Key takeaways

  • Artificial intelligence experiments are asking whether robots can truly create art.
  • AI will allow people in the arts to work with better efficiency, spending more time on being creative and less on low-level tasks.
  • The next wave of artificially intelligent robots are making their own creative decisions and have even produced movie trailers and professional photography.

“Creativity is intelligence having fun.” – Albert Einstein

These days, artificial intelligence (AI) is being applied to nearly every aspect of our working lives. Creatives, however, have been mostly left alone because, after all, a robot can’t create art, can it?

As it turns out, there’s more AI experimentation going on in artistic fields than you’d think, and it’s on its way to making the lives of marketers, designers, journalists and advertising people easier.

From the automation of monotonous everyday tasks to true autonomy, where robots may one day create their own artworks, AI is changing what it means to be creative and challenging concepts we thought were intrinsically human.

What is

Defining what we mean by ‘creativity’ is a tall order. Sternberg and Lubart, academics who wrote a handbook on creativity¹, define it as something that’s novel or unexpected and useful or appropriate to the task. It doesn’t follow a set of steps to solve a problem; it is the opposite to routine.

As artificial intelligence is a program following a set of rules, you could say that a human-designed machine can’t be creative. At the same time, it could be argued that humans are essentially the same, following a set of emotional and experience-driven neural pathways. As Steve Jobs put it,

 “Creativity is just connecting things. When you ask creative people how they did something, they feel a little guilty because they didn’t really do it, they just saw something [and] were able to connect experiences they’ve had and synthesise new things.”

Suffice it to say, defining creativity is a complicated task, as it is with creativity’s end product – art.

It could be said that there are various levels of creativity too, from a non-linear solution to a task, to a Jackson Pollock painting.

PwC’s AI report, Sizing the Prize, outlines four types of AI, from automated and assisted intelligence to augmented and autonomous intelligence. These all offer different levels of application in creative tasks.

Automating intelligence:
Automated art

Robotic process automation (RPA) is already being applied to many workplaces and the creative industry won’t be left out. In business, the automation of manual or low-level cognitive tasks, sometimes combined with other levels of AI, is beneficial as it frees up time and personnel to work on higher level tasks of greater value.

Anyone who’s worked in a creative agency will understand the appeal of having more time.

For example, Google awarded a US$806,000 grant to the UK’s Press Association and Urbs Media as part of its Digital News Initiative (which shows support for European news producers, promoting quality online journalism) to run an AI news service which will write up to 30,000 local articles a month.

While journalists will help find candidate stories, and create templates the AI will use, the amount of content and filling of those stories with local information will be done by machine. While journalists remain understandably sceptical about this development, there’s little doubt that it would take a substantial team of human writers to produce content at such scale.

Without having to write the more repetitive pieces, or spend time sourcing local data for articles, journalists will, in theory, be able to concentrate on content where they really add value.

Assisted intelligence:
Can I hold that for you?

Enter assisted intelligence. In this instance, AI is used not just to automate, but also to help humans perform their tasks faster and more easily.

Adobe’s new AI-driven framework ‘Sensei’ automates processes that designers and photographers would normally need time to set up. For example, Sensei “recognises faces, placing landmarks on eyebrows and lips so you can change expressions with one click.” Adobe claims this assistance will mean tasks that used to take longer to complete will be done in just seconds².

While the end creative is controlled by the human in this scenario, the task is now easier and quicker. Labour-intensive jobs of the past become much less painful.

Augmented intelligence:
It might work better this way

More than making things easier or faster, augmented intelligence goes a step further towards true creativity, adding data and ‘augmenting’ the human ability.

In 2015, creative agency M&C Saatchi used AI and camera-enabled digital posters in London to test audience responses to adverts³. Based on people’s reactions (happy, sad, etc) to the copy, images, fonts and colours, the poster would then change to a more pleasing design.

The experiment highlights ways in which AI could help marketers make creative decisions, by virtue of its ability to process information (and test it) at a speed and intricacy that would be prohibitive to humans.

Augmented intelligence can also help creative professionals in terms of audience analytics, making use of data to deliver useful insights that humans have neither the time nor processing power to access easily.

Autonomous intelligence:
I did it my way

Ad Agency Mcann Erickson Japan set out to create a robot creative director, capable of producing a television ad. The AI, named AI-CD ß, was given a brief of what would be required, such as visuals, music and messaging, and access to a database of deconstructed award-winning television ads.

The robot director then searched through the database and delivered its selections to a team of humans who cut together the final ad. When pitted against McCann Erickson’s human creative director, the human-created advert won – but not by a very big margin4.

In a similar experiment, IBM’s AI platform Watson created a trailer for the horror movie Morgan. Watson was given a database of horror movie trailers, then ‘watched’ the film, analysing visuals, sound and composition and handed over the most suitable pieces for the video team. A selection process that could have taken over 30 days was complete in 24 hours5.

These examples show a certain level of intelligent decision-making on the part of the machines. While augmenting the human workflow to make it quicker and easier, they’re also selecting the elements that go towards the aesthetic of the final product and thus, you could argue, being creative.

What about
true autonomy?

When asking if a robot can be creative, what most of us really mean is, can a robot be creative without being told to be? With machine learning, in which computers learn without being explicitly programmed, we certainly seem to be getting closer to such a point.

Using images taken from Street View, Google used a type of machine learning called a generative adversarial network (in which AIs learn by being pitted against each other) to coach an AI into creating professional level landscape photography. The AI wasn’t programmed to understand what made an aesthetically beautiful image – it learnt through interacting with another AI.

In 2016, an entire screenplay, Sunspring, was written by an artificial intelligence for the Sci-Fi-London film festival’s 48hr Challenge6. After feeding the AI science fiction movies and television, the bot, who named himself Benjamin, learnt by observation and wrote a screenplay that was acted out by professional actors. The resultant short film was largely convoluted and nonsensical, but then again, so are many human-written films!

If you think all this sounds a bit too contrived to be genuinely creative, heed the words of Ansel Adams, one of the world’s most well-known photographers, famous for playing with light in his compositions, who said: “You don’t take a photograph, you make it.”

CMOs, marketers, authors, musicians –
take heart

In terms of benefit to creatives, it’s likely that AI will be used mainly in the automation and assistance stages of creation, which creatives will file under ‘software’ upgrades. As CMOs and marketing teams look for new ways to stand out though, AI could provide compelling benefits, augmenting human minds with additional data and deeper analysis on ‘what works’ for audiences.

Before wondering whether a robot will one day take its place alongside Van Gogh and Monet as a great master, or whether we should hang up the paintbrushes, perhaps it’s worth asking whether a robot can be truly autonomous if it’s creating art for humans.

All the above ideas are predicated on an idea of ‘creative’ that’s purely human. Perhaps we need only worry (or rejoice, take your pick) when robots start creating art for other robots.



Amy Gibbs

Dr Amy Gibbs is a manager at PwC Australia, and the global content editor for Digital Pulse.

More About Amy Gibbs