From newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!caen!kuhub.cc.ukans.edu!spssig.spss.com!markrose Tue Jun  9 10:08:03 EDT 1992
Article 6167 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!caen!kuhub.cc.ukans.edu!spssig.spss.com!markrose
Newsgroups: comp.ai.philosophy
Subject: Re: Transducers
Message-ID: <1992Jun08.233400.7524@spss.com>
>From: markrose@spss.com (Mark Rosenfelder)
Date: Mon, 08 Jun 1992 23:34:00 GMT
References: <60790@aurs01.UUCP> <BILL.92Jun8150837@cortex.nsma.arizona.edu> <1992Jun8.221324.535@mp.cs.niu.edu>
Organization: SPSS Inc.
Nntp-Posting-Host: spssrs7.spss.com
Lines: 31

In article <1992Jun8.221324.535@mp.cs.niu.edu> rickert@mp.cs.niu.edu 
(Neil Rickert) writes:
> Perhaps.  But perhaps an argument "in principle" means that once a
>successful embodied robot has been created, a disembodied version can
>then be created *in practice* merely by copying the disk containing all
>of the learned knowledge about the real world.

Just as an operating system for a new computer system can be created by
copying some files from another system?  :)

Seriously, I don't doubt that you could do some wild new things with robots
that you can't do with humans-- copy the program and data from one robot
to another, for instance.  Only one robot of each model type needs to go
through infancy...

But beyond this, what exactly would you expect to accomplish by "copying
the disk containing all of the learned knowledge"?  In a neural-network-based
program, for instance, it might not even be possible to separate the
robot's knowledge from its operating procedures, or to extract a subset
of its knowledge rather than all of it.

Even in a traditionally programmed robot, the real-world knowledge might
be in a form which was of little use to something that wasn't a robot.

Or to look at it another way, if you build a computer system which took the
real-world knowledge disk from the robot and did something with it-- even
if that something was something as sophisticated as attempting to pass a
Turing Test-- there's no guarantee that you've transferred the "intelligence"
or the symbol grounding or even the real-world knowledge that the robot had.
You've created a *new* system and the successes of the robot can't be
simply passed over to it.


