[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: computer-go: Authenticating the identity of a remote go-playing computer program




   From: "Vlad Dumitrescu" <vladdu@xxxxxxxxxxxxxxxxx>

   If I may throw in my 2 pence worth...

   Since as someone said before, cheating cannot be prevented even when humans 
   are playing (either having a stronger player make secret signs, or maybe a 
   more high-tech solution), I'm not sure why it would be a bigger problem with 
   computer-go... except maybe that more people might try to cheat.

   A cheater might win a couple of competitions, but not all of them will be 
   through the net, and if that program never competes "in person" or if when 
   it does it performs poorly, then everyone will know it as a cheater...

   Since organizing an on-line competition would require much less effort both 
   for the competitors and the organizers, I think people will participate even 
   if there were no money prizes. And then I think few (if any) will try to 
   cheat...

   In fact, it is exactly the same situation as when playing on nngs - who 
   certifies that I am who I say I am? Or that I don't have a better player by 
   my side? Or a book? And people still compete there...

   All in all, I don't really think cheating should be a big issue...
   What do you think?

   regards,
   Vlad
   _____________________________________________________________________________________
   Get more from the Web.  FREE MSN Explorer download : http://explorer.msn.com





I pretty much agree with  you, 99% of the time,  all we want is a game
and we don't care who we are actually playing.

There have been a few interesting cases  of possible cheating in Chess
tournament, where one program has  been accused of being replaced with
another much  stonger one,  or  where one program  has been introduced
into a human only tournament but secretly in order to win prize money.
The stories are extremely interesting but I won't tell them here.  You
won't see that case any time soon in GO.

The Chess program Crafty has been a big  source of controversy because
it's public domain, and it's one of the better programs.  On the chess
servers,  there are  many versions  of Craftys being   played by chess
author  wannabees.   In one tournament  there  was a version of Crafty
being played under a different name, but the author was up front about
this and it  was allowed.   However  that didn't stop the  controversy
because he  claimed   that it  was   substantially modifed, and   some
entrants didn't feel this was  the case.  These  are problems that are
likely to occur over and over again and  having some kind of system of
authorship proof would be a blessing, not a curse.

People  have also  taken commercial  programs and  wrapped them  up as
their own.  A competent   programmer can do  this,  and you  would not
notice  by looking at the  screen.  In one  notorious case, the famous
and  powerful mephisto program  (which was a stand-alone computer) was
taken apart   and put into a  different  case and presented as   a new
program.  The  actual author of  this program  was observant enough to
notice that  the moves looked  like moves HIS   program would make and
thus this imposter was exposed.  That's why I believe that the moves a
program make  can  serve as   an   extremely good fingerprint  as   to
authorship.  I think this is much more true in Go than in Chess.

If nothing else, program authors can protect  themselves with a little
judicious  programming.   Some  programs   have a  hidden   feature to
identify their authorship, a unique set of moves that no program would
ever play  without being explicity   programmed to.  If you  hide this
feature,  you can later prove with  no question that a program belongs
to you.

Don