Hmmm...
About the measurement of general intelligence in AGI's ...
I would tend to advocate a "vectorial intelligence" approach
I tend to think that quantitatively or otherwise precisely defining
and measuring "general intelligence" -- as a single number -- is a bit
of a conceptual and pragmatic dead end.
Certainly, it is useful to (quantitatively or qualitatively) evaluate
the performance of an AGI on various tasks in various domains ...
But, combining "task performance scores" into a single overall
"intelligence metric" can be done in so many different ways, it
becomes a largely arbitrary exercise IMO.
I would place a bit more faith in a "multiple intelligences" approach,
wherein cognitive-focus-specific intelligences are defined precisely
and measured, but one doesn't focus on combining them into a single
score.
For instance, one might measure: pure-mathematics intelligence,
applied-mathematics intelligence, music-composition intelligence,
rhetoric intelligence, ethical intelligence, etc.
Defining focus-specific intelligences like this in a precise and
measurable way seems difficult but probably tractable. The value of
combining such measures into an overall general intelligence measure
seems dubious.
One might also define a "domain-transcending intelligence", measured
by supplying a system with tasks involving learning how to solve
problems in totally new areas it has never seen before. This would be
very hard to measure but perhaps not impossible.
However, in my view, this "domain-transcending intelligence" -- though
perhaps the most critical part of "general intelligence" -- should
still be considered as one among many components of general
intelligence, together with a variety of focus-specific intelligences
as defined above. Domain-transcending intelligence is "just" one
component of the multiple-intelligence vector.
-- Ben
-------
To unsubscribe, change your address, or temporarily deactivate your subscription,
please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
