Author: Graham Laight
Date: 01:39:58 05/04/01
Go up one level in this thread
On May 03, 2001 at 22:04:35, Dann Corbit wrote: >On May 03, 2001 at 19:09:10, Robert Raese wrote: >[snip] >>interesting... so you are saying that the technology explosion has really had no >>accelerating effect on chess programming? one would think that with >>signficantly faster processors, testing cycles at least would be accelerated... >>? > >Here's a funny statistic: > >Programmer productivity in the 1950's was measured in tens of lines of code per >hour. > >Programmer productivity in the 1960's was measured in tens of lines of code per >hour. > >Programmer productivity in the 1970's was measured in tens of lines of code per >hour. > >Programmer productivity in the 1980's was measured in tens of lines of code per >hour. > >Programmer productivity in the 1990's was measured in tens of lines of code per >hour. > >Programmer productivity today is measured in tens of lines of code per hour. This is a good point - but IMHO fails to take account of the fact that in each decade, 10 lines of code has become more powerful as languages have become more sophisticated, and features available to programmers have become more powerful. As a very quick example, Cobol has survived not just because there's so much legacy code, but also because modern (from the 1980s) Cobol is more productive. You can generate a well formatted report quickly and easily without all the arduous space counting, column formatting, trial and error, totalling, summarising, and so on that the 70s Cobol programmer had to do. The 1990s Cobol programmer even had visual tools for easy GUI programming. >The thing that drives the improvement of chess programs is two-fold: > >1. Enormously better algorithms. >2. Enormously faster machines. > >Neither of which help all that much. Chess is an exponential process. In >general, such problems are called "intractable" -- in other words, you can't >solve them. In fact, this is the case with chess. We can only approximate >solutions, which is usually good enough. > >Of the two factors above, the increase in hardware strength is far more >important. From 1950 to today, the computers have gained millions of times more >power. The algorithms are probably only thousands of times better. > >If someone can invent a polynomial time chess algorithm, then chess could be >solved. But I doubt if that will ever happen because chess is not a problem in >polynomial space. The "solving" of chess would mean that either every game ends in a draw, or that every game ends in a victory for white. (I personally think it's a draw, because the higher the ELO rating of the players, the higher the proportion of draws in their games). This might not be achieved mathematically, and it might not be achieved by game tree search - but there is a third way: it could possibly be achieved by generating sufficient chess knowledge. If you know enough about chess, you'd always obtain the best possible result. -g
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.