From: Patrick McCuller (patrick@kia.net)
Date: Wed Nov 22 2000 - 17:53:46 MST
>
> This sounds a bit religioid to me. We're talking, by the by, not about
> "the superintelligent supermind", but about superintelligent minds running
> on hardware we can understand - which might be a mistake in itself, but
> this is hardly avoidable. I can think of several ways to achieve an
> "intrinsic and simultaneous awareness of the past" on understandable
> hardware, although the most obvious method would require an exponentially
> increasing amount of processing power and memory storage - which might
> still turn out to be available, of course. I also disagree that
> perfection derives from awareness of past experience.
I suspect it would be much, much worse than exponential, considering the
variety of forms available 'memory' could take. I'm thinking Moser polygon
notation.
>
> Thou shalt justify thine statements.
Memory has too many forms and is too readily dynamic in machines. It isn't
like a library; it is more like a city, or the world. Making decisions based
on everything in the world would be... well, for us, impossible. It is
difficult to imagine the intelligence that could do it at all, much less in
real time.
Yes, there are forms of 'memory' that can be accessed and used without
processing (searching, indexing, retrieval) power... like holographic
crystals. But surely only a small part of an intelligent being's memory could
be constructed that way.
Patrick McCuller
>
> -- -- -- -- --
> Eliezer S. Yudkowsky http://intelligence.org/
> Research Fellow, Singularity Institute for Artificial Intelligence
>
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:35 MDT