From: Dan Clemmensen (dgc@cox.rr.com)
Date: Sun Mar 10 2002 - 07:56:19 MST
Eugene Leitl wrote:
['Gene's preliminary defense of Earth against explosive AI:]
>
> A few things off the top of my head (this is what a transhumanist think
> tank is supposed to be paid to be doing).
>
I think that your proposals might help against some catastrophe scenarios,
but the two most basic and important fixes will not happen soon, and as
you say, the window is already open. As long as the net provides free
connectivity among millions of computers that run vulnerable operating
systems, the resources needed for an AI are available. Your defenses
addressed both of these issues, but the OS base will take a long time
to fix: Microsoft will not change its practices unless it is forced,
and applying that force will take several years before they can be
made to start. Your idea of making global enhancements to the internet
will take as long, and the really effective enhancements ("white hat"
worms and AI pattern evaluators) are both really dangerous in
themselves. Corrupt a "white hat" worm to get a "black hat" worm that
the system accepts as a good guy. An AI pattern evaluator would already
be so sophisticated that it is nearly intelligent: push it over the
edge. Your other proposals are to monitor the people and to monitor
the needed resources (clusters.) The CS community is far too large
for effective monitoring. Once a monitoring beaurocracy is that
bit it is corruptible can be evaded. Even if clusters are needed,
(probably not) There already a great many clusters, and any CS
lab office building, or computer store can be turned into a
cluster.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:12:54 MST