Have you ever wondered what it would look like if Charles Darwin rode Napoleon's horse into battle against a dragon and its army of video game characters? Well, you could spend years - even decades - on artistic and design training to create the tableau yourself … or you could circumvent all that by using one of the many AI art generators that have become the subject of much public discussion recently.
Platforms such as Dall-E, Midjourney and Stable Diffusion, which emerged in 2021 and 2022, allow users to type in a prompt from which the generator lifts keywords that it uses to comb or "scrape" databases of photos and other existing artworks on the internet. A neural network then algorithmically interprets relationships between the source art using the prompt, then generates your portrait of Darwin in the style of whatever artist your heart desires - Edvard Munch, Vincent van Gogh or even Yayoi Kusama.
The applications of art generated by artificial intelligence are numerous, including but not limited to stock images required for advertising, film and fashion. Cosmopolitan championed the first AI-generated magazine cover in 2022.
Cosmopolitan’s June 2022 AI issue cover, generated by OpenAI’s Dall-E AI art generator. (Photo: Handout)
Then, there is the impact that AI has had on the art industry itself. Galleries around the world are steadily embracing AI-generated art in addition to the "artists" that use the tools. In 2018, Christie's auctioned off the Portrait of Edmond de Belamy impressionist portrait - which was signed with a portion of the algorithm used to generate the piece - for US$432,500 after it was initially valued at around US$7,000.
New York-based gallery Gagosian is this month hosting an exhibition of Bennett Miller who generated prints using text-based generator Dall-E that, according to the gallery, "engage the history and format of photography to pose questions around the nature of perception, reality and truth - an inquiry made newly urgent by revolutionary innovations in computing".
Portrait of Edmond de Belamy, a piece of art created by an algorithm by a French collective named Obvious, which produces art using artificial intelligence. (Photo: AFP)
Naysayers of text-based AI art generation rally around several key and valid concerns, such as infringement on stock image or other art copyrights held by companies or individuals. Stock image purveyor Getty Images sued Stability AI earlier this year for using more than 12 million images from its site to train its generator, while artists Sarah Andersen, Kelly McKernan and Karla Ortiz collectively filed a class-action suit in January against Microsoft, GitHub and OpenAI (which is behind Dall-E), also on copyright infringement grounds.
However, artists may not have to worry about being supplanted by AI quite yet. The New Yorker paraphrased Matthew Butterick, the lawyer who filed the suit for the above mentioned three artists, stating that though the generators "present something to you as if it's copyright-free", the resultant image is in fact the opposite and "what AI generators do falls short of transformative use". The article also asserts that generators "could not operate without the labour of humans like McK
MonoC, a virtual human designed by Hong Kong company Gusto Collective, is described on her Instagram account as a “digital creator” and woman of contradictions.ernan who unwittingly provided the source material".
MonoC, a virtual human designed by Hong Kong company Gusto Collective, is described on her Instagram account as a “digital creator” and woman of contradictions.
The very idea of digital tools relying on human input - willingly or unwillingly given - to create art has existed since the 1960s. AARON was the first significant computer art system, developed by Harold Cohen in the late 1960s in an attempt to create art via coding. AARON would generate rudimentary black and white sketches while Cohen would finish the paintings. AARON and Cohen's style eventually evolved from purely abstract pieces to more realistic rocks, plants and humans.
Generators became more independent with the advent of generative adversarial networks, which more closely resemble the generators we know today. In a 2014 paper, Ian Goodfellow introduced the concept of using data to train a generator that finishes an artwork independently and explained parameters to discriminate how "successfully" the art resembled the data inputted. In fact, the Edmond de Belamy - as well as the entire Famille de Belamy images - is a play on Goodfellow's name as the French words "bel ami" translate to "good friend".
But artists may still benefit from actively using these tools, too. For instance, NFT artist Claire Silver, known primarily for her digital pieces and heavy collaboration with AI, signed with talent agency WME and announced in March that she would exhibit her works at the Louvre … although the museum has since denied her claim.
Meanwhile, Hong Kong's own Gusto Collective takes the relationship between artist and digital tool to an extreme by backing metahuman artist MonoC. The fully digital persona relies on AI to generate artwork such as the Covid-inspired Corona Florella collection and her earlier Through My AIs collection, which was auctioned on Sotheby's as NFTs.
"MonoC is born out of data, but lives for the art," says Caroline York, Gusto Collective's director of marketing. "She is a non-human that constantly seeks to understand the human experience, driven by the curiosity to unravel what human 'emotions' and 'feelings' are." The Corona Florella series, which was exhibited at Hong Kong Digital Art Fair last November, interpreted Covid-19 data for infection, death, vaccination and recovery rates and used it to "power the generative movements of MonoC's meta-flowers".
Metahuman artist MonoC was developed by Gusto Collective and relies on AI to generate her art. (Photo: Handout)
The exhibition is a snapshot of today's digital developments: a fully digital persona using AI to generate art from big data, which viewers could purchase at a cool two ethereum per piece. However, it seems reception to the pieces is focused on their human qualities. "Some exhibition attendees found the flowers representing death to be beautiful, while others found the series deeply personal through their own Covid-19 experiences with their loved ones," York says.
For now, the human element is still key when considering the impact of AI art. "Generators can be seen as an evolutionary toolset for creators who are trained professionally, as well as artists in the non-traditional sense. This toolkit helps to level the playing field so that instead of being limited by skill level or resources, individuals can use AI to generate unique pieces of art," York explains.
Macci Ex Machina clown portrait generated by machine learning AI. (Photo: AI)
York believes in a bright future for AI art, expecting it to become more mainstream and take on more diverse applications in advertising, entertainment and other industries. "There definitely needs to be recognition if not acceptance that this technology will be a part of human discourse in the coming years," she says. "We believe it is the context, the expression and the idea behind these artworks that captivate and move the human spirit, and isn't that the very idea of art itself?"