This article is part of a series of exciting collaborative work with St Andrews Law Review. Head over to their site to read the second half of this exciting and informative history.
Art: nothing but code?
Our anxieties about the rise of algorithmic art reveals the importance of art in defining humanity
Written by: Gwendoline Choi
When Schubert composed Symphony No.8 in 1882, he couldn't have expected that two centuries later it would be finished by a smartphone.
In 2019, Emmy award-winning composer Lucas Cantor used a Huawei phone to generate the remaining movements of one of Schubert's most iconic works. Christie's auctioned off its first artwork created using AI in 2018 for nearly half a million US dollars. The advent of AI-generated art isn't coming - it's already here. So why are we loathe to admit it?
There is an instinctive resistance to conceiving of art as just a sequence of numbers. One anxiety is that if AI has finally penetrated the artistic realm, what has long been considered exclusive to us is no longer purely in our domain, signifying a concrete possibility of the singularity - the point in time where Artificial Superintelligence (ASI) becomes so dominant it escapes our control and understanding. Another reason for our hesitance to embrace AI-generated art might be that art is so tied up to our conceptions of the ineffable soul, emotions that we would like to think refuse serialisation and the mysterious subjectifies of creativity and beauty that to be 'reduced' to data would be reducing us. It seems that we fear AI-generated art because it doesn't just challenge what we understand art to be, it challenges what it means to be human.
What can, or will, AI art do?
Portrait of Edmond Belamy, 2018, created by GAN (Generative Adversarial Network). Sold for $432,500 on 25 October at Christie’s in New York. Image © Obvious
AI generates art by processing data, such as Schubert's catalogue or 16th century portraits, into the device's neural processing unit. Using that information, AI then generates a similar product by identifying patterns in shape, tone, pitch, colour and metre. Sometimes, the artist plays a role in selecting certain preferences or how the outputs are brought together. Thus, AI turns patterns into art. Schubert's symphony was composed in such a way, whereas the Portrait of Edmond de Belamy painting that was auctioned for USD 432, 500 at Christie's was created using a generative adversarial network (GAN).
A GAN employs two systems, an image generator and a discriminator, and pits them against each other to create a work of art. The first network generates a random datum, such as an image, and the discriminator compares the datum to the original set of data provided by the curator, removing options that are not tailored to the preferred aesthetic of the algorithm. The first network then strives to ‘beat’ the discriminator by providing a datum closer to the desired preference. The systems are taught by being fed information on artistic styles and existing works in the form of data. The technologies can be controlled down to the minutia - how far the end product deviates from the average pattern detected, for example. Thus, through technologies such as Amper Music, an open source AI platform, pop albums and portraits can be created.
Portrait of Le Comte de Belamy, 2018, head of the fictitious Belamy family (and Edmond de Belamy's great grandfather) created by the GAN ‘mind’ © Obvious
For example, MuseNet, a neural network developed by OpenAI, can combine different samples – such as bluegrass with Rachmaninoff – to synthesise a music sample that incorporates your desired instruments. AI Gahaku generates a portrait in a painting style of your choice that can be uploaded to Animal Crossing, through a simple upload of an image of your face.
Who are the real artists?
However, it's not as simple as inputting data – and we would be hesitant to categorise art as merely comprised of the shapes of which it is made up. AI is currently reliant on us to determine what data it is fed, and what outputs are eventually chosen. Often, we are also involved in finalising the output into the end product, by rearranging, layering and editing the 'first draft' that AI produces. Our subjectivities and choices, like in traditional modes of art, determine the final product. It could be argued that with our visions, we are still the artists in control, and that emerging technologies merely assist us in producing the final artistic product, as the Renaissance painters had their assistants. Like the guitar, Autotune or music production software, AI is just another tool.
For example, former American Idol contestant Taryn Southern created the first LP to be composed using AI in 2017. She chose the beats per minute (BPM), instruments and lengths she desired and was met with the instrumentals for her song that she could then tweak, adjust, reorder and retool. Although Amper Music produced the instrumentation, she devised the structure and lyrics for her songs, not stopping to generate more options until she was satisfied. It’s with the experience of the human artist that AI art can be corrected, finetuned and prepared for public reception.
