Learn to Code, Artists


AI art is sending ripples of fear through the art world. Artist-critic Alexander Adams explores the industry implications of computer-generated imagery.


“I’m incredibly anxious about the future of my career, more than ever before,” said Kelly McKernan, a successful painter, in a tweet a few months back. She is not alone. The rise of AI art programs like DALL-E, DALL-E 2, Midjourney, and Stable Diffusion has alarmed many artists. They aggregate digitised art and, with simple instructions, blend them to create unique combinations. Established artists worry that less-trained artists will use open-access AI to pilfer and process art by others. McKernan was shocked to discover that her artwork had been fed through the Stable Diffusion engine and had produced faithful variants of her style. “This is unethical. It feels like a violation,” she wrote.

There are increasing calls from professional image makers to boycott the use of AI programs, citing concerns about copyright infringement and lack of credit for the original artists. The looming worry is that AI-generated art could lead to redundancy for human artists, raising questions about what will happen when AI programs can create art without any human guidance. While AI has been used in the rendering of digital animation for movies and video games for many years, the latest development is that programmes are now making decisions not just about technical aspects like texture, lighting, and physics, but also artistic motifs. This is considered a form of creative input by AI, rather than just technical processing. This is a significant shift; as McKernan bluntly states: “I’m concerned for the future of human creativity.”

Overproduction, Underpayment

For the last year, anyone following public conversations among comic-book artists will have noticed two recurring subjects: artist poverty and AI art. The two are linked. Fees for comic artists have either remained stagnant or decreased in recent years, not accounting for inflation. Motherboard, now part of Vice, noted that Marvel’s rates for line art dropped from $352 per page in 2015 to $173 by 2017, with indications on social media suggesting that it has since sunk even lower. In addition, there are concerns about artists having to chase up payments from companies, with freelance artists working for major publications experiencing delays of up to 10 weeks.

For months, artists have been complaining about competitors using AI to imitate the styles of living artists and passing it off as their own work. While there has been little evidence of this practice, given the precarious state of comic-book production, the nervousness among artists is understandable. Many artists consider the use of AI to be dishonest, likening it to athletes using performance-enhancing drugs. One artist said that AI is “to art what Autotune is to music.” Up to now, AI art is not yet at a level to replace human-produced art, so this anxiety is anticipatory. However, it ultimately depends on the honesty of the person directing the AI program and those publishing the images to acknowledge sources and use of AI. Nonetheless, AI-generated art has already won in competitions, albeit in planned tests rather than fraudulent deception.

Worry about competition from AI art has come about due to the massive overproduction of artists. In the UK, there are dozens of fine art, illustration, animation, and graphics courses. These courses have different specialities, and the majority of graduates will never go into fine art or illustration fields professionally; even so, thousands of ‘qualified’ artists go on to enter an already over-supplied market. This does not account for self-trained and foreign artists. Payments for artists are so low largely due to the oversupply of artists and the increasing availability and accessibility of digital technology, which can be mastered by anyone diligent enough.

Fine artists are relatively more protected from the incursions of AI-generated art than digital artists because they produce physical objects that require craftsmanship. Notwithstanding the advent of 3D printing and the glicée print, oil paintings, pencil drawings, clay and plaster sculptures, and other handmade objects are not yet reproducible by machines. However, fine artists who work exclusively in digital formats may be forgiven for feeling more concerned about the encroachment of AI art compared to their traditional-media colleagues.

Welcome to the Uncanny Valley

Artists have always drawn inspiration from other artists. The great painter of peasant scenes Pieter Breugel the Elder, known as ‘Little Bosch’, started as a pasticheur of Hieronymous Bosch. Egon Schiele imitated the more established Gustav Klimt, earning himself the nickname ‘The Silver Klimt’. Every painter who apprenticed to a senior painter went through a phase of copying in the master’s studio as a necessary step towards becoming a master themselves. So, imitation is necessary and expected within the Western tradition.

The Lotus Eaters’ Rory Cranstoun has directed Midjourney to produce fantasy landscapes in the woodcut style of French illustrator Gustave Doré. At first glance they look plausible, impressive actually. Even if the compositions do not work technically, the style of shading is accurate. Yet once you look closely, the pictures fall apart. The architecture is jumbled; trees are disproportionately large compared to houses. Doesn’t that sheep look awfully like a cow? Should it have five legs?

These digital Dorés inhabit the Uncanny Valley, a common area for animation and AI art where images feel almost real but still unpleasantly artificial. What these images need is a human eye to detect the problems, followed by a skilled artist to solve them through tweaking.

In the Eighteenth Century, the steam engine of the Industrial Revolution revolutionised the textile industry by introducing mechanised looms in new garment factories, which decimated and, in some cases, eradicated the artisanal production of textiles by hand. This shift destroyed the economy for artisans who made textiles at home and sold them locally. However, it also created new jobs for factory workers, which then gave way to technicians who supervised highly automated fabric-weaving machines. As cheap AI-generated art becomes more prevalent in everyday illustrations, low-level artists will lose their jobs. However, the better, more adaptable artists will have a place augmenting AI production of images, such as correcting the flaws in Rory’s Doré landscapes.

AI is a long way from replacing artists of the highest calibre. However, just as Julia in Orwell’s Nineteen-Eighty-Four worked on the machines in the fiction department, today’s illustrators may find gainful employment in adjusting the products of the AI art programs of the near future.

Share:

Comments