Re: obstacles to unbounded intelligence

From: Dan Clemmensen (dgc@cox.rr.com)
Date: Sat Jan 26 2002 - 20:44:34 MST


Ben Goertzel wrote:
>[Dan wrote]

>>These are interesting examples, but as you point out, we do not yet know
>>the nature of the algorithms or non-digital techniques that an SI
>>may ultimately employ. I think we'll get modestly superior intelligence
>>using "standard" computers, and that this will get us to AGI and another
>>level of SI, but that SI can then pursue alternative strategies such as
>>holographic, quantum, and analog techniques, which have different
>>constraints.
>>
> This is a highly plausible outcome.
>
> My main point is that the alternate outcome, that there are limitations on
> superintelligence posed by physics and mathematics, is also plausible...
> we just don't know... Current science and math DOES suggest ways in which
> such limitations might emerge...
>

Indeed. We typically look at Information-theoretic limits to information
storage capacity and thermodynamic limits to computational density, but
these are useless without a bound on "computational efficiency." Which
we may define as the CPU and memory needed to reach a particular level
of intelligence. Until we know what that curve looks like, we cannot
compute the limits of (digital) intelligence. We do have the human brain
as an existence proof for one point on the curve, and as wonderful as
evolution is, I still have to believe that the human brain is nowhere
near the limit of intelligence.



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT