Author: Vincent Diepeveen
Date: 13:05:30 12/25/03
Go up one level in this thread
On December 25, 2003 at 13:25:33, Peter Hegger wrote: >On December 25, 2003 at 13:12:10, Bob Durrett wrote: > >> >>Suppose you were much richer than Bill Gates and wanted to make the biggest, >>meanest, most fantastic chess computer. Suppose you wanted a dedicated chess >>computer so big that it topped the list of the World's biggest computers. What >>would it look like? >> >>The problem IBM had with Deep Blue was that they didn't want to spend BIG money >>on the computer. But you could be different! Be extravagant! >> >>Special purpose computers are optimized for a very limited purpose, by >>definition. For example, if it were desired to create a special purpose >>computer to model the Earth's atmosphere, globally, and if it were desired that >>the model be "high fidelity" and capable of modelling repid events such as >>tornatos and lightning, then the structure and operation of that computer and >>it's software would be optimized for that purpose and would never be used for >>any other purpose. It, I suspect, would not look very much like a >>general-purpose computer. >> >>What if you were **rich** and wanted to make a really huge chess research >>machine that would blow everybody's imagination? >> >>This is a call for brilliant ideas! [If there is no response, you know what >>that'll mean. : ) ] >> >>Bob D. > > >Hi, >I've thought about a scenario where hostile aliens land and challenge earth to a >chess match for the the very existence of the planet. If earth wins or draws we >get to live and they'll quit bothering us. If we lose they'll blow the planet to >smithereens. They'll be back in exactly 5 years to play the 10 game match. >With all of mankind's resources thrown into their very survival, what kind of >chess computer could we come up with for such a match? >Regards, >Peter Oh well, after we realize that supercomputers continuesly crash and no one will make software within 5 years that will be faster than a 8-16 cpu machine will be in a year or 5 (of course all NUMA and of course none of them will run Crafty), thanks to faster processors but not faster latencies where the 8 processor machines still work great thanks to big shared L3 caches. Additionally when we realize that developing new technology for low random latency routers won't be made in 5 years of time from now, as those things just are currently optimized to perform well in certain testsets and bandwidth tests which never assume software that just needs low latency for small and very short messages ( <= 64 bytes whereas they are optimized for tens of kilobytes at a time if not more), soon we realize that it will be quite a risk using some supercomputer to save earth. So i would go for a cheap solution here to save earth. First of all i would ship Friedel to Mars to sell them deepfritz12. By then probably working in assembly for x86-64 too, the Marsmen can show up at most with some 17 ply searching quad opteron (note fritz searched also 17 ply in middlegame in 1999 at a quad Xeon 500Mhz too, so that shows again that the most important progress last few years was evaluation related, not search related). If i given enough budget i will buy myself a nice appartment in Southern of France, because i'll solve this issue cheap anyway. A location close to the games is not needed, as the games in 2008 won't be very interesting as it will be a one sided show. See below. So from what i have left after buying that apartment i'll then try to hire, the worlds number 10 player, Kasparov for $150k and tell him he just gets the money when he draws or wins the match and otherwise nothing. Of course i prefer Kasparov as he is the best entertainer. Cloaked by very good openings play and a good commenting team at television, people will fail to miss anyway that after the opening, Kasparov is not even close to the strongest player on earth. In fact he's known for not even capable to win against 2600 rated players and for getting crushed by players like Ivan Sokolov when they get that one time opportunity to play Kasparov. Basically Kasparov is good in outpreparing those few predictable opponents that he plays in all those tournaments. Nevertheless publicity is more important than quality of the match, as the quality is already garantueed not highest thanks to the software part of it. In the mean time David Levy is there to explain the top 9 players why they weren't approached as first to play the Mars software, despite having won multiple world titles in a row. David Levy, just like at the meeting in the world champs computerchess 2003 will explain that a world title is not enough to garantuee anything as there are 'other conditions that needs to be fulfilled'. In this case not money but publicity. Of course Kasparov will not decline $150k price money, but also ask an advance payment of $1 million for such a publicity match, as he already got $100k a day back in 1995, so $150k is not near enough to play DF12. That he can save earth with it, is of course no issue for Kasparov the business man, as any world top GM can perform this job of beating a program and he can earn more than $150k during that match when playing elsewhere, even when it is just a few league games. Friedel is not there to arrange a sponsor to pay for that $1 million, as he is still at Mars selling Fritz there. Of course i'll look in my pocket and tell Kasparov that i don't have that 1 million dollar anymore as i bought that nice apartment from it, which is more important to me than that match of course. So i'll then ask the strongest player by then multiple world champion Ponomariov. He will also demand $1 million, because Kasparov also asked that, and Ponomariov is no worse than Kasparov, in fact even stronger, so also needs $1 million. Knowing that Kramnik and the young Chinese 2800+ GM and the whole protected world top that play already play some 'top event', which prevents them from losing rating to all those dangerous players just under them (which makes the world champs FIDE 2004 with 128 participants very interesting of course), i clearly must move to someone who doesn't get all those nice invitations and who is ready to proof himself. So i will of course move to Ivan Sokolov, who by then still will have around 2670-2695 FIDE and not 2700 (as he hardly gets invitations for top events and the bug in the rating system is known to all titled players; you can way easier win rating or remain at your own rating by playing higher rated players). I'll explain to him he can earn $10k with the match and that it will give him perhaps some publicity which might help getting other big invitations. Also i'll explain he will only earn that $10k by not losing the match and that he doesn't get paid to lose in 19 moves by playing some opening he doesn't know (see his 1999 game). Also i'll promise him a lucrative 2 matches against DIEP the next year when he wins this match against Fritz, so he just has to show some teeth now and can relax next year in the sure know he gets another few matches after that. To avoid some disasters i will appoint some correspondence players who know all about beating computerprograms and especially know how to install other engines in chessbase9 than fritz5.32 which is the engine that the majority of Grandmasters still use in 2008 to analyze with. When Ivan argues that he doesn't need to prepare any opening very well against the computer as he still knows some sneaky gambit lines which he'll prepared a few years ago and are useless in human tournaments anyway so can be tried without problems against the computer as it won't betray his tournament preparement for the real important games and still form a considerable problem to prepare against i again need to phone for a long period of time with Ivan. I'll politely then ask him to listen to his secondants who are used to computerprograms and for sure not to read the nonsense at the internet forums. Sokolov of course wants to show then he belongs to the world top and will crush the thing hands down. Especially the games Sokolov has white will be short. With black he stubbornly still will play openings he never played before with strategies and opening pitfalls he never saw before so it will become 5-5. So for $10k the world is saved. O yes and a telephone bill of $10k, those GM's are all living at Mars so impossible to negotiate with in a normal business manner.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.