Talk:Chinese room/Archive 4
![]() | This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page. |
Archive 1 | Archive 2 | Archive 3 | Archive 4 | Archive 5 |
Instead of all this, why not this?
instead of
- The Chinese room is widely seen as an argument against the claims of leading thinkers in the field of artificial intelligence.[3] It is not concerned with the level of intelligence that an AI program can display.[4] The argument is directed against functionalism and computationalism (philosophical positions inspired by AI), rather than the goals of AI research itself.[5] The argument applies only to machines that execute programs in the same way that modern computers do and does not apply to machines in general.[6]
- The argument has generated an extremely large number of critical replies. According to Stevan Harnad, "the overwhelming majority still think that Chinese Room Argument is dead wrong."[7]
why not
- Searle claims that many proponents of AI claim that
- "if it looks like a duck, and quacks like a duck, it is a duck" YouTube
- or "seeing is believing" (positivist)
- whereas Searle claims that
- "Appearances can be deceiving".
- The argument has generated an extremely large number of critical replies. According to Stevan Harnad, "the overwhelming majority still think that Chinese Room Argument is dead wrong."[7]
This is only the lead paragraph, after all. ( Martin | talk • contribs 17:39, 7 July 2010 (UTC))
I retract this paragraph, as nice but possibly stupid. ( Martin | talk • contribs 11:07, 9 July 2010 (UTC))( Martin | talk • contribs 21:25, 9 July 2010 (UTC))
note: AI proponents need to defend the claim that computers think
The claim that a machine thinks is not self-evident. It needs to be defended. Why say a machine that speaks Chinese needs to understand what it is saying, when no one believes that a radio understands what it is saying, or a telescope understands what it is seeing. Is understanding Chinese the cause or the effect of speaking Chinese? Does the simulator suddenly achieve understanding when it passes the Turing Test?
Even if one demolished the Chinese Room, that would not establish the truth of the proposition that computers can think under certain circumstances. IMO ( Martin | talk • contribs 03:10, 10 July 2010 (UTC))
That a machine that passes the Turing test has artificial intelligence is true by definition. That artificial intelligence is the same as natural intelligence is something that needs to be defended, and seems on its face to be a behaviorist claim; ie, not respectable. The job is hard: to establish the existence of the other mind, and something I am feeling more and more at a loss to deal with. I am beginning to disagree with what Searle thinks is the meaning of his Room. His argument depends on the claim that people have only one mind. People are often of two minds, and the existence of the unconscious is well accepted, even by the men in the street. There could be two minds in the Chinese Room, and even two minds in the Englishman's head. Note that the Englishman, by agreeing to take the job as a clerk, has committed himself to a path of artificial stupidity, possibly an essential step in the emergence of new mind.( Martin | talk • contribs 18:25, 10 July 2010 (UTC)) ( Martin | talk • contribs 11:44, 10 July 2010 (UTC))
- I sympathize, but I feel a need to stress that Wikipedia articles have to reflect the range of published sources, not our own personal understanding. The things you are saying here make a good bit of sense, but they are not helpful for improving the article unless they can be attributed to published writings. Experience shows that letting a talk page turn into a debate about the topic is almost never productive. Looie496 (talk) 16:34, 10 July 2010 (UTC)
- I know. I hesitated to write that. The only defense I can think of is that this topic introduces a lot of technical arguments, and person's arguments are not necessarily unchanged from year to year. We hope that the article will be understandable, and I am trying to find a balanced position from which to recount the debate, which seems to have more to it than I thought, and not only in volume. I do see that I went over the line in the speculation about the emergence of new mind. I will
strikeout the part that I think is offending.( Martin | talk • contribs 18:25, 10 July 2010 (UTC))
- I know. I hesitated to write that. The only defense I can think of is that this topic introduces a lot of technical arguments, and person's arguments are not necessarily unchanged from year to year. We hope that the article will be understandable, and I am trying to find a balanced position from which to recount the debate, which seems to have more to it than I thought, and not only in volume. I do see that I went over the line in the speculation about the emergence of new mind. I will
CharlesGillingham, you said, above, two days ago: "The article says that believing the Turing test is adequate is part of Strong AI. Maybe the article should make this more clear." Oh yes, exactly. And the whole issue is: "Is that belief correct?" ( Martin | talk • contribs 19:14, 10 July 2010 (UTC))
After all that, do you think this is any better?
- The Chinese Room is an argument proposed by John Searle which he claims demonstrates that a program cannot give a computer a "mind" or "understanding", regardless of how intelligently the program may make the computer behave.[1] He concludes that "programs are neither constitutive of nor sufficient for minds."[2]
- The Chinese Room thought experiment assumes that a computer program has been written that has passed the Touring Test for Chinese; that is, that a person interacting with the program would assume that he is interacting with a Chinese speaker. Some leading figures in Artificial Intelligence claim that such a computer understands Chinese. Searle calls this position “Strong AI”, and is what his argument is intended to refute.
- In the thought experiment, a
manperson, an English speaker, takes the place of the computer. They follow the same algorithm, the same program, that the computer followed, and executes it manually in response to input from the outside. In this way, themanperson, is able to create the same sensible replies in Chinese as the computer did, while not understanding Chinese.
- The Chinese room is an argument directed against the philosopical positions functionalism and computationalism, which are inspired by AI, rather than the goals or practice of AI itself.[5] The argument is not concerned with the level of intelligence that an AI program can display.[4] The argument applies only to machines whose abilities are restricted to manipulation of symbols, as with modern computers, and does not argue that artificial minds cannot be created.
( Martin | talk • contribs 20:03, 10 July 2010 (UTC))
- On a first reading, this looks good. If you drop it in, I'll fix the footnotes, if you like. ---- CharlesGillingham (talk) 00:43, 11 July 2010 (UTC)
- The bit about understanding might be a little unclear. How about:
- "Some leading figures in artificial intelligence have claimed that such a computer would "understand" Chinese in precisely the same way that people "understand" Chinese – that the computer would have a subjective conscious experience."
- I think many people are confused by the word "understand" when they first read the argument. (Sometimes I wonder if Searle isn't deliberately baiting them). ---- CharlesGillingham (talk) 01:00, 11 July 2010 (UTC)
- A few days have passed. I did insert the change: "The argument leaves aside the question of creating an artificial mind by methods other than symbol manipulation", essentially from the italicized text above, but nothing else. I do think that the lead sentence should be about the significance of the Chinese Room. The second paragraph has more of that material than the first paragraph, so switch placement of some of the ideas? I do like subjective conscious experience.( Martin | talk • contribs 01:31, 13 July 2010 (UTC))
- A good deal of the material in the first paragraph is repeated in the early follow-on paragraphs. I would like to suggest that the first paragraph be deleted, or substantially reduced. ( Martin | talk • contribs 01:48, 13 July 2010 (UTC))
Gender neutral
I changed "man" in the room to "person". I'm not sure if I changed all of the references to man or he. I think I got most of them. ----Action potential talkcontribs 05:04, 11 July 2010 (UTC)
- Gender neutral usage is not to be done at the expense of clarity. There is no reason to clutter up articles to advance a political agenda. I warn you of WP:3RR. μηδείς (talk) 05:27, 11 July 2010 (UTC)
- FYI: In Searle's original description, it is Searle himself who is locked inside the room. In the systems reply, he's talking about a male "individual". Paradoctor (talk) 05:28, 11 July 2010 (UTC)
- (edit conflict) I happen to agree with Action potential as a matter of Common sense. Please see the edit summary of my revert. Dr.K. λogosπraxis 05:32, 11 July 2010 (UTC)
- In Searle's original argument he is talking about himself running the thought experiment form first person position as if he was there. Have a look how they handle it on the Stanford article on the same topic. When they quote Searle's summary of his thought experiment, they do refer to "the man in the room" but when they are referring to the argument in general they use "someone in the room". From introduction of Stanford article: "The argument centers on a thought experiment in which someone who knows only English sits alone in a room following English instructions for manipulating strings of Chinese characters, such that to those outside the room it appears as if someone in the room understands Chinese."[1] ----Action potential talkcontribs 08:30, 11 July 2010 (UTC)
- Makes sense to me. I never understood why we have to exclude the rest of humankind from this description, except if the experiment and all of its subsequent reports in literature were referring only to Searle himself but that seems to not be the case. Given the citation you provided I am convinced. The issue of political correctness has been raised to oppose your edits but I prefer to see this as an issue of neutrality and inclusivity and since this position is supported by WP:RS the argument in favour of your approach is encyclopaedically compelling. Thank you Action potential. Dr.K. λogosπraxis 12:16, 11 July 2010 (UTC)
- In Searle's original argument he is talking about himself running the thought experiment form first person position as if he was there. Have a look how they handle it on the Stanford article on the same topic. When they quote Searle's summary of his thought experiment, they do refer to "the man in the room" but when they are referring to the argument in general they use "someone in the room". From introduction of Stanford article: "The argument centers on a thought experiment in which someone who knows only English sits alone in a room following English instructions for manipulating strings of Chinese characters, such that to those outside the room it appears as if someone in the room understands Chinese."[1] ----Action potential talkcontribs 08:30, 11 July 2010 (UTC)
- (edit conflict) I happen to agree with Action potential as a matter of Common sense. Please see the edit summary of my revert. Dr.K. λogosπraxis 05:32, 11 July 2010 (UTC)
Fine use actual verbatims and change man to person if you must, but don't delete pronouns and replace then with awkward constructions.μηδείς (talk) 05:35, 11 July 2010 (UTC)
- I haven't yet seen the awkward pronoun constructions. Could you please elaborate? Dr.K. λogosπraxis 05:41, 11 July 2010 (UTC)
- He means "he or she". I think this is fixed. ---- CharlesGillingham (talk) 05:38, 12 July 2010 (UTC)
two virtual minds
The VIRTUAL paragraph contains the following <To clarify the distinction between the systems reply and virtual mind reply, David Cole notes that a program could be written that implements two minds at once—for example, one speaking Chinese and the other Korean. While there is only one system and only one man in the room, there may be an unlimited number of "virtual minds."[50]>
Hmm. I think I understand the difference between a system and a virtual system - one is virtual, and a virtual mind is a mind that is running on a virtual system. Wouldn't that be it?
