Re: How hard a Singularity?

From: Michael Roy Ames (michaelroyames@hotmail.com)
Date: Wed Jun 26 2002 - 08:32:50 MDT


James Higgins wrote:
>
> Wrong, dead wrong. For one thing, why bother to check with others if you
> just plan to ignore them anyway?

The "plan to ignore them" is your contention, which is false. But to
directly address the question "why bother to check with others" the answer
is: so that they can help improve the ideas. And, to address a related
issue: if someone offers suggestions that "make no sense" or are
"incoherent"... why wouldn't you ignore them?

>
> It is irrational to proceed unless there
> is no other choice.
>

There is always another choice. We could do nothing, and let 'fate' decide.
Not my preferred option. No-one is ever going to be 100% certain that a
Friendly Seed AI will work out perfectly... but the same thing could be said
of any complex project that involves large changes. The point is: *are* we
just going to sit around and let wild AIs just 'happen', or are we going to
deliberately create them and direct them in the best direction we can think
of?

>
> ABSOLUTELY ANYTHING that could lead to the extinction
> of the human race MUST BE checked, rechecked, checked again, etc. To the
> maximum extent possible.
>

I agree with you very much here. But the checking must be useful, and
productive, and lead to definite conclusions. Checking just for the sake of
checking is pointless. And that is why there are several sections of CFAI
devoted to "Checking"... Friendliness tournaments come to mind, as do
'stress' testing scenarios. This will be a very important part of any
project that adopts Friendly AI ideas. Further 'checking' suggestions are,
of course, encouraged.

Michael Roy Ames



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT