“Imaginative people have messier minds,” said Scott Barry Kaufman, a professor at New York University who has spent years researching the psychology of art.
According to Kaufman, creativity involves the coming together of a multitude of traits - an observant, sensitive nature, coupled with highly developed neural pathways for reasoning, memory, expression and pattern recognition.
It is this combination of traits that gave us such seminal works as Vincent Van Gogh’s The Starry Night and Marcel Proust’s In Search of Lost Time – works that philosopher Susanne Langer was surely referring to when she said that art is our “foremost laboratory for feeling and time.”
But at this particular juncture in history, as artificial intelligence (AI) becomes increasingly tethered to our daily life, the threat of art being dehumanised appears more likely. Numerous AI projects demonstrate that machines are capable of creating works to rival the great masters, with astoundingly technical paintings and music that mirrors our precise state of mind in any given moment.
But what do we lose when art becomes a computational enterprise? Will it produce a series of tepid, uninspired imitations – utterly algorithmic in their approach? Or can it be a true accomplice to artistic genius?
With the use of AI increasing by 270% in the last four years, the technology is bounding ahead in seismic leaps – and there is a league of artists game enough to run alongside with it. And its these artists who can give us a glimpse into what a computerised-art future might look like.
MUSIC
“Technology is a big destroyer of emotion and truth. Auto-tuning doesn’t do anything for creativity,” said The White Stripes lead singer Jack White, reflecting on the hyper-digitisation of music in the 21st Century. “It’s the disease we have to fight in any creative field: ease of use.”
And Nick Cave summed it up in his usual acerbic manner when he said, “Most of the time I find myself standing next to a radio thinking, “What is this garbage?”
But in 2019 – when synthesizers, loop pedals, samplers and autotuning are standard in rock n roll’s lexicon – the time for these binary arguments of art versus technology are long gone.
Traditionalists often dismiss AI generated music as rote mimicry, and in some cases they may be right. Like Cave, just turn on any commercial radio station and you’ll hear an endless procession of soulless, mass-produced pop songs.
But to vilify AI generated music entirely would be a misjudgement. The truth is, artists have been using it for much longer than you might think. Ray Kurzweil invented a program that composed a piano concerto in 1965.
More recently, Brian Eno created his 1996 album Generative Music 1 with the help of the Koan software, known for producing an infinite variety of ‘muzaks’ (ambient backing tracks). Far be it from a mechanism for ‘ease of use’, many of today’s musicians see AI as an opportunity to expand their practice, producing ground-breaking works at the nexus of technology and art.
Berlin-based artist Holly Herndon is among them. Her latest album ‘Proto’, produced in collaboration with Mat Dryhurst and Colin Self, was composed using an AI entity named ‘Spawn’, developed by the artist herself and housed in a gaming PC. Ensuring an equal partnership between humans and technology is critical to the integrity of her work, as Herndon explains.
“There’s a pervasive narrative of technology as dehumanising, but we stand in contrast to that. Choosing to work with an ensemble of humans is part of our protocol. I don’t want to live in a world in which humans are automated off stage.”
Of course, not all artists are taking the high road. There have already been entire albums of AI ‘deep fakes’, where machine learning systems scan existing music and regurgitate a variation on a theme – plonking a human avatar atop a feat of barely visible labour. And as the technology advances, commercial record labels will undoubtedly continue to capitalise on its cost cutting potential.
So perhaps what we should be concerned about then, is not so much artificial intelligence – but ourselves. “Ultimately this isn’t an AI problem, it’s a human problem,” explains artist, NYU lecturer and Herdnon’s long-time collaborator, Mat Dryhurst.
“AI is just us; meaning it learns from what we do and have done. It reflects back to us an image of ourselves.” “I am concerned that thus far, AI is reflecting something quite ugly - and that is an opportunity for those on the left to start imagining how we can improve ourselves, and by extension, our automated representations of ourselves.”
ART
A pair of translucent, spectral figures hang side-by-side in the main salon of Sotheby’s London. Their pallid skin and hollow stares are eerily familiar, calling to mind the work of the great Renaissance painters. Except that this artwork wasn’t painted by the hands of Da Vinci and Botticelli. Or any hands for that matter. It was made entirely by a machine.
Titled Memories of Passersby I, the artwork features an infinite, shifting array of portraits on two large digital displays. It hammered to an online bidder at £32,000 in 2019 – becoming the first ever AI-engineered artwork sold at a Sotheby’s auction.
German artist Mario Klingemann developed the work with an AI program known as a generative adversarial network (GAN), a machine learning system trained on huge datasets. Millions of images are bounced back and forth between the two modules until the discriminator learns to generate new data with the same statistics.
In the case of Memories of Passersby I, this training data was a huge collection of portraits from the 17th, 18th, and 19th centuries, all selected by Klingemann. From a young age, Klingemann harboured a fascination with technology, taking apart toys, radios, pocket calculators or other devices to learn what makes them tick.
