The Chinese Room, pt. 7

The Chinese Room is not merely intended as a refutation of Strong AI, but also as demonstration of the unreliability of the Turing Test for intelligence.  According to John Searle, the Chinese Room passes the Turing Test without being intelligent in the relevant sense, therefore passing the Turing Test is not sufficient for having intelligence.  It is, partially, for this reason that Searle reintroduces all the chauvinistic and prejudicial preconditions for intelligence (I will use the terms ‘intelligence’, ‘conscious’ and ‘understanding’ rather interchangeably) that Turing attempted to dispose of with his test in the form of causal powers.

There are (at least) two ways in which the Turing Test can be salvaged from Searle’s criticisms.  First, one can attempt to show that the Chinese Room actually is intelligent in the relevant sense after all.  (I personally find the virtual mind approach quite compelling.)  This, of course, would amount to a refutation of the Chinese Room argument altogether. 

While I am not at all persuaded by the Chinese Room argument, as a functionalist I see this approach as being deeply problematic.  The Chinese Room argument is aimed at showing that any and all programs that pass the Turing Test are, by themselves, insufficient for consciousness.  The denial of this conclusion does not entail that every program which passes the Turing Test is, therefore, conscious.  Consider two definitions of the “right” program which the man in the Chinese Room is running:

1. The Behavioral Definition: The right program has the proper input/output relations, regardless of what formal relations hold between the two.
2. The Functional Definition: The right program has both the proper input/output relations as well as the proper formal relations between two.

The functionalist is committed to the claim that anything which instantiated the functionally right program is conscious and any thing which is conscious instantiates the functionally right program.  Nevertheless, there are many programs which are behaviorally right, and therefore pass the Turing Test, without being functionally right.  Thus, the functionalist seems committed to the claim that the Turing Test does not serve as a reliable indicator of intelligence.

There is, however, another way for to save the Turing Test, or something very close to it, from Searle’s criticisms which not only does not deny the strength of the Chinese Room argument but also allows the functionalist to place more confidence in the test.  This approach is to accept that the Turing Test is a reliable indicator of intelligence but not a reliable locator of intelligence.

Consider the case of communicating with a friend by way of walkie-talkie.  Strictly speaking, the walkie-talkie is passing the Turing Test!  The same can be said for a telephone, instant messenger programs, mailboxes, etc.  When we communicate through these mediums, we are completely convinced that we are indeed communicating with some intelligence, exactly the reaction which the Turing Test suggests we should have.  Nevertheless, we are not, for a second, tempted to think that the mediums are where the intelligence resides.

Let us now turn our attention to the Chinese Room which, by definition, passes the Turing Test.  To deny that there is any intelligence, understanding or consciousness anyway is far too strong of a claim.  After all, how in the world is the room passing the test in the first place?  Where did Searle’s rule book come from?  The proper conclusion for Searle to draw from the Chinese Room argument is “Yes, the outside interlocutor is indeed communicating with an intelligent entity, but the man inside the room is not it.  The intelligence with which the interlocutor is communicating is the author of the rule book.”  Thus, Searle can still hold that instantiating the proper program is not, in itself, sufficient for intelligence while at the same time accepting the Turing Test as a reliable indicator of intelligence.

The functionalist, on the other hand, can tell a very similar story.  Even if the Chinese Room instantiates a program which is merely behaviorally right, as opposed to functionally right, the functionalist can claim that behaviorally right program is simply the product of some functionally right program, namely the author of the behaviorally right program.  Thus, the functionalist can claim that while passing the Turing Test does not logically entail the unidentified presence of some intelligent being, is serves, in practice, as a highly reliable indicator (but not locator!) of such.

Let us now consider the case of the Chinese Room which instantiates a functionally right program.  While the functionalist is (it seems) compelled to grant intelligence to this Chinese Room, the question of where the program came from still lingers.  Even if we grant that the room is now intelligent, is it not a sort of derived, second class intelligence?  Isn’t the real intelligence in the programmer after all?

The proper response to this question is that the difference between real and derived intelligence makes no difference since ALL intelligence is derived.  This point, I take it, is similar to Dennett’s rejection of intrinsic intentionality.  Humans, according to the functionalist, are intelligent in virtue of the program they are running in their brains.  Where did this program come from?  The answer is from our biology (by way of natural selection), our culture (by way of shared knowledge and language) and our personal experience with the world.  Human intelligence is no less derived than is that of the functionally right program in the Chinese Room.