[Wittrs] What is a computer anyway?
- From: "jrstern" <jrstern@xxxxxxxxx>
- To: Wittrs@xxxxxxxxxxxxxxx
- Date: Sat, 29 Aug 2009 17:39:36 -0000
--- In Wittrs@xxxxxxxxxxxxxxx, Glen Sizemore <gmsizemore2@...> wrote:
>
>
> ... Was it a computer? I'm guessing, by what you said above,
> you would have to say "no."
Right
> What about sort of more traditional "analog computers"?
> Are those "computers"? Where did the term "computer" come from?
> Who coined it? Is that known? The term is metaphorical - or
> originally was...No? This is brought out in an interesting
> way in the great novel "Cryptonomicon." The protagonist
> (who is fictional) invents the digital computer in an effort
> to break some Axis codes (Al Turing is one of the main characters
> in the book - the genre is sometimes called "alternate history").
> When his boss (a military type assigned to try to manage him and
> Turing at Bletchley Park) asks him what he calls the thing, he
> replies "a computer." The military guy says "A computer is a
> person!"
Yes, the point needs to be made today, but clearly the usage
when Turing wrote his OCN paper in 1936 (before Bletchley) was
for "computer" to be a person, it was a job category like
"programmer" is today.
http://www.abelard.org/turpap2/tp2-ie.asp#index
In the paper it says:
"The behaviour of the computer at any moment
is determined by the symbols which he is observing."
----------------------------------^^^^^
It is, however, the entire crux of the paper, that
this being-a-computer is what his little machines,
that we now refer to as "Turing Machines", could
do for us, mechanically. In that way the TM answered
Hilbert's question about the limits of what, not a
human computer might do, but what a mathematical,
formal system, might do.
--
Now, as to what *I* would say is a computer, it's like
this. I want to find a particular, specific device which
I am sure I would like to call a computer, and use that for
all further discussions of what a computer can do. I will
*stipulate* this as my cannonical computer, even though it
actually has no privileged or minimal position on any
metric I can think of. Now, are other things "also
computers"? Probably. Depending on just how far you want to
bend things, you can argue that this, that, or the other is
a computer. Fine by me. The only condition is that you
have to specify *exactly* the mapping from my cannonical
computer to yours, on request. Turing shows us how to do
that with the UTM. I think Searle *cannot* show us how to
do that with the paint on his wall, invalidating that as
a serious argument. Your little analog devices, I do not
think map to a UTM. You can say you don't care if it maps
to a UTM, it maps to a specific, limited TM that would
otherwise attempt to compute the same equations. OK, fine
by me, it's just not the game I'm playing.
Actually, I may well be demanding too much, in requiring
my computer to be a full UTM. That much generality is
probably not needed. OTOH, you can build a UTM (yes, yes,
yes, not including the infinite tape) as a state machine
with about a dozen states, maybe 99.44% simpler than a typical
modern Intel processor chip, so it is VERY cheap to do,
and we know a LOT of tricks for speeding it up (which is what
the rest of the stuff on the Intel chip is doing, basically)
and that's why we're all sitting in front of one at this
moment, looking at this message.
And so, I see no problem in asserting that I suppose any
one of the several BILLION processor chips, from their first
4004 to whatever they're calling the newest ones, that
Intel has ever built, "is a computer" in my stipulation
(and in spite of the aesthetic inelegance of the design
of many of them). What else is a computer, I think I've
now outlined.
--
However, the really interesting question is, given any
one of them, just what is it (besides my stipulation!)
that *makes* it a computer? That is, what makes even
Turing's TMs and UTM a computer that can do something
useful in the world, not just enumerate and evaluate
a nonconstructive, infinite set of axioms, as was the
ONLY work done in the OCN paper?
Nobody asks that. Nobody answers that.
But one of these days, I'm going to try.
(the problem being it is "obvious", but I guess we
all know just what being "obvious" is worth in
theoretical or philosophical terms. it's that old joke,
where the professor looks at the equation on the board
and begins to tell the class, "From this it is obvious
that ... um, ... " and he then wanders out of the room,
works for most of the hour, then comes back and gladly
completes the thought, "yes it *is* obvious that this
implies the result!" elaborating just those "obvious"
things that make a computer a useful device, is probably
a several hundred page exercise, which will of course
still not convince everybody.)
(LW touches on this too, in his stuff about
mathematical proofs, and being surveyable, and that
every proof is a new proof, etc)
Josh
Other related posts:
