RE: AI hardware was 'Singularity Realism'

From: Ben Goertzel (ben@goertzel.org)
Date: Sat Mar 06 2004 - 17:07:29 MST


Actually, we have a single node/link core object (the AtomTable), but
the dynamics that we carry out on this object are complex, too complex
to be put into FPGA logic

ben

> -----Original Message-----
> From: owner-sl4@sl4.org [mailto:owner-sl4@sl4.org] On Behalf Of Peter
Voss
> Sent: Saturday, March 06, 2004 7:00 PM
> To: sl4@sl4.org
> Subject: RE: AI hardware was 'Singularity Realism'
>
> Our differences here are due to our designs. Yours is a more
distributed
> (agent) architecture, while mine centers around a single node/link
core
> object that would benefit much more from massively parallel
processing.
>
> Suffice to say that optimum hardware architecture (for a given design)
> plus
> optimizing software (for performance instead of R&D) can yield many
orders
> of magnitude improvement over prototype software running on a standard
PC.
>
> Peter
>
> -----Original Message-----
> From: owner-sl4@sl4.org [mailto:owner-sl4@sl4.org]On Behalf Of Ben
> Goertzel
> Sent: Saturday, March 06, 2004 3:30 PM
> To: sl4@sl4.org
> Subject: RE: AI hardware was 'Singularity Realism'
>
>
> > I agree with Ben's comments on hardware capability. I personally do
> not
> > see
> > hardware as a show-stopper (even at this stage, and certainly not
> within
> > 5-10 years) -- however, I assume that some specialized designs will
be
> > needed (eg. FPGA).
> ...
> > Best,
> >
> > Peter
>
> I don't think specialized hardware is needed, I think we can make do
> with clusters of contemporary multiprocessor 64-bit Linux boxes.
> However, specialized interconnect fabrics like Octiga Bay's would
> certainly cut hardware costs....
>
> FPGA is a great technology but doesn't seem to have direct application
> to Novamente. Hugo de Garis did some cool things with it though ;-)
>
> Ben g



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:46 MDT