Interviews, insight & analysis on digital media & marketing

NDA Viewpoints: When it comes to creativity, nurture AI like your own child

By Alex Hobhouse, MD, Automated Creative

I’m now at the age where most of the people around me are having children.

Small, cute and (for the most part) harmless, the similarities between them and marketing’s latest craze are striking.

Machine learning and artificial intelligence are, like a baby, fundamentally simple. Both start with an empty cache and an algorithm designed to allow them to learn.

Initially these lessons are binary; good or bad and their associated causes, as time goes by this ability to learn becomes more diverse and complex, accompanied by the ability to make increasingly reliable decisions to achieve a desired output.

The only variable throughout is data, both the baby and the algorithm are reliant on and hugely affected by the quality of the data from which they learn.

While this is a largely simplified vision of AI its a useful analogy when assessing any business which places AI at its core.

Like the child of any doting parent the algorithm can do no wrong and is guaranteed to be the most developed of all the children in their NCT group. In both cases it’s always worth analysing the data before making a judgement.

The two most striking areas being affected in marketing revolve around two specific areas: performance, using data to establish the best environment to deliver a specific result; and conceptual creativity, using data to inform and elevate an idea.

When it comes to optimising creative against specific performance objectives we’re already seeing great results. Allowing an algorithm to affect changes to content throughout a campaign has seen acquisition increase substantially.

While past performance campaigns have relied on historic campaign or survey data for insight, we find the most effective data is created and used in real time, allowing a campaign to evolve according to the audience response.

Just as a human would learn and adapt to its surroundings so must an algorithm in order to remain effective.

When it comes to stretching its legs creatively, we’re already seeing some truly remarkable AI campaigns.

Synesthesia delivered its wildly successful David Beckham malaria campaign earlier this year, training an algorithm to replicate human speech as accurately as possible. It looks incredible and is certainly effective.

The technology itself is a few years old, first published by the University of Washington in 2016. It’s helped the company raise $3m in funding but this area of AI in marketing still feels more like a stunt than a campaign focused on driving effectiveness.

Similarly AKQA used machine learning to generate “Somesthetic Transfer” in Australia and Speedgate in the UK. Both are certainly interesting applications of AI but at this stage its difficult to see how they deliver against specific marketing objectives. The scope for the unexpected remains well within the human realm.

Like a newborn baby, algorithms need hard rewards and negatives to learn. While a creative director is training the algorithm we’ll always end up with Speedgate, an instilled bias.

If you train an algorithm with an audience of 10m which can either buy or not buy your product you’re betting on statistics rather than a hunch. I know which I’d prefer.