Re: Problems with AI-boxing

From: Tim Duyzer (
Date: Fri Aug 26 2005 - 16:13:10 MDT

Perhaps it would be most effective to create a perfect simulation of our own world, and then have our virtual human counterparts research and create AI so we can safely observe the result. Oh, wait...

Unfortunately to create such a perfect simulation we would probably need AI.

What I wonder about the whole box/gatekeeper thing is why the gatekeeper has to have the power to let the AI out. If the one person who got to talk to the AI didn't have that power, and had to convince a human counterpart of why the AI had to be let out, it'd at least be a firewall of sorts. I haven't seen an answer to this anywhere on the list.
  ----- Original Message -----
  From: Pilot Pirx
  Sent: Friday, August 26, 2005 5:55 PM
  Subject: Re: Problems with AI-boxing

  The virtual evaluation world would have to be completely convincing to the AI-solate, or it might decide to fool everyone and be completely friendly, pass the "test", but then raise hell once it was hooked into the real world.

  Mike Williams <> wrote:
    Rather, what kind of virtual world could be built to ethically contain a
    sentient being, so that it could be evaluated prior to releasing it?
    Its interactions with the inhabitants of the virtual world could be of
    help in determining how friendly it is.

    Pilot Pirx wrote:
> Perhaps the Moon should be considered as
> a containment site for experimental AI/nanotech research and development.
> */Mike Williams /* wrote:
> > If you[1] think it's OK to keep a human-equivalent AI in a box,
> > isolatiing it both socially and sensorily, what makes you think are
> > moral enough to be capable of creating friendly *anything*, or even
> > recognizing it when it comes along?
> >
> How big is the box? Aren't we all in a universe-sized box?> Mike
> ------------------------------------------------------------------------
> Start your day with Yahoo! - make it your home page

  Start your day with Yahoo! - make it your home page

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:52 MDT