- The systems reply means that it is not the CPU that understands Chinese, but the system; the CPU being Searle.
- The Virtual Mind means, I imagine, that the a program called a simulator is running. A simulator, for simplicity, usually would take advantage of the native operating system (that is, it runs "on top" of it) creating a virtual environment (that is, providing a different set of facilities to the programmer, rather than the set provided by the native operating system). This is always done because the virtual environment is easier to program than native environment. The reason it is easier is most often because it has facilities that look more like the problem than the facilities which exist in the underlying system. For example, it is easier to plan a rally route using Google Maps that coding a program in C++ for the purpose. Google Maps places you into an environment and gives you tool where making a route map is a trivial problem. Calling it a virtual environment is perhaps unusual, but not inaccurate.
- Multiple minds means having multiple simulators running? Does this require anything more than a system that can support multi-programming - ie virtually any home computer?
- Would a computer that understands Korean and Chinese need two minds? And would a computer that has two Chinese simulators running also have two minds? And - what's the point?
- David Cole says that one program could have two minds? That does not count as a clarification in my book.
( Martin | talk • contribs 05:30, 13 July 2010 (UTC))
- Cole's point is that "the System Reply" is wrong and the "Virtual Mind Reply" is right. The whole system can't be the mind. If the whole system were the mind, then you would need two whole systems in order to have two minds.
- The "whole system" is a set of ordinary physical objects, nothing more: a man, some file cabinets, paper, eraser, pencils. As Searle says "the conjunction of bits of paper and a man". ---- CharlesGillingham (talk) 05:40, 13 July 2010 (UTC)
- Hmm. Ok I can see that he would want to say that the system is not the mind, just as he might say that a man is not a mind, but only a man's mind is a mind. So the computer system, starting at the wall plug, is too much to identify with a mind. But commonly, the operating system (a program) is considered part of the computer system; in fact is the most important part. My Dell A90 can run XP or UBUNTU. Those are different systems, in common parlance.( Martin | talk • contribs 05:58, 13 July 2010 (UTC))
- I gather from what you say that he would say that the program running in the simulator would be the mind (the simulation), and I am not clear on what the difference between a mind and a virtual mind would be - except if it is just the label he has picked for his idea. This notion of counting minds is just going to lead to trouble I think. Again, Why does it take two minds to know two languages? Why does he say a program and then count two minds?( Martin | talk • contribs 06:06, 13 July 2010 (UTC))
- It's one program. It simulates two minds. (The room, I presume, has two slots, and people outside the room assume there are two human beings in the room, each communicating through their own slot. Searle's "machine language" commands tell him what slot to use, e.g. "Now put paper 35968584 through slot 2.")
- The difference between a mind and virtual mind (or "simulated mind") is the topic at issue. What is the difference? That's what the whole argument is about. I don't think there is an easy answer. ---- CharlesGillingham (talk) 07:05, 13 July 2010 (UTC)
- Charles, certainly you can't be saying that the way to count the minds in the room is to count the slots in the door, and expect still to be talking about minds as we know them.( Martin | talk • contribs 12:47, 13 July 2010 (UTC))
- Charles, you say <That's what the whole argument is about>. I don't think so. The whole argument is: is artificial intelligence the same as natural intelligence? This is lower down, in the branch that says it is the same, and this part tries to identify where the artificial intelligence is located. Is it located in the system or in a virtual mind? The mind is labeled virtual not because it is artificial but because it exists in a virtual machine. 没有? But how does a mind in a virtual machine differ from a mind in a real machine? One can ask in what way a virtual machine differs from a real machine, since the two are identical in function, by definition, and there is no machine that can be realized in software that cannot be realized in hardware. ( Martin | talk • contribs 12:47, 13 July 2010 (UTC))
- Google translate has only one slot in the door, and it can tell what language is coming in. Identifying the language is not that hard a trick, and maybe it would be easier than cutting a hole in the door for a second slot, and with the advantage of being easier to add to the room support for more languages. ( Martin | talk • contribs 12:53, 13 July 2010 (UTC))
- This is all the same issue: "Is a virtual mind the same as a real mind?"/"Is a simulated mind the same as real mind?"/"Is artificial intelligence the same as natural intelligence?"
- The "virtual machine" we're discussing is the mind. It's a "virtual mind". It acts just like a normal mind, in the same way that a PC emulator acts like a normal PC. ---- CharlesGillingham (talk) 20:15, 13 July 2010 (UTC)
- Really?? ( Martin | talk • contribs 22:49, 13 July 2010 (UTC))
- I assumed that it was the running program that was the mind, and that when the simulation stopped, the mind went away. I will ponder. Pondering how it differs from the systems reply. ( Martin | talk • contribs 22:56, 13 July 2010 (UTC))
Second try at the sentence:
<To clarify the distinction between the systems reply and virtual mind reply, David Cole notes that a program could be written that implements two minds at once—for example, one speaking Chinese and the other Korean. While there is only one system and only one man in the room, there may be an unlimited number of "virtual minds.">
I would not use David Cole's sentence, even if he did say that, because it is not so helpful. You could add it in addition, of course, to show the words some proponents use. Instead I would say
To clarify the distinction between the systems reply and virtual mind reply, as relates to the locus of a digital computer's understanding of Chinese: the 'system' reply identifies the locus of understanding as within the totality of the hardware beyond the wall plug. The 'virtual mind' reply narrows the locus to the hardware that is being used in the execution of the simulation. Or as David Cole puts it ... ( Martin | talk • contribs 23:39, 13 July 2010 (UTC))
Here is a supporting idea: it says that Searle characterizes Strong AI as asserting the view "that the mind is to the brain as the program is to the computer. So, the the program, the Chinese speaking program, is the mind of the computer.( Martin | talk • contribs 00:22, 14 July 2010 (UTC))
"Strong AI" no longer means "Computer functionalism" like it did in 1980
Ray Kurzweil, along with a thousand other science fiction writers and bloggers, have adopted the term "strong AI" and changed it to mean something else. Kurzweil's definition is this:
Kurzweil sells a gajillion books and this definition is used almost universally. Searle's definition of "Strong AI" is restricted to Searle and discussions of Searle. These definitions are pretty orthogonal, as the article takes great pains to point out.
Normal 2010 usage | Searle's 1980 definition | |
---|---|---|
Strong AI | Machine with full range of human intelligence | (The opinion that it is possible to create a) Machine with subjective conscious mental phenomena like "understanding" (using only a program and something that follows the program). |
Weak AI | Machine with limited intelligence | (The opinion that is is possible to create a) Machine with intelligent human behavior (using only a program and something that follows the program). |
My point is this: most readers of this article will assume that, by "strong AI", we mean "super-intelligent behavior by machines". This will cause unbelievable confusion to the reader. If they start this article assuming that this is what Searle is arguing against, the entire argument is going to seem like confused pointless nonsense. They'll think he's joking or stupid.
The good news is that we don't have to use the term "strong AI" in the introduction. We can get across exactly what Searle's targets are without having to use this confusing term. It's very simple to say "the idea that computers can have subjective conscious experience," or hide it in a less famous term, like "computer functionalism", so they'll pay attention when we define it.
I feel very strongly about this; indeed, it is one of the reasons I first started editing this article. ---- CharlesGillingham (talk) 05:32, 13 July 2010 (UTC)
- I understand the point, but this article is about Searle's argument, not about what the proper sense of strong AI is. We are not entitled to avoid or whitewash what Searle's actual words are or envaguen his statement to "one might". The solution is to attribute the words to the person who uses them (which was done with the verbatim in quotes) and, if necessary.μηδείς (talk) 05:57, 13 July 2010 (UTC)
- It's not whitewashing. I'm not suggesting that we leave anything out of the argument. I'm just saying that the term "strong AI" requires careful definition, and there isn't room for that in the introduction. The issue isn't the logic of the argument; the issue is good expository writing. ---- CharlesGillingham (talk) 06:03, 13 July 2010 (UTC)
- I feel that the primary goal should be clarity and accuracy, and someplace in there, brevity. ( Martin | talk • contribs 06:37, 13 July 2010 (UTC))
- It's extremely simple. Find a source that says Searle's usage of the term differs from the mainstream usage and add the parenthetical comment '(Searle's usage of "strong AI" is idiosyncratic,[1] see below)'.Plenty of room for that in the lead.
