be said to understand anything, because computers perform only formal operations. A computer does not know the meaning of any of the symbols it manipulates, or even whether those symbols have any meaning. It just blindly follows an algorithm that matches outputs to the inputs it is given according to a set of rules.

However, this line of attack also fails as some symbolic computation, where a computer has knowledge of what the symbols represent, can and do exist.

It can't be turtles all the way down, and there has to be a base line which is just computed without assigning meaning, but the same is true of human thought. You can't describe the intricacies of how you catch a ball; you can only layer a framework, which describes what's going on in broad terms, over the top.

[PixyMisa]

Where Searle falls seems obvious to me; I'm surprised this has ever been treated as a serious argument in philosophy. But perhaps I am missing something. His essential argument is that the person moving the symbols around inside the room does not understand chinese or the conversation, nore do the books (or whatever inanimate objects we have to assume to make the thing work like a human), so there can be no understanding or consciousness in the system.

Correct.

But of course parts of the system are not conscious. My individual brain cells are not conscious, nor does the materialist position require them to be. The system, or more accurately, the process, the sum total of the machinery, inputs, and states over time has consciousness. This, of course, is not proven yet, merely assumed (with strong evidence), but Searle's argument does not disprove the possibility. Certainly he has proven that a nerve cell is not conscious. Fortunately that is not our (science's) position.

Exactly. The entire argument is absurd. The Room is obviously functionally conscious, but Searle can't find the bit where the consciousness lives, so he decides that it's impossible. Might as well pull apart a frog looking for the jump.

I think there is some confusion here about what the Chinese room argument purports to show. The question the thought experiment asks is this: does the man in the room understand Chinese?

That's exactly the problem. The Room understands Chinese; we know that, because it can respond appropriately to arbitrary questions. It's irrelevant whether the man understands Chinese or not - and indeed, it's stipulated from the beginning that he doesn't. But we're not asking questions of the man, we're asking questions of the Room.

Asking whether the man inside the room understands Chinese makes as much sense as asking whether your spleen can play baseball.

[Freddy]

Thanks for the clarification, I did misunderstand, but you also failed to understand my counter argument. That being, that in this case, the man in the room will NOT appear to fully understand Chinese and CAN'T, because if you asked him "what were we just talking about?" the magic book won't have a satisfactory answer. It won't take context and memory into consideration. He'd HAVE to actually understand Chinese to provide a proper conversation that we couldn't tell from "the real thing".


If you asked him in Chinese, he'd have an answer.:) I see your point. You are saying that the man would not even appear to understand Chinese. I don't remember the argument that well (it has been awhile since I read it), but I think the idea is not that you are talking to the man face to face, but that you can talk to him only by inputting symbols into a hole in the wall of the room. He is not available to be questioned in English. And if you think about it, that makes sense. The argument proceeds by making the man's situation analogous to that of a computer running a program. In order to keep the analogy up, we must assume that the only way you can communicate with the man in the room is by giving him an input. After all, you can't ask your computer in English whether it knows what the symbols mean.

[PixyMisa]

We could debate this assertion until the cows come home, but I'm not sure it has anything to do with the Chinese room, where the AI is only translating Chinese. If you ask it whether it's aware/conscious, it would probably respond with "Input some Chinese, please," and you'd know it wasn't conscious.

The Room doesn't translate Chinese, it responds to questions posed in Chinese - that's the setup of the argument. So if you ask it, 你知道吗? (which is probably nonsense because I got that out of Google, but I put in "Are you aware?"), it will answer 是的,我当然 ("Yes, of course I am") or some such. If it doesn't, it's not acting as it's defined to act.

[Paul C. Anagnostopoulos]

I just don't quite understand your point, is all Paul. Why could this system not be considered conscious? Or do you just feel that translation is not enough to completely replicate consciousness?
No, I don't think it is. We're talking about human-like consciousness, right?

~~ Paul

[PixyMisa]

No, I don't think it is. We're talking about human-like consciousness, right?

Yep, human-like consciousness. And I agree that translation would not be a sufficient indicator. But that's not what the Room does!

[Ichneumonwasp]

If the human is allowed to take notes and learn from the past, the example is voided right there and you have an adaptive rather than prescriptive AI, which I'd call self aware even if someone jammed a human into the process.

A human being able to take notes and learn from the past in the way you seem to describe would be a homonculus. The ability to take notes presupposes the ability to prioritize what is important and what is not, how to apply those notes to the situation, etc. In other words, you seem to presuppose consciousness.

[Paul C. Anagnostopoulos]

The Room doesn't translate Chinese, it responds to questions posed in Chinese - that's the setup of the argument. So if you ask it, 你知道吗? (which is probably nonsense because I got that out of Google, but I put in "Are you aware?"), it will answer 是的,我当然 ("Yes, of course I am") or some such. If it doesn't, it's not acting as it's defined to act.
In the spirit of complete disclosure, I must state that I opened a fine bottle of Barbadian rum this evening.

What a twit I am! I thought this robot translated Chinese to English. But I see that it responds to questions asked in Chinese with reasonable responses in Chinese. Why does it have to be Chinese? Why can't it be English?

So please ignore every stoopid thing that I have said in this thread. Well, except that epiphenomenalism in incoherent. That still stands.

~~ Paul

[PixyMisa]

A human being able to take notes and learn from the past in the way you seem to describe would be a homonculus. The ability to take notes presupposes the ability to prioritize what is important and what is not, how to apply those notes to the situation, etc. In other words, you seem to presuppose consciousness.

