RE: design complexity of assemblers (was: Ramez Naam: redesigning children)

From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Mon Dec 02 2002 - 14:45:01 MST


On Mon, 2 Dec 2002, Ramez Naam wrote:

Mez,

> With an assembler - a totally novel and extremely powerful type of
> self-replicator

You really need to go read the papers I mentioned, particularly
Josh Halls.

An assembler does not need to be a self-replicator. In fact
a self-replicating assembler is vastly harder than an assembly
line that manufactures assembler arrays using special purpose
assembly tools. To make macroscale stuff rapidly you need
lots assemblers -- you *do not* need lots of self-replication.
Star Trek type "replicators" could be safe so long as they
are certified to only "assemble" things which are safe.
You can make apple pie but not self-replicating assemblers.

It is generally accepted that it would be unwise to build
assemblers that can self-replicate -- this is why the
concept of a "broadcast architecture" that keeps the code
separate from the execution units was invented. I think
that is covered in another one of Ralph's papers, a Google
ought to find it.

This is kind of the principle behind Zyvex's Rotapod(tm). You
get exponential growth in assembly without true self-replication.

Robert



This archive was generated by hypermail 2.1.5 : Wed Jan 15 2003 - 17:58:32 MST