Wall-E with a Paintbrush?
The cold art of artificial intelligence
Earlier this year, on February 12, science reporter Jacob Margolis tweeted a thread about the last message received from NASA’s Mars rover Opportunity, which had launched on July 7, 2003, and had been active on Mars since early 2004. In the thread Margolis claimed, “The last message they received was basically, ‘My battery is low and it’s getting dark.’”
The tweets went viral. Actress Alyssa Milano responded, “I’m not crying. Nope. My eyeballs just worked out and they are sweating, that’s all.” Comedian Patton Oswalt compared Opportunity’s valedictory message to the melancholy lyrics of Morrissey. The rover had somehow given us poetry in its final moments. Many were moved by this unmanlike machine’s manlike message.
A few days later, Margolis wrote a piece for LAist, in which he admitted that the line was his own “poetic translation” of otherwise pretty mundane messages. In poeticizing a technical transmission, he had anthropomorphized the rover, imbuing it not only with artificial intelligence, but with artificial emotions. Many had imagined Wall-E in his death throes. They wanted to believe in the machine’s existential ache as it resigned itself to the bleak fate we all will one day face: Battery low. Getting dark. Then: nothingness.
In October 2018, five months prior to Margolis’ tweets about Opportunity, Christie’s, an auction house, sold a piece of art titled Portrait of Edmond Belamy for $432,500, over 40 times its high estimate. Ostensibly, it’s a portrait of a man of the cloth, the titular Edmond Belamy, with his round, plump face, in a dark frock coat and a plain white collar. The piece appears unfinished, with exposed canvas around the dark paint that the man emerges from. It’s eerie, alluring — a bit like one of Francis Bacon’s horrified and horrific popes walked back a half-step from the edge of oblivion.
Belamy could not model for this portrait for one very simple reason: he does not exist. Never did. The painting of this imaginary man is, as Christie’s noted, “not the product of a human mind.” Instead, it was created by an “artificial intelligence” named GAN (Generative Adversarial Network) using “an algorithm defined by algebraic formula.” A fictional sitter for an automated artist.
The group behind GAN and Portrait of Edmond Belamy, Obvious, is “a collective of friends, artists, and researchers, driven by a common sensibility regarding questions bounded to the increasing advent of Artificial Intelligence and Machine Learning.”
GAN uses an algorithm comprised of two parts: the Generator and the Discriminator. Based on a set of 15,000 images given to it, the Generator makes a new art piece out of the information it gleans from the portrait dataset. The Discriminator then attempts to figure out if it can differentiate the Generator-created image from the set of man-made images. When the Discriminator can no longer tell the difference, the portrait is complete.
Portrait of Edmond Belamy is the first piece of art created by artificial intelligence sold at auction, and by all accounts it will not be the last.
Artificial intelligence (A.I.) is a burgeoning field, which we’ve seen increasingly incorporated into our everyday lives. Smart speakers or virtual assistants, like Amazon’s Alexa, are gaining popularity. Self-driving vehicles already exist and will inevitably revolutionize travel (decimating a number of industries in the process, including trucking and ride-hailing). Much of the reason an unlikely candidate for president like Andrew Yang has been able to stay on the Democratic debate stage when so many others with more conventional political backgrounds have not is because he understands that “if you go to a factory in Michigan, you will not find wall-to-wall immigrants; you will find wall-to-wall robots and machines.” It seems only natural that this technology would find a way to encroach upon the arts, as technology always has.
In looking at Portrait of Edmond Belamy, though, I was left cold. I could concede it had elements that made it aesthetically intriguing. There is much human-made art that is demonstrably worse from a solely aesthetic criterion. Yet even that lesser human-made art makes me feel more.
I was beset by a cascading series of questions: If I had seen Portrait of Edmond Belamy at an art gallery and thought a human had made it, would I have, perhaps, felt something? Would it maybe have moved me? Would I have liked it? Why does A.I. art become stiff and impersonal the moment I discover that it’s made via artificial intelligence? Will pre-singularity A.I. ever be able to make “great art”? Or does “great art” depend upon consciousness, sentience, and self-awareness? And, finally, how might my answer to these questions make me reconsider my idea of what constitutes “great art”?
Computers have already long surpassed us in their ability to generate knowledge. (I mean knowledge here in the basic layman’s definition as “the collection of facts” and not the more philosophical and epistemological meanings with regards to intentionality.) Watson is unbeatable on Jeopardy! because Watson has access to a deep and broad knowledge set that can be accessed almost instantaneously.
But knowledge is not the crucial element in art. Indeed, art depends on the opposite of knowledge: mystery, ambiguity, negative capability. Art, then, is in direct opposition to propaganda (which is defined by its ideological certainty and didacticism). Art asks questions rather than provides answers.
Watson may be able to put knowledge in the form of a question, as it did on Jeopardy!, but an answer in the form of a question is still an answer. Art necessitates an engagement with the unanswerable.
This is why Pablo Picasso famously opined, “Computers are useless. They can only give you answers.”
The story surrounding the last message of NASA’s Mars rover Opportunity is instructive here. When we anthropomorphize technology, when we make it seem human by attributing to it emotions it doesn’t have, we allow it to conjure up some of that mystery, ambiguity, and negative capability that undergirds human experience.
For art isn’t solely about not knowing, but also about empathy — connecting two or more consciousnesses through feelings and sensations. Both empathy and ambiguity are essentially human. Humans care but do not know.
Artificial intelligence is the opposite (at least thus far and for the conceivable future). A.I. knows but does not care. So while computers, robots, and other artificial intelligences may be able to mimic our art, to approximate it, there is always something missing. Where is the uncertainty? Where is the feeling? Only a coldness pervades.
But what if we were to see A.I. art without knowing it was created by machines? The director of the Art and Artificial Intelligence Lab at Rutgers University, Ahmed Elgammal, describes tests they have conducted: “We mixed human-generated art and art from machines, and posed questions — direct ones, such as ‘Do you think this painting was produced by a machine or a human artist?’ and also indirect ones such as, ‘How inspiring do you find this work?’” According to Elgammal, the results showed that there is very little difference between the human response to human art and the human response to machine art.
I can imagine this being true with me as well, even without conducting my own experiments. But if art is meant to provide connection between consciousnesses, then once I know a machine created a piece of art instead of a human being, the cord is cut, the connection severed. What am I connecting to other than myself? Like a placebo, A.I. art might work for me, but only until I am confronted with the truth of what it actually is: an empty pill in the shape of something more.
This is only true of weak A.I. though. If we reach the singularity, and strong A.I. learns to understand love and loss, joy and pain, melancholy and uncertainty, ennui and empathy — the moment an artificial intelligence becomes truly indistinguishable from a human one — I will welcome its art and, I presume, find warmth therein.
