What definition of intelligence would you like to use?
Legg's definition is perfectly fine for me.
How about the "answering machine" test for intelligence? A machine passes
the
test if people prefer talking to it over talking to a human. For example,
I
prefer to buy airline tickets online rather than talk to a travel agent.
To
pass the answering machine test, I would make the same preference given
only
voice communication, even if I know I won't be put on hold, charged a
higher
price, etc. It does not require passing the Turing test. I may be
perfectly
aware it is a machine. You may substitute instant messages for voice if
you
wish.
What does "being preferred by humans" have to do with (almost any definition
of) intelligence? If you mean that it can solve any problem (i.e. tell a
caller how to reach any goal -- or better yet even, assist them) then, sure,
it works for me. If it's only dealing with a limited domain, like being a
travel agent, then I'd call it a narrow AI. Intelligence is only as good as
your model of the world and what it allows you to do (which is pretty much a
paraphrasing of Legg's definition as far as I'm concerned). And if you're
not using an expandable model, as a calculator is not, then you're not
intelligent.
I claim that a system that can pass this test "understands" my words and
knows
what they mean, even if the words are not grounded in nonverbal
sensorimotor
experience. Its world model will be different than that of a human, but
so
what?
And I'll claim that it doesn't understand a thing UNLESS it has a model of
it's world (which could be text-only for all I care but which has the
behavior necessary for it to accurately answer questions about the real
world) that it is relating your words to. If it has that and can add to
it's world as new things are introduced to it from the "real" world, then
I'm very willing to say that it is intelligent and that it understands it's
world. If not, you just have an unintelligent program.
Its world model will be different than that of a human, but so what?
I've never claimed that an intelligence's world model has to be anything
like that of a human. All I require is that it be effective and expandable.
----- Original Message -----
From: "Matt Mahoney" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Wednesday, May 02, 2007 12:50 PM
Subject: Re: [agi] rule-based NL system
--- Mark Waser <[EMAIL PROTECTED]> wrote:
> OK, how about Legg's definition of universal intelligence as a measure
> of
> how
> a system "understands" its environment?
OK. What purpose do you wish to use Legg's definition for? You
immediately
discard it below . . . .
What definition of intelligence would you like to use?
How about the "answering machine" test for intelligence? A machine passes
the
test if people prefer talking to it over talking to a human. For example,
I
prefer to buy airline tickets online rather than talk to a travel agent.
To
pass the answering machine test, I would make the same preference given
only
voice communication, even if I know I won't be put on hold, charged a
higher
price, etc. It does not require passing the Turing test. I may be
perfectly
aware it is a machine. You may substitute instant messages for voice if
you
wish.
I claim that a system that can pass this test "understands" my words and
knows
what they mean, even if the words are not grounded in nonverbal
sensorimotor
experience. Its world model will be different than that of a human, but
so
what?
-- Matt Mahoney, [EMAIL PROTECTED]
-----
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/?&;
-----
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
