Author: Robert Hyatt
Date: 08:20:15 01/15/00
Go up one level in this thread
On January 15, 2000 at 02:17:35, James Robertson wrote: >On January 15, 2000 at 00:58:50, Robert Hyatt wrote: > >>On January 14, 2000 at 22:32:40, James Robertson wrote: >> >>>On January 14, 2000 at 19:53:34, Dann Corbit wrote: >>> >>>>On January 14, 2000 at 09:31:43, Robert Hyatt wrote: >>>>>On January 14, 2000 at 02:20:09, James Robertson wrote: >>>>[snip] >>>>>>I find it impossible to believe that he is the only human who is capable of >>>>>>doing this. >>>>> >>>>>Think about this: How long did it take him to become able to do this? At >>>>>least 12 years. How long would it take someone _new_ to catch up to his >>>>>skills at this? About 12 years. >>>> >>>>I think "anyone could have done it" is a bit of a reach. >>> >>>Who's quote is this? It certainly isn't mine. >>> >>>>There are two reasons >>>>that I say this. First, nobody else *did* it with the success that he had. If >>>>they could have, why didn't they? >>> >>>$$$ >>> >> >>That is bogus. Chiptest cost CMU practically nothing. It was built from >>spare parts in the robotics lab. Deep Thought ended up costing about 5,000 >>dollars roughly. >> >>Ed built a special board for a PC, "chessmachine". Hsu just took it to a >>whole new level. And the entire plan was in place before they left CMU for >>IBM. IBM hired the entire Deep Thought team, and bought into the 'plan' before >>hiring them. >> >>money wasn't all this required to happen. > >How much did DB cost? I don't care about chiptest or DT because they weren't >funded by IBM. > How do you want to compute the cost? Salaries for Hsu/Campbell/Hoane/Tan/etc? Count the SP or not since it was sold/shipped after the match? Or just count the fab cost for the chips? However, deep thought was the _root_ of deep blue and it didn't cost a significant fortune. 5 grand or so according to Hsu's book. >> >> >> >> >>>> >>>>The second reason is that 'could have' does not connect with 'would have.' >>>>For instance, Bill Pugh invented the skiplist in the 1980's. Hoare invented >>>>quicksort in the 1960's. The concepts are simple. Many people "could have" >>>>invented those things. But they didn't. Why not? >>>> >>>>* Lacked effort >>>>* Lacked interest >>>>* Lacked time >>>>* Lacked money >>>>* Lacked <whatever> >>>> >>>>If the Laser and Teflon had not been invented by their inventors would we have >>>>found them by now? Probably, but maybe not. In any case the proof of the >>>>pudding is in the eating. Hsu's pudding was the best tasting ever, unless you >>>>happen to be Kasparov. Apparently, he's not in any big rush for a second >>>>helping. >>>>;-) >>>> >>>> >>>>[snip] >>> >>>This is the wrong way to look at it. There is _one_ way to make Teflon. There >>>are unlimited possibilities for a chess computer. Just because Hsu invented one >>>way does not mean there are not many many other ways. >>> >>>James >> >> >>Who cares there. There is one way that certainly works, by demonstration. I >>don't want to wait for a roomful of monkeys to produce an alternative approach >>that also works. Alpha/beta has stood the test of time. DB was built around >>what works... > >You are missing the point. DB was based on alpha/beta, which doesn't require a >Hsu to make work. So therefore, if there is something that Hsu could do that >other people _couldn't_, it wasn't alpha/beta. It was the certain designs of the >chip, the exact knowledge he implemented, etc. And there are _many_ other >designs of chips that will work, and many different pieces of knotwledge that >will work too. > You miss the point that (a) yes it requires alpha/beta; but (b) only two people in the history of computer chess have built a machine that does the alpha/beta search in hardware. Thompson and Hsu. There aren't "so many other designs of the chip..." any more than there are different ways to implement the alpha/beta search. You can use different instructions, but the algorithm is _always_ the same. Someone would have had to follow in Hsu's path to reach the same point. That path wasn't easy to follow. Nor was it impossible. But it did take 12 years to traverse. I doubt anyone would have cut it shorter by any significant amount of time. >Which is the point of my response to Dann. Dann implied that writing DB was >something that took a special person to do, and compared it to the invention of >teflon. Maybe so, but there is _one_ way to make teflon. There is _not_ one way >to make a chess program. If someone sets to work, and they _don't_ invent DB, >there are many ways that will work just as well that are different... To answer >both of you: Fritz and Hiarcs both use alpha/beta, but the rest of their search >differs completely. But they are _still_ very similar in strength. > >James The designs are identical. Alpha/beta with endpoint evaluation, plus search extensions. The basic frameworks are identical. The details are different, depending on what the programmer wanted to accomplish. At the hardware level, the problems are the same. The difference is that for software we have seen hundreds of chess programs over the years. You have source you can look at. For hardware, we have had exactly 2 machines. You essentially start from scratch there. Big difference.
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.