home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: talk.origins
- Path: sparky!uunet!destroyer!news.iastate.edu!pv7440.vincent.iastate.edu!btd
- From: btd@iastate.edu (Benjamin T Dehner)
- Subject: Re: Laying a trap
- Message-ID: <btd.722108058@pv7440.vincent.iastate.edu>
- Keywords: Computer program, random, mutation, chess
- Sender: news@news.iastate.edu (USENET News System)
- Organization: Iowa State University, Ames IA
- References: <1992Nov18.133247.8546@city.cs>
- Date: Wed, 18 Nov 1992 17:34:18 GMT
- Lines: 68
-
- In <1992Nov18.133247.8546@city.cs> lionel@cs.city.ac.uk (Lionel Tun) writes:
-
- >I would like to pose a couple of questions and gauge the
- >reactions of evolutionists (and creationists if any are
- >reading). As computer programs are flavour of the month..
-
- Um, first of all, the two scenarios you post below are not
- equivalent. I'll get to why in a second.
-
- >1
- >Lets say there is a computer program which `knows' the
- >legal moves of chess - lets call it ChessMover.
- >ChessMover plays very poor chess because its moves are
- >made at random. But it does play very fast. ChessMover
- >is small, compact and extremely efficient. But it plays
- >bad chess because it has not been designed with any
- >chess playing algorithms at all.
-
- >Would it be possible to subject ChessMover to random
- >mutations, so that eventually you evolve ChessPlayer,
- >a chess program which plays very well, say at master
- >level?
-
- If you simply do random mutations, no. There is no reason
- for it to get better. IF, however, if you have all mutated chess programs
- play against each other, and kill off all of the ones that lose a lot,
- the 'winning' algorithm will survive to perhaps mutate into a more winning
- algorithm.
-
- >2
- >For those of you who are not game fans, but more business
- >oriented:
- >Consider a spreadsheet program such as Lotus123 or
- >QuattroPro. Lets say you have a small calculator program,
- >like the toy ones which pop up in some windowing front
- >ends. Would it be possible to apply random mutations to
- >Calculator until it evolves into Spreadsheet?
-
- This is not equivalent to the above: what is the driving force
- behind the change? How do you kill off 'unsuccessful' spreadsheet
- programs? In the chess example, there was a simple rule: win or die.
- In this one, however, there is no rule of success, unless you already
- have some pre-existing model of a spreadsheet that you measure things
- to.
-
- >Please note, both Calculator and ChessMover are small (by
- >today's standard) programs, lets say about 2 to 4K for the
- >executable. Shreadsheet and ChessPlayer must obviously be
- >of a more substantial size (I can't remember how big Lotus
- >is). If you like, you can randomly modify the source code
- >and compiler, rather than the executables, if you think this
- >will make it less brittle. Random modification includes of
- >course random additions. You might prefer the source code
- >approach as the compiler will reject programs which will
- >not even compile, let alone run. If you wish to modify
- >the compiler, you can keep a copy of the old compiler to
- >compile it with.
-
- >--
- >Lionel Tun (lionel@cs.city.ac.uk)
- >Vision Group, City University, London, EC1V 0HB. 071-477 8000 x 3889
-
-
- -----------------------------------------------------------------------------
- Benjamin T. Dehner Dept. of Physics and Astronomy
- btd@iastate.edu Iowa State University
- Ames, IA 50011
-
-