March 1, 2011

The Scientist: Prof. Bailey Explores Human Creativity and Machines

Print More

Computers are often viewed as technical and complex, but rarely do people associate them as being able to create art. However, the divide of art and computer science is soon fading according to Prof. Graeme Bailey, Director of the Computer Science Masters of Engineering Program and Director of the minor in Computing in the Arts.

“I wish to create a machine that could create art that is as good as humans can create.”  He cheekily explained. But in doing so one must define “what it means to be human in the sense of creativity.”

So far, his research has led to important insights on computers crossing the realms of human creativity. “How can one produce machines that are creative?” he asks. By asking these questions and exploring the creation of algorithms for the development of art, Bailey is also able to shed light on human self-awareness and perception.

Bailey’s preoccupation also lends itself toward broader questions of what humans find fascinating, or captivating in terms of art. He explained how in music for example, humans often seek an implicit sense of drama, an exercise in control and pacing; one that plays with both suddenness and subtlety. This same sense of fluidity applies to all forms of art.

“Visually we can morph one image into another, like time-lapse photography, in imperceptible steps. How can we define the same effect in music? Certainly, deforming one tune into another by sliding each note would have things very out of tune and we’d all notice!” When asked what it meant to have imperceptible change in music he remarked, “Experiments show that the harmonic and rhythmic contexts are important, so by changing different parts of a tune sequentially in discrete jumps, the listener is unaware of each change, yet has arrived somewhere very different. In psychology this is “change-deafness”. The challenge for computers is automating this process.”

It is within these ideas of structure and fluidity that the essence of human creativity lies, according to Bailey. “If we as computer scientists are hoping to build machines which can create effective art, then we must understand what psychologists and artists understand about human perception,” he explained.

In his desire to formalize the connection between arts and the sciences, Bailey, with the help of composer Prof. Steve Stucky, music, and Prof. Carol Krumhansl, psychology, created a Computing in the Arts minor in the spring of 2005.

According to the 2010-2011 Cornell Course Catalog, the concentration gives students the opportunity “to use computers to realize works of art, to study the perception of artistic phenomena, and to think about new, computer-influenced paradigms and metaphors for the experiences of making and appreciating art.” He explained that there has been a long history between the underlying ideas of computer science and the arts, especially music since the 16th century, and in more recent times, art and architecture.

The minor also allows students to combine computer science with different disciplines, or “tracks”, namely music, psychology, dance, film and art. Bailey said he hopes that by offering these tracks as a primary focus, students will develop a respect and excitement for different fields of study and realize their importance and interconnections.

The practice behind computers being able to generate original forms of art can take many forms. One could create algorithms and patterns for the computer to process, devise a grammar for the computer to work with or make the machine responsive to all manner of external stimuli.

For example, a very early exercise in the class involved creating a Shakespearean-inspired sonnet, in which students had to create an algorithm, or program, for the computer to use. This computer program would generate probabilities for the number of times certain words appear and where they appear. The computer would then use those algorithms to generate various sonnets that are completely original, and somewhat random, in nature. It then is refined by encouraging the machine to form its own preferences or biases. The same would apply to computers creating dance choreography or even paintings.

Surely, the idea of a computer being able to “think”, being able to step outside its programmed parameters to create art that is seemingly human and original in content, may come off as being almost science-fiction, explained Bailey. Though, as he said, it begs the question of what it means to think and prompts discussion on the philosophical aspects of such an endeavor. At the practical level, understanding the interactions of art, perception and computers can help create a more instinctive interface for humans and computers. “Can art move us from keyboards and mice to Star Trek? And do we want that?!”

However, Bailey claims that we are approaching a future in which the ability for humans to distinguish human art from novel, computer-created art will become very difficult indeed.

When asked about the success of these machines being able to create convincing new art through algorithms and patterns, Bailey replied “Of course we fail…but it is scary by how little we seem to fail. Perhaps looking at this gap between the best of human and computer generated art can point us to what is essentially human about human creativity.”

Original Author: Laura Comin