- If, however, your fear is that you cannot say anything that a layman might not misconstrue - they might think strong AI means powerful computers - then simply give up wikying and go live in a cave because the supply of numbskulls is endless and it is not your place to sacrifice yourself on their behalf. Good writing has nothing to do with writing down to your audience.μηδείς (talk) 06:42, 13 July 2010 (UTC)
- Well there is something in what you say: have a target audience in mind. I think it makes sense that the first paragraphs be simpler, can I say that, and later paragraphs get into more detail and sophistication? ( Martin | talk • contribs 13:00, 13 July 2010 (UTC))
Computational theory of mind vs. Functionalism
I tend to think of the classical computational theory of mind when I think of Searle's view of Strong AI, rather than Functionalism per se. - Bilby (talk) 06:06, 13 July 2010 (UTC)
- Me too. However, in Searle's Mind: a brief introduction (2004), he equates it with "computer functionalism". He also calls it that in The Rediscovery of the Mind. (Frankly, for the purposes of this article, I don't think there's really much difference between "computationalism" and "computer functionalism". You could could probably find a source that contrasts them, and then you could probably find another source that says they are same thing.) ---- CharlesGillingham (talk) 06:27, 13 July 2010 (UTC)
- They're certainly similar, although I tend to see CTM as a type of functionalism that is more immediately relevant to Searle. The key aspects of CTM relate very strongly to Searle's notion of Strong AI, which seems dependent on the idea of a program manipulating symbols (per CTM). Hence my general reading of the two terms to be mostly identical. I'll see what I can dig up, though, as maybe there's something that would help clarify Searle's Strong AI concept in that. (I finally finished moving campuses, so I've managed to unpack all my philosophy works at last - there might be something useful in there). - Bilby (talk) 06:32, 13 July 2010 (UTC)
- In Searle (2004) he gives this hierarchy (which he says is an "oversimplification")
- Dualism
- Property dualism
- Substance dualism
- Monism
- Idealism
- Materialism
- Behaviorism
- Methodological behaviorism
- Logical behaviorism
- Physicalism
- Identity theory
- Type identity
- Token identity
- Functionalism
- Black box functionalism
- Computer functionalism
- Functionalism
- Identity theory
- Behaviorism
- And he says that computer functionalism is the view that "mental states are computational states of the brain". (Which sounds to me like "computationalism", but what I do know.) My guess is that he's trying to (obliquely) make the point that computationalism is a form of functionalism. But why doesn't he just say that? ---- CharlesGillingham (talk) 06:54, 13 July 2010 (UTC)
- Based on that, computer functionalism sounds very much like CTM, and that would be in keeping with other views I've picked up that CTM is a subset of functionalism. It is interesting to read his thoughts on connectionism in The Failures of Computationalism, which tends to support my thought that CTM is his idea of strong AI, although not well enough for anything here. (Mind you, much of the paper is a criticism of Harnad's TTT, which I'm always happy to see more of - I'm not a TTT fan). - Bilby (talk) 12:35, 13 July 2010 (UTC)
- Charles - do you mind if I add to your listing of Searle's hierarchy. A few words of detail ( Martin | talk • contribs 16:39, 13 July 2010 (UTC))
- Based on that, computer functionalism sounds very much like CTM, and that would be in keeping with other views I've picked up that CTM is a subset of functionalism. It is interesting to read his thoughts on connectionism in The Failures of Computationalism, which tends to support my thought that CTM is his idea of strong AI, although not well enough for anything here. (Mind you, much of the paper is a criticism of Harnad's TTT, which I'm always happy to see more of - I'm not a TTT fan). - Bilby (talk) 12:35, 13 July 2010 (UTC)
- They're certainly similar, although I tend to see CTM as a type of functionalism that is more immediately relevant to Searle. The key aspects of CTM relate very strongly to Searle's notion of Strong AI, which seems dependent on the idea of a program manipulating symbols (per CTM). Hence my general reading of the two terms to be mostly identical. I'll see what I can dig up, though, as maybe there's something that would help clarify Searle's Strong AI concept in that. (I finally finished moving campuses, so I've managed to unpack all my philosophy works at last - there might be something useful in there). - Bilby (talk) 06:32, 13 July 2010 (UTC)
- Just don't change the structure; I copied this verbatim from Searle's The Mind: a brief introduction. ---- CharlesGillingham (talk) 19:41, 13 July 2010 (UTC)
- No no no I wouldn't, and may not do anything. ( Martin | talk • contribs 22:46, 13 July 2010 (UTC))
- Just don't change the structure; I copied this verbatim from Searle's The Mind: a brief introduction. ---- CharlesGillingham (talk) 19:41, 13 July 2010 (UTC)
Implementation independent, again
The VIRTUAL paragraph contains the sentence: A virtual machine is also "implementation independent" in that it doesn't matter what sort of hardware it runs on: a Macintosh, a supercomputer, a brain or Searle in his Chinese room.
- This "characteristic" of a virtual system is hard won; it is an ideal, and I would suspect not often reached, since each implementation will have different bugs. And in the ideal case, I imagine that you could not determine the underlying hardware, to protect from exploits.
- In any case - so what? What of significance is this statement in there for, and how does it advance any argument?
( Martin | talk • contribs 02:14, 13 July 2010 (UTC))
- I agree with you that this sentence is unclear. It could be struck from the article, without any harm done. It's not a necessary part of the argument.
- The point that this is supposed to make (and it makes it poorly) is that, at a certain level, a program doesn't care what hardware it runs on. Word for Windows looks very similar on various models of Dell and Compaq computers. You can hardly tell the difference, and in theory, programmers could make there was literally no difference between Word for Windows on Dell on Compaq, even though these computers may use different computer chips to get the job done. Similarly, functionalism claims that the human mind can exist on various types of hardware, an idea called multiple realizability. ---- CharlesGillingham (talk) 04:40, 13 July 2010 (UTC)
- Do no harm. It's hardly an unimportant point. It is the point of functionalism. The sentence is neither false nor misleading. If you can't improve it, at least don't mutilate it. Wikipedia is a work in progress. The proper response to a point you think may be unclear is to clarify it, not to delete it. Or else wikipedia will become a tale told by an idiot, full of sound and fury, signifying nothing.μηδείς (talk) 05:03, 13 July 2010 (UTC)
- I should have said "it's not necessary part of the paragraph", rather than argument. The important point of the paragraph is to explain what "virtual" means in computer science. It's supposed to explaining Minsky's thinking, not functionalism. It's okay if the paragraph doesn't get all the way to functionalism. Functionalism is covered in it's own section.
- We could use a paragraph on "multiple realizability", I'm just not sure where. We can only say so much about it 10 pages. This is not an article on functionalism. We just need to define it accurately, simply and clearly. ---- CharlesGillingham (talk) 05:55, 13 July 2010 (UTC)
- Medeis, you say <The sentence is neither false nor misleading> but it seems false to me, I don't think I am being too strong. A wave, that travels through a mix of liquids, continues and maintains an identity even when the material it travels through changes. This is not true of a computer program. Can you explain what the sentence means to you or why it is important? As I said above, a virtual machine in the sense used in the sentence is not a program, it is a design. And even then it does matter what hardware it runs on: the underlying hardware has to meet minimum requirements set by the design. ( Martin | talk • contribs 12:23, 13 July 2010 (UTC))
- Charles, you say <program doesn't care what hardware it runs on> which in one sense is true by definition, since a program doesn't have cares. If you mean that any program will work on any hardware, that is obviously false. The fact that there is a design spec, originally called IBM-compatible, which programmers coded to, only means that they all coded to the same virtual machine, and it was up to the manufacturers to make that happen, to ensure that sameness, which they did originally via the Phoenix_BIOS, and possibly other things as well. The idea that the human mind can exist on various types of hardware is a claim, and it doesn't really matter if you call it multiple realizability. What matters is why you say it, how you came to know it, and what you mean by it. ( Martin | talk • contribs 05:45, 13 July 2010 (UTC))
- Okay. Yikes. I guess I should have said "the program looks the same no matter what hardware it runs on." ---- CharlesGillingham (talk) 05:55, 13 July 2010 (UTC)
- But that doesn't work for minds. I should have said "the program acts the same no matter what hardware it runs on". ---- CharlesGillingham (talk) 06:00, 13 July 2010 (UTC)
- You cede ground grudgingly. ( Martin | talk • contribs 12:23, 13 July 2010 (UTC))
- Sorry, I don't mean to. We agree the sentence is no good. We just have to figure out what to do about it. We can't really say "it's false", because even if that's true, we can't write that. That would be our opinion, not Searle's or Minsky's. Minsky (and Cole, etc) think that this is going to work. So we have to figure out why it works. I'm just doing a poor job of explaining why it's supposed to work. ---- CharlesGillingham (talk) 20:41, 13 July 2010 (UTC)
- Okay, here's my fix:
- "A virtual machine is also "implementation independent" in that the emulator can be written for many different machines: a Macintosh, a supercomputer, a brain or Searle in his Chinese room."
