[ExI] The nature of intelligence (was: predictive neurons?)

Spencer Campbell lacertilian at gmail.com
Sun Jan 31 02:52:06 UTC 2010


Mike Dougherty <msd001 at gmail.com>:
> I think calling intelligence a feature or property of the environment
> makes it more like the situation where gravity is the deformation of
> spacetime topology (or vice-versa) - so how might we improve
> intelligence?  Alter the environment?  Is it a recursive process?

That's precisely the principle behind formal education, so... inconclusive.

My definition of intelligence is relative to the possibly-intelligent
object in question, and presupposes that the object has a definite
purpose:

Intelligence is inversely proportional to the average time required
before the object achieves its purpose, assuming that it encounters
every possible obstacle on the way and that none of those obstacles
are insurmountable to the object.

I'll note that "obstacle" is a hypernym for "problem" in my mind. This
implies that "problem-solving ability" is a less complete, but still
correct, interpretation of intelligence.

Things get a little more complicated for humans and other animals,
whose purposes tend to change rather frequently. The easiest solution
is to say that my measured intelligence changes depending on what goal
I have in mind, which seems consistent with reality: I'm brilliant
when it comes to eating three meals a day, yet functionally retarded
when it comes to navigating the public transit system. Behavior may be
either very stupid or very ingenious, depending on what you assume
it's meant to accomplish.

A corollary: unless the number of possible teleological states of a
given system is finite, it is not possible to measure the absolute
intelligence of that system. My only problem with IQ tests is that
they don't specify what they're measuring, save for "the ability to
score well on IQ tests".



More information about the extropy-chat mailing list