Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!dali.cs.montana.edu!milton!cphoenix@csli.Stanford.EDU
From: cphoenix@csli.Stanford.EDU (Chris Phoenix)
Newsgroups: sci.virtual-worlds
Subject: Re: personal simulations
Message-ID: <15027@csli.Stanford.EDU>
Date: 23 Aug 90 21:58:13 GMT
References: <1990Aug21.120733.2752@world.std.com>
Sender: hlab@milton.u.washington.edu
Organization: Center for the Study of Language and Information, Stanford U.
Lines: 48
Approved: hitl@hardy.u.washington.edu


In article <1990Aug21.120733.2752@world.std.com> fhapgood@world.std.com (Fred Hapgood) writes:

>The animation would be generated by voice, movement, appearance,
>and discourse synthesizers all running in parallel.  Each would be
>tuned to the style characteristic of the subject in that domain:
>his pitch and tones, phonemic pronunciation, physiogomy, gait,
>posture, body language, and conversational habits.  I grant you
>some of the software problems here are non-trivial, but nothing
>in that list is technically _impossible_, right?

It sounds like you're talking about something that can pass the Turing test.
Maybe not here, but the rest of the article leans further and further in 
this direction.  It's a neat idea, and I'm sure people will try to do it,
but the part about conversational habits alone may be impossible.

>Will there be an issue of misusing animations?  Of programming
>your boss to be gang-raped by a gang of street thugs?  Or of
>living a dream life as one of the talented, famous, rich, and
>beautiful?  Will people get off on watching themselves make
>passionate love with the most desireable sex objects on the
>planet?

There's a definite problem here.  You're talking about not only programming
simulations to act like you in some settings, but to act like your boss in
totally unexpected settings.  Take the issue of making love.  Either 1) you'd
have to have a good enough model of human physiology and mentation that the 
computer can decide how to act when presented with such an unusual setting,
2) you'd have to let the computer watch you making love (in which case the 
simulations you let other people see would probably be *very* limited) or 
3) if people are similar enough, you could have one "making-love" paradigm 
that could be plugged into any simulation.  Note that this is different from
1, in that you can "hard-code" such a paradigm rather than making the computer
extract it from its modelling of humans--but it would be very impractical to
hard-code paradigms for each situation, for each person, so people would have
to be similar enough that one paradigm produced a believable simulation no 
matter which person it was run for.  IMHO, 2) is the only way it will work.
But if you want to have anything better than a computerized video recorder--
if you want to have the computer extrapolate new behavior, which is what I 
think you're saying--then you'll solve half the problems in AI along the 
way.  A correlary of the previous sentence is that if half the problems in 
AI can't be solved, what you're proposing is impossible.

-- 
Chris Phoenix               |  "I've spent the last nine years structuring my
cphoenix@csli.Stanford.EDU  |      life so that this couldn't happen."
...And I only kiss your shadow, I cannot see your hand, you're a stranger 
now unto me, lost in the dangling conversation, and the superficial sighs...