Re: Dynamic ethics

From: Russell Wallace (russell.wallace@gmail.com)
Date: Sat Jan 21 2006 - 08:58:48 MST


On 1/21/06, Kevin Osborne <kevin.osborne@gmail.com> wrote:
>
> But what I am suggesting is that an AGI may find itself unable to
> accept such malignant acts, in their entirety. Not one child, not ever
> again; forget about a million. So what does it do? Raise them itself?
> Add nano/chemical restraints into the air/water supply? Surreptiously
> implant sense-recorders? Decide adults are damaged goods, and keep us
> all pre-pubescent children of the corn?

Sorry, I was half asleep when I wrote my last reply to this; I think it
needs a more considered reply, which I'll try now.

First, let's separate out the issues.

Child abuse (rarely by the archetypal stalker, more commonly by mothers'
boyfriends, peers and government and church authority figures) is a real
problem. There are ways to address it, which have nothing to do with
technology, much less artificial intelligence, and everything to do with the
use of human intelligence to look dispassionately at the real causes and
statistics and work from there. However, this isn't even close to being on
topic for this list, so I'll refrain from further discussion of it here.

"A witch cursed my crops!" "The Jews are conspiring with the bolsheviks!"
"My co-worker is a communist!" "Children will be molested unless Big Brother
watches and controls everyone!" History is full of cases of mass hysteria
whipped up and encouraged by authority figures for their own purposes; human
psychology appears to be sadly vulnerable to this sort of thing.

Then of course we get projection behavior. Folklore is full of stories about
fairies cursing people. The modern fairies (aliens in flying saucers) were
abducting and brainwashing people in the paranoid 1950s; then they were
preaching New Age fluff when that was in vogue; we're seeing more abduction
and brainwashing popping up now that we're back to being paranoid.

And on to AI, which is on topic for this list. But let us try to keep
separate the real challenge posed by Friendly AI (which involves among other
things mathematics, computer science and moral philosophy) from the use of
the concept to project the currently prominent form of mass hysteria (which
is a matter of politics and psychology). AI will have its limitations, like
everything else that must work in the real universe; but there is nothing in
the logic of the problem that requires it to inherit those forms of paranoid
irrationality peculiar to the human mind.

- Russell



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:55 MDT