Author: Dann Corbit
Date: 14:51:48 02/17/04
Go up one level in this thread
On February 17, 2004 at 10:36:28, Janosch Zwerensky wrote: > >>>but there is Rule 2: The Computer-SpeedUp-Curve has a saturation. >>>This 1.5- factor will be decreasing with years and years. >> >>I think the opposite is true. It used to be that every 1.5 years the speed >>doubled. Now it has shrunk to every 1.0 years. I have seen a study that shows >>that compute speed has been growing superexponentially for the last couple >>centuries. > >Well, no exponential is going to last forever, as they say.... :) That's OK, because forever never gets here, no matter how long we wait. >At any rate, a hundred years might be about right after all. As far as I know, >by the mid-nineties, which is ten years ago, the top programs were said to be >around 10-kyu, which would be two or possibly three stones weaker than they are >now. Given that the top programs of today are around 20 stones weaker than the >strongest humans..... > >>(...) >>I expect that they might be 2^150 faster. > >Are you sure that such a thing would even be compatible with known laws of >physics? Possibly not. But then again, we may simply discover things that are not currently known. For instance, nuclear fusion was not possible according to the known laws of physics in 1900. >Aside from that, though I'd tend to agree with you that likely lack of >suitable hardware is not going to be a problem for computer go researchers (as >might well be the case today), one should notice that there's no such thing as a >Moore's law for software development... Are you sure? Since information is expanding in a superexponential manner over time, I suspect that algorithm advancements are also. Also, as capabilities expand, the techniques we can use expand with it. For instance, if the level 0 cache on your chip is 100 Terabytes, you can design software differently than if it is 256K.
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.