Saturday, May 27, 2006

Ermacora: No Rest For Chips

Today 26 May is the first rest day at the Olympiad. However, there is no rest for computers. The World Chess Computer Championship is still on today (as are all the Open Tournaments). Zappa is the favourite computer in the absence of Hydra. Others are Junior, Shredder, Fritz, Rebel, all ex world champions.

The question was raised by an Italian programmer whether it is fair when testing software to have differences, sometimes very great, in the hardware used. Zappa, backed by commercial sponsors, has 512 processors which examine 100 million positions in a flash while poorer programmers have the backing of much more modest machines. There are other tournaments where the chess programs all have to have the same hardware. That seems fairer.

Well, those who defend the way the world championship is run say that such differences don't matter: ultimately it is how good the software is, i.e. the chess knowledge programmed into it. The example is given of the Formula 1 car races and the relative importance of the driver (software) and the car (hardware). You can have a more powerful car but if the driver is no good...

They make a similar analogy with the America's Cup. Should all software be tested using the same hardware? No, powerful hardware doesn't win over smart software they say. Do you find that convincing?

I went to a lecture and discussion panel on computer chess today at the university of Turin, IT department. On stage were Cianciarini, Levy, Schaeffer and Van den Herik. Is there less interest in computer chess (artificial intelligence)? The former International Computer Chess Association is now the International Computer Games Association, having accepted a few other games also (bridge, go, draughts, etc).

Yes, while there is still a quest for the best chess computer, the academic interest seems to have waned. Commercially there is a lot more money in the pop computer games young people play nowadays and lucrative prospects in that market for chess programmers. Levy told me that there could be a revival in future in human vs computer challenges with the computer giving odds, e.g. a pawn less vs the human World Champion (with all his pawns and perhaps a $1M purse inducement to overcome the possible humiliation).

In fact it was said that more and more, the play by computers resemble that of humans (games become hard to distinguish). But as computers will be playing better than human world champions, perhaps the time has come where humans should learn to play like computers!
And we still don't have computers that learn by themselves from their games. It will be even worse for humans then.

Ah, but it seems that an Australian researcher has concluded that humanity is also getting more intelligent. He did this apparently by a close study of movements in Elo ratings of top chess players over the years. He found that younger and younger players achieve very high Elos. So there is hope for humanity yet.

- Larry Ermacora

For the next couple of weeks, during the Olympiad, we will feature a guest blogger - Larry Ermacora. He will be providing reports all the way from Torino after having been officially accredited by Olympiad officials as a journo for this blog.

Larry is a well known personality in Australian chess circles. He was an administrator in both Western Australia and New South Wales. And, in 1990, he was awarded the highly coveted Koshnitsky medal.

No comments: