Newsgroups: comp.ai.alife,comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!torn!watserv2.uwaterloo.ca!watdragon.uwaterloo.ca!bpvanstr
From: bpvanstr@yoho.uwaterloo.ca (Brian Van Straalen)
Subject: Re: [Q] Robotic 'ants' for Pest control...
Message-ID: <Cz2641.6xE@watdragon.uwaterloo.ca>
Sender: news@watdragon.uwaterloo.ca (USENET News System)
Nntp-Posting-Host: yoho.uwaterloo.ca
Organization: University of Waterloo
References: <39aof0$5ds@todd-06.cs.strath.ac.uk> <39gu8mINNf89@life.ai.mit.edu> <39l68m$654@todd-06.cs.strath.ac.uk> <39m9dvINN6de@life.ai.mit.edu>
Date: Thu, 10 Nov 1994 15:38:23 GMT
Lines: 37
Xref: glinda.oz.cs.cmu.edu comp.ai.alife:1268 comp.ai.genetic:4225

In article <39m9dvINN6de@life.ai.mit.edu>,
Marvin Minsky <minsky@transit.ai.mit.edu> wrote:

  [deletia]

>
>As for the COG project of Brooks et al., I disagree with them about
>the value of building the real-time hardware machinery.  My personal
>opinion is that more would be learned by making simulated robots
>operate in relatively simplified simulated worlds.  They object that
>if you do this, you might overlook serious real world problems.  I
>don't agree: in my view, it is not important precisely which kinds of
>noise , you encounter -- or otherwise unpredictable friction effects,
>etc.  You'll run into the same basic cognotive problems whatever you
>do, so you might as well introduce cheap, easy to compute types of
>variation.  I'm not winning that argument, though.
>
>So I don't agree that important AI research requires big hardware
>resources.  There is still the problem of salaries, though.  

 Ah, big EGO, not big hardware :)

seriously.  Wasn't there a student of prof. Stein's that did a paper on
this topic ? Despite having a bachelors degree in mechanical engineering
and liking hardware, the paper had some good arguments for simulation.

Simulation allows for `controlled complexity' (my own term) . How a user
/researcher manipulates their `control' has a strong bearing on the
type of results and insights generated.  Brooks' approach takes a lot
of the control out of the researcher's subjective hands.

There's also the point that Brooks' Subsumption Architecture approach
doesn't lend itself to current simulation techniques.


Brian Van Straalen
University of Waterloo
