Vladimir Kramnik, who otherwise played an excellent and aggressive game against Deep Fritz today, blundered in Game 2 at move 34 by not seeing a very obvious mate in one.
Kramnik played black and obliged Fritz with a Queen's Gambit Accepted. He forced much of the play early on, with Deep Fritz eventually rebounding to equalize the board. Had Kramnik seen Fritz's threat at move 34 the game would have most likely ended in a draw.
In the diagram to the right, Kramnik moves his queen to E3 threatening a mate of his own and assumes an exchange of queens. At this time he's completely oblivious to the threat in the far right corner, as Fritz's next move is queen to H7: mate.
Interestingly, even one commentator missed the mate-in-one threat. After making his ill-fated move, Kramnik was about to stand-up and take a break when Fritz quickly calculated the checkmate. Kramnik put his hand to his forehead in shocked disbelief.
Blunders happen, even among the grandmasters.
But not computers.
lol@ human error.
So AIs can consistently play even with the best chess players in the world. Now all I need is an AI smart enough to check my groceries out without constantly pestering me to re-scan stuff that I've already scanned. :)
False positives are a bitch.
BTW, has anybody else noticed that chess is, in a sense, like glorified tic-tac-toe? Two competent players can play to an endless series of draws in tic-tac-toe. You have to go out of your way to screw up in order to lose.
Of course, chess is far more complex. While there are precisely 765 unique positions in tic-tac-toe, there are ~10^118 game trees for chess (http://en.wikipedia.org/wiki/Game_complexity http://en.wikipedia.org/wiki/Shannon_number). That means there are way more possibilities for error, which is why we "normal" players will eventually screw up leading to one clear winner.
However, two exceptional players can essentially steer the course of the game to avoid most errors so that a large number of draws result. Anyway, in that sense, it's like glorified tic-tac-toe.
BTW, does that observation imply that we are reaching the theoretical limits of chess expertise? IE, if all (or all major) errors can be avoided, then the best you can hope for between two superlative players is an infinite series of ties. Maybe chess ratings don't just scale up forever. Maybe somewhere around 3000 is the theoretical limit, because two players with a rating of 3000 would never or extremely rarely screw up in a way that could lead to a clear winner.
So maybe there won't ever be a day when machines destroy even the best humans in chess every time.
Just a thought.
Domin8 MySpace with unique myspace music. Check out www.domin8myspace.com
Domin8 MySpace with myspace proxies.
Post a Comment