From: David Picon Alvarez (eleuteri@myrealbox.com)
Date: Thu Sep 14 2006 - 15:56:28 MDT
> I figured this type of thing had been tried or at least contemplated. It
is a simple idea. I wonder if organic chemistry is really more complex then
the x86 instruction set? I am not qualified to answer that. When things got
‘stuck’ they may have just needed more time or maybe more randomness
inserted. If intelligence could be created in 4GB of x86 instructions then I
think this has a possibility of working given enough time. If not then all
current attempts at creating AI are in vain and we need to wait for quantum
computers.
The thing is, evolution is a terribly slow and inefficient optimization
process. I guess it's ok when there's no other choice, but ... I've heard
evolution described as a blind man randomly fiddling with a Rubik's cube and
asking each time if it is solved.
Besides, say you get an AI out of it. Its perceptual modality is going to be
based on quite a weird environment, a binary 4gb buffer and the x86
instruction set. I can't even think how you could communicate with it, or
for that matter know if/when it is intelligent and to what extent, since you
don't have a mechanism to observe the environment in which it develops (we
don't have modalities to observe bitstreams). All that aside, sounds like a
fun project.
--David.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT