Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Self-Healing Computers - :)

Author: Matt Taylor

Date: 12:57:05 12/17/02

Go up one level in this thread


Science fantasy is always enjoyable.

I'm not sure what the author means by "self-modifying code is mostly
experimental." People were writing self-modifying code in the 70's. (Anyone
remember Data General?) They were still doing it in the 80's. I worked on
un-self-modifying some self-modifying code that was written in the early 90's.

Just last night I was suggesting to a friend that he employ self-modifying
techniques to optimize a project of his once he has data. (e.g. don't do a
calculation that always results in a multiplier of 0 or 1, or an addend of 0).
It's not new. In fact, I have thrown around the idea many times with other
people, and I'm currently implementing such a project.

I must also raise an eyebrow at the thought of a natural selection algorithm
optimizer. The number of permutations to modify a single instruction are
immense. To change things and make them work right, you need a fair number of
modifications all at the same time. The probability sinks dramatically low as
you add more complexity.

The Biological debate has always raged over irreducibly complex systems. Since
it's the same question here, I would argue that code would be an irreducibly
complex system. You have a bunch of registers. When one changes, all the others
need to as well. Changing one instruction won't even produce code that works.

Of course, in simpler instruction sets it is not as necessarily complex, but the
difference between Biochemistry and Computer Science is that you can get similar
compounds by flipping genes in Biochemistry. Computers are designed so you
-can't-, and even if you could, you still wouldn't be doing anything useful in
most cases.

-Matt



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.