- Sound good? ---- CharlesGillingham (talk) 21:40, 13 July 2010 (UTC)
- Okay, here's my fix:
- Re grudgingly: I wasn't being serious except in the grumble to myself that I thought there was more in that reply than "a computer doesn't have cares". I really don't expect you to listen to grumbling if the argument does not persuade. ( Martin | talk • contribs 22:08, 13 July 2010 (UTC))
- If the people involved actually talk like that, who am I to say don't put it in there. My comment is what they heck do they think follows from this multiple realizability, from the fact that a spoon can be made from wood or metal? If they are saying that a Turing machine can be made of brains or silicon, that seems true enough. If they are saying that a mind can be made of tissue or metal, I can't deny it, but I think the way it is put is obfuscation, dressing up a poor idea in rich clothes. If they mean that a mind can be implemented on a digital computer, they are just begging the question. ( Martin | talk • contribs 22:28, 13 July 2010 (UTC))
- If multiple realizability means implementation independent, then it also means multiple implementation: a virtual machine must be implemented anew for each platform it is to run on, but you know that. This is the idea that is glossed over - maybe you cant implement the virtual machine on an array of beer cans. ( Martin | talk • contribs 22:37, 13 July 2010 (UTC))
New paragraph for "Finding the Mind"
I want to add this paragraph at the top of the "finding the mind" section. It would be a new second paragraph. I think it would help us with completeness and also help to put the system reply and virtual mind reply into a larger context. However, I don't have time to dig through the literature to find sources and make sure the list is somewhat comprehensive and fair. ---- CharlesGillingham (talk) 21:40, 13 July 2010 (UTC)
All of the replies that identify the mind in the room are sometimes called "the system reply". They differ in how they describe "the system", i.e. they differ in exactly who they ascribe intentionality to. According to these replies, the "mind that speaks Chinese" could be either: the "software",[2] the "program",[3] the "running program",[4] a simulation of the "neural correlates of consciousness",[5] the "functional system",[6] a "simulated mind",[7] an "emergent property",[8] the "the set of physical objects" (Searle's version of the system reply, described below), or "a virtual mind" (Marvin Minsky's version of the system reply, also described below).
- Looks good to me (except isn't it: they differ in exactly to what they ascribe intentionality). But I think that "to ascribe intentionality" is to identify its locus, and to do that, you have to identify some subset of space-time; something that might not be thought to contain intentionality. Like a frog that jumps suddenly; Oh my! I didn't see that. The "virtual mind" answer does that only in an evasive way; it does not unambiguously stake a claim for the locus of that virtual mind. "an emergent property" yea but of what? "A simulation" is another name for "the running program". Does the "a simulated mind" answer mean that the intentionality is ascribed to "a simulated mind"? That's too circular for my taste, but I assume it means, again, the running program; ie in the hardware that is devoted to running the program. How would the article handle it if someone ascribed intentionality to something called "simulated intentionality". ( Martin | talk • contribs 10:37, 14 July 2010 (UTC))
New Introduction
Here's the situation. I wrote an introduction for this article about a week ago. Martin wrote an introduction which he politely placed on the talk page above. And μηδείς wrote the introduction that appears in the article as of today.
I'm just going be frank, if you all will forgive me. My favorite of the three is Martin's above. My least favorite is the current one, by μηδείς. The biggest problem with μηδείς's version is organization. The paragraphs don't seem to stay on topic very well and this makes it a little hard to read. I suggest we use Martin's above.
(Of course, I'd prefer it if he used my suggestion about "subjective conscious experience", rather "strong AI" "understanding", but let's talk about that two six sections above.) ---- CharlesGillingham (talk) 06:43, 13 July 2010 (UTC)
- Then address the order. I did not delete any material from the lead, I added relevant material that hadn't been addressed. The prior leads didn't even mention where the argument was published. Try reading WP:LEAD. The point is to address the relevant, essential and controversial points, not to dumb an article down because something MIGHT be misunderstood. μηδείς (talk) 06:51, 13 July 2010 (UTC)
- I am addressing the "order" (if by "order", you mean "organization"). I like Martin's order better. You're responding to criticisms I didn't make. ---- CharlesGillingham (talk) 07:09, 13 July 2010 (UTC)
- I liked and like "subjective conscious experience". ( Martin | talk • contribs 12:54, 13 July 2010 (UTC))
- In the "note" paragraph, and in the "instead of" paragraph, I hinted at my thoughts about the lead paragraph: it lacks punch. Explain the controversy, explain the idea. The publication date of the idea is less important than the idea, and the date is mention, I think appropriately, as the lead sentence of the paragraph History. I think a good first step would be to remove the entire first paragraph, then sit for some time and breathe, then look again and see that that was not as bad as it felt. The original version has a nice simplicity.( Martin | talk • contribs 15:40, 13 July 2010 (UTC))
- This wikipedia article on phenomenology doesn't get down to business until after the lead paragraph, and I found it very readable. So possibly there is room at the top for preliminary matter. ( Martin | talk • contribs 21:23, 13 July 2010 (UTC))
Systems and virtual mind, modified
I modified the systems and virtual mind paragraph along the lines of "second try".( Martin | talk • contribs 17:32, 14 July 2010 (UTC))
- I think your changes are fine. I want to move the link from virtual machine to virtual (computing) because I think this is what Minsky means. Virtual in a sense that starts with machine simulation but also in the same sense as virtual reality, etc.
- I think Minsky is trying to say it's the simulation, not the simulator. It's very hard (for me, anyway) to get this ontology precisely right. ---- CharlesGillingham (talk)
- Thanks. I also first thought of virtual reality as an accessible use of the term virtual, but as I was typing, it came out that way, which I ended up liking. Yes, yes. the simulation. ( Martin | talk • contribs 22:07, 15 July 2010 (UTC))
- Oh! I assumed that you already moved the link because you said that you wanted to move the link. ( Martin | talk • contribs 22:34, 15 July 2010 (UTC))
Steve Harnad - huh?
In "What's Right and Wrong in the Chinese Room", Steve Harnad says a number of things that I don't understand. Can anyone explain them to me? They are:
- He claims Searle claims that Strong AI proponents claim: <(2*) The brain is irrelevant.> In what sense could Harnad think that Searle imagines that anyone could claim that? Irrelevant to what? Or why does Harnad say it? Or has it been clarified already? Martin | talk • contribs 21:05, 15 July 2010 (UTC))
- I can answer one of these. Brain is irrelevant (to Strong AI). Searle definitely says this. (This article gives the page number.) Why? Because the structure of AI programs (especially symbolic, knowledge-based programs of the sort that existed in 1980) does not typically have any connection with the structure of the brain. There's no pineal gland, no distinction between grey matter and white matter, no amygdala, no corpus calloseum. Early AI research typically never looked at the brain. Why is that? Because they didn't need to. The structure of their programs came from the structure of logic and problem solving; it came from the function. To understand the function of the brain, you don't need to understand the brain. You just need to understand what it does. You don't need to understand what it is. This is functionalism. ---- CharlesGillingham (talk) 22:52, 15 July 2010 (UTC)
Changes to lead
- <Searle, who can not read Chinese — "To me, Chinese writing is just so many meaningless squiggles." [2]> An explanation of "cannot read Chinese" is not needed, nor a fotenote (76.180.164.161 (talk) 18:41, 19 July 2010 (UTC))
The article is noted for its style as well as its content and brief illustrative quotes interest the general reader and make him want to read more.μηδείς (talk) 19:17, 19 July 2010 (UTC)
Paragraph defining the question
I argue that this version:
- The question Searle wants to answer is this: does the machine literally "understand" Chinese? Or is it merely simulating the ability to understand Chinese?[9] Searle calls the first position "strong AI" (see below) and the latter "weak AI".
introduces the central question of Searle's thought experiment with utter clarity and simplicity and as such belongs towards the top of the article. In addition, it is attributed directly to the primary source and paraphrases Searle's exact words.
I also argue that this version:
- Searle claims that there are some proponents of artificial intelligence, who hold a functionalist position,[10] who would conclude that the computer "understands" Chinese.[11] This conclusion, a position he refers to as strong AI, is the target of Searle's argument.
is poorly written, in several ways. It is repetitive and convoluted. It introduces jargon that is not defined until later in the article (i.e. "functionalism"). It attributes a philosophical position to an unnamed source ("some proponents"). And finally, it is redundant with respect to later sections of the article which carefully identify "functionalism" and give a name for the unnamed AI researchers. ---- CharlesGillingham (talk) 22:28, 2 November 2010 (UTC)
Futurama
The question "Am I just an automaton, or can a machine of sufficient complexity legitimately achieve consciousness?" is about Artificial Intelligence, as is the Chinese Room thought experiment. However, that question is not a mention of the Chinese Room thought experiment nor is it in any respect whatsoever an allusion to it. It is not therefore appropriate for this article. Dlabtot (talk) 23:38, 29 November 2010 (UTC)
- In the episode, The robot-Leela is explicitly described as a computer simulation of the original. The computer replica says: "Am I just an automaton, or can a machine of sufficient complexity legitimately achieve consciousness?" What could possibly be more relevant to the argument? μηδείς (talk) 23:46, 29 November 2010 (UTC)
- It is simply an mention of the question of whether "a machine of sufficient complexity legitimately achieve consciousness" and does not in any way reference this thought experiment, mention it, or allude to it.
