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

[David Stafford <david_stafford@xxxxxxxxx>]

You could send a list of a dozen problem positions and if it got
any of them right you would immediately disqualify it as a fake. 
:-)

This is something of a reverse Turing test: Prove the identity is
not human or human-assisted.  Unfortunately, there is no way to
do this over a network.  A program could always be made that
would prompt a human for help.  Repeatability is too strong a
requirement since the move selected may depend may depend on
factors that can not be exactly duplicated (e.g. machine speed,
free memory, hash table contents, side-effects of non-critical
bugs.)

Someone else suggested the possibility of having the programs
submitted to you and run locally.  This approach has its own
difficulties (chiefly that some programmers might be reluctanct
to distribute their program) but it does guarantee the program is
operating on its own.  Even this is not enough to guarantee
honest behavior as it is possible for someone to plagiarize
another program.

I don't see an iron-clad way to enforce an honest outcome but it
might be sufficient to have a trusted and knowledgeable committee
to resolve disputes in cases where there was doubt.

-Dave


--- Bob Myers <rtm@xxxxxxxxxxxxxxxxx> wrote:
> Is it possible to authenticate the identity of a remote
> computer program
> playing a game of go?
> 
> In Guiyang last month, I was talking with a go programmer,
> discussing the
> possibility of having computer go tournaments on the net
> instead of in one
> physical location.  I know that there are already tournaments
> being held on
> the net, but the stakes are not high enough to motivate anyone
> to cheat.  If
> a large amount of prize money were at stake, on the other hand,
> unfortunately there might be people who would try to cheat (for
> instance, by
> having a 7-dan human pretend to be the computer).  Thus, we
> need a way to
> reliably authenticate the identity of a remote computer program
> playing a
> game of go.
> 
> I know just enough about public-key encryption techniques and
> digital
> signatures to believe that this is possible.  I hope that
> experts on this
> list can shed some light on how it might work.  It would be
> ideal if a
> standard could be established that eventually all go programs
> might adhere
> to and thus allow more tournaments to be held on the net.
> 
> Bob Myers
> IntelligentGo Foundation
> www.intelligentgo.org
> 


__________________________________________________
Do You Yahoo!?
Yahoo! Shopping - Thousands of Stores. Millions of Products.
http://shopping.yahoo.com/