Putting the art in artificial intelligence – can machines be creative?

It has been said that art can only truly be art when it serves no practical purpose. A piece of art; whether it is a painting, a song, a sculpture, a poem, or an endless list of other possible mediums, is means quite easily defined: it is the end product that emerges from a creative endeavour.

But getting people to agree on what constitutes a piece of art or music is, in practice, notoriously difficult. And now another spanner has been thrown in the works; art (or ‘art’, depending on your view) is being produced by machines.

Famed British auction house Christie’s recently announced that for the first time, a machine-created painting will be going under the hammer in London. The painting is the result of Obvious, a French collective, feeding its ‘Generative Adversarial Network’ a collection of 15,000 portraits painted between the 14th and 20th century.

The Portrait of Edmond Belamy, as it is called, is a partially faceless and slightly eerie man. But if you hadn’t been told otherwise, it would be very reasonable to assume it was a human creation.

Artificial intelligence in art and the creative industries has been growing as a concept in recent years. Some projects mess with ideas of human-machine relationships, including one by a New Zealand university lecturer in which drawings were created that means nothing to a person, but are possible for a computer, with its “algorithmic gaze”, to make sense of.

Conversations about artificial intelligence often covers well-trodden ground: the distinction between artificial general intelligence that most people understand, and the more practical applications of machine learning; the ethics of it; and the potential for job loss.

But when we look at creativity, a more fundamental question arises. Some might say that creativity is one of our defining features: it’s fundamental to storytelling, for instance, something which has played a huge role in our evolution and development as a race.

Even if we reached a point where we had artificial general intelligence, it is still possible to imagine that these machines could not muster creativity from within themselves. But a number of projects lay doubt on that. The question is, where is the line? People already use technology to enhance their creativity; at what point does the warmth of the human touch end, and the cold code begin?

Musical AI

Classical-style art and expensive auction houses are fascinating, but they tend not to enter the public consciousness very often. Pop music, on the other hand, is everywhere. Music is one of the oldest forms of creative expression, and it too is seeing the impact of the AI revolution.

Jukedeck is a startup that’s received considerable media attention for its attempts to ‘democratise music.’ Its AI-based product can independently create jingles and pieces of music based on your choice of genre, mood, tempo, instruments and track length. Put it to work and you will have your own piece of music that you can either license or buy the entire copyright for.

The company recognises the role it might play in eliminating people that are currently paid for their musical skills. The company works with creatives, it says, and its CEO Ed Newton-Rex has said that human music “will never die out.”

For now, Newton-Rex acknowledges that there’s a lot of research left to do. But he also cites DeepMind’s famous AlphaGo success, which beat a human player nearly a decade before it was expected to.
DeepMind has itself had a go at letting machines create music, with varying levels of success. It changed the format in which machines were fed music to learn from, and has produced some relatively coherent, if jumbled, jingles.

People Power

Other attempts at creative work have illustrated the nascent stage artificial intelligence is at. Rather than just writing about it, magazine The Economist got a machine to write an article for its science and technology column, with nonsensical results.

Difficult though it is to draw comparisons between the written word and music, there are similarities between the jumbled, confused nature of that writing, and the discordant tunes created by DeepMind. What it illustrates is that so far, without the human touch, machine-made creative efforts might create grammatically correct sentences, or notes played in order, but an end product that ultimately doesn’t mean anything.
It’s clearly imperative to let technology augment human input; amplify and improve it. But as with so many technologies, without a person involved, a creative endeavour is rudderless and destined to fail. And as Forbes writes; creativity is the skill of the future.