From newshub.ccs.yorku.ca!torn!utcsri!rpi!uwm.edu!ogicse!psgrain!percy!nosun!techbook!szabo Thu Jul  9 16:20:10 EDT 1992
Article 6391 of comp.ai.philosophy:
Xref: newshub.ccs.yorku.ca rec.arts.sf.science:2635 comp.ai.philosophy:6391
Path: newshub.ccs.yorku.ca!torn!utcsri!rpi!uwm.edu!ogicse!psgrain!percy!nosun!techbook!szabo
>From: szabo@techbook.com (Nick Szabo)
Newsgroups: rec.arts.sf.science,comp.ai.philosophy
Subject: Re: Putting the pratice effect to work
Summary: Computer simulation of practice effect universe?
Message-ID: <1992Jun29.205441.14071@techbook.com>
Date: 29 Jun 92 20:54:41 GMT
Article-I.D.: techbook.1992Jun29.205441.14071
References: <1992Jun29.060742.4765@ccu1.aukuni.ac.nz>
Organization: TECHbooks --- Public Access UNIX --- (503) 644-8135
Lines: 37

In article <1992Jun29.060742.4765@ccu1.aukuni.ac.nz> ecmtwhk@ccu1.aukuni.ac.nz (Thomas Koenig) writes:

>[Dave Brin, _The Practice Effect_]

>For those which have not read the book:  this effect means that things
>get better with use (changing materials, shape etc.), but cannot
>depart from the original principle of operation; for example, a glider
>will never develop an engine.  It also takes a lot of man - hours
>both to make things better, and to keep the effect up; things which
>are not 'practiced' revert to their original state.

I've pondered on how we could implement a subset of this effect in
our universe.  Imbedded neural nets might give a pretty good practice 
effect for machine behaviors.  Even better, we might implement a 
a practice-effect unverse in software.  It would contain extensive
knowledge of design principles, and the ability to simulate most physical 
processes.  The basic structure is a genetic algorithm, where designs
are replicated or rejected based on judgement of its performance in the
simulation.  The genetic code specifies the design being evolved: 
materials, shapes, etc.  The human operator would judge "good" and 
"bad" designs as they are simulated, a la Richard Dawkin's "Blind 
Watchmaker" program.  Most of the judgements should be made by the 
simulation itself, however.  For example, it could throw out glider 
designs that crash too often. 

>look like with this strange reversal of usual physixcs.

I don't see how mere "reversal" would arrive at such laws.
There is a large dose of teleology here -- the objects seem to "know" 
what constitutes "better", and have some way of altering their very 
atomic structures to fit it.  Does Brin go into any detail about how 
he has altered our universe's physical laws to arrive at this effect?


-- 
szabo@techbook.COM  Public Access User --- Not affiliated with TECHbooks
Public Access UNIX and Internet at (503) 644-8135 (1200/2400, N81) 


