Newsgroups: comp.robotics
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!agate!darkstar.UCSC.EDU!nic.scruz.net!earth.armory.com!rstevew
From: rstevew@armory.com (Richard Steven Walz)
Subject: Re: What's a ROBOT? (was Re: consumer robotics)
Organization: The Armory
Date: Fri, 9 Dec 1994 11:43:21 GMT
Message-ID: <D0JKKB.7K6@armory.com>
References: <3bvfq4$ecm@ixnews1.ix.netcom.com> <1994Dec5.215916.10683@jarvis.cs.toronto.edu> <D0FB4q.7tC@armory.com> <1994Dec7.025246.12559@jarvis.cs.toronto.edu>
Sender: news@armory.com (Usenet News)
Nntp-Posting-Host: deepthought.armory.com
Lines: 95

In article <1994Dec7.025246.12559@jarvis.cs.toronto.edu>,
Roger Barry Hertz <rbh@wolverine.utias.utoronto.ca> wrote:
>In article <D0FB4q.7tC@armory.com>,
>Richard Steven Walz <rstevew@armory.com> wrote:
>
>>Actually, I think humans started using robots when they invented traps, and
>[snip]
>
>I don't buy that, sorry.  Most `devices' or `contraptions' by nature
>are designed to work without a human, that's what separates them
>from `tools'.  Simple devices that react to their environment are
>simply responding to an input, just like a clock would respond to
>the input (torque) of a wound spring.
>
>Don't get me wrong, your free to call a robot anything you like,
>but this definitely is the most general definition I've seen.
>It's almost a game of one-ups-manship, were people try to find
>`older' and `older' examples of `robots', simply be leaving out
>commonly understood restrictions of the term.  After a while,
>the term begins to loose its meaning all together, since almost
>everything in the history of man is now described by it. 
>
>>It is often assumed that man was a hunter,
>>but actually what we really excelled at was traps and enclosures to
>>facilitate the capture of animals and then the keeping of animals
>>(husbandry).  [snip]
>
>I think we agree that people have been using devices for quite
>a while, but disagree exactly what to call them.  What's wrong
>with just plain ol' device?  Although the buffalo jumps used by the 
>plains indians were very effective, I wouldn't call them robot's.
> 
>If your interested, a whole issue of Mechanism and Machine Theory
>was dedicated to terminology of devices, mechanisms, robots,
>manipulators, etc.  It was organized by the International 
>Federation for the Theory of Machines and Mechanisms, and
>gives definitions in four languages: English, French, German,
>and Russian.  (MMT, Vol 26, No 5, 1991)  For what its worth,
>they used the following definition for `robot':
>
>   A robot is a mechanical system under automatic control 
>   that performs operations such as handling and locomotion.
>
>which I am not particularly fond of either.
>
>>-Steve Walz   rstevew@armory.com
>Roger
>rbh@wolverine.utias.utoronto.ca
---------------------------------------
As I have said before, an early and venerable popular book on the subject
of robots states: "A robot is any device which senses, decides, and acts,
all in the absence of human attention or control." And recall that Turing
insisted that a computer must be able to do conditionals or it is not a
"computer". Surely a robot is not merely a sequenced device, like a clock
or a timer, and certainly not merely a tool which must be controlled by
a human ("Waldo"-type manipulators, for example). A robot is instead a
device which can, or is controlled by a device which can sense, decide,
and act, all in the absence of human attention or control. A robot can be a
trap, and the book in which I found that definition it clearly cites a trap,
specifically a mousetrap or any of the neolithic variety of active traps,
(not pit traps or enclosures then, and nothing which cannot physically act
for a human who is absent), as the first robot. Now you may see a trap
release mechanism as something trivial, but it is as much a "gate" as
anything made of silicon! And some traps had very convoluted logic,
actually representing two or more gates in their design features, upon
closer examination. And with Turing's definition of a computer, having a
computer as the "controller" within a robot certainly makes it a robot, as
long as conditionals are used, that is, as long as the robot "senses"
(actuates a sensor in the world which feeds a computer input, and which
input value figures conditionally in the computer's programming), "decides"
(processes input values in its program which causes an output), and "acts"
(whose output controls an actuator for some purpose), without requiring
human presence to do any of the above. Actually, in the data processing
realm, a computer is technically already a robot, when running a program
which accepts real world input (keyboard), and produces real world output,
(textual screen or printer), since it must use conditionals to know when
and how to store and retrieve and output that data for humans to use later.
A spreadsheet or numeric cruncher program in science qualifies the computer
with program as a "robot", technically! The difference being that we are
often the cause of its inputs and, in the case of the screen, are acted
upon by its outputs. But it does serve a purpose that was preprogrammed and
which we could not do without it and which the user does often not even
know how to program! But from traps to a Unimate system with sensors, they
are bothg robots! The book does clearly state what are NOT robots as well!
They fall into the categories of tools and devices, or sequencers. Tools we
control, from a flint knife to a bulldozer. Sequencers do no decisions or
conditionals, examples are clocks and timed functions with no input except
an ON switch. A washer which does not have any sensing is not a robot,
whereas one with a level sensor, or other sensor, IS a robot. The criterion
then is really conditional logic, and acting in the real world from real
world sensing. Your definition above is the same, really, but it is
awkwardly worded by someone who was NOT inspired at the time, perhaps a
textbook hack or such, a "publish or perish" sort of reluctant author!
-Steve Walz   rstevew@armory.com

