Blogs - Capital Advertising

Could a robot start an advertising agency? - Capital Advertising

Written by Capital Advertising | 7 april 2017 10:00:00 Z

Human creativity is a neglected topic in psychology, which seems to focus more on intelligence tests. Yet creativity is an essential part of human intelligence. Therefore, when trying to answer the question of whether computers can think, it also makes sense to ask whether computers can think creatively. Computers and robots have a reputation for excelling in structured problem solving. Could they possibly do more? And thereby, for example, render the advertising agency broody? Creativity is often associated with unexpected connections, can robots make them too?

In 1965, Polish science fiction writer Stanistaw Lem published the story collection Cosmic Adventure. In it, two constructors outdo each other by building insane machines. The highlight was the Elektribald; a device that can write poems. In the process, Lem figured that the robot would have to know the entire universe in order to do so. He suggested back in 1965 that creativity is impossible without the billions of years of universe history flowing through our veins. If we want to make unexpected connections, our brains must be filled with enough subjects between which those unexpected bridges can be built.

What was science fiction is now more or less possible
After all, Google is mapping world history and knowledge for us, and that offers the prospect of computers and robots actually containing some form of creativity. For example, PhD student Michael Cook of the Computational Creativity Group (CCG) at the University of London recently built Angelina: a computer programme that designs its own games. Cook: "For example, Angelina watches Google complete the sentence 'the police hate...' and then finds 'skaters', 'weed' and 'lawyers'. She then concludes herself that, in the game, police should go up against lawyers."

From cuckoo to plane
Several years ago, researchers had three agencies AND a computer come up with an advertisement for an airline that flies punctually according to its flight schedule. The results were presented to the airline's management, which practically unanimously chose an advertisement showing a cuckoo clock.

In the place where the cuckoo normally comes out of the clock, we now saw a plane appear. The ad came from the creative mind of... a computer. So yes, a computer can indeed be creative, and we can assume (or fear) that it is becoming increasingly creative.

After all, a rapidly emerging component in computing is 'machine learning': the ability of computer systems to let themselves 'learn' how to perform a task. IBM fed its Watson computer thousands of recipes for barbecue sauce. Watson analysed on a 'molecular level' which ingredients apparently go well together and made suggestions for a new sauce. Gerard Smit, technology director of IBM Benelux: "If you are able to come up with recipes that a chef wouldn't normally come up with, because they work with a maximum of ten ingredients, I think that's creative."

Computers can do a lot. In some ways even more than us humans. Based on data, they can come up with products that we can call creative: an advertisement, a sauce or a piece of music. What they can hardly do, if at all, is then judge their own piece of work. It was a human who judged which computer ad went with it to the customer, and plenty of tasters do not find Watson's barbecue sauce very tasty. So perhaps that's where the future lies for us 'human experts'. We will judge whether what computers develop from a mash of creative data actually matches what we want.