Not if the note-taking is a mechanical process fully defined in the Book. (When you see this character, write down this entry in row M of column N on page X. If it's followed by this character, write down this in row M2...) The Book contains all the instructions, and there's no reason why those instructions could not include writing down symbols and looking them up again.

Yes, the Book is preposterously impractical in reality, but this is a question of philosophy. Just being preposterously impractical is nothing to a philosopher. ;)

[PixyMisa]

In the spirit of complete disclosure, I must state that I opened a fine bottle of Barbadian rum this evening.

Ah, then you're in the perfect state of mind to discuss the Chinese Room. :)

What a twit I am! I thought this robot translated Chinese to English. But I see that it responds to questions asked in Chinese with reasonable responses in Chinese. Why does it have to be Chinese? Why can't it be English?

Because then it wouldn't be the Chinese Room!!!! :eek:

(Of course, it could perfectly well be English, or any other natural language that the man inside doesn't speak.)

So please ignore every stoopid thing that I have said in this thread. Well, except that epiphenomenalism in incoherent. That still stands.

Indeed.

[Z]

It seems little more than an attempt at forcing a homonculus to be the center of consciousness. Enforced duality, if you will.

The fact that you could remove the human is pretty damning to their point, though.

[Paul C. Anagnostopoulos]

(Of course, it could perfectly well be English, or any other natural language that the man inside doesn't speak.)
Did he choose Chinese just so that English-speaking readers would be forced to think about a language they don't understand?

What if part of passing the Turing test is to have a somewhat random response to questions?

~~ Paul

[PixyMisa]

Did he choose Chinese just so that English-speaking readers would be forced to think about a language they don't understand?

Probably something along those lines.

What if part of passing the Turing test is to have a somewhat random response to questions?

Then that will be encoded into the instructions in the book. The Room is defined so that if you give it a question in Chinese, it will respond appropriately, but that this is accomplished by a man who speaks no Chinese simply following the instructions in a (preposterously impractical) book. If "appropriately" means "as would a native speaker", then that is part of the requirements of the book.

(The question Searle posed was whether the Room understands Chinese - which it clearly does - and not whether the room was self-aware, but it works the same way.)

[Paul C. Anagnostopoulos]

It seems little more than an attempt at forcing a homonculus to be the center of consciousness. Enforced duality, if you will.
Yes, but would the AI have a "feeling of understanding Chinese," which a human certainly has? There may be no homunculus, but there is certainly an overarching feeling of awareness about the things we do.

I'm not a dualist, but there is certainly an unanswered question here: Why does awareness of an activity feel the way it does?

~~ Paul

[Ichneumonwasp]

Not if the note-taking is a mechanical process fully defined in the Book. (When you see this character, write down this entry in row M of column N on page X. If it's followed by this character, write down this in row M2...) The Book contains all the instructions, and there's no reason why those instructions could not include writing down symbols and looking them up again.

Yes, the Book is preposterously impractical in reality, but this is a question of philosophy. Just being preposterously impractical is nothing to a philosopher. ;)

Which is all fine. I thought Searle included this whole contingency in his argument, which is why I brought it up.

It doesn't solve the original problem, though, which is included in the issue of what we mean by "meaning", which is why emotion is central to this. Part of consciousness is prioritization, which involves emotion. Searle left emotion and feeling out, which is part of why, I think, he can't see that the room is conscious (or that consciousness exists in the whole process -- which must include emotion/motivation).

[Paul C. Anagnostopoulos]

Then that will be encoded into the instructions in the book. The Room is defined so that if you give it a question in Chinese, it will respond appropriately, but that this is accomplished by a man who speaks no Chinese simply following the instructions in a (preposterously impractical) book.
But does he need the book and a true random number generator?

~~ Paul

[PixyMisa]

But does he need the book and a true random number generator?

Humans are notoriously bad at generating random numbers, so a list of badly skewed pre-generated numbers that index into different variants of the response patterns would suffice. If the Room is defined to produce human-like responses, then it would start swearing at you after you asked the same question half-a-dozen times anyway, so the list doesn't even need to be very long.

[Paul C. Anagnostopoulos]

Humans are notoriously bad at generating random numbers, so a list of badly skewed pre-generated numbers that index into different variants of the response patterns would suffice. If the Room is defined to produce human-like responses, then it would start swearing at you after you asked the same question half-a-dozen times anyway, so the list doesn't even need to be very long.
But what if certain brain mechanisms produce true random activity? That room guy might need a true random number generator. I wonder if the guy asking the questions could tell?

Remember, rum good.

~~ Paul

[Jekyll]

By definition, there's not going to be a difference between random numbers and sufficiently good pseudo-random numbers. At least not in a finite time span.

[Paul C. Anagnostopoulos]

Okay, so AI might need a good pseudo-random number generator. Check.

~~ Paul

[chriswl]

Where Searle falls seems obvious to me; I'm surprised this has ever been treated as a serious argument in philosophy. But perhaps I am missing something. His essential argument is that the person moving the symbols around inside the room does not understand chinese or the conversation, nore do the books (or whatever inanimate objects we have to assume to make the thing work like a human), so there can be no understanding or consciousness in the system.

But of course parts of the system are not conscious. My individual brain cells are not conscious, nor does the materialist position require them to be. The system, or more accurately, the process, the sum total of the machinery, inputs, and states over time has consciousness. This, of course, is not proven yet, merely assumed (with strong evidence), but Searle's argument does not disprove the possibility. Certainly he has proven that a nerve cell is not conscious. Fortunately that is not our (science's) position.


