From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!swrinde!gatech!ncar!noao!amethyst!organpipe.uug.arizona.edu!NSMA.AriZonA.EdU!bill Tue May 12 15:50:15 EDT 1992
Article 5539 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!swrinde!gatech!ncar!noao!amethyst!organpipe.uug.arizona.edu!NSMA.AriZonA.EdU!bill
>From: bill@NSMA.AriZonA.EdU (Bill Skaggs)
Newsgroups: comp.ai.philosophy
Subject: Re: AI failures
Message-ID: <1992May10.184816.19179@organpipe.uug.arizona.edu>
Date: 10 May 92 18:48:16 GMT
References: <1992May7.152447.7930@waikato.ac.nz> <727@ckgp.UUCP> <uc2m8INNn5d@early-bird.think.com> <1992May9.013815.24244@nas.nasa.gov> <MORAVEC.92May9113029@turing.think.com>
Sender: news@organpipe.uug.arizona.edu
Reply-To: bill@NSMA.AriZonA.EdU (Bill Skaggs)
Organization: Center for Neural Systems, Memory, and Aging
Lines: 38

In article <MORAVEC.92May9113029@turing.think.com> 
moravec@turing.think.com (Hans Moravec) writes:
>
>Extrapolating from the ratio of neural tissue in the retina to the amount
>of computing required to produce a similar end result with computer image
>processing programs, I've concluded in past that a whole human mind should
>be implementable in real time with a processing power of very roughly 10^13
>(ten trillion) operations/second.  I've also guessed that with reasonably
>efficient encoding (including some data compression) the state of a human
>brain should be encodable in (even more roughly) 10^14 bits.  

  Applying information theory to the existing data about communication
between neurons in the cerebral cortex, I come up with numbers about
two orders of magnitude smaller than these.  (Which only reinforces
the argument.)  In fact, if the numbers I get are accurate, the
processing power of a human brain is not a whole lot more than
that of a 64,000 processor Connection Machine.
  If this seems implausible, note that the source code for the
brain (i.e. the genome) would come to about ten million lines
of C code (though it's unclear how much of this has been
"commented out" during evolution), and that a human brain
requires twenty years of intensive training to reach full
functionality.  I don't think its at all obvious what the
limitations would be of a Connection Machine with an equivalent
base of software.

>
>With continuing persistent 1000 fold/ 20 year growth in computer power
>and memory, within 100 years it should be possible to have a BILLION
>human-scale AIs running simultaneously at human rates in a (vintage 2090)
>workstation-class machine!  

  Unfortunately, what looks exponential now is probably just the
rising phase of a sigmoid, and nobody can tell when the inflection
point will come.  I don't take this prediction too seriously,
but it does seem possible.

	-- Bill


