How AI Is Driving the Next Generation of Ad Creative
Most industries have a healthy fear that artificial intelligence is going to render humans obsolete (and even possibly take over the world). But for the digital advertising industry, if applied the right way, AI has the potential to improve production and performance by making more room for original ideas without stripping away reliance on humans.
What is “Creative AI?”
Creative AI is a type of AI that can generate videos, images, audio, or text. But to generate creative assets, humans must train Creative AI by providing reference data and one or more sets of rules and operations or deep-learning algorithms (similar to how humans learn).
The streaming service Netflix, for instance, uses Creative AI to automate A/B testing of hundreds of thumbnail images for all their shows and films, using variations of key story elements, character moments, photography, and text to grab the viewer’s attention.
Although AI may be able to generate derivatives of title images, it is far from being able to create entire movies or series like “Stranger Things.”
Will AI Speed Up The Creative Process?
How long would it take a person to create hundreds of thumbnails for “Stranger Things”? They would have to vet around 60,480 frames from the title per episode. What if the title has multiple episodes or seasons? That person is going to need a jug of coffee.
Creating digital ads is not unlike creating thumbnails for Netflix titles. In fact, often digital ad creation is more complex since creative agencies promote multiple campaigns on multiple platforms. And with the cost of digital advertising increasing every year, advertisers will only be under more pressure to make efficient use of their marketing dollars.
The solution just might be artificial intelligence. AI can do in a fraction of the resources what would cost an agency a great deal of time, money, and people. The headlines, call-to-action (CTA) buttons, and imagery needed to build creatives could easily be done by trained, computer-generated intelligence and, let’s face it, we will always need creative minds to come up with the concepts and ideas we see in campaigns today.
It is common to serve one creative piece in various sizes across thousands of websites. In the future, however, AI could expand the possibilities to automate A/B testing within the creative itself. This, paired with complex data and targeting options (also powered by AI), could seismically alter the digital advertising landscape.
Although digital advertising has not reached this stage yet, many of the processes are already automated, and will soon be improved by AI in greater increments. For instance, we have already started learning how AI can help ad exchanges to better read all the design data points of a site while it is loading like font, size, color, and positioning to render an ad that better fits in with the page and leads to better advertiser performance.
AI Can Improve the Performance of Ads
In the milliseconds it will take to generate an ad, AI will make educated decisions based on the algorithms and templates set by advertisers. And with every generated ad, AI will become better and faster. But how?
“Are you a robot?”
Similar to the “I’m not a robot” prompts that are used to verify the delivery of ad impressions to real humans by asking them to click on images of bridges, buses, or cats, AI is trained through user interaction.
When people click on a cat picture, AI uses a process called computer vision to learn what cats look like. With a lack of interaction (no click), the AI determines the other pictures are not cats. Now, AI knows what cats do not look like, and will not generate images of a bus.
Creative AI could determine which ads are performing the best using a similar method, with the added benefit of automatically A/B testing and generating a new ad.
Thanks to deep learning algorithms, every interaction (or lack thereof) gives the AI more information on which ad is performing better. That learning could then be used towards displaying a more effective ad with every cycle and impression.
Do Designers Still Have A Role in the Creative Process?
Should ad designers start handing in their resignations, packing their bags, and heading out to pasture? No.
Artificial Intelligence is (un)fortunately kind of… dumb. Without human guidance, Creative AI is efficient but ineffective. The people behind the creative work, whether it’s the groundbreaking idea, the branding, or the play on words will always have their place in the creative process. At the end of the day, advertising often plays on human emotion. Plus, we still need people to train AI, including Creative AI, for it to be effective and succeed. On an exciting note, instead of training humans to do the repetitive work of building ads in different sizes for multiple campaigns over the year, we will be training humans to teach AI how to do it but better and faster.
Without Designers, Content Writers and Campaign Managers, AI is at risk of:
Generating nightmare fuel
AI is not yet good at generating images on its own. For instance, people know what a car looks like because there are references filling the streets, freeways, and parking lots. However, tasking Creative AI to generate an image of a car without showing it references of real cars could instead fuel nightmares, as evidence from the attempts of AI to date demonstrate.
Designers are needed to create original visual references, like Toyota’s vehicle inventory, for Creative AI to generate variations of car ads.
Descending in a downward spiral
Like how AI requires visual references to generate an image, scripts are needed for AI to say the intended message. Without a script, AI will parrot what users say. Microsoft found out the hard way what would happen if AI goes unchecked when it created Tay, a Twitter chatbot, that turned into a racist in less than 24 hours.
Content writers are vital in drafting the copy and providing keywords that AI can then alter and iterate on.
Providing suggestions that traumatize
Marketers and campaign managers are imperative for AI to understand context and improve ad performance. AI may understand that a centipede is a bug, but recommending “The Human Centipede” after watching “A Bug’s Life” will traumatize kids (and some adults).
Marketers can apply deep-learning algorithms to calibrate where, when, and what ads are displayed in which contexts.
In the end…
Creative AI will not overthrow the creative process of the digital advertising industry. Instead, Creative AI could free advertisers from busywork and emphasize creating original and meaningful content.
Rob Fan is CTO of Sharethrough.
