Kaj Sotala wrote:
On 4/26/07, Richard Loosemore <[EMAIL PROTECTED]> wrote:
Can you point to an objective definition that is clear about which
things are more intelligent than others, and which does not accidentally
include things that manifestly conflict with the commonsense definition
(by false negatives or false positives)?
(Disclaimer: I do not claim to know the sort of maths that Ben and
Hutter and others have used in defining intelligence. I'm fully aware
that I'm dabbling in areas that I have little education in, and might
be making a complete fool of myself. Nonetheless...)
In thinking about human rationality, I've found it useful to consider
intelligence and goals as two separate things. Goals are what you want
to do, intelligence is how you achieve them. An intelligent system is
one that can be given a wide variety of goals, which it will then
figure out how to achieve.
Therefore, I suggest the following amendment of Ben's formulation of
intelligence:
The intelligence of a system is a function of the amount of different
arbitrary goals (functions that the system maximizes as it changes
over time) it can carry out and the degree by which it can succeed in
those different goals (how much it manages to maximize the functions
in question) in different environments as compared to other systems.
This would eliminate a thermostat from being an intelligent system,
since a thermostat only carries out one goal. Humans would be
classified as relatively intelligent, since they can be given a wide
variety of goals to achieve. It also has the benefit of assigning
narrow-AI systems a very low intelligence, which is what we want it to
do.
Kaj,
Please let me explain my only problem with this definition, and with the
many other definitions that keep being offered.
There are literally hundreds of such definitions (we have already had,
what, a dozen? just in this thread alone), but they are all INFORMAL,
which means that when you look into them carefully (push them to their
limits by trying to apply them to weird cases) their basic terms turn
out to require a judgment call by a person (i.e. an intelligence)
applying the definition. I mean, sooner or later someone says, "Well,
maybe I can't exactly say what a 'goal' is, but *I* can tell the cases
that are real goal seeking from the cases that are just dumb parameter
optimization." Either that, or every time a weird case is brought up,
another refinement needs to be added to the 'definition' until the
so-called definition looks like an infinitely expanding list of features.
But all these attempts to squeeze intelligence into the straightjacket
of a definition start to look ridiculous if they always end up that way.
Why are we not capable of jumping up a level, observing our own
behavior when we try to generate these definitions and realizing that
there is a pattern here, and that ALL these informal definitions of
intelligence eventually break when they are pushed too far?
So long as we keep generating useless definitions that have no function
except entertainment, we are behaving like the Sphex wasp that
Hofstadter described in GEB: it takes a paralysed grub to the edge of
its hole, then checks inside to see if the egg is still there, then puts
the grub in the hole...... unless a human observer pulls the grub away
while it is in the hole, in which case it brings the grub back to the
edge of the hole again and pops inside for a quick look at the egg.....
and again, and again and ..... it will repeat the same cycle without
EVER realizing that something fishy is going on.
If all definitions of intelligence end up with a judgment call, we might
as well throw away all of them and just say "Intelligence is what a
human would consider intelligent". That is useless, of course, but no
less useless than longer definitins that say the same thing in the end.
The only situation where definitions becomes a serious issue is when
some people claim that they really can find a completely objective
definition that does not involve a judgment call: that is the ONLY
THING I have been attacking: in my opinion, those so-called objective
definitions sacrifice accuracy for the sake of objectivity. In other
words, they do not really define intelligence because they allow all
kinds of anomalous things to be classed as intelligent, or they are
incapable of being evaluated in the real world. Some of the people who
do this have a wider agenda, as I said before.
So: it certainly is fun to play around with definitions (and if it is
fun you are after, I ghave no problem), but it is frustrating after you
have watched people argue about the first 200 different definitions,
none of which seem capable of being phrased without implicit
subjectivity.... and still nobody notices the fact.
Me, I am just impatient. Want to get things done, not argue in circles.
Richard Loosemore.
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?member_id=231415&user_secret=fabd7936