This is the so-called "systems reply". Searle answers it as follows: suppose that instead of relying on books of instructions you memorised them. This would be an impractically gigantic feat of memory but not logically impossible. Now, the room has been internalised, its all inside your brain, but you still don't understand Chinese, you merely follow rules that you don't understand.

I think the answer here is that the "you" that doesn't understand Chinese is the you that has memorised the Chinese Room rules. But there is now another you, kind of like a virtual machine, that "runs" on top of the original you. He understands Chinese.

[Z]

Yes, but would the AI have a "feeling of understanding Chinese," which a human certainly has? There may be no homunculus, but there is certainly an overarching feeling of awareness about the things we do.

I'm not a dualist, but there is certainly an unanswered question here: Why does awareness of an activity feel the way it does?

~~ Paul

I think, if it were made properly, it would.

In the Chinese room, there is something which 'understands Chinese' - the book - and something which 'has a feeling that ((the whole)) understands Chinese'- the man, in regards to the whole (room, book, man).

Our brains are like a collection of individuals, in essence - some that understand Chinese input, some that reference appropriate responses, some that regulate outputs - and some that have feelings of understanding Chinese. If we isolate that part of the brain - the one that thinks that we know Chinese - from the parts of the brain that DO the language processing, this is no different from the man being removed from his room. He has the feeling that as a whole, the Chinese Room knows Chinese, but removed from the room, HE certainly can't.

[PixyMisa]

I think the answer here is that the "you" that doesn't understand Chinese is the you that has memorised the Chinese Room rules. But there is now another you, kind of like a virtual machine, that "runs" on top of the original you. He understands Chinese.

Yep, that's exactly right. Your conscious mind is acting as the computer for an AI.

How Searle can fail to understand this - how this argument can be taken seriously by any philosopher - is beyond me.

[Paul C. Anagnostopoulos]

What does it mean to "understand Chinese"? The word understand here seems to imply some sort of "feeling of understanding" or "being convinced of understanding." Would the AI have that? I think so, but it's not patently obvious.

~~ Paul

[PixyMisa]

What does it mean to "understand Chinese"? The word understand here seems to imply some sort of "feeling of understanding" or "being convinced of understanding." Would the AI have that? I think so, but it's not patently obvious.

Well, that's the thing. We can question the Chinese Room about its feelings, and it will answer the questions as would a conscious speaker of Chinese - that's part of the definition. Not only will it claim to have a feeling of understanding, but it will be able to describe that feeling of understanding, and answer questions about it.

Searle would suggest that somehow there is a qualitative difference between the way the Chinese Room processes these questions and the way a human would do so, but he presents no evidence for this.

[Paul C. Anagnostopoulos]

So if we say that there is no question we could ask the AI that would not produce an answer compatible with that from a real human, then I see no way that we could ever know if it is "experiencing" its understanding of Chinese the same way a human does. This is true even if all human mental experience is a function of brain processes, with no dualism involved whatsoever.

So then, do we just define the AI as conscious because it's indistinguishable from such, or do we reach a dead end?

~~ Paul

[Taffer]

No, I don't think it is. We're talking about human-like consciousness, right?

~~ Paul

Yes. And I don't see why it wouldn't be. Although perhaps I misunderstand the thought-experiment.

Remove the person from the equation, I do not feel it is important. Instead, replace the person in the room with a computer program which follows a set of rules to give outputs based on inputs. How is this not a turing machine?

[Dark Jaguar]

I speak not a word of English. I merely repeat these sound phonetically.

I saw a sketch about just this thing on Kids in the Hall. However, that illustrates my point.

You see, you are now adding things like a random number generator and taking notes about previous events and that is outside the book itself. We've gone beyond that and now have some big ol' self checking system. However, my point was merely that the book by itself could never actually provide these answers like deviating from specific responses to exact stimulii. Indeed, I hold that if an AI program IS in fact programmed to take previous responses and questions and weave them into something coherent, it's aware. The very act of taking down those notes is one of noting one unique instance which is not actually IN the book but the rules are able to cover only because it is using memory and can consider certain situations.

[PixyMisa]

So if we say that there is no question we could ask the AI that would not produce an answer compatible with that from a real human, then I see no way that we could ever know if it is "experiencing" its understanding of Chinese the same way a human does. This is true even if all human mental experience is a function of brain processes, with no dualism involved whatsoever.

So then, do we just define the AI as conscious because it's indistinguishable from such, or do we reach a dead end?

Well, this is the difference between philosophers and people who have to work for a living.

[PixyMisa]

You see, you are now adding things like a random number generator and taking notes about previous events and that is outside the book itself. We've gone beyond that and now have some big ol' self checking system. However, my point was merely that the book by itself could never actually provide these answers like deviating from specific responses to exact stimulii. Indeed, I hold that if an AI program IS in fact programmed to take previous responses and questions and weave them into something coherent, it's aware. The very act of taking down those notes is one of noting one unique instance which is not actually IN the book but the rules are able to cover only because it is using memory and can consider certain situations.

The only difference is whether you want the Room merely to be able to answer questions - for which the Book alone will suffice - or to conduct conversations.

[RandFan]

Anyway, I just wanted to say I think I've finally put into words my biggest problem with this thought experiment. It presupposes that such a book is actually possible, one that so perfectly accounts for set stimulii that it can always give a rational answer. It also supposes that all stimuli can be converted into a set of codes for the man to recieve. I think this critical assumption is false, and I conclude this from my experience with all the "chat bots" I've ever tried out online. It's not enough to just take CURRENT stimulii and respond to them. I used to be a dualist. Thanks to Mercutio and Darat I've given up the homunculus. Those wre some Great discussions. It can work and people can change. I'm sure there were others btw who were involved in my conversion.

