I'm including the comments of a friend of mine, a parallel programming
whiz.

On Tue, May 13, 1997 at 18:03:57 (-0700) Louis Proyect writes:
> The RS/6000 SP's role in the defeat of Kasparov will be used by IBM as a
> marketing wedge against Microsoft.

This is downright silly.  At conferences, people from IBM have acknowledged
that the main benefit of the Deep Blue project is the free, positive PR it
generates.  Somebody at work said somebody at IBM estimated the equivalent
advertising cost of this PR at on the order of $100M.  The project isn't 
terribly expensive to IBM (maybe $10M per year?).  The PR is IMHO "brand
recognition", kind of like the reason Coke and Pepsi want their names
appearing on every surface in sight.

IBM doesn't really compete with Microsoft directly, anyway.  It has
subsidiaries like Lotus that do, in applications software, and it has
relatively minor (for IBM) products like OS/2 that compete.  But overall,
IBM is mostly selling complete systems into specific corporate markets.

> In fact the only difference between the SP2 that I work on and
> the one that defeated Kasparov is that mine has 8 processors while Deep
> Blue has 516 processors working in parallel.

This is quite wrong.  The Deep Blue SP has far fewer CPUs (32?), but a huge
number (about 2,000?) specialized chips attached to these CPUs that are
designed to do nothing but search chess game trees and evaluate game
positions.

> Microsoft's emergence as a successful corporation
> is a study in the benefits of privatizing technology created in the public
> sphere. Gates developed a proprietary operating system that was modeled in
> the final analysis on the publicly available Unix for use on IBM personal
> computers.

Actually, DOS was modeled on CP/M, a commercial system, while Windows was
(as observed) a rip-off of the Mac OS.  Where Chairman Bill has benefitted
from government support is in the utter lack of enforcement of anti-trust
and other "fair play" legislation.  As I'm fond of saying, he should be
rotting in prison rather than building an extension to his mansion.

> What computers like the RS6000/SP can do best is assist human beings make
> *intelligent* decisions. The spread of this type of technology in
> capitalist society has been mostly to keep track of financial transactions.
> In a socialist society, they could be used to monitor resource allocation
> worldwide. Banks of RS6000s in a global network could surely be used to
> calculate the impact of the substitution of railways for automobiles. Or
> the cost of replacing inorganic fertilizers with natural ones, etc. 

I've often thought this would be a good idea, and would demonstrate how
effective a planned economy could be.  Besides the obvious problem that
tying policy decisions to scientific analyses rather than capitalist bribes
is anathema at present, I'm not sure that the state of computer simulation
is up to it, because the supercomputer applications that have received
funding up to now have been primarily useless things like simulation of
nuclear weapons (the current cash-cow for buying big machines).


Bill (speaking for his friend)

-- 
William S. Lear | Who is there that sees not that this inextricable labyrinth
[EMAIL PROTECTED] | of reasons  of state was artfully invented, lest the people
quid faciendum? | should  understand  their own  affairs, and, understanding,
quaere verum    | become inclined to conduct them?    ---William Godwin, 1793


Reply via email to