Author: Dann Corbit
Date: 11:39:24 01/13/05
Go up one level in this thread
On January 13, 2005 at 14:31:10, Anthony Cozzie wrote: >On January 13, 2005 at 13:41:12, Dann Corbit wrote: > >>I want to say something clearly. >> >>I cannot possibly prove that chess will be solved even one billion years from >>now. >> >>I cannot possibly prove that chess compute power will double every year. >> >>I cannot possibly prove that memory storage techology will double space every >>year. >> >>We have seen in the past that this is a trend, and the trend has continued for >>150 years or so. >> >>But this is an argument from induction and not deduction. It is like a man who >>sees a white duck every morning for 6 years. He looks out his window and sees >>flocks and flocks of white ducks. And so he concludes (sensibly) that ducks are >>white. But induction can never prove a thing, because it is only a projection >>based on past experience and not any sort of a proof. Someday, he may see a >>mallard duck and it will spoil his whole theory. >> >>Therefore, anyone else's opinion is just as good as mine about it. >> >>And any sort of projection of the future is always a guess. For instance, I may >>see that my shoe is untied and decide to tie it. What is the probability that >>my shoe will get tied? Before the action occurs, it is something less than one. >> >>There may be an asteroid that strikes at my lat/lon just as I bend over to tie >>it. And so all projections of the future (which are clearly made with imperfect >>knowledge) are only educated guesses at best. >> >>In a sense, it is like arguing over how many angels can dance on the head of a >>pin. > >Moore's Law is _dead_. I don't think this is in question any more. Even at >.13u we saw considerable gains, but EVERYONE is having problems with 90nm: AMD, >IBM, Intel, TSMC . . the list goes on. The power requirements are just too >high. In terms of MHz, you can buy a 2.4GHz Opteron today. Last year you could >buy a 2.0GHz Opteron. We are just not going to see the performance gains from >processors that we saw in past. There are some chips near 4GHz. Intel has more experience at high clock rates than AMD and some of the other competitors. I remember a prediction in this forum a long time ago that said 1GHz was not going to be doable, because of trace creep, emanations from the chip, and other related problems. >Now, this doesn't mean that we won't see other techniques. There are a lot of >interesting things on the horizon: IRAM, multiple cores, reconfigurable chips, >VPGA, etc. But none of these are going to provide the speed increases we have >seen in the past. > >Of course, there are also completely non-silicon possibilities as well. Quantum >computing. Biological computing. None of that stuff is going to be on my desk >in the next 10 years, though. Kurzweil has some sound arguments: the more >people on the planet, the more cool ideas we'll come up with. But we have just >about maxxed out the CMOS superscalar processor. The Vacuum tube was used to construct the first computers. When they reached the point where you could not build a better or faster computer, someone invented the transistor. When discreet transistors no longer did the job, someone invented the integrated circuit. When the integrated circuit can no longer be improved, someone will invent something better. And we won't have long to wait to see the improvement on our desk. Squids can do petaflop switch rates. They require superconductors, but perhaps a room temperature superconductor can be found. Perhaps integrated circuits can be grown in three dimentional configurations like tubes or some other interesting shape. For every problem, there is a solution that can be found. With knowledge doubling at a faster and faster rate, it is inevitable that the solutions will be found.
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.