Re: Plagiary, CGF view

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

Last year Prof Chen started rewriting Handtalk to enable it to
get stronger in the future.  At FOST he said that his new program
was still a little weaker than Handtalk.  That is why he changed
the name to Goemate, since it has a new engine.  The old Handtalk
can also beat Goemate.

This plagerism was also not fair to the other competitors.  It's as
if Prof Chen had been allowed to enter Handtalk 3 times, and took
the top 3 places.  The second best program comes in 4th instead.

Even between evenly matched programs there is a lot of variation from
game to game.  A year ago I did a lot of testing between Many Faces and 
Handtalk.  Many Faces won about 30% of the games, but there was a huge
variation is score from game to game.  Most games were in the range of
Many Faces wins by 10 to Handtalk wins by 30, but there were some 100
point wins by each side.

The claim of plagerism is based on the other programs copying Handtalk's
data and code directly, rather than just mimicking its moves.  The other
programs made the same moves as Handtalk because they had the same engine
inside, not because they were designed to mimick Handtalk.

I talked to a lot of go programmers and they never said anything about
mimicking Nemesis.  And Nemesis wasn't the top program after I started
participating.  I had never heard of it before the first year I competed,
and came in 4th (to nemesis' 5th).  Bruce always thought that a program
should be able to play well without reading.  Since the other programs
all were developing good tactics, I suppose you could say they were
designed to beat a program that didn't read :)

David

At 10:30 AM 3/15/99 -0500, P. Shotwell wrote:
>Hello,
>
>Just a curious question.
>
>If Hamlet and Silver Igo were 'copies' or 'partial copies' of Handtalk,
how did
>they beat Handtalk? They must have improved on Handtalk or else they were
>specifically programmed to beat Handtalk. From what I know, in the early
years
>computer go programs were specifically designed to exploit weaknesses in
Bruce
>Wilcox's Nemisis program. Had they, in effect, 'copied' Nemesis? They lacked
>strength in other ways and thus looked better than they actually were. But
they
>still won. But then, of course, these winners had to take on the problem of
>beating other programs which had done the same thing. Therefore there was
>Darwinian 'progress' and evolution in the most fundamental way--survival
of the
>fittest.
>
>Is this somewhat the same situation? I don't remember Wilcox complaining
except to
>say that his program mimicked 'human thinking' better than the others.
However, in
>a money tournament, we are dealing with money and go programs, and not
Artificial
>Intelligence (meaning, who says that humans know how to play go the best
way?) so
>the final proof at this stage of our ignorance, as in war, would seem to
be in the
>winning or losing, not how it was done.
>
>In other words, with all due respect to Dr. Chen, it would seem to me that
his
>task will be to figure out through 'reverse engineering' why Handtalk lost
and
>correct the situation.
>
>Although I can claim some knowledge of the history of Asian war
stratagems, I am
>not a computer expert so I may be missing some of the point of this
discussion.
>Please correct me if I am wrong.
>
>Thank you,
>Peter Shotwell
>shotwell@xxxxxxxxxxxxxxxxx
>
>