The idea is that a human, contrary to the computer, has the idea of "worthless" moves and a special preselection. So a grandmaster only considers the "good" moves (of which there are maybe 5 or 10), while the computer has to scan through all possible moves and rate them to even find out what "good" moves are. Vast libraries for the opening help the computer reduce the initial tree of possible moves down to a set of "traditional" moves.

Another side point might make this phenomenon more plausible to you :
If you show chessboards with regular play situations on them for a short time to people, those who play chess will more accurately reconstruct the board than those who don't. But if you show them pieces randomly placed on a board, both groups will show a similar rate of errors. This could be interpreted that the human does not see the places of all pieces on the board separately but in relation to the other figures, remembering patterns he has seen/experienced (in play) before.

A game that solely relies on patterns and their recognition is Go, a game that is considered very hard to write computer opponents for.

perl -MHTTP::Daemon -MHTTP::Response -MLWP::Simple -e ' ;    # The  
$d = new HTTP::Daemon and fork and getprint $d->url and exit;#spider
($c = $d->accept())->get_request(); $c->send_response( new   #in the
HTTP::Response(200,$_,$_,qq(Just another Perl hacker\n))); ' #  web
[download]