AI: Creativity

David Musick (
Thu, 23 Jan 97 06:34:45 UT

Here's an excerpt from a message I sent to my father (who lives in
Switzerland). We are working on a science fiction book together, and right
now, we are basically brainstorming and discussing what we want to do with the
book. We are planning to have a very powerful artificially intelligent being
in the book, and we were discussing whether it would be creative, or would
depend on humans to come up with ideas for it. He said it would be totally
rational and non-creative, and I said that if it was so rational and had so
much knowledge, surely it would know about all the ways that have been
developed to make a computer creative (such as genetic algorithms), and would
use those methods and be creative itself. He countered that evolution and
genetic algorithms were not really creative since they were only fiddling
around with what was already *there*. He takes a sort of mystical slant on
things (although he is quite rational and highly intelligent; he's just had
some very strange experiences), so he is viewing creativity as something
beyond the physical world. My response follows:

My belief is that creativity is a physical process, and that our creative mind
is similar in many respects to our thinking mind. I also believe that the
processes of creativity (and also of consciousness), being physical processes,
can be carried out on a machine which humans build.

I believe creativity is inherent in the processes of variation and selection.
The variation is where newness comes in, and the selection is where only the
"good" patterns are retained. I believe a sort of variation and selection
happens in our minds during all of our thought processes, and I believe that
the state of mind that we call "creative" is one in which more radical
variation is happening.

You said that just fiddling around with what's already there is not creative,
but is more like what intellectual thinking is. Creativity, according to you,
is when something entirely *new* is created. When I consider the various
creative ideas that people have come up with, and when I consider my
experiences with creativity, I realize that what was created was very related
to what had been worked on before. In other words, nothing *completely* new
was created; in every case of creativity I am aware of, the creation was
basically a rearrangement or revision of old ideas. Do you have any examples
where someone suddenly started getting creative ideas that were completely
unrelated to things they had thought of or experienced previously? Even if
there *are* /some/ examples, nearly all examples of human creativity are
basically modifications of already held ideas. Even the most creative of us
aren't all *that* original. We're all borrowing each other's ideas and our
own old ideas.

Of course, when one is *experiencing* the creative state, one is basically
open to anything, and the creative ideas just seem to come pouring in. During
the experience, there is no awareness that these ideas are simply
modifications of ideas one already had. The new ideas seem so fresh and
creative, but if they are analyzed later, one can see how various past
experiences and thinking helped shape the mind to be open to the creative
idea, and how nearly all of the elements in the idea are ones that one already
had in one's mind, and very little of the idea is actually *new*, except for
the particular *arrangement* of the elements of the idea.

So, it is certainly good advice to tell someone to relax their mind and
suspend thinking and critical judgement when they wish to enter a creative
state. This allows more radical ideas than usual to form before they are
dismissed by critical judgement too early. It's a state that allows much
greater *variation* of thought than usual. The normal thinking state allows
some variation, just not as much as the creative state, and I believe that is
the only real difference between the two states. To the person experiencing
these different states, they can seem quite different, but to a cognitive
scientist, I believe they would appear to be *basically* the same thing, with
one state simply allowing for more variation and less critical selection than
the other.

This type of behaviour is often programmed into computer software which use
genetic algorithms to find optimal solutions to problems. If the computer
doesn't vary the parameters enough, it runs the risk of finding a solution
which is as good as it can find by making only small variations. Imagine a
bumpy landscape, with bumps of all sizes. The computer's job is to find the
highest peak of the highest bump, or at least on a bump *close* in height to
the highest. If the computer is on a fairly small bump, and it keeps making
small steps up the bump, it will get to the top of that bump and stop,
thinking it had found the best solution. But if it makes large enough
variations, it will often end up on completely different bumps, which it can
explore with small steps, to find the top. If it does enough of these big
jumps, followed by little steps, it will likely find the optimal solution, the
highest peak of the highest bump, or at least something close enough to the
best not to matter too much.

The variations going on in the computer software is a process of creativity,
and the selection going on, of the best solutions, makes it rational
creativity. Creativity that works.

Of course, as you say, someone has to tell the computer what problem to work
on, and once they describe the problem sufficiently to the computer and give
it all the criteria for what type of solution they are looking for, the
computer will be off and searching for a creative solution to that problem.
Things are done this way because we haven't built computers yet which have
their own sets of goals and create their own problems to solve. I see no
reason, though, why it is not *possible* for a computer to be made which can
pursue its own goals and make up its own problems to solve. If there was a
computer doing this, would you say that it was not really creative? If so,
why is it creative when humans come up with their own goals and pursue them,
finding ingenious solutions to their problems?

You ask for one creative idea a computer has come up with. I don't know all
the details of this particular example, but essentially, there was some sort
of game that people played on the computer, a kind of space game with trading
going on and battles between groups and so forth. The computer was given the
rules of the game and so forth, and played by itself for a while, tyring out
all kinds of different strategies, evolving them as it went, and eventually it
developed a very elegant and clever strategy that blew away every human
player; it was unbeatable.

Well, of course, *that's* not creative! We understand exactly *how* the
computer went about finding that strategy. It was completely rational and
logical, so it must not be creative. If we understand it, then it must not be
creative! Creativity is *beyond* understanding!

I don't really think so. I think we *can* understand creativity, and I think
we can program software that is creative once we do understand it. I don't
think it's as mysterious as some *want* it to be.

- David Musick

-- All progress is a result of variation and selection and spontaneous
organization. --