[ExI] intelligence and generalization
William Flynn Wallace
foozler83 at gmail.com
Mon Jan 7 15:30:21 UTC 2019
If we accept the definition of intelligence as the ability to adapt to
novel situations, then it seems to me that we can break that down into two
categories; creativity at dealing with the unique characteristics of the
new situation, and the ability to recognize in the new parts of the old -
that is, to see that some already learned things can be applied to the new
situation, which is generalization.
It could very well be that generalization makes a bell curve, with most
people of middling ability and so on.
The highly intelligent have, according to this plan, either a great
creativity at dealing with the new, a great ability to use old learnings,
or both. A person with greater experience would have a bigger database
from which to generalize, but that doesn't guarantee that a person can use
it well. A person with a much smaller database could have more creativity
and do better than a more experienced person. The ratio of
experience/creativity will vary a lot from person to person, I will assume.
I don't have any better understanding of creativity than anyone else, or a
better way of measuring it, but I don't think we can leave all the weight
of the ability to adapt to new situations to generalization. If we could,
then I think that building an intelligent AI would be fairly easy.
How do we teach creativity to an AI when we don't even know what it is or
how to measure it? This is a big challenge.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat