Re: Safety of brain-like AGIs

From: Ben Goertzel (ben@goertzel.org)
Date: Thu Mar 01 2007 - 17:03:04 MST


Shane Legg wrote:
> On 2/28/07, *Ben Goertzel* <ben@goertzel.org
> <mailto:ben@goertzel.org>> wrote:
>
> Shane Legg wrote:
> >
> > If we consider informal definitions, then clearly some humans are
> > friendly and intelligent.
>
> Yes but no humans are **guaranteeably** friendly and intelligent...
>
>
> Wouldn't such a strong guarantee of friendliness require a formal proof?

No absolute guarantee is ever going to be possible....

Even a mathematical proof only applies to the physical world based on
certain assumptions about the physical world, which are necessarily
known only by induction, which can never provide guarantees...

However, there are degrees of certainty; and the degree of certainty
obtainable using a humanlike architecture seems quite low...

-- Ben

>
> Shane
>
>
>
>
>
>



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT