Re: guaranteeing friendliness

From: H C (
Date: Tue Nov 29 2005 - 21:44:34 MST

>From: Robin Lee Powell <>
>Subject: Re: guaranteeing friendliness
>Date: Mon, 28 Nov 2005 23:44:23 -0800
>On Tue, Nov 29, 2005 at 07:08:13AM +0000, H C wrote:
> > It's not so rediculous as it sounds.
> >
> > For example, provide an AGI with some sort of virtual environment,
> > in which it is indirectly capable of action.
> >
> > It's direct actions would be in text only direct action area
> > (imagine it's only direct actions being typing a letter on the
> > keyboard, such as in a text editor).
>Oh god, not again.
>Quick tip #1: if it's *smarter than you*, it can convince you of
>*anything it wants*.

And I am describing a way of verifying *what exactly it wants*

>Quick tip #2: what you're describing is called "slavery"; it has
>teensy little moral issues.

First of all, describe me a necessary and sufficient moral system.
Second of all, this kind of ties in to what someone else commented about:
slavery is in the eyes of the slave. We are slaves to our genetic desires
programmed by Evolution Himself, etc.

>Quick tip #3: Search the archives/google for "ai box".

If you are going to suggest readings, then I suggest you read everything on
the Singularity Institute website.

> ***
>Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
>Proud Supporter of the Singularity Institute -

As far as "AI boxes" go, yes the answer is deadfully obvious. However,
perhaps in the future you might give me a little more credit, because on
this occasion I wasn't referring to the classic problem.

In this case, the programmer is capable of directly accessing and observing
the unconciouss motivations, concsiouss intentions, thoughts, plans, etc,
and is essentially left in complete control of any real-world effectual
action. The AI must, as necessary for any action to be carried out, submit
its actions, in algorithmic form (along with comments) to a panel of human
judges. If these judges are the least bit uninformed about the AI in any
conceivable manner, then they would simply deny the approval of the action
and gather the necessary information, by means of directly accessing AI
memory or by means of asking the AI specific questions, or even running
"temporary" instances of the AI, just to see what it would do, or running
dream sequences or some similar psychedelic experience where the AI would
tend to abandon reason and allow it's environment to flow to its desires,

In this specific situation, the AI is, presumably, quite engrossed with very
complex "projects" or goals and has been so over a period of time. Not only
that, but ALL of its actions, even in the simplest virtual environement,
must be established in the proper algorithmic form and approved before the
action even occurs, no matter how simple of trivial. If, over the course of
it's history, it has demonstrated a very open and easy to understand
personality in which it enthusiastically embraces the expected goals, and is
very cautious and questioning about subjects that go against the designed
goals, and over the course of all the observation and testing we have found
very little evidence of deviance from its expected goal system, then there
will be plenty of evidence to say that we can verify that is it following
its designed goal system.


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:53 MDT