The Meaning of Life (was Ai vs. general intelligence)

James Rogers jamesr at best.com
Sat Oct 25 01:51:48 PDT 2003


On 10/20/03 8:28 PM, "jbone at place.org" <jbone at place.org> wrote:
>
> The most important "war" humanity is waging isn't against Iraq, or
> terrorism --- it's against entropy.


Not really a correct characterization, so let me expand on it.  And the
issue is more about Kolmogorov complexity than entropy; entropy is just
correlated with Kolmogorov complexity, a symptom if you will.

Humans are not fighting entropy in general, we are fighting Kolmogorov
complexity that is beyond the predictive limits of our computational
machinery.  From the standpoint of intelligent machinery *entropy is our
friend*, as it allows greater effective intelligence on a given
computational substrate.

We lossily reduce entropy (and hence Kolmogorov complexity) because we
cannot increase our intelligence.  For complex processes that are beyond our
capabilities to grok directly, we reduce the effective entropy in our models
until the Kolmogorov complexity *does* fit within the limits of our brains.
It is the old 80/20 rule, except that in this case we have no choice.  We
either coarsen the resolution and increase the noise floor, or accept not
being able to conceive a usable model in our brains.

Intelligent machines will always be fighting Kolmogorov complexity on some
level, but what that level is will be determined by the basic predictive
limits of a given finite state machine.  It is easy to conceive that we will
build machines in the near term that find a level of Kolmogorov complexity
to be acceptable that humans are physically incapable of modeling with our
current computational substrate.

Reducing entropy is a workaround for dealing with high Kolmogorov
complexity.  Without this, we'd be hosed.


Cheers,

-James Rogers
 jamesr at best.com



More information about the FoRK mailing list