Author: Landon Rabern
Date: 09:47:35 06/05/01
Go up one level in this thread
On June 05, 2001 at 04:27:31, Graham Laight wrote: >On June 04, 2001 at 19:00:55, Landon Rabern wrote: > >>On June 04, 2001 at 18:48:40, Marc van Hal wrote: >> >>>On June 04, 2001 at 17:44:56, Christophe Theron wrote: >>> >>>>On June 04, 2001 at 16:03:36, John Hatcher wrote: >>>> >>>>>Here is a news report today from Reuters which may be of interest: >>>>> >>>>>********************************************************** >>>>> >>>>>(Reuters) >>>>> >>>>>Scottish University Sets Up First Chess Doctorate >>>>> >>>>>EDINBURGH, Scotland (Reuters) - A Scottish university is setting up the world's >>>>>first chess doctorate which its creator hopes will lead to the development of >>>>>supercomputers capable of beating even the greatest of grandmasters. >>>>> >>>>>``My computers will be as clever as 1,000 Einsteins,'' course director Peter >>>>>Vas, professor of artificial intelligence at Scotland's Aberdeen University, >>>>>told Reuters Monday. >>>>> >>>>>A keen chess player himself, Vas is looking for around 40 graduates for the >>>>>three-year PhD course, which also aims to push back the boundaries of artificial >>>>>intelligence, creating computers that can think and learn by themselves. >>>>> >>>>>He hopes former Russian world chess champion Garry Kasparov will become a >>>>>lecturer. >>>>> >>>>>Only the brainiest need bother applying -- prospective candidates must be highly >>>>>proficient at math and computing and be in the top flight of their national >>>>>chess rankings. >>>>> >>>>>``There will be a grandmaster entrance exam,'' Vas said, although he denied >>>>>reports that they had to beat the grandmaster in a game to get in. >>>>> >>>>>``Obviously we can't ask for them to beat a grandmaster because if it was >>>>>Kasparov playing 100 people simultaneously, he would still beat them all,'' Vas >>>>>said. ``Just showing the potential will be enough.'' >>>>> >>>>>Vas said the new supercomputers would have the combined intelligence of the >>>>>world's finest human minds. >>>>> >>>>>However, they would not live up to the apocalyptic fantasies of some Hollywood >>>>>film directors and get so smart they tried to destroy humanity. >>>>> >>>>>``An intelligent thing will always stop short of destroying itself,'' Vas said. >>>>>``There's no chance of that happening.'' >>>>> >>>>>Even Kasparov, who has met tough electronic challengers in IBM's 'Deep Blue' and >>>>>its bigger, better cousin 'Deeper Blue', stands to gain. >>>>> >>>>>``Playing something better than him will help him improve his game,'' Vas said. >>>>>******************* >>>>>END >>>> >>>> >>>> >>>>The tone and the content sound like an April 1st joke. >>>> >>>> >>>>"Only the brainiest need bother applying -- prospective candidates must be >>>>highly proficient at math and computing and be in the top flight of their >>>>national chess rankings." >>>> >>>> >>>>What a major mistake. Strong chess players are not the people of choice to >>>>create a strong chess computer. >>>> >>>> >>>> >>>> Christophe >>> >>> >>>Well then this is where your wrong human chess players are the only beings who >>>are capabale to see the positional moves better then the programs of today will >>>And i do not think that keeping the same course of programing ever will be able >>>to see the positional moves (a combining with neural networks shall be an >>>extremely improvement for positional moves and planning >>>Today many people still think that every opening is just playable I already >>>showed many times that this is not the case >>>Even many openingsbooks from programs contain these mistakes >>>Then again I have seen programs of today ruin winning positions >>>and even where glad to draw that winning position. >>> >>> >>>For the matter tactical moves are moves wich capture a piece,threaten to capture >>>a piece giving check ,threaten to give check mate,threaten to give mate. >>>giving pat threaten to give pat >>> >>>positional moves are moves wich are the moves when there is nothing to do and >>>enable a tactical move (Tarrasch). >> >>I have done some testing will a neural network evaluation in my program for my >>independent study. The biggest problem I ran into was the slowness of >>calculating all the sigmoids(I actually used tanh(NET)). It drastically cuts >>down the nps and gets spanked by my handcrafted eval. I got moderate results >>playing with set ply depths no set time controls, but that isn't saying much. >> >>Regards, >> >>Landon W. Rabern > >It is my personal experience (outside of chess) that when you have what looks >like an insoluble intelligence problem, you can effectively attack it by >combining different AI tools, rather than relying on a single one. > >For example, in chess, you might use CBR as the over-arching method, NNs to >classify positions, GAs to create evaluation functions within classifications >etc. > >-g I was not relying on a single AI tool, it was a NN eval function on top of a mini-max search. I have had some experience with GA's as well and the conclusion I have come to(which fits with many people in the field) is that they don't work well on all but the simplest toy problems. Regards, Landon W. Rabern
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.