Could a computer write a prize-winning ad campaign?

Much of what is written about artificial intelligence (AI) is either hype or a variation on the theme about the robots coming to take all our jobs.

However, aside from worrying about the threat of the machines developing self-awareness and Turing tests, what about AI and marketing? Just like AI discussions in general, it can be hard to work out what is going on.

But what is AI? AI is often used as an umbrella term to cover a wide range of different technologies. A formal definition of AI is the “development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages”. Machine learning is closely related – and, in fact, I believe machine learning is generally what marketers mean when they talk about AI. Machine learning is a branch of AI that allows software to learn from experience, to modify its processing on the basis of historical data.

Anything that can learn directly from examples, data and experience can be perfect for a huge range of optimisation and categorisation capabilities. For example, machine learning can improve digital advertising campaigns and email campaigns by ‘learning’ what works best on the fly. It can also take customer queries and service questions, categorise them and build conversational voice experiences using chatbots.

We have all seen AI and machine learning in action with recommendation engines like Amazon’s ‘you-might-also-like’ product suggestions, which learn from masses of user data to create personalised content based on how we automatically label the data we browse.

Clicking or not clicking; staying on one web page longer than another; hovering over a Facebook video to see what happens at the end. These activities mean we are all tagging our data and helping build a fine-grained picture of our habits, demands and desires.

And recommendation engines are just the start. There’s more to come. So, as the machines outsmart us in lots more domains, we can comfort ourselves that the one area that will never be touched by AI is human creativity.

Time will tell whether AI-created advertisements will be troubling the Cannes Lions awards soon. However, they won’t become mainstream for other reasons.

Given the number of column inches devoted to creativity in every marketing publication, you can convincingly argue that creativity is the last bastion of protection against the robots -everybody knows machines don’t have a creative spark. Crunching big numbers, spotting patterns in data, and creating customer models, yes, but creativity is the real marvel of the mind.

Creativity is about ideas, imagination and ingenuity, we say. Creativity is about appearing to conjure ideas out of thin air, the flash of inspiration. Creativity is about the new and the novel. Surely boring, inanimate machines cannot compete with the domain of creativity, no matter how powerful or complex the software might be?

Or can they?

Marcus de Sautoy’s book, The Creativity Code makes the case for why algorithms could match Beethoven or Picasso – or for that matter, the creative departments of advertising agencies. By understanding the nature of creativity, du Sautoy questions many of our assumption around creativity. Much of what we consider to be creativity consists of copying, combining and synthesising rather than the supposed flash of inspiration.

De Sautoy uses the three examples of creativity described by Margaret Boden, professor of cognitive science at University of Sussex. The first type of creativity – combinatorial – is about mash-ups of different types of ideas or unfamiliar combinations of familiar ideas. The second type of creativity is exploratory – simply exploring what’s possible using existing structures, for instance, painting in the impressionist style. The third type of creativity – transformational – is the game-changing type: a radical change of thinking that breaks from the past, such as the early 20th century modernism movement that gave birth to seminal works by Picasso, Joyce and Matisse.

Professor Boden reckons the first two types of creativity account for 98% of what we call creativity.

And, this is the crunchpoint: all three types of creativity can be modelled by AI – in some cases, with impressive results. If you take Arthur Koestler’s definition of creativity as “the ability to make connections between previously unconnected ideas”, you can see why. Pattern recognition is what computers do very well.

Machine learning systems carry out complex processes by learning from data, rather than following pre-programmed rules. And, as pointed earlier, machine learning is already used today to help identify trends or common occurrences and effectively predict responses and reactions. Turns out that is very useful in the context of creativity.

Du Sautoy also points out how much of our emotional response to art and what we call creativity is a product of our brains reacting to pattern and structure. And the creation of advertising, and in particular copywriting, is something that has a definite pattern and structure. So, it’s not surprising that today AI is being used to write email subject lines, content for Facebook ads and reasonably useful written analysis of financial reports.

Time will tell whether AI-created advertisements will be troubling the Cannes Lions awards soon. However, they won’t become mainstream for other reasons.

As Professor Boden points out: creativity means lots of things to different people. Whether a piece of creativity is valuable is a whole different thing as the notion of value varies by person, by country, by culture, by era. And that’s before we get to talk about advertising. Discussions about creativity are rooted in our views about value. Getting AI to identify and know what constitutes value, well, we are still a long way from that. So don’t panic, the robots are not coming to take our jobs.