In any event, I came up with the Chinese room long before I knew of the Chinese room. Mine wasn't a Chinese room. Just a group of people who processed data. Some were registers and some were processors some were busses. I could theoreticaly perform any computer function (current or future) with my human version.

I can't, intiutively, reconcile the Chinese room with my newfound materialism. I say this because I'm personaly not moved by your "problem". Please don't take this as a critisism. It could very well be a blind spot or bias on my part. I'm not dismissing you at the present. I'll think about it some more.

Thanks.

[Corpse Cruncher]

Yep, human-like consciousness. And I agree that translation would not be a sufficient indicator. But that's not what the Room does!
But what are the parameter that state human consciousness or even like-human-consciousness. Who defines what it is and what is to be defined as?

[RandFan]

But what are the parameter that state human consciousness or even like-human-consciousness. Who defines what it is and what is to be defined as?This is the classic unanswered question. It was the theme of Artificial Intelligence: AI (http://www.imdb.com/title/tt0212720/) and many other sci-fi movies and books. To date, AIU, the best we can do is come up with tests that might indicate that a machine has human like consciousness (see turing test (http://www.aaai.org/AITopics/html/turing.html)).

[Jekyll]

The only difference is whether you want the Room merely to be able to answer questions - for which the Book alone will suffice - or to conduct conversations.

Well as Beerina was saying you can produce conversation quite easily with a much larger book that contains pre-computed responses to every conversation possible. It's not very feasible but that's not stopped a thought experiment before.

[PixyMisa]

But what are the parameter that state human consciousness or even like-human-consciousness. Who defines what it is and what is to be defined as?

Neuroscientists.

We all have an instinctive understanding of what consciousness is, but some of the results of neuroscience have shown that our instinctive understanding differs significantly from what is actually going on. One famous example is that conscious decisions have been shown (for certain cases) to come a couple of hundred milliseconds after the action supposedly resulting from that decision has begun.

So you actually make the decision on some level, motor neurons begin to fire to act on the decision, and only then do you become aware of having made the decision.

[PixyMisa]

Well as Beerina was saying you can produce conversation quite easily with a much larger book that contains pre-computed responses to every conversation possible. It's not very feasible but that's not stopped a thought experiment before.

Well, yes, but you still need to write something down - or remember it - to track where you are in the conversation. So it's not a reset-to-zero process for every new question.

[Jekyll]

Well, yes, but you still need to write something down - or remember it - to track where you are in the conversation. So it's not a reset-to-zero process for every new question.

Sure, but you could do it with a bookmark or even your finger. It doesn't require any more awareness than answering a single question does. It's just that the input and output are being passed in chunks.
I was aiming this more at D.J. who said that the awareness was coming from the guy writing stuff down.

[Corpse Cruncher]

Neuroscientists. And they what exactly? The whole or some very small fragment of the brain/mind? Is the conscious apart of that gooey grey matter? As far as I was concerned, the brain and its workings, like the ocean, are largely unknown.

We all have an instinctive understanding of what consciousness is, but some of the results of neuroscience have shown that our instinctive understanding differs significantly from what is actually going on. One famous example is that conscious decisions have been shown (for certain cases) to come a couple of hundred milliseconds after the action supposedly resulting from that decision has begun.

So you actually make the decision on some level, motor neurons begin to fire to act on the decision, and only then do you become aware of having made the decision.

Instinct. Are instincts correct and can we trust them to be accurate or provide a basis for accuracy. I don't know why, the same as I don't honestly think I can explain, but I don't think this is the case or can be basis.

A thought or an action thought, 'I think therefore I am conscious'. How does that explain the unconscious or thoughts, for instance, that I have no idea about?

If I consciously observe a problem, such as boiling kettle and my arms proximity to it, meaning if I leave it here I am scalded. Just because I make the decision to move my arm, can that in all honestly be interpreted to being conscious or is it a body reaction?

What it seems to me is, conscious= problem solving. If I can evaluate a situation and remedy it therefore I am conscious and ultimately be consciously self aware.

No - yes?

[Paul C. Anagnostopoulos]

Well, this is the difference between philosophers and people who have to work for a living.
And a warning to those who work for a living that they should be careful not merely to define away the problem of consciousness. This may be the correct tack, but one should be careful nonetheless.

~~ Paul

[Paul C. Anagnostopoulos]

So you actually make the decision on some level, motor neurons begin to fire to act on the decision, and only then do you become aware of having made the decision.
Even more astonishing is blindsight, where a person can react correctly to visual stimuli even though not conscious of them.

I think the Chinese Room may resolve itself once we have some sort of grasp on why mental experience feels the way it does. Then we'll be able to state what it is about mental experience that makes us claim to be "conscious."

~~ Paul

[Z]

I think it's no different from asking if the section of grey matter responsible for passing language bits between the senses and the memory is conscious. Maybe the router is the conscious part, but lacks any knowledge, itself, of Chinese.

(I lean towards the 'many individuals' model of the brain, myself.)

