How Happy Finish Used AI to Generate a Controversial ‘Perfect Mum’

When Dove unveiled a huge billboard in London’s Waterloo station that claimed to depict the ‘perfect mum’, hackles were, understandably, raised. Beleaguered mothers took to Twitter to criticise the brand for piling even more pressure on mums struggling under society’s critical gaze. 

But all was not quite as it seemed. The cherub-faced, blonde, white woman featured on the giant poster was not a real mum… or even a real person. She was an artificial intelligence-created image, synthesised by a computer that had been fed 1800 images of mothers from advertisements, magazines and social media. Aimee was an illusion – and therefore a commentary on the fact that a ‘perfect’ mum doesn’t exist. She’s also a reflection of the non-diverse and stereotypical representations of mothers in the media. When the stunt by Dove’s PR company, MHP, was revealed, the response in the press was positive, with approving write ups in the likes of Stylist and The Pool 

And where the project lets really fascinating is where you speak to Happy Finish, who were behind the AI-generated image – the Dr. Frankensteins to Aimee’s ‘monster’. 

Stuart Waplington, the CEO at Happy Finish, explains that the project is something of a creative ‘first’ when it comes to AI. As far as they can tell it’s the first megapixel image created by AI – most photoreal images have been fairly small to date. What’s more, the image itself was built up pixel by pixel by the AI neural network. 

“For the image in this project, the whole face was generated by an AI. And when we say that, we mean pixel by pixel. It’s not like bringing a nose from there or an eye from here. It’s very much like an AI imagining,” explains Stuart. “That opens the door to a lot of other interesting possibilities, once you can get a photoreal image that’s completely imagined by AI.”

The image was made possible by developments in both computational power and AI techniques. In this case, Happy Finish used an approach known as ‘generative adversarial neural networks’. In layman’s terms, that’s basically where two parts of the AI compete. Each suggests an image, which the other rates on how well it looks and what it’s supposed to look like. The team built their own neural network (a computer system based on the human brain and nervous system) and also used an opensource software library from Google called Tensorflow.

These advances are allowing for a more creative use of AI. “AI hasn’t been able to do these sort of things until fairly recently,” says Stuart. “It’s been used in the industry for quite a few years in terms of delivering ads to the right people in the right context, but use in the creative side of things is fairly new.”

And this creative side to AI has given Stuart and the team a lot to ponder about the nature of creativity – and how machine-driven creativity differs from human creativity.

“After this campaign, we looked at the neural network we developed and we thought ‘let’s have some fun with it’. We had loads of images of trainers lying around that we happened to have, we fed it all these images and it came back with some amazing trainer designs. Some of them were totally out there!” he recalls. 

“What I find interesting, when I look at the things it comes up with, is that there’s definitely a creativity there but a different type of creativity. It’s not a human creativity, it’s not based within the human experience – it’s kind of otherworldly. I can see that it’s really interesting, but it’s really weird too,” says Stuart. He references last year’s landmark Go tournament that saw Google DeepMind beat Lee Sedol and the ‘flashes of creativity’ that saw the AI make moves and use strategies that were so strange that no human would have thought of them. 

Happy Finish has been working in the world of creative AI for about a year now (“In terms of the latest AI revolution, that’s a relatively long time!” laughs Stuart), and they’ve built a team of specialists. For a company that was built on a business of digital retouching and CG imagery, the last few years have taken it on a crazy journey, through virtual reality and now artificial intelligence. How does Stuart draw all those elements together? How does he see and describe the company now?

“It’s a little bit harder for anybody in this industry to define exactly what they are these days because the platforms move so quickly. I think even larger network agencies have struggled a little bit with how to integrate digital and how that changes them,” he muses.

“We see ourselves as unified under one thing – artistic and creative use of technology. That started with digital retouching, CG stills, CG animation all the way through to AI. I think that Silicon Valley is, without doubt, the global leader in technology by a long shot, but I think what we do so well in London, is we use technology creatively. I think that’s where we’re looking to position ourselves: taking that technology and creating really artistic and interesting user journeys and experiences – whether that’s creating a really beautiful image created by a re-toucher or a virtual reality experience or an AI experience, it’s all united by this one thing.”

The rate of technological change generally – and AI evolution specifically – is accelerating and, no doubt, in a year’s time Happy Finish will be working on AI projects that make Aimee look prehistoric. And while she may be an illusion – what she represents in terms of AI as another creative tool is very much real.