Information Technology Reference
In-Depth Information
However, there is a smaller book that would do the job for the twenty numbers.
Consider a book that contained the procedures PROC0, PROC1, PROC2, and PROC3
from chapter 1. This topic can certainly exist. In fact, it would be more like a small
pamphlet, and the instructions there would allow the person in the room to produce
the correct sums.
However, notice what happens this time when the person in the room memorizes
the instructions in the pamphlet: he learns how to add . In fact, most of us learned how
to add by being given instructions like those in the pamphlet. So the man in the room
does get the right answers, but now he also knows how to add.
This is enough to cast some doubt on Searle's Chinese Room argument. Perhaps
adding two ten-digit numbers is simple enough that a big table of answers would
allow someone to fake it. But adding twenty ten-digit numbers appears to be complex
enough that a book that does the job would also end up teaching the person in the
room how to add. Beyond some level, if you need to get the sums out, you will not be
able to fake it; you will need to do the addition. So why should we think that Chinese
(however that book turns out) is going to be any easier to fake? It might be more
sensible to conclude that once a behavior is complex enough (on the order of adding
twenty ten-digit numbers, say), insisting that the behavior be right will be enough to
rule out various forms of tricks or fakery.
12.5 A final word
So does this settle the question as to whether a computer can think? No, not at all.
At best, it casts doubt on Searle's refutation of Turing's position, which was that we
should not even ask this question in the first place. And new and improved arguments
on this topic may yet appear.
But no matter what one's position on the philosophical issues, we are still left with
what might be called the AI question : If it is indeed true that tricks and fakery are not
sufficient to generate intelligent behavior such as passing some form of the Turing
Test, then what is ? In the end, it is this question that is perhaps the most profound one
to emerge out of the entire discussion, and one that will not be resolved by merely
arguing one way or another.
And where to look for a possible answer is what this topic has been about.
 
 
Search WWH ::




Custom Search