[roger]

How Searle can fail to understand this - how this argument can be taken seriously by any philosopher - is beyond me.How a lot of philosophy gets written and taken seriously is beyond me. A few years ago, not being a linguist, I bought an interesting looking book on language usage and what that might say about the brain. I of course was in no position to judge the quality of the arguments, but it seemed interesting. But then he got talking about computers. He argued that is was obvious that computers could never attain intelligence. He used the example of an excel spreadsheet, and said that it was programmed to be precise, and get the same results everytime. Thus computers must be precise, and since humans react to things fluidly, and with creativity, we must not be using algorithmic processes.

That's about the logical equivalent of saying "a pool of blood is not alive. Therefore, blood cannot be part of, or compose a living system." I threw the book against a wall.

[Mercutio]

Neuroscientists.

We all have an instinctive understanding of what consciousness is, but some of the results of neuroscience have shown that our instinctive understanding differs significantly from what is actually going on. One famous example is that conscious decisions have been shown (for certain cases) to come a couple of hundred milliseconds after the action supposedly resulting from that decision has begun.

So you actually make the decision on some level, motor neurons begin to fire to act on the decision, and only then do you become aware of having made the decision.
Actually, there is a very good argument to be made against Neuroscientists being the ones to define consciousness. I am getting ready for class right now, but I will try to throw together some references and put together an argument later. One paper worth reading (though a bit dated) on it is Skinner's "Cognitive science and behaviorism" (1985, British Journal of Psychology), which may be available in pdf format online. I am blanking on the author of a much more recent one. Anyway, the argument is one of level of analysis. Like you said before (was it you?), neuroscientists defining consciousness is a bit like dissecting a frog to find its jump. Consciousness is what we do, more than what our bits and pieces do.

Argh--gotta run...

[Dark Jaguar]

The only difference is whether you want the Room merely to be able to answer questions - for which the Book alone will suffice - or to conduct conversations.

I think the idea of the room is that it is in fact supposed to be able to conduct a conversation that can't be discerned from the real thing, so yes, I want it to do more than just answer questions like a chat bot.

[Dark Jaguar]

Sure, but you could do it with a bookmark or even your finger. It doesn't require any more awareness than answering a single question does. It's just that the input and output are being passed in chunks.
I was aiming this more at D.J. who said that the awareness was coming from the guy writing stuff down.

A bookmark you say? Where do you place the bookmark? In THIS conversation set, or THAT conversation set, as the conversation has yet to reach certain criteria perhaps to let you know? Which one do you use for this or that conversation? Do the room know the person it's talking to, or is it a stranger? How can it tell?

[cyborg]

e argued that is was obvious that computers could never attain intelligence. He used the example of an excel spreadsheet, and said that it was programmed to be precise, and get the same results everytime. Thus computers must be precise, and since humans react to things fluidly, and with creativity, we must not be using algorithmic processes.

As I pointed out earlier this is clearly not an accurate model.

You are right reject the notion that we compute without computation as nonsense. Unless one invokes spooky immaterialism then one must accept that we are performing computations whenever we decide to do anything. Limitations in current computer technology cannot be used as an argument against this.

Our computational systems are incredibly complex and certainly pose quite a challenge of modelling since it is quite intractible currently.

[Jekyll]

A bookmark you say? Where do you place the bookmark? In THIS conversation set, or THAT conversation set, as the conversation has yet to reach certain criteria perhaps to let you know? Which one do you use for this or that conversation? Do the room know the person it's talking to, or is it a stranger? How can it tell?

Place it at the last instruction read.
You can perform different instructions depending on who is talking you, if this an acceptable cue, else it must be inferred based on the conversation.

I'll describe how to manufacture the no writing involved book if that makes it clearer.

Assume a life span of k years for your device.

Take the man in the box with memory and pass him every possible set of inputs that could occur over these years. Record the responses, and index them appropriately to allow for branching navigation.

This can be done as the possible input set is finite, and unlike general programs on Turing machines, we have imposed a halting constraint on output (we don't care what happens after k years).

Hey presto, we have a conversation machine that relies on nothing more than output instructions, conditionals on the current input statement (if the last letter they said was X goto page Y), and a bookmark to keep place.

[Dark Jaguar]

Even that limit makes for infinite possible inputs over that time. I really don't care how long the room is around anyway. Basically what you are suggesting is some massive set of algorythms that can take into account past events and calculate a proper response, and not just a tit for tat "this for this" thing. How could it not be aware?

[PixyMisa]

Consciousness is what we do, more than what our bits and pieces do.

Yeah, consciousness is what we do (good old functionalism!), but what we think we do turns out not to be what we really do. And its the neuroscientists who found that out. The people who know what consciousness actually is are neuroscientists, pretty much by definition.

[Mercutio]

Yeah, consciousness is what we do (good old functionalism!), but what we think we do turns out not to be what we really do. And its the neuroscientists who found that out. The people who know what consciousness actually is are neuroscientists, pretty much by definition.

Well...no. They find out how we do what we do, which is a different level of analysis altogether. And consciousness as "what we do" is much better explained via a behavioral identity theory than through a physiological identity theory. I agree 100% that our introspective accounts are nearly worthless, and that neuroscientists are doing incredible and worthwhile work in this area, but because of the level of analysis difference, their explanations are unsatisfactory.

If you have access to it, read Howie Rachlin's (2005) "What Muller's Law of Specific Nerve Energies Says About the Mind" (Behavior and Philosophy, 33, 41-54). If you don't have access, PM me an email addy and I can send you a pdf. If I have time later, I'll condense some of it to what is appropriate for this thread.

