Author: Uri Blass
Date: 15:21:47 05/07/05
Go up one level in this thread
On May 07, 2005 at 17:32:27, Peter Fendrich wrote: >On May 06, 2005 at 13:17:44, Uri Blass wrote: > >>On May 06, 2005 at 13:08:31, Alex Schmidt wrote: >> >>>Hello Uri, >>> >>>>I think that we need to generate secret test that can be run automatically to >>>>detect if a program is clone of another known program and send it to Leo and >>>>Frank quisinsky so they can detect clones fast when they get one. >>> >>>a nice idea. The only problem is that the evaluation can be changed and the >>>moves can be different. So it would be necessary to write an own program that >>>can send special commands to the engines and seeks for identical replies. >>> >>>Maybe inbetween could be a good basis for such a program? >>> >>>Best, >>>Alex >> >>In the cases that were detected the evaluation was also similiar. >> >>similiarity in evaluation and in moves can be one thing to check automatically >>but not the only thing and it is better if cloners will not know exactly what is >>checked automatically. >> >>Uri > > >Uri, >You talk about secret positions but _someone_ have to know about these >positions. If for instance you or some other programmer develop these positions, >you can be a cloner yourself. I'm not accusing you of anything, it's just a >theoretical reasoning. Furthermore, if I was accused of that Terra or Alaric >were clones, based of some secret positions, I would demand to see these >positions and claim that someting is wrong. I do not think that you need to be afraid of secret positions. The secret positions will need to satisfy some conditions. They will need to show that all the free source codes are different in many positions in every stage of the game and that all the known clones can be detected by them. The target is that people who generate a clone and do small changes in the evaluation will not be able to escape from clone detection. Maybe if you do big changes in the evaluation you may escape from clone detection by the evaluation part of the test but I think that in that case the program will usually be significantly weaker than the original program. I think that after detecting clones based on secret positions it may be easy to generate different positions that prove that the program is a clone and expose them so the positions that are used can remain secret and in the worst case you may expose part of the positions after the clone is detected that is better than exposing all of them even before detecting clones. Uri
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.