From: David Clark (clarkd@rccconsulting.com)
Date: Wed Jan 18 2006 - 11:49:18 MST
If the following test were given to humans, who most humans would consider
intelligent (not necessarily extremely intelligent), how many humans would
pass?
I would NOT pass the test you propose so therefore I would NOT be
intelligent according to your test? Is this reasonable?
I would propose another kind of test. Propose some questions that an
average human could answer correctly a large percentage of the time.
Exclude all questions that would be specific to human beings or our physical
experience. Exclude all questions that require a context that only humans
would be knowledgeable of. (The Turing test is obviously a very human biased
test of intelligence if that was the purpose of it being proposed.) Make
sure that all the questions, if answered incorrectly, show a lack of
intelligence as the reason for the incorrect answer rather than a lack of
knowledge of our culture or our creations (including Math). Many minorities
get low IQ scores in the US because of the cultural bias that exists in most
IQ tests. They argue that IQ tests don't accurately reflect their
intelligence because IQ tests test knowledge (not intelligence) which can be
different among different groups. There is also the problem of having
different kinds of intelligence. (Musical IQ, Reasoning IQ, Math IQ etc)
If a program (software, hardware or some combination) then successfully
answers these questions then we should assume that it has some intelligence
similar to what some humans have. Why should human level intelligence
assume University or greater knowledge in an esoteric field like physics?
Most people that walk the streets don't have this kind of knowledge so why
should that be a criteria for intelligence?
I agree that intelligence doesn't imply consciousness.
David Clark
----- Original Message -----
From: "Richard Loosemore" <rpwl@lightlink.com>
To: <sl4@sl4.org>
Sent: Wednesday, January 18, 2006 6:28 AM
Subject: Re: 3 "Real" Conscious Machines
> What I am interested in is WHETHER YOUR SYSTEM IS INTELLIGENT. I would
> like to know if you think it is going to talk (with a voice?) in such a
> way that it clearly knows what it is talking about.
>
> To ensure that you don't cheat, I propose the following test.
>
> I will ask your machine several questions. These will be designed to
> stretch it to the limits of its cognitive powers. The first set of
> questions I will give you now, so your machine has time to prepare.
> Others must be answered in real time, and they will be of a similar
> character as these.
>
> 1) Give an introduction to Heim's theory of quantum gravity, in
> sufficient detail to allow a Physics graduate to understand it.
>
> 2) You are living in a hovel in northern England some time in the Dark
> Ages, and when you open your door and look outside one morning, you hear
> a fragment of a T.S. Eliot poem come to you (suitably translated into
> your vernacular) through a time warp. You smile with recognition.
> Explain why you smiled, what you heard and what the context is.
>
> 3) Might Igor Stravinsky have had any opinion on the subject of motor
> scooters? Why?
>
> When the remaining questions are asked in real time, your machine will
> be required to respond in any human language that is currently spoken by
> more than 1 million people. I will specify which language, but I will
> give only one second's notice of the language.
>
> Regards
>
>
> Richard Loosemore.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:55 MDT