[Freddy]

Another interesting thing to read would be the work of Roger Penrose. He argues that the human mind cannot simply be a computer program. He is interesting because, despite believing this, he is a materialist. His argument is based on Godel's incompleteness proof. I've only read about his argument; I have not had time to actually read his work. But I hope to find the time, as it sounds interesting.

[Jekyll]

Even that limit makes for infinite possible inputs over that time.
No, it doesn't. It makes for a really big but finite number of inputs you can tell apart.

I really don't care how long the room is around anyway. Basically what you are suggesting is some massive set of algorythms that can take into account past events and calculate a proper response, and not just a tit for tat "this for this" thing. How could it not be aware?
I'm not arguing that it's not aware. I'm just saying that it's not the writing down or random number generation that makes this thing aware, as you can effectively get the same result from reading a giant do-it-yourself adventure book.

[PixyMisa]

Well...no. They find out how we do what we do, which is a different level of analysis altogether. And consciousness as "what we do" is much better explained via a behavioral identity theory than through a physiological identity theory. I agree 100% that our introspective accounts are nearly worthless, and that neuroscientists are doing incredible and worthwhile work in this area, but because of the level of analysis difference, their explanations are unsatisfactory.

I disagree. The fact that conscious decision making is actually a playback of lower-level function is critical to the definition of consciousness, and we would have no idea that this is what we are really doing if it were not for neuroscientific research.

[PixyMisa]

Another interesting thing to read would be the work of Roger Penrose. He argues that the human mind cannot simply be a computer program. He is interesting because, despite believing this, he is a materialist. His argument is based on Godel's incompleteness proof. I've only read about his argument; I have not had time to actually read his work. But I hope to find the time, as it sounds interesting.

Sadly, it's complete drivel. Godel's Incompleteness Theorem applies to formal systems. There's no evidence that the brain is a formal system, and even if it is, there's no evidence that the brain isn't bound by Godel's Incompleteness Theorem.

[cyborg]

If the human mind is not bound by Godel's Incompleteness Theorem then the implication is that we should be able to compute things that are supposedly incomputable by some non-computational mechanism.

Since we can't I have no reason to believe that the human mind is not a computational system.

[PixyMisa]

If the human mind is not bound by Godel's Incompleteness Theorem then the implication is that we should be able to compute things that are supposedly incomputable by some non-computational mechanism.

Since we can't I have no reason to believe that the human mind is not a computational system.

Godel's Incompleteness Theorem only applies to consistent systems.

The human mind is bound by the theorem precisely to the extent that it acts as a consistent (and sufficiently powerful, the other rule for applicability) formal system.

So we can come up with propositions that cannot be reached by a consistent formal system, but when we do so we are not acting as consistent formal systems, so it remains impossible for us to prove those propositions.

[Dark Jaguar]

I'm not arguing that it's not aware. I'm just saying that it's not the writing down or random number generation that makes this thing aware, as you can effectively get the same result from reading a giant do-it-yourself adventure book.

Nope it's the guy having to understand on some level what's going on, and understanding all those rules and their relation to what page he needs to turn to in the book (adventure book, like D&D?) sort of needs a concious mind as a prerequisite.

[nescafe]

Nope it's the guy having to understand on some level what's going on, and understanding all those rules and their relation to what page he needs to turn to in the book (adventure book, like D&D?) sort of needs a concious mind as a prerequisite.

Not at all -- consider the lowly computer, like this one I am typing on. It is capable performing several tasks at a time and reacting properly to a wide variety of signals from it's environment, thanks to the software that tells it how to respond to the environmental changes it perceives through its various input channels. The parallel between the software of this computer/the gigantic book in the room and the CPU/person in the room is too easy to pass up.

Translating the Chinese Room experiment into an identical experiment that uses a computer instead of a person is easy -- just stipulate into existence software for the computer instead of a book for the person.

From the outside view, we don't have to care at all what in the room enables it to speak fluent mandarin or cantonese -- Searle making it into a man/book combination and then implying that something special is just misleading -- sort of a reverse anthromorphization. I do not see his thought experiment leading to anything but hot air, wasted time, and misconception.

[Dark Jaguar]

Not at all -- consider the lowly computer, like this one I am typing on. It is capable performing several tasks at a time and reacting properly to a wide variety of signals from it's environment, thanks to the software that tells it how to respond to the environmental changes it perceives through its various input channels. The parallel between the software of this computer/the gigantic book in the room and the CPU/person in the room is too easy to pass up.

Translating the Chinese Room experiment into an identical experiment that uses a computer instead of a person is easy -- just stipulate into existence software for the computer instead of a book for the person.

From the outside view, we don't have to care at all what in the room enables it to speak fluent mandarin or cantonese -- Searle making it into a man/book combination and then implying that something special is just misleading -- sort of a reverse anthromorphization. I do not see his thought experiment leading to anything but hot air, wasted time, and misconception.

Yeah, I thought "computer" the whole time myself, but that's the point. A computer, as they are running now, can be said to have environmental awareness to a limited extent. To get it to respond to a conversation dynamically, it would have to have self awareness to some extent. This room used a man instead of a computer to operate those instructions, but the book isn't the processor, the person is. The book is more like the hard drive.

[RandFan]

I've heard what seems like contradictory statements from Daniel Dennett on the subject. In one of his books he said that a light switch is aware. On the other hand, in an interview, IIRC, he made a distinction between current machine capacity and human awareness. Dennett said something to the effect that humans are capable of drawing conclusions from unrelated facts. I don't remember his example but an alternate would be, I don't like getting wet from the rain. This tar repels water. Hey, if I coat my roof in the stuff I won't get wet.

