Reconstructing the Chinese Room (III)

Replies and objections*

The ‘System reply’

The reply: The person in the room is not a computer or software it is a CPU (“as it were”) running a software. So of course the person does not understand Chinese, but the whole system does.

Searle’s reply: (1) The system has no more means of attaching content to the symbols that the “CPU-person” does. Give him the walls, pen, paper etc. Conjunctions of that person with bits of paper and whatnot will not give cognitive features to non-cognitive entities, that is, a program (2) And even if it does, consider that the person internalizes everything by memorizing every single bit of what’s outside the Chinese room. (3) This reply would have the absurd consequence that “mind is everywhere.” For instance, “there is a level of description at which my stomach does information processing” there being “nothing to prevent [describers] from treating the input and output of my digestive organs as information if they so desire.

This in fact a very common reply to functionalism and it is based on the idea that “non-standard realizations” of cognitive phenomena will not be cognitive.

The ‘Robot reply’

The reply: There is no understanding involved because there aren’t right kind of causal relations between the symbols, the program and the referents. An appropriately programmed robot, in rich causal contact with its environment (capable of reacting to stimuli, memorize sense-date, moving etc.) would indeed have genuine understanding.

Searle’s reply: (1) First of all, this concedes the soundness of CRA. To insist that syntax plus external causation would produce semantics is to agree that it otherwise wouldn’t. (2) Intentionality is irreducible, that is to say, you cannot reduce it to non-subjective phenomena (such as the action of stimuli upon some receptors). (3) But give the man in the Room all the senses and still no understanding of Chinese would occur.

The ‘Brain Simulator reply’

The reply: What Searle rules out is the possibility that what the computer does is simulate the actual activity of the brain. Thus, classic Strong AI aside, an appropriately configured and trained connectionist network would have the genuine psychological properties and would do so “solely in virtue” of having the properties in question.

Searle’s reply: (1) “I thought the whole idea of strong AI is that we don’t need to know how the brain works to know how the mind works” (2) The computational power of a simulated neuronal network is – so the Church-Turing thesis says – the same as that of a Turing-machine. In other words, ‘any computation you can do on a parallel machine you can do on a classical machine” (3) The Chinese Gym: Imagine many-many-many human beings doing whatever a single individual firing of a neuron does inside a brain. Still, none of the individuals, nor the whole room would understand Chinese. (3) Imagine a man operating water-pipes instead of symbols. And let’s assume that the water pipes work, formally, in the exact same way as neurons do. Then he would be simulating a brain and still not have understanding of Chinese.

Preston (2002) identifies further objections which – although haven’t received a name or a tag – might be of equal importance as the ones above. I will note here their main points also because neither Damper (2006) nor the two internet articles from Stanford Encyclopedia of Philosophy & Internet Encyclopedia of Philosophy say anything about them. I’ll also have a go at some labels…

The “running program” reply

The reply: The deal with the Chinese room was that the man inside was implementing a program and that we make decisions about this man by regarding the properties of implementing (or running) a program. However, Searle’s premise is that Syntax is not sufficient for semantics – in other words that the program (not a program run) is not made up of anything else that syntax. Although this might be the case, the question still remains whether a running program is doing anything different. Hauser (2002) says – not very convincingly, if I may say – that it does, it must do, and recent empirical studies had shown how this could be the case. Whether one chooses to call this intentionality or not is a matter of terminological dispute.

The no-rule reply

The reply: Even the simplest rule-following operations require agents capable of exhibiting the capacities characteristic of normativity: they must understand the rules being followed, be capable of explaining, justifying, and correcting what they (and others) do by reference to the rules in question. Both Turing and his followers mis-characterized his achievement. What they had done was to identify the set of functions which could be processed without intelligence. In this view, what weak & strong AI is doing is in fact showing how ultimately mechanical objects can appear to be performing what humans are actually performing. They only appear to be doing it because they don’t do it in the same way and by using the same means.

______

*The series “Reconstructing the Chinese Room” follows some of the articles in:

Preston, J. & Bishop, M. (2002). Views into the Chinese Room: New essays on Searle and artificial intelligence. Oxford: Clarendon Press.

The first three posts will follow:

Preston, J. (2002). Introduction. In Preston, J & Bishop, M. Views into the Chinese Room: New essays on Searle and artificial intelligence, pp. 1-51. Oxford: Clarendon Press

Other works cited here are:

Damper, R. (2006). The logic of John Searle’s Chinese room argument. Mind, 16, 163-183
Hauser, L. (2002). Nixin’ Goes to China. In Preston, J & Bishop, M. Views into the Chinese Room: New essays on Searle and artificial intelligence, pp. 123-144. Oxford: Clarendon Press