Right.  I thought I'd ask here before bothering Leonid Broukhis since
people often have ideas about how to exploit weaknesses in contest
rules that they are willing to share.  I'm sure his experience will
turn out to be invaluable.

I ran across your poster session paper  "Text Compression as a Test for
Artificial Intelligence" which is one of the things that inspired me to
look into the prize award approach.  My original motivation is my
rather strong sense that Ockham's Razor is more than a nice rule of
thumb for scientists.

Your proof is elegant, simple and convincing however people still don't
"get it".  When people start arguing about intelligence tests they can
get pretty squirrley and I think the Turing test is a poison that has
impeded thought about artificial intelligence more than it has advanced
it.  Your paper is crucial to overcoming that handicap.

However, since the 1960s there have been other arguments advanced,
primarily by the algorithmic complexity guys, that should have taken
root in the AI community but didn't.  I don't understand exactly why
that intellectual failure occurred in the AI community but it most
obviously did.

As for the text corpus I would choose:

That was one of the main questions I wanted to discuss because it is so
crucial to the success of such a prize.  Right now if I had to choose a
single text corpus I'd pick the 1 terabyte corpus that Peter Turney
used in his recent accomplishment of human level performance on the SAT
verbal analogies test -- a feat that is particularly interesting given
the exceptionally high correlation between that test and general
intelligence aka the 'g' factor.

In this regard, you may be interested in my recently published article:

"AI Breakthrough or the Mismeasure of Machine?"

http://www.kuro5hin.org/story/2005/5/26/192639/466

I'm thinking of writing another article on the C-Prize.