Re: Friendliness, Vagueness, self-modifying AGI, etc

From: Michael Roy Ames (
Date: Wed Apr 07 2004 - 14:29:55 MDT

--- wrote:

> Formal design methods, patterns, etc. are no
> substitute for intuition trained by experience.

Correct. Good formal design methods are not intended to be a
substitute for good coding, they occupy another level of
creative space.

> Personally I consider them an end result,
> useful documentation which may be written down
> to describe the design you've implemented to
> others.

Your comment seems to discount the purpose of, and the value in,
formal software design. Formal software design has nothing to
do with the *how* of a solution: the techniques used during
implementation, the algorithms or the coding language, the
software or hardware infrastructure. It has everything to do
with the *what* of the solution: the data inputs and outputs,
the conceptual data flows, the functional purposes of the
software solution. A good formal design could be used *without
change* to implement a software solution on an IBM 3090
mainframe or with a warehouse of young women & notepads.

In business software design today the *designing* activity
extends down into great detail. This tendency to
'overspecification' in the design stage has occured for a
variety of good reasons, most of which nolonger apply when
creating fundamentally new software. Creating a Friendly AI is
not going to be like creating business software, even complex
business software. We are going to *need* the separation of
design-space from implimentation-space. That separation will
work to improve the content of *both* spaces through feedback in
*both* directions.

Michael Roy Ames

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:46 MDT