That is an awareness that we don't as yet understand how to program into a computer. To solve problems that are completely out of the scope of the computers capabilities because it occurs to the computer that the solution would be beneficial somehow.

I think we will figure it out though.

[Dark Jaguar]

Here's an idea for that problem Rand. Basically the idea of finding matches between our data, and even though we humans CAN do it, we aren't exactly pros at it, as compaired to an idealised form of it. There are many times when I know two seperate things but, even though in retrospect it's obvious, I never thought to "cross reference" that information to notice something. For example, a new fence is built seperating backyard and front. I know that's the case, but I'm not thinking about that when I decide to drag something back there. I think "hey it's easier than going through the house", and then I see the fence, and I'm forced to consider that information I knew but hadn't considered in my plan.

Knowing everything is one thing, cross referencing all the data to see how it relates to all the rest is another. Writing a computer program to do that, well basically it would require the ability to do so across otherwise unrelated databases by always looking for commonalities and then experimenting until an internally consistant "idea" is formed, and then from that cross referencing with other data that has to do with the newly formed "idea" to make sure it is as consistant with the stored information as possible. There's a possible system of realizing painting a roof with tar might repel water. Programming something like that would certainly be a major project though.

[RandFan]

Here's an idea for that problem Rand. Basically the idea of finding matches between our data, and even though we humans CAN do it, we aren't exactly pros at it, as compaired to an idealised form of it. There are many times when I know two seperate things but, even though in retrospect it's obvious, I never thought to "cross reference" that information to notice something. For example, a new fence is built seperating backyard and front. I know that's the case, but I'm not thinking about that when I decide to drag something back there. I think "hey it's easier than going through the house", and then I see the fence, and I'm forced to consider that information I knew but hadn't considered in my plan.

Knowing everything is one thing, cross referencing all the data to see how it relates to all the rest is another. Writing a computer program to do that, well basically it would require the ability to do so across otherwise unrelated databases by always looking for commonalities and then experimenting until an internally consistant "idea" is formed, and then from that cross referencing with other data that has to do with the newly formed "idea" to make sure it is as consistant with the stored information as possible. There's a possible system of realizing painting a roof with tar might repel water. Programming something like that would certainly be a major project though. A couple of problems come to mind. Not necassarily insurmountable but significant nonetheless. Let me touch on the one that I think most significant. What to cross reference for?

Bird > Birdman of Alcatraz > Incarceration.
Bird > Flight > Freedom
Bird > Chick > Birth
Bird > Raven > Death
Bird > Pheonix > Rebirth
Bird > Dove > Peace
Bird > Hawk > War

There is a virtual infinite number of cross references that can be made. What do we cross reference ahead of time? I don't think humans are pre-wired with such built in cross refernces. In fact I know we are not. We build cross references ad hoc, as needed.

The question becomes, why do humans make the connections that we do? The answer is that many, if not most, are made as a result of emotions including wants. I.E. When it rains I am uncomfortable. I know that it will rain. When it does my roof will leak. I can envision the discomfort I will have. I recognize that a problem exists. I then recognize that there is a solution based on disparate data I have stored in my brain. Tar will seal the roof.

[Jekyll]

Nope it's the guy having to understand on some level what's going on, and understanding all those rules and their relation to what page he needs to turn to in the book (adventure book, like D&D?) sort of needs a concious mind as a prerequisite.

It's probably a simplified version. They don't have rules or things to collect, just instructions at the bottom of the page like:

"If you said 'Hi, how are you?' please turn to page 14435436346"

[nescafe]

This room used a man instead of a computer to operate those instructions, but the book isn't the processor, the person is. The book is more like the hard drive.

That is what I said -- the book is the software, the person is the CPU.

The parallel between the software of this computer/the gigantic book in the room and the CPU/person in the room is too easy to pass up.

A computer, as they are running now, can be said to have environmental awareness to a limited extent. To get it to respond to a conversation dynamically, it would have to have self awareness to some extent.

Not at all. Searle postulated a gigantic book that, if you think about it a bit, must have knowledge of all possible conversations in Chinese encoded in it. If Searle can postulate into existence such an artifact, then the parallel experiment that postulates into existence software that has knowledge of all possible Chinese conversations is equally justifiable. Self awareness on the part of either the software/book or the CPU/person is not required, yet the CPU/software/IO system in the experiment will act as if it is self aware (as also stipulated in the thought experiment, IIRC).

In the absence of a well-grounded technical definition for self-awareness, Searle's Chinese Room experiment says more about how easily people fall for the homunculus fallacy more than anything else.

[chriswl]

There is a virtual infinite number of cross references that can be made. What do we cross reference ahead of time? I don't think humans are pre-wired with such built in cross refernces. In fact I know we are not. We build cross references ad hoc, as needed.

The question becomes, why do humans make the connections that we do? The answer is that many, if not most, are made as a result of emotions including wants. I.E. When it rains I am uncomfortable. I know that it will rain. When it does my roof will leak. I can envision the discomfort I will have. I recognize that a problem exists. I then recognize that there is a solution based on disparate data I have stored in my brain. Tar will seal the roof.
I think this is one of the ideas motivating the recent trend in AI towards "embodied cognition." It is just impractical to expect a computer to make and weigh up all possible associations and would be asking it to do something far in excess of anything a human can do. But if robots have goals that drive their learning and their decision-making this allows them to prune their databases and avoid the otherwise inevitable combinatorial explosion of information.

The Chinese Room is sometimes cited in support of this view, as an example of the kind of isolated, box-in-the-corner kind of computer that could never really understand things in the same way as a human (or robot) that has real world experiences that "ground" its symbols and make its activities more than just empty shuffling of paper. But that's not what Searle intended - he actually thinks that all computers are inherently incapable of really understanding things, however they interact with the world and he thinks his argument somehow proves this.

[cyborg]

But that's not what Searle intended - he actually thinks that all computers are inherently incapable of really understanding things, however they interact with the world and he thinks his argument somehow proves this.

It only proves he does not really understand computation or computers. I suspect he sought such a conclusion anyway - most people have a need to feel they are inherently doing something special in their heads that a 'mere' machine could never duplicate. Such hubris.

[RandFan]

It only proves he does not really understand computation or computers. I suspect he sought such a conclusion anyway - most people have a need to feel they are inherently doing something special in their heads that a 'mere' machine could never duplicate. Such hubris.Then I'm guilty as charged. I don't see it your way. And I certainly don't see Searle your way. I came to the same initial conclusion as Searle.

I have to conced that in the end the experiment doesn't prove anything but it is a damn interesting thought experiment and I think we are all the better for the idea if for no other reason than to challenge our conceptions about AI. I've not yet reconciled the chinese room on an intuitive level. On the other hand I have not yet reconciled relativity on an intuitive level either. Life's tough. Timespace is even tougher.

I'm a materialist. That said I don't think we are anywhere close to understanding human cognition as to suppose that we could create a chinese room that would be aware. To be sure I prefre my example but that's probably a bias.

[cyborg]

Unless something informationless is going on - and since I for one don't know of anything physical that can be informationless - a computer is adequate to model the processes occurring in the human brain - or indeed any other physical phenomena.

So from my perspective it is quite simple - either a computer is adequate to model all physical phenomena (ignoring issues of tractibility) and hence computer cognition is merely a matter of the correct model or it is not and someone needs to explain what physical process it is that cannot be modelled.

[RandFan]

Unless something informationless is going on - and since I for one don't know of anything physical that can be informationless - a computer is adequate to model the processes occurring in the human brain - or indeed any other physical phenomena.

So from my perspective it is quite simple - either a computer is adequate to model all physical phenomena (ignoring issues of tractibility) and hence computer cognition is merely a matter of the correct model or it is not and someone needs to explain what physical process it is that cannot be modelled.I'm having some difficulty parsing your statement. Please forgive me.

Human consciousness is a physical process. When we understand how the physical processes of the brain give rise to consciousness we can then devise computers that will be adequate to model the processes occurring in the brain.

Like aerodynamics was to flight, I suspect that we will need to gain a new paradigm as to consciousness before we can replicate it. Our computers today are no better than those devices that were used before the Wright brothers in an attempt to fly. IMO.

[cyborg]

Human consciousness is a physical process. When we understand how the physical processes of the brain give rise to consciousness we can then devise computers that will be adequate to model the processes occurring in the brain.

We certainly can and neuroscientists have made extensive use of computer models of neurons in trying to understand human brain processes.

Like aerodynamics was to flight, I suspect that we will need to gain a new paradigm as to consciousness before we can replicate it. Our computers today are no better than those devices that were used before the Wright brothers in an attempt to fly. IMO.

There was no paradigm shift necessary to build the 747. Our computers are entirely sufficient to compute any problem - as of yet we do not know what problem it is we are trying to compute. It is an ill-defined problem with fuzzy parameters - that is what makes it difficult. Unless you would care to have a shot at making a concise and precise definition of what something has to have/do to be conscious?

[RandFan]

There was no paradigm shift necessary to build the 747.I'm sorry? Are you saying that consciousness is simply a problem of size and complexity? I don't think so but hey, I could be wrong.

Our computers are entirely sufficient to compute any problem...Having been interested in AI for two decades and having been a programmer and avid follower of computer trends I find the statement a bit odd. I don't know of any programmer that would agree but then that is anecdotal. I don't at all agree but that's fine. We can disagree.

...as of yet we do not know what problem it is we are trying to compute. It is an ill-defined problem with fuzzy parameters - that is what makes it difficult. Unless you would care to have a shot at making a concise and precise definition of what something has to have/do to be conscious?Well, I said it would require a new paradigm. You pose the classic quandary. I don't know and I think that is part of the problem. Don't get me wrong, I'm not saying our ignorance proves anything I'm simply saying that our ignorance prevents us from replicating consciousness at this time.

[cyborg]

I'm sorry? Are you saying that consciousness is simply a problem of size and complexity?

I can see no reason not to think so.

Having been interested in AI for two decades and having been a programmer and avid follower of computer trends I find the statement a bit odd.

It is not. There is nothing fundamentally new we need to discover about computation. Show as such and there would be a Nobel in it.

As such it is a matter of degrees. If it can be computed then it is merely a matter of whether or not it is practical to compute.

Well, I said it would require a new paradigm. You pose the classic quandary. I don't know and I think that is part of the problem. Don't get me wrong, I'm not saying our ignorance proves anything I'm simply saying that our ignorance prevents us from replicating consciousness at this time.

Then one cannot make a statement on what is required or what cannot achieve consciousness. One can only make appeals such as that this thread is based on that for some reason it is not reducible to computation.

I cannot see any good reason to presuppose that what we do does not boil down to a series of computations. An introspective appeal that 'I understand' doesn't mean much at all. It's just pushing the problem to some other area that is fuzzy and ill-defined - what it means to understand.