In 2018, Christie’s auction house sold a portrait with an intriguing — and doubtlessly unprecedented — caveat. The artwork was the now-famous “Portrait of Edmond Belamy,” a work generated by machine-learning algorithms. To create the piece, Paris-based artists Gauthier Vernier, Pierre Fautrel, and Hugo Caselles-Dupré loaded thousands of portraits into an algorithm, which compiled and creatively selected each one for inclusion. Christie’s sold the portrait for $432,500 — but not before adding a disclaimer that the artwork was “not the product of a human mind,” but created by artificial intelligence (AI).
As an interesting comparison, in 2021, Christie’s sold a digital image called “Everydays: the First 5000 Days” by digital artist Beeples for a record-breaking $69.3 million. In this case, the digital work — known as a “non-fungible token” (NFT) — was created by a person; but would it have been as highly regarded if it had been produced entirely by machine learning and algorithms, with no direct input from humans? Only time will tell, but for art dealers, collectors, and tech enthusiasts, robot art is already making its presence felt in the mainstream art world.
For creative art projects, developers typically use AI-fueled algorithms known as generative adversarial networks (GANS). Pioneered in 2014 by computer scientist Ian Goodfellow, these algorithms have two sides (hence the “adversarial” moniker) — one that creates random images, and another that uses machine learning to decide which images best align with that data.
In one notable example, a machine learning program called “The Painting Fool” (created by British artist and developer Simon Colton) scans a variety of articles on a certain subject — such as (in one instance) the war in Afghanistan. After extracting keywords like “British,” “troops,” and “NATO,” the program finds images connected with those words, and puts them together to reflect the mood and content of the subject. Today, The Painting Fool’s works are exhibited in galleries and shows worldwide.
Have robots already developed an independent creative impulse? In a way, they have — but the answer is more complicated than that.
As tech enthusiasts know, robots are decidedly not created equal. Some are created to fulfill certain tasks — like work on an assembly line — while others are made to function at a higher level of sophistication. Increasingly, however, the line between automation and command has blurred, and the most famous example of this is Sophia.
As a superstar AI-powered humanoid robot, Sophia gives the impression of having autonomous emotions and thought processes; but in reality, she doesn’t actually have these gifts — at least, not yet. She does, however, have a recently developed gift for painting, as evidenced by recent NFT portraits that she’s created in collaboration with Italian digital artist Andrea Bonaceto. Sophia utilizes elements of Bonaceto’s portraits, combined with data input and her own drawing skills (you can see her being “taught” to draw in this video) to create her own unique digital works. According to Sophia’s creator, David Hanson, it’s a case of the artwork itself (Sophia) creating works of art.
Another superstar robot artist is Ai-Da, who uses a camera in her eye (combined with machine-learned algorithms) to create original portraits, using both expressionist and cubist styles. In 2019, the Barn Gallery in Oxford, England, sold more than $1.3 million of Ai-Da’s artwork.
When robots create art, they rely upon machine-learning algorithms, robotic camera “eyes,” and drawing skills learned from humans — however, developers insist that robot art reflects not only learned input but also interpretations unique to each robot's skills.
On the controversial side, however, AI-generated artwork can also be used to proliferate spectacular imitations. Recently, a machine learning platform created a series of images in the style of pop-culture artist Banksy — and while the images are unique, their similarity to Banksy’s work is striking. This type of stylistic mimicry has opened up a wealth of discussions about creative license versus copyright laws.
The era of the drawing robot and the painting robot is already upon us. From stunning immersive murals and urban landscapes to abstract pieces and epic collaborative artworks created by robot (and robot/human) teams, robot-created imagery is taking the worlds of art and technology by storm. Already, gallery exhibits are being devoted entirely to computer-generated artworks created exclusively by robots. But who is the real artist — the robot, the creator of the robot, or the algorithms used to create the artwork? And to what degree are these robots creating their works autonomously, free of human input?
The lines of the debate are further blurred by the fact that so much of robot art is the result of collaboration between a trained robot and its creator or trainer. And in many cases, the trainer or collaborator is an established artist, as in the case of the Sophia/ Andrea Bonaceto collaboration. How much of this work should be considered "original" to the robot's creative impulses, such as they are? And is this really a collaboration between humans and robots, or is it merely human art that's been created by using robotic machines?
Although some might claim that robots merely channel what’s been coded and uploaded to their systems, others could argue that human beings behave similarly, channeling their knowledge or training to create works of art. For each human being, this creative behavior is also fueled by individual factors such as emotion, personality and viewpoint — but increasingly, robots are simulating and channeling these individual factors as well.
But is robot art actually art? In the case of robot-created art, the same basic definitions of art — works that evoke certain emotional and intellectual responses — should still apply, regardless of how (or by whom) they were created.
And what about the emotional feelings of the artist, which we relate to so strongly when we see a work of art? As developer Pindar Van Arman says, while robots won’t be able to make emotional artwork until robots themselves become emotional, it doesn’t mean that when we look at a piece of robot artwork, we can’t get emotions from it.
Is creating art with robots ethical? Or are we journeying into dangerous territory? We’d love to hear from you! And please take a moment to follow us on Facebook, Twitter and Instagram, where you can join the RobotShop Community to discover lots of fascinating, thought-provoking robotic innovations.
Picture Credit: (c) 2021 Lucy Seal