Author: Robert Hyatt
Date: 14:19:58 10/14/99
Go up one level in this thread
On October 14, 1999 at 12:03:50, Shep wrote: >On October 14, 1999 at 11:12:39, Joshua Lee wrote: > >>I know that even if you had a 2Ghz machine it wouldn't solve Nolot pos 1 in >>under 3 Min. In fact 10 complete in under 3 min you would have to have a 35Ghz >>Machine. so the only solution seeing how this isn't going to happen for hmm >>maybe 9 years you would think Moore's Law. So how can you make programs faster >>annd more intellegent? > >I don't think a program has to solve all Nolot positions (or any other test >suite) within tournament time control to have GM strength. >I bet several GM's would not solve all of them (provided they didn't know them >already which they do...) either. >[Remember the notorious Qe3 which both Kasparov and Deeper Blue missed, yet most >PC programs find it.] Deep Blue didn't 'miss it'. It was expecting it. It just didn't see that it led to a forced draw. However, _no_ micro has any clue that it leads to anything at all, much less a forced draw. It is a tough position where the repetition is some 60 plies into the future. On a different note, test suites are not a good way to compare humans to computers. Crafty, for example, just tries to find the 'best move' in any position. Where a human would say "hmm... test position... must have a strange tactical move as the key so I am going to look for tactics and not worry about deep strategy." So the computer and the human are solving such positions in totally different ways, unless you take a program like Paradise and use it, since it is only a problem-solver and not a game player... > >In fact, if a program found just 3 or 4 of them in 10 minutes, it would probably >be at par with many GM's (unless it was tuned for the suite). I agree. I think any of the following are possible: 1. A GM doesn't find any of the right moves. 2. he finds some of them. 3. he is unlikely to find all of them. 4. a gm-level computer (when they exist) doesn't find any of the right moves. 5. it finds some of them. 6. it might even find all of them. > >--- >Shep
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.