Can AI be creative?

The artwork was created using a machine learning algorithm created by a French student collective called Obvious. This used a type of AI called GAN (generative adversarial network), which is two algorithms that work together: one creating ideas while the other judges them in a loop that continues until both agree that what the machine has generated is acceptable.

While certainly a watershed moment in dollar terms, this wasn't the first time that machine algorithms have been used to generate art. Harold Cohen's 20+-year AARON project showed machines can generate pleasing images of stylistic integrity.

More recently, researchers from Facebook and Rutger University used GAN to create art, while apps that use machine intelligence to adjust photos to make them look like works of art already exist. In 2016, IBM's Watson AI was even used to create a movie trailer for 20th Century Fox's horror flick, Morgan.

In 2016 when Go champion Lee Sodol was defeated by the AlphaGo AI, he said: "It made me question human creativity."

The field of machine intelligence and computer imaging is evolving rapidly. Big tech is immersed in using AI to augment human-made photography, while the digital health sector sees multiple illustrations of innovative technologies capable of doing everything from assessing cancer risk to measuring blood loss.

What's at stake?

It is important to consider how AI builds these creations. In most cases, the machines are programmed using some model of neural network learning combined with huge data sets – thousands of artistic images may be scanned to give the machines a set of reference objects on which to base what they make. Control of the data used may impact the creations produced – and the fundamental philosophical ideas they represent.

Most of the creative AI examples mentioned so far reflect the data used to train them. This means that while their creative expression may seem authentic, the actual content is limited by earlier choices taken when the AI is built. This makes it possible to control the creative process by defining what is and is not artistic. Germany's infamous Degenerate Art Exhibition in 1937 shows the danger when entire schools of artistic expression are outlawed. How can AI be trained to create art free of such limitations?

"Many studies have shown that if you just train a machine learning system from data that is randomly selected from the Web, you end up with a system that is racist, misogynist, and sexist, and that's just a mirror to our society," said Apple's Senior Director of Machine Learning and AI, Carlos Guestrin at Geekwire's Cloud Tech Summit in 2018.

The impact of data on results is well illustrated by the infamous Shirley cards used by photo processing companies in the '60's as reference items when developing images. These cards depicted a very light-skinned model, which meant images of anyone with darker skin tones tended to be poorly exposed during the printing process.

"It's not enough just to think about the data that we use but also how that data reflects our culture and values that we aspire to," Guestrin observed.

Augmented reality

AI is already seeing use in the news room: Associated Press uses AI to parse and write reports based on company earnings releases and to create sports reports.

The story templates are built by human editors, and the machine-written stories are verified by humans before publication, but the idea is that humans don't have to write these formulaic stories, enabling talented journalists to focus on research, interviews and more.

"It's about the augmentation of creativity. In the end, the human really is the one being creative, and it's more about how can you get better efficiencies," said John Smith, Manager of Multimedia and Vision at IBM Research.

Justin Hendrix, Executive Director of NYC Media Lab told PR Week: "Journalists' job functions will change. People will spend their time training computers to do things that formerly only humans could, like sifting through documents, looking at data sets, or reviewing footage."

Such creative augmentation is perhaps where we are most likely to see AI impact the creative industries, at least in the next few years. AI may help find the best images captured during news events, could assist in identifying usable film footage or to assess audience reaction to creative works in order to identify which artists are most likely to become successful.

AI may also generate interesting new forms of creativity. The computer science lab at Toronto University created an AI that makes songs in response to scanned images. "Instead of buying a karaoke machine with certain tracks on it, you can create your own karaoke at home by throwing in some interesting photos and inviting the machine to generate music for you," said Toronto’s Sanja Fidler, Associate Professor in Machine Learning and Computer Vision.

What is art?

"It's easy for AI to come up with something novel just randomly. But it's very hard to come up with something that is novel and unexpected and useful," says IBM's Smith.

Can machines be creative? To what extent can the separation between human emotions and societal values and the way machine intelligence functions be used to define the difference between one created item and another? At what point does the person who creates the code that drives an intelligent machine become the artist, and at what point does the creative expression become art?

These aren't questions I feel qualified to answer, but it seems inevitable that the next decade will see increasing use of AI across the creative industries. Their contribution should leave cutting-edge creatives free of many mundane tasks in order that they can focus their soft skills on creating art that defines what it is to be human in an age of increasingly intelligent machines.

Discover how Orange can help you turn your data into intelligence.