- Are you saying that it is a mention of or reference to this thought experiment? In what way? I am having difficulty understanding why you think it should be included in this article. Perhaps an WP:RFC or WP:THIRDOPINION would help. Dlabtot (talk) 01:14, 30 November 2010 (UTC)
Have you seen the episode? A computer simulation of Leela is created. The question arises whether that computer simulation is actually conscious or just appears that way. To what other philosophical issue would you relate this? The analytic-synthetic dichotomy? I really don't understand your objection.μηδείς (talk) 04:17, 30 November 2010 (UTC)
- My objection is what I have repeatedly stated. Yes, it is an expression of the question of whether that computer simulation is actually conscious or just appears that way. However it does not mention the Chinese Room thought experiment nor allude to it in any way. That is if it has been accurately described here. Do you really not understand the distinction?
- My asking you to explain in what way this is a mention of the Chinese Room thought experiment was a sincere request. Please respond. No, I have not seen the episode. If it actually did refer to the Chinese Room thought experiment or allude to it in some way, then it would be appropriate for inclusion here.
- What do you think about an WP:RfC or WP:THIRD? Dlabtot (talk) 04:39, 30 November 2010 (UTC)
- The question of artificial consciousness is a valid and interesting topic. That is what is referenced in the Futurama quote. The Chinese Room, in contrast, is an argument flawed on so many levels and so poorly premised, that unsurprisingly, the artificial consciousness article, like the Futurama episode, doesn't mention it. Dlabtot (talk) 04:50, 30 November 2010 (UTC) Although it does mention John Searle. Dlabtot (talk) 05:55, 30 November 2010 (UTC)
Yes, it is a rather direct allusion to the dilemma. From the article:
"It addresses the question: if a machine can convincingly simulate an intelligent conversation, does it necessarily understand? . . . The experiment is the centerpiece of Searle's Chinese Room Argument which holds that a program cannot give a computer a "mind" or "understanding", regardless of how intelligently it may make it behave.[1]"
From the description of the episode:
"a robotic replica of Leela, programmed to simulate her personality, asks 'Am I just an automaton, or can a machine of sufficient complexity legitimately achieve consciousness?'"
This is set in the context of Frye having a sit down dinner conversation with the simulated Leela.
μηδείς (talk) 01:44, 1 December 2010 (UTC)
- So the Futurama episode doesn't mention the Chinese room, or a room, or the Chinese language, or any of the premises or alleged conclusions of the Chinese room argument, which is the subject of this article. Thank you for clearing that up. This is not an article about a 'dilemma' it is an article about a specific thought experiment and argument. I am disappointed that rather than directly respond to my actual questions, you simply repeated yourself. I will start an RfC. Dlabtot (talk) 03:12, 1 December 2010 (UTC)
- As to the wording of the RfC, I asked you twice about posting an RfC, and you ignored both inquiries. While I am sure you are engaging in a good faith effort to improve the project, you don't seem to be engaging in an effort to collaborate with me. You have ignored every specific question I've asked you, simply repeating your original assertion instead of responding on point. I don't believe the two of us will be able to achieve consensus without the input of others. Please do not edit my comments again. Dlabtot (talk) 05:21, 1 December 2010 (UTC)
- RfC wording, like all section heads, is not a "personal comment" - I have not touched your "personal comments". You seem to think I edit and debate at your convenience. Guessing at the vague intent of an editor unwilling to refer to the source is not my idea of collaboration. If you have simple concrete yes or no questions, please ask them. I suggest in the meantime that you watch the episode in question. μηδείς (talk) 05:35, 1 December 2010 (UTC)
- Since I accept what you said occurred in the episode as 100% accurate, we have no dispute about that.
- I will repeat one of the "simple concrete yes or no questions" I previously asked, slightly rephrased - the most basic one:
- Do you understand the distinction I am making between mentioning the question of artificial consciousness' and mentioning the Chinese Room argument? Dlabtot (talk) 06:01, 1 December 2010 (UTC)
- RfC wording, like all section heads, is not a "personal comment" - I have not touched your "personal comments". You seem to think I edit and debate at your convenience. Guessing at the vague intent of an editor unwilling to refer to the source is not my idea of collaboration. If you have simple concrete yes or no questions, please ask them. I suggest in the meantime that you watch the episode in question. μηδείς (talk) 05:35, 1 December 2010 (UTC)
- Again, you asked for "simple concrete yes or no questions". I have repeatedly posed such questions and you have ignored them. I again ask:
- Do you understand the distinction I am making between mentioning the question of artificial consciousness' and mentioning the Chinese Room argument?
- It is a simple question. It is concrete. It is 'yes or no'. A respectful discussion of this question could form the basis of discussion that could move towards consensus. What is your answer? Dlabtot (talk) 07:30, 1 December 2010 (UTC)
RfC: Should the Chinese Room article mention the Futurama episode Rebirth [in the Popular culture section]?
Edit battle
Several reversions recently about the use of 'human' versus 'person'. Come on guys - stop playing around. It's too much. Ask for mediation or something. Myrvin (talk) 08:06, 19 January 2011 (UTC)
Unable to learn
The critical problem with the Chinese room is that it excludes the rule-writers entirely.
This means that the room can only react -- it cannot interact or learn.
It also means that it cannot truly communicate, since communication requires developing an understanding of the other party involved.
Include the rule-writers in the system, and it becomes very different. —Preceding unsigned comment added by 121.133.123.198 (talk) 15:34, 28 January 2011 (UTC)
- We assume Searle has a pencil and an eraser. ---- CharlesGillingham (talk) 17:37, 28 January 2011 (UTC)
- Let me add a reminder that anything that goes into our article needs to be based explicitly on the published literature -- there's not much value in making points unless one can show where similar points have previously been made in the literature. Looie496 (talk) 19:51, 28 January 2011 (UTC)
- I think you are right, and that the article should address this issue. I think the best way to handle it is in a footnote to the first paragraph of the section Computers vs. machines vs. brains.
- The question is this: is the Chinese Room Turing complete? Several points: (1) It's clear that Searle thinks that the Chinese Room is Turing complete, i.e. he thinks that he's doing exactly what every other CPU does, and he dismisses any argument to the contrary with a rhetorical eye-roll. (2) If the room (as described in some version of the argument) isn't exactly Turing complete, it's easy to fix (just give him an eraser). (3) Critiques that are based on the assumption that room isn't Turing complete are attacking a straw man.
- But the difficulty is the sources, of course. You are right that Searle is less than clear about whether he has that eraser or not. (I speculate that he didn't understand the importance of showing, step-by-step, that the room is Turing complete.) We need to find the places where he touches on this and try to stitch together a proof that Searle thinks the Chinese Room is Turing complete.
- The footnote should make two points: Searle is less than clear, but (according to this recent description) the Chinese room is Turing complete. (The footnote could also mention that there are a large number of "replies" that require that the room is not Turing complete, but this issue is covered in more depth in the section of the article called Redesigning the room.)