Since then, he’s become a pioneer in the field of neural networks, computer learning, and AI art – working with Google’s Art and Culture Lab, and exhibiting all of the world including New York’s Metropolitan Museum of Art and Paris’s Centre Pompidou. Perhaps the most interesting thing about Klingemann’s practice is his focus on getting the machine to make mistakes - a quality we would normally associate with humanness.
“The beauty of neural networks and one of the reasons they are my favourite tools is that they are imperfect and make many mistakes or misinterpretations of the data I feed them,” Klingemann explains. “It is exactly those mistakes (or accidents as I prefer to call them) that I am searching for and which hold the promise of showing me something unexpected or surprising – something beyond the limits of my imagination.”
Despite these inventive approaches to AI art, there are certainly no shortage of naysayers, including philosopher Sean Dorrance Kelly, who vehemently argued that if AI art becomes the norm, “it will not be because machines have outstripped us. It will be because we will have denigrated ourselves.”
But art has always evolved in tandem with technology – from Jan Van Eyck experimenting with oil-based pigments in the 15 th Century to the introduction of photography at the end of the 19 th century. In this sense, AI is simply another evolutionary tool to express ideas.
Perhaps the hive of negativity surrounding deep-learning art has little to do with the technology, and more to do with our lack of understanding. And mainstream media’s portrayal of AI as some kind of menacing, all-seeing robots certainly doesn’t help. For those in the upper echelons of the contemporary art world, it seems AI will remain a supplementary medium to realise their ideas.
“AI will surely change the way we create and appreciate artistic work, but it will not be a full replacement for human imagination,” Klingemann explains. “Machines will be able to replicate and automate cookie-cutter art and might dominate markets where success is measured by sales or likes.” “But history has shown that humans are very adaptable to challenges, and we will find ways to distinguish our work.”
LITERATURE
Margaret Atwood was moved to write The Handmaid’s Tale after a lively debate with a friend. A passing joke made by his wife prompted Kazuo Ishiguro to create The Remains Of The Day, and JK Rowling dreamed up the Harry Potter series on her daily commute in London.
But these ideas don’t exist in a vacuum, or strike out of nowhere like lightning. They float to the top after years of gestation – years fettered with arguments and broken hearts, earning degrees, moving house and having children. All of those very human markers that constitute a life. This begs the question: how does a computer conceive of a novel or write a love song? And, more importantly, would it be any good?
Thanks to recent advances in artificial intelligence, computers are now able to ingest tons of information at a rapid pace – roaming a vast landscape of information, learning and encountering the world much like a child does. When the literati world piled into the University of Oxford’s Sheldonian Theatre for best- selling author SB Ekhad’s first ever public appearance – they weren’t greeted by a smiling face, but an avatar on a computer screen.
SB Ekhad’s books were a product of the Oxford University’s mathematics department. They devised a code that had allowed the computer to learn from thousands of books in the Bodleian Library. By mimicking the logical connections between neurons in the human brain, the computer quickly gaining a nuanced understanding of what it was humanity valued.
In January of 2019, the Elon Musk funded technology lab OpenAI released GPT2, a digital language model which creates ‘deep fakes for text’, trained on a data set of 8 million web pages. Once prompted with an opening line, it spits out a series of coherent paragraphs – much like the technology behind Gmail auto-replies. If that wasn’t enough to get fiction writers shaking in their boots, OpenAI announced they would not release the fully trained model, due to their “concerns about malicious applications of the technology.”
But perhaps we needn’t worry too much yet. While AI generated prose can be impressive, it often misses out on the complexity and ambiguity of natural language, offering up clumsy sentences and reminding us that is, in fact, just data and not true artistic expression.
As Oxford University mathematician Marcus du Sautoy explains, “In my exploration of the impact that AI is currently having on the creative arts, I must admit I was surprised that the written word — in contrast to music, say, or the visual arts, seemed to be the most difficult medium for AI to demonstrate true creativity.”
“This is in part because the meaning of a sentence often depends on so much more knowledge than just the meaning of the words.”
Technological advances aside, perhaps the most important question we can ask ourselves as a human race is, what is art? What does it constitute? Why do we value it? Trawl through thousands of pages on the philosophy of art, much like GPT2, and you might arrive at this: art is an arbiter of psychological truth, a marker of feeling and time, and a profound expression of our inner life.
When is AI utilised for rote mimicry alone, it would be a stretch to call what it produces ‘art’. But when artists collaborate with AI to create something based in human emotion – something transcendent and original. Is that art? Unquestionably.
But robots becoming sentient is a matter of when, not if, with researchers honing in on AI models that can develop a sense of self, where they can adapt to new scenarios and self- regulate. And then, they will be able to produce their own, original stories, from their own place of truth.
As Marcus du Sautoy explains, “And it’s the stories they will tell that will give us our best glimpse into what it’s like to be a machine.” And whether the story is good or not? Well, that’s up to us. Art is made for humans. And we are the ones who infuse it with meaning.