Selected Newsgroup Message

Dejanews Thread

From: "Sergio Navega" <>
Subject: Re: AI & Creativity
Date: 10 Dec 1998 00:00:00 GMT
Message-ID: <>
References: <74mlmn$mnm$> <>
X-Notice: should be reported to
X-Mimeole: Produced By Microsoft MimeOLE V4.71.1712.3
X-Trace: 10 Dec 1998 20:14:54 GMT,
Organization: SilWis

William Cumming wrote in message <>...
>As children we aren't creative. We are learning the basics: vision, speech,
>hearing, touch, taste, stuff like that. Once a certain level of confidence is
>reached we become what our mothers (those who have no choice but to love every
>little thing we do) call creative. We start to explore. Some of us take things
>apart and put them back together again. Some of us mix the associations we've
>learned are natural, like a patch of snow on a desert plain, stuff like that.
>Sometimes we learn about music, about how wonderful it sounds, and we catch a
>glimpse of why it sounds wonderful, no longer listening to the sound but
>listening to the ears, and playing with that.

Allow me to disagree?
When we're children we are *very much* creative, more in relative terms
than when we're adults. We think children are less creative just because
they know less and the result of their creations, in terms of what we
adults see as creative, seems to be poor.

But it is exactly the opposite, children are very creative.
The problem is that our education not only adds knowledge to our brain
but also adds *restrictive* patterns of thought. They tell you to think
right from the first time, they tell you to be ashamed of failure, to
be afraid of making experimentations in which you don't know the final
result. How are we supposed to learn without doing experiments and
without letting our mind flow to uncharted territories?

This is the opposite of what we should receive from our teachers.
They should yes, emphasize rigorous forms of thinking, but these forms
must be applied only *after* we manage to "play" with ideas, and this
must be done without silly restrictions.

This affects profoundly our adult's creativity. To recover the characteristic
ingenuity of children we've got to exercise a lot, remembering how we
let our minds delve into dreams without prior judgments and restrictions.
No surprise most of creative ideas came from dreams: it is a moment in
which our minds suspend those "critic teachers" we have inside.

Sergio Navega.

From: "Sergio Navega" <>
Subject: Re: AI & Creativity
Date: 10 Dec 1998 00:00:00 GMT
Message-ID: <>
References: <74mlmn$mnm$> <> <>
X-Notice: should be reported to
X-Mimeole: Produced By Microsoft MimeOLE V4.71.1712.3
X-Trace: 10 Dec 1998 20:14:56 GMT,
Organization: SilWis

Casey Marks wrote in message ...
>In article <>, William Cumming <>
>> NigelJ wrote:
>> > Hi everyone!
>> > I am working on a AI project and the issue is CREATIVITY, to be precise
>> > the question is : could machines be creative?!
>> > I have done some research into this and so far I am still confused because
>> > there are many convincing arguments answering yes and no! I need more
>> > opinions and reasonable explanations to make up my mind.
>> > At the moment I believe they can't because as good as you can program the
>> > machine it is still missing that little "something"!!!!
>> > Help out please - what is it?????
>> Creativity is not, to me, such a great mystery, at least not from my human
>> perspective. From an AI perspective I think we have to learn the basics first,
>> then we have to experiment with it, flirt with it. What the senses tell humans
>> aren't really "rules" but more like "building blocks", and really that's all
>> the world is.
>I would agree, and expand on that.  It seems very true to me that
>creativity cannot occur without a body of experience to draw from.  Having
>this experience allows one to create creatively instead of randomly.  For
>example, if I want to write a love poem, I will use all my romantic
>experiences in deciding which words to use and how to put them together.

Sorry, I disagree from both of you. Creativity cannot be postponed until
"one have the knowledge", and that's because creativity is what produces
valuable knowledge in the first place. Sure, one cannot create *valuable*
things unless drawing from a lot of previous experiences. But the point
I would like to raise here is that we often associate creativity with
*valuable* results. Valuable result is something utterly dependent on
context. If a kid of 5 years old discovers accidentally the principle
of induction by moving a coil near a magnet, he will not have the same
impact as Faraday's original discovery, some centuries ago.

This kid is creative for me. I also associate creativity to obtaining
results that are not always awesome. The good point in seeing creativity
as a matter of degree is that you start analyzing all human cognition
in terms of *creative processes*, with different measures of significance.
This is very useful to lead us to find out how to implement these
mechanisms. Douglas Hofstadter puts this creative mechanism in the
center of the human cognition and this seems to be an important point
to keep in mind when one designs an AI.

>  But something else is missing too.  The machine really needs to have a
>semantic understanding of what it is creating.  Think about trying to
>write a poem in a language that you don't know.

Yes, this seems to be correct. Of course, this poses another age-old
problem: what is semantic understanding? I think the answer to this
question is very close to the answer of how to build that "mechanism
of intelligence". No wonder this is a problem haunting researchers
for more than four decades: a bunch of interconnected concepts, one
needing the other to provide its definition (intelligence, meaning,
creativity, understanding, knowledge).

>  Perhaps this could be rephrased: the machine needs to be able to
>appreciate it's own creation,

That is a very good phrase...

> or at least abstract to a level such that
>you can train it to be creative generally rather than specifically.

...and this is related to what should be, IMHO, the central concern of
any AI researcher: general methods for intelligence. Oh, yes, this is a
very old idea, one that often is thought as being "the illusions of the
first naive researchers". After all, everybody knows that, as Howard
Gardner would say, we have multiple intelligences and this, coupled
with the usual "modularity of mind" argument, should be enough to
justify specific mechanisms for each kind of "processing tasks".
Is this really tenable?

I have serious doubts, and what feeds those
doubts are the discoveries neuroscientists are producing. Although
we have different kinds of neurons and neural organizations in our
brain, most of those differences don't seem to interfere on the
basic functional method of work. This, obviously, would lead to a
common substrate to all human cognitive abilities. Unfortunately,
this is mere speculation. But some results may alter this situation.

In particular, there are some recent fields of investigation (temporal
synchrony and oscillatory temporal coding) that appears to indicate a
substrate to the "neural code", the way groups of neurons work to
"represent" things. An important result have just been published regarding
locust's brains (that may not seem very much, but I'm very excited).
I can't believe I'm excited with locust brains!
How nice to live these days :-)

Sergio Navega.

Back to Menu of Messages               Sergio's Homepage

Any Comments?