All I'm saying is that when you say that a computer "can't"
do
something, you need to be clearer that you're talking
about the hardware level
in a general-purpose computer,
Yes I confirm this is what is
meant. I thought I was clear but it seems I was not clear enough. Thanks for
pointing this out.
and it's not so much that it "can't", it's that
it doesn't need to.
No the issue is deeper than this. The fact
that a computer can't process the semantics at a hardware level imposes a
constraint on the type of procedures which could be programmed. Anything which
requires a judgment call based on meaning is out of question. Only procedures
which happens to be consistent with the semantics without using the semantics in
any of the steps will work. These procedures are exactly those mathematicians
call algorithms. This is why symbol processing is crucial. The ability to
recognize symbols is part of what computers can do at the hardware level. A
computer has no trouble recognizing the series of bits representing the letters
FROG. Going to the next abstraction of semantics and telling which animal these
letters stand for is what is out of bounds.
When you say that the ability to
be consistent with semantics is the same as the ability to understand semantics
you miss this crucial point. The inability to make a judgment call based on
semantics is a limitation humans don't have. Humans may apply procedures which
require judgment calls. Courts do this all the time. Legal tests are not
mathematical algorithms and they cannot be programmed in
computers.
Actually, the confusion was within the phrase
"what
computers can do".
Yes, this is where the confusion
lies.
If the computer processes the symbols without using their meanings, on
what basis do we say the computer can process the meaning? The limitations on
judgment calls is real. It affects the work of the programmer. He must abstract
away meanings in order to get a program which is machine executable. How can we
ignore this limitation to say the computer can recognize meanings?
The
computer never uses the meanings. It is like the printing press. The press can
print a novel which carry some meanings. Is this the function of the press? If
not, why should meanings be the function of the computer? Both technologies work
at the hardware level on symbols. None use meanings.
I will give you an
example of why this matters. Suppose we have an algorithm for calculating the
shape of a parabola in three dimension. An engineer may use this to compute the
shape of an antenna. The semantics of the symbols is this engineering
application. But there is nothing in what the computer does which depends on the
symbols to mean something about an antenna. A mathematician may as well use the
same calculation to compute the pure geometric shape. This is what algorithms
are. Procedures which work independently from the semantics. In this example
there is some degree of semantics in terms of geometry but even this is not used
by the computer. It just pushes the bits around.
[ Reply to This | Parent | # ]
|