Re: HAL program..

[Subbarao Kambhampati <rao@asu.edu>]

At 08:02 PM 11/28/2001 -0700, Timothy.Warnock@asu.edu wrote:

> Regarding the HAL program on PBS,
> it seemed that the AI research was pretty
> disparate with conflicting perceptions of what direction the field
> should
> take. They seemed to dance around the idea of consciousness, refering to
> HAL,
> (talking about emotions) but without any defined concept of 
> computer
> consciousness (or human consciousness for that matter).  I wasn't
> sure if that
> was just for the purposes of the show, to entice the audience, or are
> AI
> researchers really concerned with creating a conscious computer
> mind?  And if so, do they have an idea of what that consciousness
> would be
> like? 

This is an interesting observation. Speaking for myself, I don't really
care whether or not AI creates/emulates consciousness... however, neither
do I 
see any reason why that is ipso facto, unattainable. I think this is true
of a large number of active researchers (as against philosophers).

 The consciousness debate, to a large extent, is a lightning rod.
Many otherwise brilliant people have taken very anti-AI stance 
just because they think it is not possible to emulate consciousness-- a
sort of
 "If human brain is so simple that we can understand it, we
will be so simple that we cannot"
stance. An interesting case in point is the set of books written by Roger
Penrose--a brilliant mathamatical physicist--trying to somehow show that
one can never achieve computer consciouness, and that consciousness is
somehow inherently biological. 

[[as an aside, there are so many anti-computer-consciousness experts that
some irate AI folks have started a (somewhat ungracious) award called the
"Simon Newcomb" award. Simon Newcomb was a 19th-ear;y 20th
century astronomer, one of whose claims to fame were a series of rather
passionately argued articles on how "heavier than air" flight
is "scientifically" impossible.]] 


The discussion on emotions on that program is quite practical, IMO. It is
easy enough to see that it is important for computers to "fake"
some emotions if they were to interact with humans. There are all sorts
of studies showing that people spend more time working on computer
quizzes etc. if the computer shows appropriate (if canned) emotional
response when people falter. The whole Cynthia Brazelton bit with Kismet
the cutie-pie robot in the HAL program is another point. Those big red
lips and funny eyebrows of Kismet can have a utilitarian purpose beyond
general cuteness. Rodney Brooks once said that they decided that since
ultimately robots have to be "trained" by people and people are
not too patient with random ugly windows pcs, it is better to take the
evolutionary biases that people have about teaching kids and make the
robots behave like kids. 

What is more suprising is that understanding emotions may be important
even to make the computer agents perform well on their own--even in the
absence of other human interactions.  Check out this mail I sent to
a previous class:

http://rakaposhi.eas.asu.edu/cse471/emotions.html

Rao
-----------------
   I'd rather learn
from one bird how to
sing              
   than teach ten thousand stars how not to
dance           
                                                           
             
--You shall above all things be glad and young
               
EE
Cummings