One of the most popular tools used in the attempt to define consciousness and intentionality is John R. Searle's Chinese Room argument. While the Chinese Room is intended to show that intentionality can't be induced merely by sufficient complexity, I believe that there are significant flaws in the argument, as well as the Turing Test it is derived from. The Chinese Room seeks to separate man from machine regardless of their respective functional levels, but in this paper I hope to show that the argument is flawed by the very assumption that a creation of man has no consciousness in it. While not one of the five provided replies to the Chinese Room argument, I would like to examine it for this assignment. I call this argument "the Proxy reply."
Briefly, the Chinese Room argument describes a hypothetical room with a person who doesn't know Chinese, and a computer or other device that has some programming such that it can respond to Chinese. Messages written in Chinese are passed into the room, where they are entered into the device by the person, and the response is sent outside the door. In an effective Chinese Room, the device produces results that are indistinguishable from those that might be given by a person who knows Chinese. Searle states: "But all the same, neither the person inside nor any other part of the system literally understands Chinese; and because the programmed computer has nothing that this system doesn't have, the programmed computer, qua computer, does not understand Chinese either." In so saying, Searle claims the room doesn't understand Chinese, and so no part of the room can either.
Before I can progress to the heart of the Proxy reply, there are several areas that require greater specificity for this to be an effective argument. The most obvious is: "What is understanding?" Searle differentiates understanding from mere knowledge by saying that a computer has a syntactic knowledge of the words being said, but that understanding requires a semantic knowledge as well. He ties this with to the mind by saying that consciousness is required for semantic understanding.
Consciousness itself is an underspecified model. So-called 'neural chauvinism' and 'feature bias' are examples of how, in a black box situation, we define consciousness by similarity to ourselves. Even worse is that the philosophical arguments above treat consciousness as a binary trait; something is either conscious or it isn't. At the same time, the current scientific community accepts the gradiation of brain complexity, and the gradiation of 'levels' of consciousness, from human to dog to mouse to flea to microbe. At some point on this continuum a benchmark id drawn for consciousness, and yet the Chinese Room argument ignores the gradiation of consciousness, not allowing that something could be a 'little bit' conscious. Part of the reason for this is the task specified: Linguistic communication is the most evolved conscious trait humans have, and so to pass the Chinese Room (or the Turing Test) based on that trait would require an incredible simulation, or a 'created consciousness' of the highest degree. In comparison, Turing tests based on Chess ability are entirely different, as Chess is not a semantic problem, or to be more specific, it doesn't have to be.
Because the criterion for the Chinese Room is so richly semantic, in order to create a device that passes for a Chinese speaker, a system would have to 'learn' Chinese in one form or another. The mechanism for this learning could be by example, by instruction, or by an experiential reinforcement process, such as a neural net, for example. The basis of the Proxy reply is that the system encompassing this learning process contains a consciousness and an understanding of Chinese, and that this is taken into the room when the device is taken into the room.
For example, say the 'chinese computer' was created by taking a native Chinese speaker (or an arbitrarily large number of Chinese speakers), and posing him with questions, to which he writes replies. The results of the 'black box' testing are programmed into the Chinese computer. This computer is then placed in the Chinese Room. The conventional argument is that the computer doesn't understand Chinese, because it doesn't have semantic content. My contention is that the greater system does have semantic content, because the programmer (or Chinese speakers used as the data source) has semantic meaning associated with their responses.
The fallacy inthe Chinese Room argument is the perception of the room as a 3-D space, instead of a 4-D space. In a 4-D capacity, the Chinese computer is connected to the programmer (or data source), which has consciousness and semantic representation. To put it another way, the Chinese Room argument can successfully prove that my mouth doesn't have consciousness, even though it can speak English. The argument, true though it may be, is irrelevant in any philosophical capacity.
I would assert that it is not possible to create a device, conscious or otherwise, that could pass the technical requirements of the Chinese Room, without imbuing the device with the semantic meaning shared by a conscious entity. If the machine learned from point zero, by speaking gibberish and learning via interactions with Chinese speakers, then it is their semantic knowledge that is in the Chinese Room. If it learned by being programmed by a Chinese speaker, then it is that programmer's semantic knowledge inside the room. Most importantly, if the machine learned Chinese through interactions with another Chinese machine, then it still has semantic meaning imbued within it, because the 'teaching' machine acquired the semantic knowledge from its teacher.
By the Proxy response, the Chinese computer in the room is merely acting as a mechanical proxy for the intelligence that gave rise to its syntactic knowledge. To reduce the problem, if the Chinese Room was limited to three possible 'input statements' and the programmer, knowing what those three statements were, provided three envelopes, each with a response to a statement, then the system does indeed know Chinese, but the system includes the programmer, though he is only represented by proxy in the room. It is important to note that this is fundamentally different than the 'Systems reply' because the consciousness in the Proxy reply is the creator of the device, not the person inside the room who interacts with the device. In the Proxy reply, the person physically inside the Chinese room is inconsequential to the argument.
The most obvious response to the Proxy reply would be that the room, as defined, still doesn't know Chinese, that the semantic knowledge isn't inside the room, and therefore the problem has been reduced to a simple syntactic response. My reply to this response is to ask where semantic consciousness lies in a human being. It's not inside the neurons, just as in the computer's case its not inside the circuitry. The semantic knowledge is a property of the whole, and in the computer's case, the creator is a part of that whole. To cut the person out of that equasion by the rules of the Chinese Room is equivilant to cutting the lips off a person. While it is true that you could use the Room to prove the lips don't have a mind, when you consider that the lips didn't create the words, that proof is not of any value. In the Chinese Room, the computer doesn't create the words, it merely processes them in a fashion directed by a mind that has semantic meaning.
The Proxy reply has ramifications on the mind-body problem as well: The arguments for the consciousness of the machine could just as easily be turned around to argue that a human is merely a mechanical device. If a computer could be said to have semantic knowledge because it learned from an entity with semantic knowledge, then the same could just as well be said of man. Whether semantic knowledge and consciousness are passed along by way of interaction or procreation, there seems to be a strong sentiment that a purely constructed brain would not have semantic meaning. Note that in this case, 'purely constructed' means that the hardware and the software would have to be constructed form scratch, ie not relying on any interaction or knowledge gained from a conscious entity. This is black box engineering in the strictest sense. Ironically, this means that the Proxy reply is an argument against Strong AI, and yet against the Chinese room as well. It attempts to show that Strong AI is flawed, because the software is a creation born of consciousness, and thus is not completely artificial.
In conclusion, the Proxy reply seeks to demonstrate that the Chinese Room argument is flawed by underspecification. The Chinese Room argument assumes without sufficient justification that knowledge can be transferred into a device, but that semantics are lost in the process. While the Chinese Room focuses on the abilities of the machine, it ignores how those abilities were obtained, and the possibility that semantics might give rise to consciousness, instead of the other way around. Further, it makes the assumption that by cutting off a conscious entity from its packaged means of expression (as in the envelope example), the expression itself is no longer tied to consciousness. While it's easy to say that the conduit of communication doesn't have semantic meaning (though words do have semantic meaning, right?) in so stating, any means of communicating semantic concepts is also eliminated.