- So what we need are the sources to put this footnote together. I'm too busy at the moment to fix this, but I hope to get to it someday. ---- CharlesGillingham (talk) 21:16, 3 February 2011 (UTC)
- First, learning can be achieved by adding rules the govern the addition of new rules, he still doesn't need to understand any of this; only that certain symbols coming through mean to do things different. In fact, a simpler model would be to use conditonals in the right way to be equivalent to rule modifications, etc. Second, there is no reason that the room needs to be turing complete, the room only needs to be able to run the given program; if I can write a program that can pass the turing test, then I only need enough computational strength to run that program. —Preceding unsigned comment added by 209.252.235.206 (talk) 10:55, 23 February 2011 (UTC)
- The program used in the Chinese Room can simulate all the (verbal) action of a human being. A human being can simulate the action of Universal Turing machine. Therefor the Chinese room can simulate the action of a Universal Turing Machine. Therefor the program is Turing complete. ---- CharlesGillingham (talk) 06:23, 5 June 2011 (UTC)
Finding the mind
This section does not give a comprehensive list of all the "system" replies. See Talk:Chinese room/Archive 4#New paragraph for "Finding the Mind" where I outlined a the kind of paragraph we need.---- CharlesGillingham (talk) 21:13, 10 February 2011 (UTC)
- Done, although I'd like to do more research and find sources for the system replies that are mentioned only in passing. ---- CharlesGillingham (talk) 23:33, 28 February 2011 (UTC)
What they do and not prove
This title is too chatty for an encyclopedia. Perhaps we could structure these sections better. Should we indent the description of each reply? Then the final paragraphs would clearly refer to all the replies in the section. This was briefly discussed here: Talk:Chinese room/Archive 2#Original research. ---- CharlesGillingham (talk) 21:16, 10 February 2011 (UTC)
- Done ---- CharlesGillingham (talk) 07:53, 25 February 2011 (UTC)
Split "Other minds" and "Speed and Complexity"
I don't think that the "other minds" and "Consciousness (as Searle understands it) is epiphenomal" replies are "appeals to intuition" as the article states. Rather, I think are they are demonstrations that Searle's argument is meaningless, which is a different thing. I'd like to split these out into their own section. ---- CharlesGillingham (talk) 21:37, 10 February 2011 (UTC)
- Done. Although I need to add some more citations to really get this right. ---- CharlesGillingham (talk) 21:43, 28 February 2011 (UTC)
Open issues / Todo
The bot is archiving things a bit faster than we are fixing them. I thought I would refresh a few threads here. These are some research and writing projects that I would like to do someday to improve the article. If anyone else is interested, then by all means, feel free. ---- CharlesGillingham (talk) 21:37, 10 February 2011 (UTC)
Flawed logic
Symbol manipulation is an abstraction for voltage manipulation, which is what a computer "really" is doing. Voltage manipulation is also what neurons are doing. If doing "symbol manipulation" somehow means there cannot be "semantics", then brains cannot either. Who cares if the cpu does not "understand", nobody thinks a neuron "understand" anything either; but extrapolating that the sum of its parts cannot exhibit what he calls "understanding" is just silly. Qvasi (talk) 07:31, 3 July 2011 (UTC)
- Does that mean a mechanical or optical computer would not have genuine intelligence, since voltage manipulation is missing 1Z (talk) 14:46, 3 July 2011 (UTC)
- Let me insert a reminder that talk pages are for discussing ways to improve articles, and any improvements must be based on reputable published literature. Personally invented arguments do not belong here, however compelling they may be. Looie496 (talk) 17:13, 3 July 2011 (UTC)
Well, some criticism in the article instead of just saying how great it is would be nice. — Preceding unsigned comment added by 99.236.136.206 (talk) 12:17, 3 October 2011 (UTC)
- Eh? The article contains a very extensive discussion of criticisms. Looie496 (talk) 14:53, 3 October 2011 (UTC)
The introduction
The introduction is not very well organized. The material it presents is more or less right, but the paragraphs do not have clear topics and some sentences are needlessly convoluted. See these alternatives:
Here is an example of a one paragraph introduction. This is approximately the right amount of material:
This is a discussion where I aired my criticism:
--- CharlesGillingham (talk) 21:11, 10 February 2011 (UTC)
- Done Finally got around to rewriting the lede. ---- CharlesGillingham (talk) 09:58, 11 December 2011 (UTC)
Rebuttal to "system reply"
It would be nice to have a more fleshed out reply to the system reply, although I may be missing something obvious. Uninteresting retort: "he asks what happens if the man memorizes the rules and keeps track of everything in his head? Then the whole system consists of just one object: the man himself. Searle argues that if the man doesn't understand Chinese then the system doesn't understand Chinese either and the fact that the man appears to understand Chinese proves nothing." Then, the obvious reply: "Critics of Searle's response argue that the program has allowed the man to have two minds in one head." In other words, the program with "real understanding" of Chinese is being "run" by Searle's real consciousness as if Searle's consciousness is the CPU. We've all had the experience of mindlessly following instructions that we didn't really understand, right? Stevan Harnad, one of the sources used in the article, says this in reply to the other "conscious Chinese-understanding entity inside his head" reply: "I will not dwell on any of these heroics; suffice it to say that even Creationism could be saved by ad hoc speculations of this order." Um, ok. Has anyone "dwelled on these heroics"? I feel like this is a pretty basic and obvious problem with CRA, so it would be nice to have a good response to it in the article so people can understand why anyone learned takes CRA seriously. Does anyone know of a good retort that we could cite? Something by Searle himself? ErikHaugen (talk | contribs) 07:54, 24 November 2011 (UTC)
- I'm not sure I understand what you are asking for. The retort described in the article does in fact come from Searle himself. For those of us who think the Systems Reply is the correct reply, the fact that there is no strong retort is not disturbing. Looie496 (talk) 16:18, 24 November 2011 (UTC)
- Yes, I know that the retort about memorizing the rules/database is from Searle. My question is: is there a retort (from Searle or anyone else) to the obvious response to it that there's another "conscious Chinese-understanding entity inside his head" that is not Searle's conscious? This feels like a very obvious response to CRA and way of understanding what is going on. The disturbing thing is either 1.) why does this article have this obvious reply but not a response to it, or if there is no response, then 2.) if CRA is so broken why has it been taken so seriously over the years? I'm assuming that there is some reply to this out there, but I looked and I can't find it. All I can seem to find is Harnad's dismissal, which doesn't help but gives me hope that there is something to be found. If (2) is reality, then I suppose there's nothing to be done here in this article about it, but wow. I'm no cognitive scientist, so maybe this is obvious to the initiated, but as a non-expert reader I gotta say I don't get it. ErikHaugen (talk | contribs) 05:52, 25 November 2011 (UTC)
- I think it is important to realize that the CRA is an intuition pump (in Dennett's terminology), not a logical proof. To Searle and other people who share his intuitions, the idea of a separate mind in the head of the agent is not plausible -- his response is basically to laugh. People who favor the Systems Reply generally don't want to push that line of argument because the idea that a person could memorize the rules for simulating a Chinese speaker is not plausible to begin with. Looie496 (talk) 17:35, 25 November 2011 (UTC)
- Searle seems to think it is a proof. Oh dear; are you saying there is no reply? ErikHaugen (talk | contribs) 21:59, 25 November 2011 (UTC)
- He doesn't say on that page that the Chinese Room is a proof, he says it is a thought experiment that illustrates a crucial tenet of a proof, namely, "Syntax by itself is neither sufficient for nor constitutive of semantics". I am not saying that there is no reply, only that the most useful reply to a misleading illustration is to show how it can be made better. Looie496 (talk) 16:43, 27 November 2011 (UTC)
- Ok fine, but doesn't the "conscious Chinese-understanding entity inside his head" reply expose a hole in the proof? Does anyone take the proof+CRA seriously as a proof? If so, my point remains that this article ought to have something to address this, since it appears to be a very obvious flaw. Let me put it this way—even if it isn't taken seriously as a proof, the analogy of a running program is just wrong, since Searle assumes it is the person in the room who should have understanding, when really the person in the room is just the CPU and we wouldn't expect the CPU itself to have the understanding. Is there a way to clarify this article so that the thought experiment makes any sense? Or does it just not make any sense despite how seriously people seem to take it? ErikHaugen (talk | contribs) 05:43, 28 November 2011 (UTC)
- He doesn't say on that page that the Chinese Room is a proof, he says it is a thought experiment that illustrates a crucial tenet of a proof, namely, "Syntax by itself is neither sufficient for nor constitutive of semantics". I am not saying that there is no reply, only that the most useful reply to a misleading illustration is to show how it can be made better. Looie496 (talk) 16:43, 27 November 2011 (UTC)
- Searle seems to think it is a proof. Oh dear; are you saying there is no reply? ErikHaugen (talk | contribs) 21:59, 25 November 2011 (UTC)
- I think it is important to realize that the CRA is an intuition pump (in Dennett's terminology), not a logical proof. To Searle and other people who share his intuitions, the idea of a separate mind in the head of the agent is not plausible -- his response is basically to laugh. People who favor the Systems Reply generally don't want to push that line of argument because the idea that a person could memorize the rules for simulating a Chinese speaker is not plausible to begin with. Looie496 (talk) 17:35, 25 November 2011 (UTC)
- Yes, I know that the retort about memorizing the rules/database is from Searle. My question is: is there a retort (from Searle or anyone else) to the obvious response to it that there's another "conscious Chinese-understanding entity inside his head" that is not Searle's conscious? This feels like a very obvious response to CRA and way of understanding what is going on. The disturbing thing is either 1.) why does this article have this obvious reply but not a response to it, or if there is no response, then 2.) if CRA is so broken why has it been taken so seriously over the years? I'm assuming that there is some reply to this out there, but I looked and I can't find it. All I can seem to find is Harnad's dismissal, which doesn't help but gives me hope that there is something to be found. If (2) is reality, then I suppose there's nothing to be done here in this article about it, but wow. I'm no cognitive scientist, so maybe this is obvious to the initiated, but as a non-expert reader I gotta say I don't get it. ErikHaugen (talk | contribs) 05:52, 25 November 2011 (UTC)
You are not the first person to suggest that we add a paragraph here describing the strengths and weaknesses of the "two minds in one head" argument. We have yet to find someone who has the time to research and write this paragraph. There is a lot of literature on this, but it is (in my view, anyway) very technical and convoluted. Currently, the article just bounces over this topic to avoid getting bogged down. The "Virtual Mind" version of the system reply is much more clear and direct, and then we get to the discussion of "simulation", which is the heart of the matter. So, to answer your question: This swamp has been edited out for brevity.
