Are computers more creative than humans?

Perhaps -- at least when it comes to developing ad concepts, according to an experiment comparing ideas that a specially programmed computer generated with those dreamed up by humans.To convey that a particular brand of computer is user-friendly, for instance, the computer came up with the notion of having a bouquet of flowers come out of a computer screen. The human concept, on the other hand, called for a lipstick imprint on the screen, as if a woman had kissed it. The computer's idea was judged better by a panel of ordinary people, not trained in advertising.

An essay titled "Creative Sparks" reports the findings in the current issue of the journal Science.

The point of the exercise was not to minimize human creativity, or to compare computers with humans, but to explore the mechanical aspects of creativity, so that what makes humans human can be better understood, said Sorin Solomon, a physics professor at the Hebrew University of Jerusalem. He co-authored the report with marketing colleagues Jacob Goldenberg and David Mazursky.

"Deep in ourselves, we hope that humans are really something special. I still believe we are," said Solomon, a particle physicist by training, in a telephone interview Friday.

And the way Solomon and his colleagues are trying to prove that is through a process of elimination, by seeing what aspects of human endeavor can be performed just as well by computers and which can't.

"All these efforts of reductionism of mine are not because I really believe that everything can be reduced to physics, to mechanics. It's in order to take away what can be reduced, and leave in all the splendor the things that cannot be reduced," Solomon said.

As a result of his work as a particle physicist, studying the natural rules that govern microscopic particles, Solomon became intrigued with trying to find similar rules that might explain human activities, if they are broken down into small enough parts to be studied. He calls this effort "microscopic representation."

So for the advertising experiment, he and his colleagues worked with advertising professionals and analyzed award-winning ads (designed by humans), to come up with discrete formulas to explain what makes a great ad.

These formulas, or algorithms, were then plugged into a computer, which then came up with ad concepts of its own.

One of the most common formulas for an award-winning ad, for instance, turns out to be what the researchers call the "pictorial analogy template," in which a symbol, like a bouquet of flowers, is put on a product, like a computer.

The humans who came up with ad ideas for the experiment were highly educated people like professors and scientists, but not necessarily advertising professionals.

Then the ideas from both computers and humans were given to artists to turn into advertisements. The judges compared those ads with each other, as well as to some of the human-made award-winning ads that had served as the basis for some of the computer formulas.

View Comments

All hope is not lost for humans, though. Even though the computer-generated ideas beat out those dreamed up by humans in the experiment, the judges found the best ads to actually be the award-winning ones (created by humans) that served as the basis of the computer-formulated ones.

And advertising officials in Boston scoffed at the notion that a computer could ever take the place of the most talented professionals.

"I'd put our best creative minds against computers any time," said Bethany Kendall, president of the 7,000-member Advertising Club of Greater Boston.

Alan Holliday, one of the founders of ad agency Hill Holliday Connors Cosmopulos and now an assistant professor of mass communications at Boston University, said, "There's an awful lot of dumb, boring advertising that a computer could probably do just as well, but the really great advertising takes in a lot of factors that I don't think a computer can."

Join the Conversation
Looking for comments?
Find comments in their new home! Click the buttons at the top or within the article to view them — or use the button below for quick access.