[alicebot-aiethics] Re: No more eithics concerns

[Christopher Fahey [askrom]]

Ryan wrote:
> If consciousness is the difference between a zombie and a human 
> then can't we define conciousness as having the following qualities:
> 
> 1)  The ability to act without being given external stimuli.
> 2)  Unpredictability.

Two hypothetical questions:
 - Could you prove that a zombie really lacked these qualities?
 - Could you prove that a human posessed these qualities? 

You're including "that which we can never know" as a part of your
definition. That's where I see the hocus-pocus in the word
"consciousness". If a definition includes a term that by its nature
cannot be proven or disproven, then it collapses. 

Are you defining "consciousness" or are you defining a zombie (by, in
turn, defining "two qualities a zombie does not posess")? I'm not even
sure that your idea of a zombie and mine are the same. To a person like
me who considers "consciousness" to be a non-concept, a zombie (if such
a thing existed) would be just as "conscious" as you or me, although a
lot dumber. 

You even go on to argue that a human mind might, in theory, be fully
probe-able and explainable using an advanced MRI device. On the one hand
you assign the quality of "unpredictability" to consciousness, but on
the other hand you seem to suspect that a human brain might, in fact, be
predictable. Which begs the question, "Can consciousness be explained?:
;)  


> Suppose for a second that an advance form of a MRI could 
> record the activity of the electrons in the human brain.  
> Further suppose that we had enough background information to 
> "decode" the thought process that this electrical activity 
> would produce, thus "debugging" the human brain.

That's an interesting image - I like it a lot. I can imagine the
headlines ("Machine Proves that Human Beings are No More than Just
Machines") People will wonder if your "advanced MRI machine" isn't just
trying to spread anti-human propaganda! :)

 
> Would this proceduce be legal?  Should this procedure be 
> legal? And if a robot has acheived consciousness (equivolence 
> to humans), should it be legal to disassembling its code?

This is a perfect example of why I think the *legal* aspects of
consciousness will be the most important as robots approach human-like
behavior and intellect. Reverse the terms of your question, and I think
you will see the sequence in which history will likely answer the legal
and metaphysical questions of consciousness: "If it is illegal to
disassemble a robot's code, does it mean it has achieved consciousness?"

In other words, someday in the far future (after a bunch of robots
demand the right to vote, for example) "consciousness" will be defined
as a legal quality, and the legal definition will be the only one that
really counts - the legal definition will dominate our understanding of
the metaphysical concept. There will likely be different legal
definitions in different countries and states, but in the end it will
become a word like "manslaughter" or "pornography", a term whose
definition will be shaped by lawyers and juries.

I wonder if we will "give" the legal right of consciousness to robots
without a second thought, will we "give it up" after a long cultural
upheaval, we will have it taken from us by the robots in a revolution,
or will we never give it up at all. Your answer to this question should
match your answer to the question "can machines think".


> > Well, okay, I suppose we _could_ build a robot with the 
> > followingcharacteristics that would by default circumvent all
> > metaphysical *and*legal arguments:   1) Complete autonomy from human
> > control.   2) Invulnerability to probing (through any number of
means
> > includingencryption, overwhelming physical/military might)#1 
> > is easy, #2 is hard. Such a robot (even if it failed a Turing 
> > Testand even if we understood it's brain fully) could in 
> > theory insist on usgiving it legal rights. Regardless of 
> > whether or not it has a soul, sucha robot could force us to 
> > treat it with more respect and dignity thanmost humans ever 
> > get. 

> I'd argue that neither is actually possible.  We, as 
> humans, do not live in complete autonomy from human control.  
> The belief in such autonomy is merely an illusion.  Every 
> exposure to human activity influences one's actions.  

I simply meant that there was no remote control connected to its brain.
The kind of control you're speaking of is social, not physical. My point
was that a robot with the capability for unstoppable violence would make
all metaphysical and legal question silly. I think we would treat such a
robot like an equal human being if that's what would make it stop
killing us. Violence creates truth and winners write the history books
-- and robots can theoretically be capable of great violence.


> In the case of 2) I refer to my previous hypothetical.

An un-probe-able machine is just as fanciful as a unstoppable-probe, so
you gotta give me some slack here. I'm talking about a comic-book-style
megalomaniacal invincible robot, for cryin out loud. :)

Come to think of it, all really good artificial intelligence ethics and
theory sounds like comic book plots. I like that!

-Cf

[christopher eli fahey]
art: http://www.graphpaper.com
science: http://www.askrom.com