From flower wreaths and love hearts to dog ears and a lolling tongue, filters that map and transform a face have become a common occurrence on social media channels. It’s a phenomenon artist Adam Ferriss draws on in his latest installation Likeness, which reimagines face-filtering and remoulds people’s appearances in real time. “The end goal was to take a person’s face, and make it appear as another – but not in a face-swappy kind of way,” Adam tells us. “This is a bit of a unconventional project in that I didn’t know what kind of output to expect until I was well into the making of the project.”
Adam first started learning how to code seven years ago when he was nearing the end of his undergraduate degree in photography. Likeness was made for Google IO 2018 in the Museum of Developer Art, curated by Alex Czetwertynski, and was commissioned by the UCLA Conditional Studio. “The brief for the commission was pretty open: it had to have something to do with AI, data, or privacy. I felt particularly inspired by Mario Klingemann’s works so I reached out to him. He was really helpful in providing advice in getting this project off the ground. Lots of my projects use faces so it’s a familiar place for me to be working,” Adam says. The outcome of Likeness is a series of distorted, uncanny portraits that challenge the boundaries between collage, photography and art.
Adam started off by introducing the computer to two sets of images; the first set was made up of generic photographs of people; the second was a set of, what Adam calls, “label maps” that identify and highlight the different facial features, like eyes, eyebrows, jawline, mouth and nose, present in the photographs. “You then let the computer compare the two data sets and learn correlations between the labels and the photos—this is called training, and can take hours or days. Mine was trained over the course of a week for around 30 to 40 hours,” Adam says. Once training is complete, Adam feeds a label map to the computer, which then alters and reshapes the face. The output can change anything from small characteristics to gender or skin colour.
“I wanted to find some balance between photographic reproduction and AI weirdness. Dealing with these neural networks is kind of like trying to predict the behaviour of your dog. You know it’s trained, you know it knows how to do some very specific things, but it doesn’t always obey or react in the way that you wanted or predicted. It’s simultaneously frustrating and exciting to be dealing with neural networks; they don’t always perform in the ways that you anticipate, but that same property also makes them interesting artistic tools,” Adam says. Likeness was installed as an eight by ten-foot-high LED wall equipped with a camera that invited people to interact with and pose for the AI face-generator. Although the majority of people have unknowingly come across neural networks, Adam says, the artist wanted to shine a light on the playful aspect of AI; to highlight how technological advancements shouldn’t be seen as either apocalyptic or the solution to all world problems.
Through his work, Adam not only reimagines portraiture and how people perceive themselves, he also rethinks the potentially revolutionary role artificial intelligence can play in the creative industry. “I try to keep up with the current developments in machine learning because I think there’s often good potential there for new artworks. Most of the projects produced by the computer scientists doing this research are very technical in their nature and demonstration. They tend to shy away from creativity in favour or empiricism. But I think that if they’re making new tools for generating images and video, we may as well explore what they’re capable of.”