Re: The Future of Chess and Getting It "Unstuck" (potential heresy

On Apr 5, 7:14 am, richardhut...@[EMAIL PROTECTED]
 wrote:

> What will be interesting is if you start doing games where players can
> fundamentally alter the game rules to hinder their opponent, and then
> challenge the AI to do that.

Do you mean inside the game? If so, that's just another rule for the
game. You can define the movement of the chess pieces in terms of rule
changes, if you really want to. I don't understand why you care about
this.

Or do you mean outside the game? Then it's politics, and using
politics to redefine the game as to "hinder your opponent" is just an
ugly metagame I'm not interested in playing - or writing an AI for.

But for AI for a game with "rules that change" (or more properly,
there is one position in the game which has an infinite number of
possible moves), that can be done. MC programs in principle need only
a description of the game in order to play it, these descriptions can
be arbitrarily large.

I recently learned that Cameron Browne (game designer and author of
"Hex Strategy" and "Connection Games") has already done this,
developed a language to describe games and and MC/UCT player that can
play them well. He used it to evolve the game of Yavalath.

There are games which are hard for MC prgrams. There are games which
are easy for MC programs, but which are even easier for humans, like
Go. But this proves nothing. I'm sure it would be easy to evolve an
anti-human game with Browne's system. Indeed, I suspect the computer
will play better on most large randomly designed rulesets.