You have to try to look at it from Searle's point of view, or else it will make no sense. The "retort" refutes the (naive) "Systems Reply" by removing the system. (As Searle writes, he "got rid of the system".) The only system that had been described so far was built from a particular set of physical objects. He got rid of the system by getting rid of the physical objects. The system, as it was described, is gone.
It may be possible to have "two minds in one head", but this is not the same system that was originally described. This is moving the goal posts. (This may be why Harnad dismisses these replies as "ad hoc".) The "two minds in one head" argument doesn't directly defend the naive set-of-physical-objects systems reply; it's just a consequence of functionalism itself. If you're a functionalist, it makes perfect sense. If you're not, it seems off the point. (And, as you noted, no one from the biological consciousness side bothers to even discuss it.)
Remember that Searle argues that there may be some specific biological process needed for phenomenal consciousness. This is the position the systems reply needs to refute. If you believe in consciousness and you think it has a biological basis, then you need to know how the System is implementing this "biological basis" before you will agree that the System is conscious. It doesn't really matter if the system is in a room or if it's in some guy's head. If you don't agree that the Chinese Room is conscious, why would you agree that some "simulated chinese room in a guy's head" is conscious? What does this really prove? ---- CharlesGillingham (talk) 09:22, 1 December 2011 (UTC)
- There is a lot of literature on this, but it is (in my view, anyway) very technical and convoluted.—Can you give me a pointer to some? I'd love to hear some kind of reply to what seems like an obviously fatal flaw in Searle's argument—something to help me understand why this wasn't DOA and why we are talking about it decades later. Anything. I'd love to take a stab at adding it to the article if there is anything. To be clear, that is my complaint about the article as it stands. Essentially, it is "here's a thought experiment/associated proof that understanding can not be achieved simply by a digital computer running a program (ie a Turing machine), IOW, a proof that computationalism is bogus. Here's a mind-numbingly obvious fatal flaw in that argument. Nevertheless, people take it seriously." The reader is—well, I am at least—left wondering why that is? Perhaps the people that take it seriously don't understand what a Turing machine is, and don't understand that Searle's conciousness does not map to the state diagram and the tape of the turing machine, but rather maps only to the thing reading the state diagram and moving the head/etc. No "computationalist" would ever dream that that part would have understanding by itself in any way. Absurd. But it is very diffcult for me to believe that so many top-shelf academics would make such a silly mistake, so I *must* be missing something—am I misunderstanding the intent of the CRA? Is there a reply to the obvious "two minds" comeback to the "removing the system" argument that I can not seem to find?
- He got rid of the system by getting rid of the physical objects.—But then he made a new one. The components of his new one map, 1-1, to the components of the old one. I don't see how this is interesting or can be seen as an effectual reply?
- It may be possible to have "two minds in one head", but this is not the same system that was originally described. This is moving the goal posts.—You mean Searle moved the goal posts? What? "Move the goal posts" implies "cheating" of some kind. Can you elaborate?
- Remember that Searle argues that there may be some specific biological process needed for phenomenal consciousness. This is the position the systems reply needs to refute.... If you don't agree that the Chinese Room is conscious, why would you agree that some "simulated chinese room in a guy's head" is conscious? What does this really prove?—No, I don't think that's right, and this confusion comes up a lot. First, the systems reply does not disagree that the room might be conscious—in fact, the whole point of the systems reply is that the room might be conscious. Second, I'm reading the CRA (and associated "proof"—I'll use CRA as a shorthand for both) to be a demonstration that there must be a biological (or at least non-Turing machine/VNA) thing going on. The article says CRA "holds that a program cannot give a computer a "mind"". (emphasis added) The systems reply says "no, you have not proved this because your analogy is completely wrong". The systems reply does not claim a proof that the system understands—just that the part that understands might be independent of or at least not the same as Searle's character. In other words, there seems to be some confusion here about whether CRA is trying to disprove computationalism or merely whether it is trying to argue that a program that passes the turing test might not really "understand". Searle thinks its the former. ErikHaugen (talk | contribs) 23:31, 1 December 2011 (UTC)
- I think you might have misunderstood me on your last point above. The "you" I was referring to (in "why would you agree") was Searle, or someone who agrees with Searle. Sorry that was misleading.
- The point is, if the systems reply is not persuasive (to Searle), then two-minds-in-one-head is certainly not going to be persuasive (to Searle). Both arguments assume an abstract functional system can understand. Searle doesn't agree that anything abstract can understand; he thinks only physical things can understand. So when his critics say "system" he thinks (or pretends to think) they are talking about a physical thing: the room and everything in it. So he takes away the physical objects that make up the system. But what his critics meant was an abstract functional system (i.e., something that can be equated by using a 1-to-1 correspondence, as you said.) Both sides have these deep seated intuitions about consciousness (as Louie mentioned above): for Searle, it must be physical. For his critics, it must be information, living out in the platonic realm with the numbers. Basically, as I see it, Searle is responding to a position his critics don't hold, which makes absolutely no sense to his critics and then they respond with pretty much the same argument they started with, at which point Searle & co. decide they're wasting their time. Both sides have begun wondering if the other side is joking or crazy. (No offense to Searle or his critics.) Anyway, that's my take on it.
- This is why, as an editor, I think that covering any more of this argument is a waste of the reader's time. If there were a way to responsibly skip Searle's response all together, I would prefer to do that.
- I think the article needs to move forward to the other four arguments in the section. They are very cogent, directly on point and fairly easy to grasp. (1) Minsky's "virtual mind" analogy is much more specific than the ambiguous word "system" and Searle can't deliberately misunderstand it. (2) "Simulation verses reality" gets to the heart of the matter: what is the ontological status of information and of the mind, and do they have the same status? This is the question at issue. (3) The system reply proves that the man's understanding is irrelevant, so the Chinese Room is at least misleading. (4) The system reply does not provide any evidence that the system is conscious (as you mentioned, above), and therefore hasn't given Searle any reason to think there is consciousness in the room. These are solid, comprehensible arguments for and against the Chinese Room.
- You asked for sources and I wish I had time to give you a tidy list. Is this helpful? ---- CharlesGillingham (talk) 13:32, 7 December 2011 (UTC)
- Charles; thanks for writing back, this is very much appreciated. That does make more sense than my misunderstanding, but still: Both arguments assume an abstract functional system can understand—No, neither argument assumes that (I'm assuming that by "abstract functional system" you mean turing-equivalent machine?); the systems reply is not a proof—or even evidence for—computationalism. It merely points out that the analogy that the CRA fully relies on is 100% flawed: it is not Searle whom we would expect to have any understanding, it is the entire system that would have the understanding, if there is any understanding. So the observation that Searle still doesn't understand Chinese is missing the point in a very fundamental, obvious, easy-to-see way. At least at first pass, for someone who knows what a Turing machine is, it seems that way. So nobody is asking anyone to agree that the room is conscious. In other words, my bewildermint is not "how can anyone support weak AI?" Instead, it is "how can anyone who understands basic things like what a turing machine is think that the analogy in the CRA is valid?"
- Basically, as I see it, Searle is responding to a position his critics don't hold, which makes absolutely no sense to his critics and then they respond with pretty much the same argument they started with, at which point Searle & co. decide they're wasting their time.—That is very depressing. Are you essentially saying that anyone who can wrap her head around why the CRA analogy is broken (ie, computationalists of course assume that the program and data would be "part of" any consciousness) can see how silly the CRA is and all these other accomplished academics simply don't get what a Turing machine is? Is that what you mean by "responding to a position his critics don't hold"? I find it really hard to believe that there is such a basic misunderstanding here. But if that's the case I guess there's not much we can do to the article about it. ErikHaugen (talk | contribs) 20:10, 7 December 2011 (UTC)
- So, as an editor, I agree that the "virtual mind" section is more interesting, and you're right, it sort of mitigates the problem I'm describing. But only because it ends with essentially a statement that the CRA is broken. The reader is still left wondering why it is so famous and why some serious academics still take it seriously.
- "Is this helpful?"—That is fascinating! What a fun site. But unfortunately none of the coherent replies to the "internalization reply" seem to themselves have any replies. ErikHaugen (talk | contribs) 20:10, 7 December 2011 (UTC)
- I said earlier that "Searle is responding to a position that his critics do not hold", and (from your post above), I think I need to be little more specific about what I mean. His critics meant to say "of course the man in the room doesn't understand, it's the information processing system encoded in this pile of stuff that's doing the understanding." Searle takes them as saying "of course the man in the room doesn't understand, it's this pile of stuff that's doing the understanding." He responds to this position, which is (a) not the position his critics hold, and (b) ridiculous. Why would some random pile of stuff have understanding? You need to give him some reason why this list of things has understanding, or else he's not going to take you seriously. It's just a pile of non-biological stuff and a guy who doesn't speak Chinese. He writes "I feel somewhat embarrassed to give even this answer to the systems theory because the theory seems to me so implausible to start with. The idea is that while a person doesn’t understand Chinese, somehow the conjunction of that person and bits of paper might understand Chinese. It is not easy for me to imagine how someone who was not in the grip of an ideology would find the idea at all plausible." (Searle, 1980) I get the feeling that he knows that "the internalization response" is (in a word) lame, but he doesn't think he needs to try very hard because he thinks it is obvious that the systems reply misses the point.
