Re: Seed AI milestones (types of self-modification.)

From: Christian Szegedy (
Date: Thu Feb 28 2002 - 09:57:19 MST

Gordon Worley wrote:

> Not entirely regardless of Ben's possible mathematical argument, just
> in case you aren't noticing, the intuition that it's like this (broken
> up into different levels of self modification) comes from reading GEB
> and the fundamental limitations found in computers (programs without
> loops, programs with loops, and some meta programs that don't really
> exist that can solve the halting problem (in programs with loops)).
> These levels of incompleteness imply similar intuitions about other
> computational systems, like ones that self modify.

I am afraid that his classes are not based on such
theoretical intuititon. If they would, perhapse he would
have made more thoughts about whether they
are distinct at all - the first thoughts of a theorist.
My most important point of critics is that he
(subconsciously or not - I don't know) makes
the impression that he has some theoretical
justification for it. I am basically quite sceptical,
so I don't fall for it, but people having less
background knowledge, would take this implicit
message for granted.

I may be wrong, but I would guess, that his statements
are rooted in a purely experimental base, They are
the manifestation of considerations put it into a plausible,
but very special architecture completely disregarding
any theoretical background.

I don't crititicize his work done as an "experimental
mathematician.", I think, there are chances that one
constructs something great on minimal theoretical
knowledge. The cathedrals were built
without knowing the laws of statics, also.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT