From: Petter Wingren-Rasmussen (petterwr@gmail.com)
Date: Fri Feb 27 2009 - 01:09:35 MST
I think dividing intelligence into different properties makes it easier for
different research teams to compare their results. ("Wow! That AI is both
quicker and more enduring than ours!" "Yeah, but it makes more mistakes
too.")
I'm not saying that Eysenck's model is very good for our purpose, but its a
step in the right direction imo.
On 2/26/09, Johnicholas Hines <johnicholas.hines@gmail.com> wrote:
> We need ideas of "intelligence" that can be applied to rocks,
> bacteria, single-celled creatures, plants, animals, calculators,
> single humans, the human race as a whole, weakly-superintelligent
> emulated humans, computers with many more MIPS than the human race as
> a whole and novel strongly superintelligent entities.
>
> Evolutionary "fitness" might be a candidate measure of intelligence,
> for all that it's uncomputable.
The IQ-tests themself are meaningless for anything except human beings and
AIs trying to pass a turing test.
I think the model can be used on anything though.
Take the biosphere as an example and view it as one unit. Let's say the
mission is to produce as intelligent and adaptable species as possible:
1. Mental quickness: as far as we know every decision is random, so IŽd say
very low.
2. Carefulness: most organisms are tested against a lot of others, so pretty
high.
3. Endurance: Been at it for billions of years, extremely high.
End result: the biosphere has developed more advanced organisms/units than
we are able to do and is more intelligent, given enough time.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:04 MDT