Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Some facts about Deep Thought / Deep Blue

Author: Robert Hyatt

Date: 07:42:08 08/31/01

Go up one level in this thread


On August 30, 2001 at 18:43:39, Bruce Moreland wrote:

>On August 30, 2001 at 15:30:59, Robert Hyatt wrote:
>
>>I agree.  There are three positions someone can take on the DB issue.  I will
>>list them and then pick the one I like:
>>
>>1.  DB sucks and is worse than today's micros.
>>
>>2.  DB is invincible and is so far above today's micros it is not worth
>>    discussing.
>>
>>3.  There is ample evidence that older versions of the thing were invincible
>>    when they were playing.  And the newest version did something nobody else
>>    has repeated, yet (beating Kasparov in a match).  This leads me to believe
>>    that they certainly are ahead of today's machines, until one of today's
>>    machines shows some evidence of catching up to them.
>>
>>I fall in category 3 above.  Several fall in category 1.  Category 2 isn't
>>really worth talking about.  I would personally be just as happy as anything
>>if the (1) group would just remain silent.  Because (1) is not supportable by
>>any evidence other than prejudice.  I think there is a lot to be learned from
>>the machine, and it will be learned over time...
>
>There are some people in and around this field who are in the first camp.  These
>people will waste your time if allowed.  Please don't accuse me with these
>people.

I hope I didn't.  I consider you "simply skeptical" since we can never get our
hands on the thing to see what it is really capable of.  That is ok.  But the
ones that say "it is obviously weaker than today's micros" have their heads
someplace dark and stinky.  :)



>
>I think that your third category is also too much of a stretch.  You are trying
>to extrapolate along too many axes.
>
>Older versions of DB were not invincible.  The thing in Hong Kong lost a point
>and a half playing against micros running on 100 mhz Pentiums.

I don't disagree.  But overall, it had an alarming record against _everything_
from micros thru supercomputers.



>
>The Kasparov match was very weird and I don't think it makes sense to take
>anything from that.  If there were more data-points against humans, it might
>make sense to make conclusions about that, but we have none.

There are dozens of games played in public, as I have mentioned.  I am not
aware of any PGN from them, but the games were played and a few recorded the
moves for whatever reasons.  I watched some of them but didn't feel like
writing something down.  These were "DB Jr" games of course.


>
>I think it is logical to suspect that DB would be very strong against modern
>micros on modern hardware.  But lack of *any* concrete evidence precludes any
>conclusion.
>
>IBM would have you believe that this event had something to do with science, but
>it had nothing at all to do with science.  It was a giant publicity stunt, and
>once they got what they wanted they weren't willing to do even one tiny bit of
>science with the thing.  No data came out of that project.  All that it
>generated was PR.

There were some interesting points from the experiment.  special-purpose
hardware is a killer in terms of performance.  I think the most interesting
thing is that the _really_ didn't expect to win match two.  I think they
really thought that the third match would be the one where they scalped
Kasparov.  I got a sense of "let-down" more than once from conversations with
them... they were mentally ready to continue tuning as the DB2 machine was
barely ready in time, and it was missing several important pieces that were
planned for (hashing in hardware support was there, but no memory to do it,
new eval hardware that had 0 for the weights, etc.)

IE suddenly it was over, before they were really ready and prepared for it to
be over.



>
>I refuse to even entertain the notion that it's a good idea to use science or
>math or logic to somehow manufacture evidence that can be used in order to
>proclaim them as permanent kings of the computer chess world.  They could have
>provided a huge amount of data in a single afternoon, but instead they provided
>nothing.
>


I don't want to proclaim them "kings" either.  But I do want to avoid
proclaiming them "dunces" which many here want to do.





>What should be concrete and provable is turned into a matter of religion, where
>nothing is provable and everything must be taken on faith.
>
>This is not how science is supposed to work, and scientists are not supposed to
>condone that kind of crap.
>
>IBM dicked over this field, all because they wanted to save money that would
>have otherwise been spent on a single commercial shown during the Super Bowl.
>
>bruce


IBM does it.  Microsoft does it.  Firestone/Ford does it.  That is just a
way of life, it seems.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.