Re: Implementation

From: SLP <SLP.domain.name.hidden>
Date: Tue, 27 Jul 1999 23:24:21 -0500

HM:
I dare you to demonstrate to me that a HLT isn't exactly as
conscious as any other implementation of the same behavior.


SLP:
The simplest and best argument for this may be an evolutionary one. An
HLUT is a very inefficient way of generating complex behavior. We
evolved to have self-aware internal states precisely because it is a
computationally economic way of producing highly complex and responsive
adaptive behaviors. And neural net systems work much more efficiently
in adaptive applications than things like lookup tables. In a very
primitive way, you can see what types of layered, hierarchical
structures brainlike systems must have if you look at work by people
like Hopfield, Tank, Grossberg, etc. Thus gives a feel for what must be
going on computationally. Non-conscious neural modules, interacting in
highly complex ways, and showing non-local degrees of freedom, are the
probable building blocks of consciousness. The HLUT has a transparent
and very elementary structure and shows none of this. It is a huge,
clunky, dumb thing, far too inelegant for biology, which evolves
streamlined adaptations. And consciousness IS, above all, an
adaptation.

This, of course, is not a demonstration. Proving a negative is a very
difficult thing to do. Generally, the burden of proof is one someone
who makes a positive claim. Since we have no evidence of consciousness
in nonbiological systems, the person claiming such consciousness exists
would have the normal initial burden of proof.

Although surveys prove nothing beyond mere opinion, I wonder how many
people on this list would be willing to upload their consciousness into
a system based on a HLUT--and take the chance that, from the standpoint
of qualia and selfhood, they would be dead.

I think we are entirely agreed that there WILL be conscious robots in
the coming century. And your books superbly outline the general shape
of things to come in a powerful and compelling manner. The picture you
paint is a very exciting and, I think, largely realistic one, and you
express it wonderfully. We only disagree on what will make robots
conscious. We each have our areas of expertise, and perhaps this
creates certain biases in each of us. I rather strongly suspect that
the first conscious robots will be based on evolvable hardware, a la
Hugo deGaris, and will have strong similarities to biological brains.
Later, nonbiological consciousness will exist in much more abstract
forms. Then Mind Fire can begin in earnest.


Steve Price, MD
Received on Tue Jul 27 1999 - 21:35:21 PDT