We need to stop thinking of machines as rivals, says Ben Jones, Google’s global creative director. Instead, we should see them as an opportunity to take our creativity to the next level.
As advertisers in the age of machine learning and artificial intelligence, it’s easy to think of ourselves in an epic face-off with these machines. But I believe we do ourselves and our work a disservice with this “human vs. machine” mentality. There’s much more to be gained by embracing machine learning as an accelerant for our creative powers.
To explore these opportunities, it’s important to understand what machines do well. First, they’re able to identify patterns that can reveal insights and shape our sense of creative possibilities. Second, they can automate tasks we already do at intense speed and scale, saving time and improving outcomes. Finally, they can bring together sets of data in ways that open up entirely new kinds of creative expression.
And while each of these strengths can help accelerate our creative capabilities, they also help expose what machines can’t do well – and why humans continue to wield a significant amount of creative power.
Pattern recognition at scale can reveal new insights
Machines can churn through millions of videos and correlate creative elements with effectiveness outcomes. That lets us answer questions like, “Will the use of this font or colour improve creative effectiveness?” This is how we learned, for example, that lens flare’s effectiveness peaked in 2013 and that you might therefore be better off avoiding this stylistic technique.
This kind of analysis is not expert; it’s a massive application of something called basic element identification. That’s why Ben Evans compares the power of machine learning to having “infinite interns” – immense but non-expert capacity. For that reason, we certainly shouldn’t see it as a quick and easy solution for creating great ads.
While machines can surface patterns, human intelligence is required to sort and apply them.
For example, imagine that you analysed hundreds of YouTube ads and identified a factor that was highly correlated to effectiveness, like top-performing ads were set in living rooms. Would you set all your ads in living rooms? Or edit your existing footage to include more living room shots? Clearly neither of these would automatically improve performance. While machines can surface patterns, human intelligence is required to sort and apply them.
Improving outcomes with automated marketing intelligence
Automation can also simplify parts of the creative development process, bringing new speed and efficiencies. An example of this is a situation where you might want to explore hundreds of different copy lines across many different audiences to find the best combinations. Rather than spending months listening to bored focus groups debate two or three variations over a plate of stale doughnuts, machine-learning-driven tools can use existing data, audience signals and a library of assets to reveal the best combination.
The building blocks of this process are already in place. Between robust audience signals and tools like YouTube’s Director Mix – which can take basic video assets and create thousands of versions tailored to different target audiences – brands are taking advantage and driving impact. Recently, Caesars Entertainment used the tool to quickly create and serve more than 150 different variations of contextually relevant ads to improve brand perception.
Enabling new types of creative expression
If these opportunities are starting to sound exciting, think how much more potential we can unlock by teaching machines to answer more complex questions, like what kinds of stories to tell in the first place and how to tell them.
For example, what if machines could tell us what narrative structure is most effective for achieving a certain goal? What if they could tell us how fast certain creative trends are changing, so that we could easily understand what’s simply a fad and what is an enduring best practice? And imagine a world in which machines could reveal the “unknown unknowns” – the questions we don’t even know to ask. The possibilities are endless and thrilling.
Where humans will always have the edge
Machines – our infinite interns – clearly come with their own unique potential to help. They also have limitations.
Machine learning does well with low-level processing tasks, which we as advertisers can use to improve our work and save time. But it’s important to recognise that this is still a partnership: the machine is optimising from a set of variations. It is not conceiving of a campaign platform and writing an ad from scratch.
A great example of this is Mondelez chocolate brand Lacta’s Valentine’s Day campaign this year, which sought to drive interest in its product and an upcoming branded film, “The Taste of Love”. Using contextually targeted TrueView ads, Lacta invited users to submit photos representing love through a campaign microsite. It then used machine learning to analyse and organise the photos around common representations of love, such as “kiss”, “smile” and (of course) “dog” and “cat”. Users could explore these love-related concepts by visiting an interactive experience online.
Machines are partners that can help us accelerate our creativity and explore its possibilities more deeply than ever before.
Lacta’s approach successfully drove interest in the brand and proved an important point: while machines can analyse, organise and facilitate a new campaign approach, they can’t conceive of the campaign or provide the insights and emotive elements that people connect with. Only humans can do that.
So as technology continues to advance, keep your dystopian fears in check. Instead, embrace the idea that machines are partners that can help us accelerate our creativity and explore its possibilities more deeply than ever before.