Can robots write adverts?

Barely a week passes nowadays without news about the latest developments in artificial intelligence, quickly followed by a deluge of think pieces declaring the demise of human workers and their imminent replacement by a legion of robots.

Nowhere has this been more true than in marketing – AI now plays an important role in a wide variety of marketing functions, from providing purchase recommendations to identifying the leads most likely to convert. Until recently, there was one area in our industry considered the sole preserve of humans – creativity. Machines may have proved themselves adept at data processing and analysis, but they couldn’t ever come up with their own ideas, could they? Surely the job of the agency creative was safe from the AI revolution? Actually, a world in which AI is capable of creating its own ads may be closer than we think.

What is creativity?

In a recent agency workshop, True’s Creative Director Mark Dearman offered this definition of creativity, from Sir Ken Robinson:

“Creativity is applied imagination - the process of having original ideas that have value”.

In an advertising context, the “value” of a piece of creative work lies in its ability to provoke a response in people. It may be that the response leads an individual to making a purchase, or it could simply be about encouraging the individual to form a positive association with the brand behind it.

In his book The Runaway Species, author and neuroscientist David Eagleman makes the point that machines are at a disadvantage against people when trying to elicit these kinds of responses. As humans we spend our lives engaged in conversations with others, and trying to interest, amuse and surprise each other with our thoughts, ideas, jokes and stories. Machines don’t have this level of social interactive experience, and so do not share our instinctive understanding of the emotional reactions that different messages can provoke. Witness the efforts of AI to come up with original jokes (ok, some of these are pretty funny – just not for the reasons intended).

AI in Direct Response Marketing

As Eagleman argues, AI creativity is currently held back by its limited ability to understand human responses – but one thing machines are very good at is understanding feedback in terms of clicks, sign-ups and other easily measurable metrics. This means there is a big opportunity for AI in the field of direct response marketing, particularly in channels like email and pay per click advertising.

One example of this is Facebook’s Dynamic Creative tool. This allows an advertiser to upload a batch of different imagery and copy options – it will then use these assets in different combinations to create adverts and identify the ones that work best.

Taking this a stage further is Phrasee, a machine-learning tool for creating small, structured pieces of copy such as email subject lines. It analyses the performance of your previous copy and creates new subject lines which fit with your brand’s tone and style and will (so Phrasee says) outperform a human’s copy 98% of the time.

Phrasee received £1m in seed investment in 2016, and is now being used by major brands like Domino’s, DFS and Virgin Holidays.

Untapped Potential?

So AI is fast proving its worth when it comes to short-form advertising with a limited number of elements – but can machine learning technology be used to create something bigger and more complex, like a TV advert?
McCann Japan believes it can, and has created the world’s first AI Creative Director. This technology was tasked with creating a 30 second tv advert for Mondelez brand Clorets Mints, which would be pitted against an ad made by a human creative team, to see which was rated higher in a poll of consumers.

The AI uses a database of award-winning ads from the last 10 years, which have been deconstructed down into their essential elements – target audience, tone, music, context, key messages etc. The client filled out a form to outline what they wanted from the ad, and the AI analysed the database to produce a new set of ideas.

In the end, the human-made advert narrowly won the public vote 54% to 46% (you can see both adverts here). However, when the ads were shown to 200 advertising executives at a conference to see which they preferred, they voted in favour the AI’s ad.

It’s impressive, but can machine-generated work like this be considered truly creative, when it leans so heavily on analysis of previous work? I asked Mark Dearman for his thoughts on this point:

“Rehashing existing work or picking from a bank of assets isn’t technically creative. In the same way AI that creates something from reference material isn’t creative in its truest sense. I would imagine someone will come up with a new term to define computerised creativity which as it stands is still reliant on source material and parameters being set by humans.”

Criticism of AI Creative

The increasing encroachment of AI into the creative realm has not been universally welcomed. Saatchi’s Group Chief Creative Officer Justin Tindall attracted attention recently with his strident criticisms of what he saw as the problems plaguing the modern advertising industry, with AI one of the main targets of his ire:

“Mother’s Moneysupermarket.com campaign is an example worth following too. It’s one of the most defiantly distinctive campaigns of the past five years. Show me an AI algorithm that can come up with Skeletor and He-Man "dirty dancing" and I will happily shallow fry my own lightly buttered testicles”.

While easy to dismiss as the “kids these days with their new-fangled technology” grumbles of an ageing ad man unwilling to learn new tricks, there is perhaps a point here about human creativity being lost in our eagerness to adopt AI.

It seems there is a sweet spot between the human mind’s capacity for invention and originality, and AI’s vast ability to process data and understand what consumers are likely to respond positively to. Viewed in this light, AI is here to assist us in bringing our best ideas to life, not replace us.

As for the long term future, Eagleman suggests the solution to creating AI that can genuinely rival human imagination could be lay in building robots that can talk to one another -

“I really love the idea that if you want to actually build artificial intelligence in a meaningful way, what you should do is build a series of AI agents that are all trying to impress one another. And so you have all of these computers constantly competing to say the thing that will make all the others think, “Fuck, I never thought of that.”

This points to the essence of true creativity - the ability to generate ideas that entertain and surprise people and make them see things in new and exciting ways. AI is redefining so much of what we do, but the impact of strong, original creative work is still as powerful as ever.

To find out more about our creative approach and how we’re using AI to help in our work, talk to Sam.