Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!rutgers!sun-barr!apple!versatc!mips!prls!philabs!linus!mbunix!bwk
From: bwk@mbunix.mitre.org (Barry W. Kort)
Newsgroups: comp.ai
Subject: Re: Free will and responsibility.
Summary: Driving under the influence
Keywords: Observer, Feedback Loop, Values, Goals, Dead Reckoning
Message-ID: <53228@linus.UUCP>
Date: 13 May 89 12:47:07 GMT
References: <10333@ihlpb.ATT.COM> <3850@uhccux.uhcc.hawaii.edu> <387@nbires.nbi.com> <17754@cup.portal.com> <52019@linus.UUCP> <881@maestro.htsa.aha.nl>
Sender: news@linus.UUCP
Reply-To: bwk@mbunix (Barry Kort)
Organization: The Gallimaufrey, Atsea, UK
Lines: 33

In article <881@maestro.htsa.aha.nl> fransvo@htsa.UUCP
(Frans van Otten) writes:

 > Barry Kort writes:
  
 > > [controlling a system requires...] that the output of the system
 > > must be observable. ...  Try driving a car blindfolded.  If you
 > > cannot see the output of the system, you cannot control it.
  
 > In general, I agree with this.  But if you have a very accurate map,
 > and if there is no one else around, you can drive anywhere you want,
 > even if you can't see the world in which you are driving, or feel the
 > motions of the car.  You don't need to see or feel anything: you can
 > calculate exactly what will happen when you perform a controlling
 > action, like turning the steering wheel.  Or is the driving not
 > called "controlling" in this case ?

I claim that after a while, you will no longer have an accurate
estimate of your actual location on the terrain.  One problem with 
dead reckoning is the precision of your calculations.  You need
infinite precision arithmetic.  Any errors inexorably accumulate
according to a random-walk model.  The statistical variance of the
error grows linearly with time.

Another problem is with the quantum nature of your travels.  No
theoretical model can predict below the precision of quantum
indeterminacy.  

The only known solution to these limitations is timely observation
out the windshield.  We once tried to send a rocket to the moon
without mid-course corrections.  It missed.

--Barry Kort