From: blsonne at rogers.com (Byron L. Sonne)
Subject: Empirical data surrounding guards and firewalls.

First off, it bears remembering that I said 'computer programs' not 
'artificial organisms'.

 > You clearly don't know very much about AI, or sentience. You clearly
 > were unable to exercise understanding of the definition of sentience.

On the contrary! I have defined sentience and can therefore confidently 
operate using that word. Your point is therefore proven incorrect, but 
if you wish I could post a delicate dissection of why I think you are 
wrong. I believe it is you who have failed to grasp the definition and 
more importantly the spirit of the word 'sentience'.

 > To perceive by the senses. Well not that there are for one common
 > example, PS2 games which actively do this with vision

That is not perception. Perception is a faculty reserved for those with 
awareness. Without awareness it is merely operating on data using 
techniques/algorithms that approximate the behaviour of life, if not 
operating in an altogether simpler fashion. Even if something were 
alive, that would not necessarily mean it could perceive.

 > Of course maybe you should learn a little about AI and its techniques
 > before you try and bad mouth it. There are far more running
 > applications of various fields of AI than you are aware of.

Oh, I've learned of it and that is why I dismiss the lofty claims that 
many of its proponents advance. It is of incredible value, yes, but 
c'mon... if people have trouble getting simple vision recognition 
systems to work as well as a healthy human being, than I feel I can 
safely say the mysteries of the mind lay outside the reach of mankind 
for the time being and into the foreseeable future.

They've become overawed of their simulacra, and mistake the quacking of 
a duck for the proof of it's very duckness. I understand completely as 
it is a very seductive field and extremely hard work; I believe this 
leads them to overestimate the true magnitude of any advance.

 > Neural Networks exhibit sentience by their very nature. There exist
 > (logical computer) programs which utilize a data structure
 > representing and functioning as a neural network.

Neural networks are nothing more than behaviour models that can learn. 
'Learn' doesn't mean anything more in this context than 
predictive/future behaviour based on trial and error. It's guestimation! 
Considering that until recently the role of glial cells in brain 
operation and mental processing was vastly underestimated, I would find 
it unlikely (yet amusing) if the AI researcher's model/approximation of 
living neuronal behaviour somehow proved to be a stunning vindication of 
neural networks as the sole, or most important, fundament for sentience.

Ok, let's define 'sentience' for once and for all. I'll take the output 
of 'dict sentience' and pick the definition that best encapsulates what 
I think sentience is:

 From WordNet (r) 2.0 [wn]:
  sentience
       n 1: state of elementary or undifferentiated consciousness; "the
            crash intruded on his awareness" [syn: {awareness}]
       2: the faculty through which the external world is apprehended;
          "in the dark he had to depend on touch and on his senses
          of smell and hearing" [syn: {sense}, {sensation}, {sentiency},
           {sensory faculty}]
       3: the readiness to perceive sensations; elementary or
          undifferentiated consciousness; "gave sentience to slugs
          and newts"- Richard Eberhart [ant: {insentience}]

"state of elementary or undifferentiated consciousness" might be 
construed as lending credence to your argument but in this case I think 
it concerns the division, or lack thereof, between the id, ego, and 
superego of Freudian interpretation. Freud might be considered by many 
to be somewhat specious or overrated, but if I could just find that damn 
Scientific American article, I'd show you that evidence exists that 
correlates Freud's interpretation/model with empiric physiological 
evidence. I can't believe that I'm admitting Freud was anywhere near 
being right, or even useful, but what can I do but accept the evidence 
as it stands? :) Ah, here it is: 
http://www.sciam.com/article.cfm?chanID=sa006&articleID=00074EE5-1AFE-1085-94F483414B7F0000 
you'll have to subscribe to read the full article, but as you seem 
interested in this topic I'd suggest trying to track down a copy of the 
May 2004 issue. It's worth it! But I digress.

You are clearly a novitiate in the deeper understandings of philosophy 
if you think that an object, by merely exhibiting the properties of 
something, means that it is equal to something. I would enjoin you to 
familiarize yourself with the Chinese Room argument.

The extraordinary claims you may make regarding AI require extraordinary 
proof, not just your conjecture that because they exercise capabilities 
or qualities similar to sentient, and therefore living, beings that they 
must, by definition, be living or sentient beings. Furthermore, there is 
nothing rigorously scientific about the Turing Test. It is entirely 
subjective as it is based on the opinion of the interrogator. As far as 
an objective test of intelligence it *is* a sham. Sham, by the way, 
defined as "False; counterfeit; pretended; feigned; unreal; as, a sham 
fight.". Now I'm just being pedantic.

Minds such as Marvin Minsky and Roger Penrose have tackled this debate 
in far deeper and rigorous fashion, and have not, in my opinion, 
advanced significantly further than the Greek philosophers did millenia 
ago. While that statement could be considered fallacious as an appeal to 
authority, it goes to show that the problem is far harder to tackle than 
most (all?) people, myself included, realize. Minsky practically comes 
right out and says this at the end of his essay "Why People Think 
Computers Can't"

 > "Fail" the Turing Test? I was not sure this could be applicable to a
 > human

If something is not subject to falsification then it can not be 
considered an arbiter of truth. At the very least, it can not be 
considered scientific as it cannot be used to establish a hypothesis 
that can be confirmed or denied.

I say this all as a former hard core machine intelligence and AI 
believer. To make an even simpler reduction of the argument, if people 
ever create something that is truly sentient, self conscious, aware, 
alive... then certainly it is no longer artificial, now is it? Life is 
life. Whether man created a scaffold to entice, house and tend that 
spark of life is another matter, but that life I do not think is a 
property we can claim to have endowed it with since we cannot even 
accurately describe what it is or how it came to be. I think therefore I 
am, regardless of whatever I was before or what I was made of.

School's out.

Regards,
Byron

Powered by blists - more mailing lists