Re: ART/NEURAL NETS Creativity

Robert Schrader (rms@cts.com)
Tue, 4 Mar 1997 14:23:39 -0800 (PST)


{ I modified the header on this to differentiate it from sibling threads }

On Thu, 27 Feb 1997, Natasha V. More(f/k/a/Nancie Clark) wrote:

>>>
>>> Creativity is often looked upon as a mysterious magical quality of
>>> artists.
>>> Yet, the word creativity covers a wide range of skills. For example,
>>> Do Bono writes:
>>>
>>> "A few people, a very few people, now know that there is an absolute
>>> mathematical necessity for human creativity because of the way human
>>> perception works as a self-organiziang information system. Such systems
>>> demand creativity and also provocation."

Then I wrote:

>>I'd quibble over words here, though I basically agree. Learning neural
>>nets demand _feedback_. Creativity is an artifact, not a neccesity.

Natasha invited:

> Let's quibble. Please explain in what ways you deem creativity to
> be an artifact.

Learning neural nets learn by a process of trial and error. A new
link is added, or an existing link is weighted differently, or some
similar change is made. Then the net gets feedback and judges it's new
connection accordingly. The whole process takes place with no creativity
involved. A simple exanple is on the shelf at your local software store
in the section labeled 'optical character recognition'.
Thus, having an element of the set 'neural nets' without the characteristic
'creativity', I feel confident in stating that neural nets don't demand
creativity.
But then what is one to think about the fact that creativity does arise
in some neural nets, especially the more complicted ones? So I called
it an artifact.

And QM requested:
> ...define artifact

Off the top of my head: an artifact is something that appears from a
system, but is not essential to its functioning.
I was surprised that some people took this term to be pejorative. It was
simply intended to be discriptive. I readers prefer, I could substitute
the phrase 'emergent behavior' without significantly changing the meaning.

Natasha's pre-quibble post continued:
>>> {snip} Learning how to be creative requires an ability to train thought
>>> patterns to take new routes. {snip}

And my pre-quibble response continued:
>>That's part of it. Letting one's train of thought jump the tracks is
>>merely daydreaming or smoking dope. Anyone can do that. The essential
>>part of creativity is doing something with that meandering thought;
>>getting it to alight on another - perhaps new - track.
>>
>>The extreme of letting one's thoughts take new routes is insanity. ( Many
>>notable artists and musicians are known to strayed over the edge. ) The
>>trick is to let the thoughts loose, but not too loose.

Whereupon QM admonished:
>...your hypothesis that artists in general tend more toward insanity than
>other groups...is not supported by any evidence... Can you tell me how
>you came to this conclusion?

Oops! I got caught making assertions that I can't support. How embarrasing!
I guess I just tapped into our cultural mythology about artists because it
seemed convenient to confirm my ideas. But I know of no study that
demonstrates the converse either. I still suspect that it may be true.
Anybody got any hard data on this, pro or con? [ Please hurry so KA can
cancel her reservations :) ]

I stumbled to the conclusion mainly through my experience - both professional
and recreational - with neural nets on computers. I have seen that allowing
a net to alter its pathways a little is part of its learning process, and
that sometimes this results in useful but unexpected constructions that can
best be labeled 'creative'. And I have further seen that too much altering
can result in behavior that can only be described as 'insane'.

The whole point of this post was to try to get a grip on the mechanism
of creativity, not to offend artists. I'll do some more thinking
and post more on it later.

Robert Schrader rms@cts.com