November 20, 2003
HOW ABOUT SOME HEALTHY SPECIESISM?:
Man vs. Machine (Arnold Kling, 11/20/2003, Tech Central Station)
The results of the latest match between Gary Kasparov and the top computer X3D Fritz ended in a draw, vindicating Tyler Cowen and Jeff Sonas, who believe that computers have not yet overtaken the top human player. Sonas argues that "Although computers obviously must be improving in recent years, the strongest humans seem to also be improving at about the same rate."As a fan of Moore's Law, I am disappointed by the outcome of the latest chess match. As Cowen implies, the strength of the computers should be doubling each year or two. Yet they seem to be improving no faster than the best humans.
Cowen describes the following process by which computers beat humans at chess.
"The human grandmaster carries a significant advantage out of the opening or early middle game, where it is harder for the machine to calculate all relevant possibilities and positional judgment is at a premium. But as the game progresses, the machine plays perfect defense and the human cannot convert the advantage into a win."
The implication is that the computer only wins by wearing down the human opponent. In Cowen's view, the human is really the better player.
In fact, it is often the case in games between humans that the inferior player takes an early advantage and appears to be "worn down" by the better player. But that is the way games proceed between players of slightly uneven abilities.
How can you be disappointed when your own species retains superiority? Posted by Orrin Judd at November 20, 2003 8:31 PM
That's wrong, anyway. The speed of computers doubles about every year or two. The strength of computers at chess is presumably a function of their lookahead depth, which should be proportional to log(speed).
If you have to consider ten responses to each possible move, ten times the speed lets you look only 1 move further ahead, not ten times as far.
I understand Fritz sees about 10 moves ahead on high-end conventional hardware; he'll get another 2-3 a decade from Moore's law, and maybe another 5-6 from running on large custom hardware. It may be a while.
Posted by: mike earl at November 20, 2003 9:05 PMRight you are, Mike Earl!
The gain for computers vs. computers is proportional to the log of the increase in speed - roughly about 50 rating points per doubling. There is also some research showing a decreasing gain for further search depth.
For lots of reasons, the gain for computers vs. humans is _far_ less for each doubling. One thing Kling's article didn't address was tablebases. There now exist databases (called tablebases) for perfect play, all the way up to six men (including 2 kings). Trouble is, it occupies about 6 GB already, and each additional piece requires ~64 times as much space. If a phenomenally cheap way of having *lots* of memory becomes available, we could have 32-man tablebases and perfect play.
To address Orrin's point, Kling's AI boosterism was a bit creepy. He actually expects increasing computer speeds will immanatize the eschaton. I dunno about that...
Posted by: Bruce Cleaver at November 20, 2003 9:24 PMWhat's remarkable is that we're no longer talking about special purpose machines like Deep Blue, but just high end PCs. And Grand Masters are only getting draws against these programs. If you programmed the Fritz algorithm into the latest supercomputer you might have something that can beat even Kasparov consistently.
One day there'll be a war between humans and machines... and I, for one, will welcome our new electronic overlords!
Humans might someday make war on machines, but I don't think that the machines will start it.
Human/machine conflict makes for good SF, but really, where's the overlap in interest ? The machines don't want our land or women...
Posted by: Michael Herdegen at November 21, 2003 2:03 AMThey'd want our oil.
Posted by: M Ali Choudhury at November 21, 2003 5:07 AMThey'd want our oil.
Posted by: M Ali Choudhury at November 21, 2003 5:08 AMThey'd think us a virus.
Posted by: oj at November 21, 2003 7:53 AMAll;
Mr. Herdegen is right. Intelligent computers wouldn't fight with us, they'd leave us behind as they spread out in to the Universe. Would that be the meek inheriting the Earth?
As for oil, what use is that to machines? There's a gigantic nuclear reactor spewing free energy just 100 million kilometers away.
(1) Like I said before, this is rigged. Kasparov fifteen years ago woulda been better than any of these. Chessplayers lose their edge at 26. TS squared.
(2) Oh, and that last paragraph was bull. Having been the superior and the inferior player in thousands of matches, for the record: It is almost never the case that the inferior player looks to have a lead, only to be worn down. A case like that is called "a blunder by the superior player, compensated for over the course of the game." Sometimes, it is called a "gambit," but only someone who knows jack about chess would think the guy launching the gambit was in the inferior position.
(3) I'm all in favor of speciesism. The computers get too smart, we blow them up. They are tools, nothing more, nothing less.
Posted by: Chris at November 21, 2003 10:38 AMBruce -
Don't hold your breath on tablespaces. The space required expands exponentially with the number of pieces. I'm pretty sure you can demonstrate there isn't enough matter in the known universe to hold a 32-piece tablespace.
Posted by: Mike Earl at November 21, 2003 12:17 PM