Diana McPartlin (diana@asiantravel.com)
Tue, 14 Oct 1997 10:30:39 +0100
Bodvar Skutvik wrote:
> Diana (rightly) believes that an AI can be found out by confronting
> it with a question it isn't programmed for.
Almost. Actually I believe that it is the human ability to make value
judgements that separates us from computers. Humans can make value
judgements on the spot (whadye think of this post so far?), but computers
can't. Consequently the way to suss one out is to present it with some new
phenomena and ask it for its opinion.
As the AI intellect vs intelligence debate is rather large, I want to
limit myself to one aspect for now. In this case I am/was referring
specifically to the Turing Test.
Here is a description of the Turing Test from Roger Penrose's book "The
Emperor's New Mind"
" According to the Turing test, the computer together with some human
volunteer, are both to be hidden from the view of some (perceptive)
interrogator. The interrogator has to try to decide which of the two is
the computer and which is the human being merely by putting probing
questions to each of them. These questions, but more importantly the
answers that she receives, are all transmitted in an impersonal fashion,
say, typed on a keyboard and displayed on a screen. The interrogator is
allowed no information about either party other than that obtained merely
from this question and answer session. The human subject answers the
questions truthfully and tries to persuade her that he is indeed the human
being and that the other subject is the computer; but the computer is
programmed to 'lie' so as to try to convince the interrogator that it,
instead, is the human being. If in the course of a series of such tests
the interrogator is unable to identify the real human subject in any
consistent way, then the computer is deemed to have passed the test."
Lars suggested that you present the computer with a piece of poetry and
ask it for its opinion.
> Lars Marius Garshol wrote:
>
>
> >
> > JUDGE: What does this make you think:
> > Snowfall,
> > unspeakable, infinite
> > loneliness.
> >
> > How do you list an answer to that?
>
I then agreed that the computer would be unable to come up with a
convincing answer. I said:
> All it takes to spot the
> different between a human and a computer is to ask a few questions that
> require judgement, ask them about likes and dislikes, opinions and
> experiences.
And the key point was supposed to be:
> Unless the
> computer has its own sense of value it cannot make value judgements
> about new phenomena. You could always catch it out by presenting it with
> an original idea and asking it for its opinion.
>
In summary my objective with that post was to demonstrate that it would
not be possible to build a computer that would pass the Turing Test. I
believe that a computer would not pass the test because it would not be
able to make value judgements.
It occurred to me that it would be possible to make the computer appear to
be making value judgements by programming it with "opinions" that would be
trotted out in response to predetermined stimuli. That is why I gave the
example of the Japanese virtual idols.
I said:
> I wonder if anyone else in the LS is familiar with the Japanese virtual
> idols - computer generated pop stars. IMHO these computer beings are far
> closer to humans than Big Blue. They might not have such impressive
> processing power but they are programmed to have likes and dislikes,
> histories and opinions. You could have a convincing conversation with
> one of them -- as long as you stuck to questions such as "what's your
> favorite color?" "favorite food?" etc. But ask them something that they
> haven't been programmed to respond to and the illusion falls apart.
Then Bodvar responded:
> But ask a cat or a dog
> to do a task they aren't programmed for and their "illusion" will
> also fall apart. Yes I would say that so will Diana's if confronted
> with a unexpected situation. At least I can become quite moronic if I
> haven't prepared myself.
I don't understand why you are suggesting that a cat is an "illusion". The
computer generated pop idol is an illusion of a human being. What are you
suggesting that the cat and the Diana are illusions of? Ask a cat what it
thinks of the poem and it will be unable to respond. That is in accordance
with cat behaviour so I would conclude that this is in fact a cat. Ask a
computer what it thinks of the poem and it will also be unable to respond.
As the ability to make value judgements about words is normal human
behaviour I would suspect that the computer is not human.
Of course it may require some more probing than this. The computer may
respond by saying "I don't understand", which could conceivably be a human
response too. But the Turing Test allows for probing. If the computer
refuses to understand poetry you could tell it a joke, a story, a song, a
nursery rhyme, a moral dilemma. If you keep on getting blank or
non-committal responses then you would know that that was the computer.
(Remember the human is required to be truthful and to try and persuade the
interrogator that he is the human.)
Could you give me an example of how my "illusion" will fall apart when
confronted with an unexpected situation. I don't understand this part.
Sorry for repeating some of my earlier posts, but I feel I have made a
strong case and proved an interesting point. I have shown that a computer
cannot pass the Turing Test. Perhaps my argument is wrong, but I can't
find anything in Bodvar's comments that refutes it.
Diana
-- post message - mailto:skwok@spark.net.hk unsubscribe/queries - mailto:diana@asiantravel.com homepage - http://www.geocities.com/Athens/Forum/4670
This archive was generated by hypermail 2.0b3 on Thu May 13 1999 - 16:42:05 CEST