An art exhibition launching at the HG Contemporary gallery in New York this month showcases the joint efforts of Ahmed Elgammal, a computer scientist who heads up the Art and AI lab at Rutgers University and the AI programme he created, AICAN.
"For the exhibition I wanted to make faceless portraits, taking the portrait out of the context of a particular person, and looking at the formal form of portraits, then abstracting them and pushing them into the Surrealist realm," Ahmed Elgammal told Techworld.
But how does this collaborative effort play out? AICAN presents a vast array of images to Elgammal and he selects the most aesthetically compelling results. Some of the results have a blurred, expressive appearance, like a pale, misshapen face against a black backdrop.
"Usually portraits capture something about the people depicted," says Elgammal. "Here, the image has no reference to a specific person or a historical point. It's totally faceless. The portrait becomes a very abstract concept that doesn't have a particular meaning or context."
The lab at Rutgers was launched seven years ago, with the intention of examining how to advance AI by looking at art. Today, AICAN's work has been been exhibited around the world and one piece was even sold for $16,000 at an auction.
Elgammal says they are pursuing answers to questions such as: "How do we make the machine look at an art work and go deeper than just realising is it a man, or a woman, or a cat, or a dog?
"Can it look at an art work and understand the iconology, the iconography? Can it understand the element or tell us something about style, something about depth, something about the contours and texture?"
The next steps he foresees for artificial intelligence are algorithms that better understand the creative products of people - "basically looking at things like making art, and making music, and creating stories and jokes" - which ultimately could lead to a symbiotic working relationship between humans and their AI collaborators.
AICAN is programmed with an algorithm called the 'creative adversarial network', that allows the software to learn from viewing other artworks and when producing its own, to create something novel enough to catch attention, but grounded enough in art history to be comprehensible and somewhat familiar to viewers. This formulation was based on research into the principles guiding whether human-created pieces of art are well-received or not.
AICAN has scanned over 100,000 pieces of art created over the past five centuries. The software can judge how creative its efforts have been using an algorithm the group developed for quantifying creativity. AICAN even names its pieces: one blurry seascape is called The Beach at Pourville, another is cheekily entitled Orgy.
Research by the group indicated that people couldn't differentiate AICAN's work from a human's. To the question, 'Was this art created by a robot or human?' 75 percent of respondents answered the latter.
Elgammal is certain efforts such as these "really open up a whole new horizon for creativity" and that in future AI will find its place as a collaborative tool for creativity.
"The machine will explore and generate a lot of images following the constraints that I put in place, and gives me the luxury to choose what I want to show," he says of his artificially intelligent partner. "I don't see it as something independent from human creativity. I see it as a creative tool what will become part of how art is made."
But can AI-created art secure a place in the popular canon? This may be down to people's perceptions of what art is and how they engage with it.
"A very important question in our study is the psychology of the perception of art and building conditional models to simulate feelings," says Elgammal. "We came to the belief that the visual formal element of art is very fundamental in affecting the viewer. So that's in conjunction with the semantics, and intention."
By this, he means that despite knowing the artist's original intention or emotional state when creating the piece, we can still be affected by the qualities of art, and because of this, he imagines people can still be moved by art created in conjunction with AI.
Of course, whenever AI encroaches on typically human spheres of expression, there are fears that it will reduce the space for human creators. Elgammal says these concerns are akin to fears that the camera would end artistry. "And, this is not true," he asserts. "It might have taken the jobs of some artists who made portraits, but there are now more artists than ever. So, it never killed artists."