Computers and Creativity
Since the beginning of the computer revolution people have been asking:
That's a question best left to the philosophers. It probably all depends on what we mean by computer and what we mean by creativity - and, in the end, what we mean by human. Fascinating subjects but beyond the scope of this website.
A more useful question might be:
The answer to that is a definite yes.
Generating Ideas
If creativity is about generating, recombining and sifting ideas then computers can definitely help. You can buy many software packages to help with all aspects of the creative process from ideas capture through brainstorming and organisation. But can we go further still?One of the most interesting advances in software technology and AI in recent years has been so-called genetic algorithms (GA) that borrow from the ideas of evolution. Algorthims are "mutated" and "recombined" to form new possibilities in the hope of arriving at a solution that the original programmer would never have considered. That sounds like an excellent basis for computer aided creativity.
Genetic Algorithms
Use of Genetic Algorithms consists of three main elements: mutation, selection and recombination. (There is also a one-off "initialisation" phase). In a "pure" GA all of these will be handled programmatically.An example of a situation where pure GA can work well is developing a program to play a strategy game. Here the goal is clear - winning is well defined. The program can generate, mutate and recombine many thousands of possible algorithms quickly and test them by playing a game either against itself or against an established computer program.
Such a pure approach fails in the case of creativity. There is no clear evaluation function for art! How is the program to decide which of the randomly generated abstract designs is "best"? Human interaction is clearly needed in the selection phase.
Such an approach - human selection, everything else programmatic - works in theory. Unfortunately it can take a long time and human fatigue often sets in. After you've rated the first hundred blurry abstracts they all begin to look the same!
Human Based Genetic Algorithms (HBGA)
This problem can alleviated by also allowing human interaction in the recombination process. Instead of just saying "That's best" a person can add "... because of the symmetry" etc. Or a human can say "I'd like to see what happens if we combine this one with that one". Such a process takes more time per iteration but can potentially reach a better solution in less generations.Use of HBGA is essentially about opening up any and all of the elements of the GA process to human interaction as and when appropriate. It also allows for specialisation - the people doing the selection might not be the same as those suggesting the recombinations.
In some systems it is even possible for humans to have input directly with ideas, possibly at the initialisation stage. Different people can contribute their own ideas to different parts of the system at different times, enabling knowledge exchange. Thus the use of HBGA can be seen less as a computer program and more as a collaboration between computers and humans.
Returning to the beginning of this article, a good question in the case of HBGA might be:
I for one like to think so.