We can now summarize some
conclusions from our shared experiences.
In Part 1, we optimized a ballistic missile to achieve a given range for
a minimum lift off weight. This was a straight forward optimization made
possible by the existence of a computer program which could evaluate
each proposed design.
In Part 2, randomness helped to find magic squares of any size. We used
a swapping technique to modify each proposed square until we found one
that met our criteria of having each row, column, and diagonal add to
the same total. We also had to “shake things up” from time to time, as
we would get trapped in some path that would not lead directly to an
acceptable solution.
In Part 3, in the Black Box problem, we found that learning was
possible, but also that overlearning could lead to a loss of randomness
and then a loss of creativity as a result.
In Part 4, we played poker on the computer, and used an evolutionary
approach to find the best strategies. In a surprising result, we
observed a “strategy bifurcation” where two distinctly different
strategies emerged, both able to survive in the same environment.
In Part 5, using the classic “Traveling Salesman Problem,” we learned
that the human has unique capabilities to process visual information. We
can process image data as images, without converting them into digital
form, and use these capabilities to create and evaluate ideas. It is
called imagination.
In Part 6, Neural Networks, a statistical optimization technique, looks
for patterns in data. It has been used successfully to create models of
human judgment, in cases where the problems can be stated in some
digital form. In one case, this meant using Fourier Transforms to
digitize sound.
In Part 7, we considered Genetic Algorithms and Evolutionary Computing,
which are still in early stages of development.
So here we are. What have our toy problems told us? Remember, we are
trying to answer the question: “can computers be creative?”
The answer is, within limits, yes. The computer is effective in
searching through millions, billions or more of possible potential
solutions to problems if they can be stated in digital form. The model
that the computer uses looks more like evolution than does the use of
human imagination.
The common denominator in everything that we tried revolved around using
some method of evaluation. Here is where the computer’s capability is
most critical. The need for evaluation goes beyond just deciding whether
something is good or bad. It is critical to know when something is
better. The creative process follows a path; new ideas derive from old
ones. But the connection from something to something better does not
need to be, nor should it be smooth or continuous. It is necessary, at
some point, to know, or to guess, when we have gone as far as one
direction will take us, and make a leap to a new one.
So, it is vital to have some means by which we can determine, with some
probability, that changes are moving us in a better direction. We do not
say “the right direction” because there may not be one, but some
directions may lead to better results, and we need to be able to measure
that. Nor is there always a best solution, just as there no perfect
plant or animal in the kingdom of life.
In the biological world, better is measured by survival and
reproduction. The political world is similar. In the world of worldly
goods, better is often measured by the buyers in the market place. In
engineering, as in the design of a ballistic missile, an airplane wing
or an electrical circuit, we can use the laws of physics for our
evaluation model. In situations calling for a creative strategy, it may
be possible to create games that do a good job of replicating a real
world situation.
In biochemistry, the idea of creating a model of the protein folding
process, and then creating a model of that molecule after it has folded,
is daunting both in its scope and in its implications.
Art, literature, and music, are evaluated by survival in the world of
viewers, readers, listeners, (and also buyers). The computer does not
make aesthetic judgments—that is, we have found no instances where that
is the case. If we could really model the human mind, we could create
aesthetics, such as, music or art or literature that would be considered
good, or beautiful. But we can’t. We are limited to finding rules that
seem to have worked for things that we have already heard, or seen and
make evaluations based on the rules. Today, the creative power of the
computer is limited to those situations where evaluations can be
measured. That is a serious limitation, but not a destroying one.
So we can draw a somewhat fuzzy line that defines the limits of the
computer’s creative capabilities. Its powers are limited to areas where
results can be measured. It is not true, or at least not a fair
statement, to say that creativity is something which a computer can not
do. It is fair to say that the computer has some powers of creativity
not possessed by the human, and the human has powers that are different
and not possessed by the computer. Imagination is a very powerful
creative force, but it is not the only one. Perhaps a more important
force is selection.
An interesting question that remains is whether a computer will be
designed or invented that can process images as images. It would scan,
store, evaluate and create images based on image content, not digitized.
Now there is a problem that requires creativity, doesn’t it?
|