There are also the software developers, engineers and scientists to consider - those who are creating the different AI algorithms and developing them for specific creative uses. The software themselves rely on artistic choices made subjectively by humans. It could very plausibly be argued that the software developers are doing the bulk of the artistic work, and are themselves stretching creativity to its limits by developing systems that can be creative themselves. Is this a collaborative process, or are we still using AI networks like we would use a keyboard? These issues could lead to copyright and crediting issues in the future as more artistic AI technologies emerge on the market.
Moreover, AI art forces us to question how much of art is reliant on the perceiver’s values, tastes and assumptions, rather than the creator’s. If we can find art in nature, which is a culmination of successful randomisations, surely we can impose our understanding of art onto computer-generated sequences, just like evolutionary art or BioArt, where art is made in partnership with living tissues and organisms.
This plays into broader questions on creativity. For example, if I splash paint at a canvas enough times, I am likely to eventually produce something that others would generally agree is artistic - an artistic counterpart to the infinite monkey theorem. AI could just be seen as a further derivative of randomly splashing paint at a canvas - an endless generator just finetuned to only include certain combinations with novelty – a key facet of creativity. Randomness to this degree, for many, precludes artistic value – potentially contradicting the key role spontaneity, luck and inspiration has in the creative process. However, some still resist seeing this as art - art, to many, requires a certain emotional intent. If this is the case, will naysayers finally accept AI art if AI does develop consciousness, when the definition of good art may no longer be defined by us?
How alien is AI art, really?
The notion of creativity as distinctly human, or at the very least tied to animal consciousness, is instinctive to many of us. In a poll conducted by NBC News, 31% of respondents believe that art is exclusive to humans. In a poll conducted by NBC News, 31% of respondents believe that art is exclusive to humans. Art has connotations of emotions, spontaneous inspiration and imagination. How could extracting patterns and principles from existing art pieces and generating more be art, when a psychological connection is required? Can human aesthetics be formalised? Critics declare AI cannot recreate the human reasoning used to reach a final product – AI represents the objectivity art seeks to reject.
However, creativity is involved in many processes beyond our artistic experiments. Biological processes of self-creation and self-organisation, such as ontogenesis, require creativity. AI just looks to be another way that the boundaries of creativity are further expanded. Moreover, how different is AI-generated art from other kinds of software, such as Vocaloid software or the Autotune that makes Cher’s Believe so distinct? And AI is certainly not neutral – it is subject to the systemic biases of the human developers as is anything else.
AI art also challenges our perceptions of what it means to be an artist and what counts as art. Even today, elitism persists in the art world in giving primacy to hyperrealistic work that is harder to achieve due to socioeconomic barriers in training, and preferences in what is considered ‘real’ art as compared to other forms, such as street art. AI can break those barriers in giving more people the tool to immediately create new forms in a participatory manner, where the concept of the universal artist – long assumed to be wealthy men – can now be disturbed. As we create AI art, we are both artist and audience. AI art centres the human experience in technology, and is an opportunity for collaboration across the artistic and technological spectrum. Our definitions of ‘human’ shift as we push the boundaries of art, with AI opening up possibilities unknown to the human eye.
But despite our misgivings, one thing is certain – AI art is not leaving anytime soon. The Victoria and Albert Museum has a growing Digital Futures art collection, and research in the Digital Humanities continues to grow as technology enables museum practices that are increasingly accessible and far-reaching. If AI is what defines our current age, it is only natural for AI art to follow as just another step in the ultimate creative process of making art from our experience of the world. As Walter Benjamin put it, ‘even the most perfect reproduction of a work of art is lacking in one element: its presence in time and space, its unique existence at the place where it happens to be’.
Lastly, the differences between mathematics and art are not so great after all. In the words of GH Hardy, ‘the mathematician’s patterns, like the painters or the poets, must be beautiful…beauty is the first test. There is no permanent place in the world for ugly mathematics.’ If fractal art seems a far cry away from the Fibonacci sequence, it has less to do with the process and more to do with our own anxieties.