kjm@UTEXAS-11@sri-unix (10/21/82)
How many computer programs do you know of or have written in which one wrong instruction gives some inherent slight "advantage"? It seems to me that program mutation is more a fast way to get inert or malfunctioning pieces of junk rather than better robots. Ken Montgomery -------
REM@MIT-MC@sri-unix (10/22/82)
From: Robert Elton Maas <REM at MIT-MC> Date: 21 Oct 1982 at 1009-CDT From: kjm@UTEXAS-11 How many computer programs do you know of or have written in which one wrong instruction gives some inherent slight "advantage"? It seems to me that program mutation is more a fast way to get inert or malfunctioning pieces of junk rather than better robots. The first time I read this I took it to be a rhetorical question, as it was probably intended, but upon second reading I happened to remember an actual case to point. At SU-AI about ten years ago we were actively writing display hacks for the III (Information International Incorporated) display processor. Somebody wrote a program that was supposed to draw random stars (points of light that apeared in random positions), and then wink out later to be replaced by new random stars (there was a queue of stars that initially grew until it was full, then stars were recycled to new points in strict revolving order). The first time he ran the program it had a bug, one instruction was wrong, instead of drawing random points it drew random rays from the center of the screen. It was a wonderful accident, much better than the original idea. I'm sure if people in this community think hard they can think of other examples, though rare, of single wrong instructions being improvements over the original conception. This would seem to be the correct forum since we have a reason for considering this question (who would have believed it?).