- So here is the key point: the system reply is ridiculous unless you have some plausible explanation of (1) what "the whole system" is and (2) how it might have a mind. It's not the pile of stuff. Well, then, what is it?
- And question (2), hopefully you will recognize, is not a trivial question at all. We don't know how the brain produces consciousness, not exactly. How could we tell if the "system" can or can't produce it? To computationalists, it seems plausible that an information processing system could have consciousness, but they can't prove it, not completely. To "biological naturalists", it doesn't seem plausible at all. Human-like consciousness is an experience that people have. Why would anything else have this experience?
- As Louie noted above, the Chinese room depends on "intuitions" (as Dennett calls them) or "the grip of an ideology" (as Searle called it above). If you have computationalist intuitions, then you hear the word "system" as meaning "information processing system" and (because you're a computationalist) you believe that an information processing system can have a mind, so you think it's plausible that a "system" can have a mind. But of course, you've just used computationalism to keep the system reply from being ridiculous. But computationalism is what the CRA was trying to refute. You can't refute a critique of computationalism by assuming computationalism. The only people who will be convinced are your fellow computationalists, who didn't need to be convinced in the first place.
- You said above that the systems reply is not a proof—or even evidence for—computationalism. But, this is precisely the problem with the system reply. You need computationalism just to make sense of the system reply. You must begin with the intuition that it's obviously plausible that a "system" could have consciousness, or else the reply is incoherent.
- Searle doesn't have computationalist intuitions, and I think it's important to acknowledge that Searle's "biological naturalism" is not completely crazy. It's perfectly reasonable (in our modern materialist world view) to think that consciousness is the result of the action of a particular biological device. As I said at the beginning, you have to try to see it from this point of view, or else the CRA will make no sense.
- On both sides, we need to escape "the grip of ideology" in order see the real issues: What causes consciousness to exist? What machinery is required? Is a computer program enough, or does consciousness require more than this? In my own view, neither side has successfully answered these questions without assuming the conclusion. And this is why I think that CRA remains interesting: the central issues remain unresolved. ---- CharlesGillingham (talk) 21:43, 9 December 2011 (UTC)
- Again, it feels like you're putting a burden of proof on the systems reply. It isn't attempting to prove anything, though; it's just pointing out that the CRA is a broken analogy, and that the part where it says "Since he does not understand Chinese, Searle argues, we must infer that the computer does not understand Chinese either" is a total nonsense non-sequitur. That's all. You must begin with the intuition that it's obviously plausible that a "system" could have consciousness, or else the reply is incoherent.—I don't think it's obviously plausible that a turing machine-type system could have consciousness. I don't think anything I've said supposes that? I'm just saying that if it did, then of course it wouldn't be the "CPU" part of the computer that has the understanding by itself, so the "proof" part of this is nonsense. It sounds like the reply is just "Well you can't prove your side either." Sure—but is there anything left of the CRA after that exchange? This reply doesn't seem to do anything to salvage the CRA itself. My complaint with this article is not at all that computationalism should be obvious to everyone. I'm very sympathetic to the skepticism that a computer program can understand like a human mind can; my complaint is simply that it is unclear to me from reading this article how the CRA can possibly be seen as helping to bolster that skepticism in any way.
- I don't doubt that neither side has successfully answered the questions, and I don't think weak AI is crazy. I just think the CRA is crazy or the article is missing something important. I'm holding out a faint hope that the computationalists are wrong; the idea that it might be possible to build a super-turing machine and solve undecidable problems is fascinating. It's just difficult to see how the CRA sheds any light on the question. ErikHaugen (talk | contribs) 09:28, 11 December 2011 (UTC)
- Well, I guess we can leave it this: the vast majority of critics agree with you that the CRA is crazy. So you're certainly not alone. Pretty much everyone also feels that the simple systems reply is enough to defeat the argument, and can't make any sense out of Searle's position (or Harnad's, or Jeff Hawkins', etc.). Hopefully, the article is getting this across when it says "most people think the Chinese room is dead wrong". ---- CharlesGillingham (talk) 19:01, 11 December 2011 (UTC)
It is all very simple. Is it not?
Proposition 1. Thinking is problem solving, else if it is not this, there is nothing worth saying about it.
Proposition 2. No human has ever been proved to have solved an incomputable problem. Does anyone know different?
Proposition 3. A computer program can in principle be written to solve any computable problem, and in principle any computer can solve this - given enough time.
Conclusion 1. There is no problem a human can solve that a computer cannot solve.
Conclusion 2. If thinking is problem solving, implementing a program enables thought to take place.
I guess this does mean I rule out metaphysics, which indeed I am happy to do. — Preceding unsigned comment added by Froh (talk • contribs) 17:44, 16 December 2011 (UTC)
- "Does anyone know different?"—Well, that's sort of the whole question isn't it? Maybe understanding itself as humans do it is not computable. It's pretty fantastic to imaging solving incomputable problems, so it seems unlikely, but your dismissing it here seems like begging the question to me. Are you proposing some kind of addition/change to the article? ErikHaugen (talk | contribs) 19:41, 16 December 2011 (UTC)
I guess I am suggesting the whole thing is a non question, and proposing the entry say something to that effect. But as this is - to say the least- a long standing argument and this is my first contribution, I am somewhat tentative about this. I think that discussion of 'understanding' in terms of something other than problem solving is metaphysics. I am not prepared to accept that metaphysics is anything at all. If this is 'begging the question' we are at a stand. Can you explain 'understanding' in terms of something other than problem solving, without metaphysics?Froh (talk) 21:16, 16 December 2011 (UTC)
- Are you talking about what Norvig is getting at in the quote in the article: "as long as the program works, they don't care whether you call it a simulation of intelligence or real intelligence"? As far as I am aware, the issue that the CRA is trying to address (although see above, I feel I must be missing something) is whether everything the brain does can be represented by a turing machine–equivalent device. "Can you explain 'understanding' in terms of something other than problem solving, without metaphysics?"—Maybe it's fine if we want to think about understanding as problem-solving. The question is is it a computable problem, or is it an incomputable one, like, say, the halting problem or finding the nth busy beaver number. I don't think this has anything to do with metaphysics. Maybe the brain has some mysterious machinery that allows it to be a hypercomputer; I've never seen a shred of evidence or any reason to believe this is so, but I also am not aware of any reliable sources that would justify saying what you propose here. Are you? ErikHaugen (talk | contribs) 00:10, 17 December 2011 (UTC)
I am following Dennett in describing consciousness as a user illusion. If consciousness is an illusion then we don't need to worry about what it is. Human understanding (=thought) allows us to solve problems, and only (I think) computable problems. I guess that leads back to functionalism. I don't know how to prove that consciousness is a user illusion, but I don't know how Searle can prove it is not. I think that is a problem for Searle to address. Until he does, I see no merit in the CRA. But on reflection I am not saying anything not said before.Froh (talk) 01:20, 17 December 2011 (UTC)
Flawed Logic
The only thing Searle's Chinese Room thought experiment proves is that Searle does not understand logic. He says that since the components executing the program do not understand Chinese, the system does not understand Chinese. There's no logical connection between these two; applying the same argument to the human brain, we would conclude that since the individual neurons do not understand anything, (which they don't), the human brain does not understand anything either. This may be true, but it doesn't follow.
It is amazing that a piece of so obviously flawed logic should still be discussed. And yet, the only person I've seen who actually dismisses the Chinese Room argument is Douglas Hofstadter, who quite clearly stated that the argument is drivel from start to finish. The fact that he did it an article written entirely without using the letter 'e' is a feather in Hofstadter's cap. Gnomon, February 2012
- Unfortunately I have to repeat something that needs to be said on this page over and over again: the purpose of the talk page is to discuss how to improve the article on the basis of reputable published sources, not to discuss the topic itself. Regards, Looie496 (talk) 00:05, 21 February 2012 (UTC)
- However, there may be a case for including Hofstadter in the article. Myrvin (talk) 09:33, 21 February 2012 (UTC)
- Gnomon: Perhaps you could say where this Hofstadter article is to be found? I have re-read his Reflections following the reprint of Searle's paper in The Mind's I. The criticisms there are much wider and more detailed than a mere argument about logic. In fact, I can't find what you say about logic in there. There must be another article somewhere - I would like to read it. Myrvin (talk) 10:24, 21 February 2012 (UTC)
- It's in Le Ton beau de Marot -- Hofstadter translated the Chinese Room argument into what he called "Anglo-Saxon", which is a version of English that does not use the letter E. Looie496 (talk) 22:14, 25 February 2012 (UTC)
- ^ source
- ^ in progress
- ^ In progress
- ^ in progress
- ^ in progress
- ^ in progress
- ^ in progress
- ^ in progress
- ^ Cite error: The named reference
Searle1980StrongAIQuote
was invoked but never defined (see the help page). - ^ The Rediscovery of the Mind, "The Recent History of Materialism" (pp 40-45, "V. Black Box Functionalism", "VI. Strong Artificial Intelligence")
- ^ Searle 1980, p. 1.