[ExI] Unfrendly AI is a mistaken idea.

[Lee Corbin]

> As an interesting sideshow to the current world championship candidates
> match, the two top commercial chess programs will go at it starting tomorrow
> in a six game match. 
> 
> http://www.washingtonpost.com/wp-dyn/content/article/2007/05/11/AR2007051102050.html

> look at the games afterwards and figure out which of the games was played by
> computers and which by humans.  I can't tell, however I am a mere expert,
> and this only on good days.  This is a form of a Turing test, ja?

I strongly suspect that only a grandmaster would have much
chance telling human grandmaster play from machine play.
Even then, I suppose that it would help to have specialized
in the study of computer played games.  It might be boring;
for example, it might turn out that the best way was to watch
how the program handled the endgame.

But that is a good question!  I wonder if anyone has a collection
of "computer-program combinations".  One or two I've seen
definitely have an inhuman quality to them.  They start with 
extremely unlikely looking moves, moves that any good player
would never investigate (because it was so improbable that
anything lay in them).  But a program often just looks at all the
possibilities, and so discovers those outrageous things.

Lee