SPOOFED AI GENERATION: A PAINTED PORTRAIT

Self Portrait: From My Dataset, Oil on 3 canvas boards, 2024

Generative AI is the talk of the town. 

I am particularly interested in gen-AI image production. 

Fast AI generated 'paintings'! 

Gen-AI image production has provided the vehicle for tech-bros types, opportunists, and wannabes to become 'artists'. After a prompt, the speed of generation induces exclamations about how fast images can be 'created', 'painted'. I have seen these words used, and yes, I have pointed out, for example, that a generated image may have elements of various paintings, due to data scraped from online sources, but this does not make the generated image a painting. Rather, it falls into contested areas of copyrighted image use! Yes, the generated image might simulate painterly-textural elements, but again this does not make it a painting, because a painting is a material object, created with actual materials, such as, canvas, brushes - and paint! 

The clincher is that the gen-AI 'painting' has been GENERATED using algorithms - it has not been created using paint. If the generated image was to be made into a material object, it would need to be printed, and then it would become a print. And, there's a huge range of print types, from monoprints, limited edition artist prints (lithographs, etching, screen prints etc) to the mass produced prints found in places like IKEA, news agencies, KMART etc. And, of course, the use of canvas giclee prints (not a fan!).

Last year (2023) I made all these comments (above) to someone on TWITTER (X), in response to their excited tweet about their speedily produced series of gen-AI 'paintings' - and - their response was something like, "Hey, good points"! It was like a revelation to this person! This is why artists and art historians with a broad experiences and understanding across multiple creative practices and histories need to engage in current debates about generative-AI image production. Some of the tools list 'styles' in which a prompter can choose for their image generation. This is akin to a list of font styles - but choosing a font style to write text, does not make a person a good writer, nor a creative writer, a poet, a speech writer.... 

On the issue of speed and generative AI, I have many concerns about the lure of speed and its effect on contemporary world-building activities. These concerns are based on my PhD research, an examination of increasing military interest in the electromagnetic spectrum. But, that's for another blog post! In the meantime, here's an article I wrote, published in Digital War Journal, Light-Speed, Contemporary War, and Australia's National Defence Strategic Review 

Is 'cultural product' a better description?

Although my comments above may not indicate it, while I am concerned, I also find the explosion in generated AI image production to be extremely interesting, for many reasons. I do prefer, however, to view the phenomenon as cultural activity, with image outputs as cultural product. I propose that these descriptions avoid the binary question - is it art or not? I hasten to add that art is also a cultural activity. Thus, the topic can be opened up to embrace wider cultural and social issues, for example, the influence of tech platforms on increasingly homogenised visual aesthetics across various aspects of cultural production - design, fashion, advertising, football paraphernalia. and yes, art too? One of my hobby-horses is the homogenisation and standardisation of car design!

I'll add here, to avoid being called a technophobe 😁, some images and videos derived from gen-AI tools, could be called art. It's what someone, an artist, does with the tool or the tool's product generation? For example, if an artist develops their own datasets and experiments with them, or 'breaks' the technology by introducing glitches or other interstices, or does something with a generated image, for example - prints it and then draws or paints over it, or rips it up and weaves pieces into other ripped up pieces. There are so many possibilities. 

One needs to keep in mind that art, whether AI assisted, painted, sculpted, digital or other, can be good or bad, fashionable or cutting edge. Critical appraisal is another subject for another day. 

'Art' and advertising
Simply ascribing the term 'art' to a product/service/tool does not elevate the product, but some opportunistically seem to think so. The term 'art' has become a marketing word or medium for companies that develop gen-AI tools. This reeks of desires to legitimise and elevate their products. But, this reeking exposes ignorance, opportunism, carelessness. 

John Berger's observation, in Ways of Seeing (book and TV show-1970s), that during the mid 20th century famous paintings/art were used to advertise completely unassociated products, such as alcohol and cars, provides a way to critically think about how the term 'art' is used to market contemporary tech products. The 21st century twist is that the so-called 'art' and the tech being promoted/sold are less divisible than in the mid 20th century promotions. 

Divisibility is illustrated in the image below, a photo from a 1967 edition of a French magazine, Réalités, that my mother used to subscribe to. A detail of Emil Jean Horace Vernet's painting Bataille du pont d'Acole' (1826) is used to promote a product, Courvoisier Cognac. AND, while the advertisement in effect cheapens art as an advertising motif, please notice that the advertiser places a caption acknowledging the artist, the painting, and the context of the painting. This does not happen with images now scraped from online sites where artists works are posted. In the twenty-first century the cheapening continues at speed! Ironically, this is at the same time people aspire to be called artists, many claiming the anybody can be an artist, and that all artists are inspired by other artists, so copying is ok etc! Regarding the latter, the art theory term appropriation is another topic - for another post! 

From Réalités, November 1967. 

Self Portrait: From My Dataset, 2024

SO, my self portrait at the top of this post, is a painting, an actual painting - a triptych painted over a number of months. It is a spoof of gen-AI processes and outputs. This self-portrait seems to be grappling with how it might emerge. Each piece of Self-Portrait: From My Dataset conveys something about me; my likeness, my personality, my geeky sense of humour. Therefore, each piece is a portrait/self-portrait. Combined, the three paintings are another self-portrait.

The first piece of the triptych makes my 'prompt', To paint a portrait of Kathryn, appear like the headline for a fabulous show! Is it a gameshow, a carnival oddity, or....? This is a critique of the importance placed on prompting, now a 'profession' eg: prompt engineer. 

The middle piece is my image 'dataset', but it includes images chosen for reasons no scoping/scraping algorithm could detect. 

The third piece is the final 'generated image', but there seems to have been a glitch because the 'AI' has not been able to pixelate a final 'perfect' image. Have correlations from the dataset and statistical probabilities failed to sequence? Did my painted dataset of characteristics no algorithm could scope, trick the tool? Or, maybe it was because, over the few months I took to paint Self Portrait: From My Dataset a few insects got stuck in wet paint. While I did attempt to remove them, and repaint areas, I am pretty sure insect body parts may still be embalmed in the paint. Pretty sure an AI tool would not have to deal with a stuck insect! But, for an artist - a painter - this kind of occurrence is all part of the process of problem solving. 

NEWS

• My commentary piece, "Light-Speed, Contemporary War, and Australia's National Defence Strategic Review", published in Digital War Journal.
  
  
• Visual essay chapter "Imaginational Metaveillance: Revelations in the Drone Age", published in Drone Aesthetics: War, Culture, Ecology  Open Humanities Press. 

BOOK LAUNCH for Drone Aesthetics: War, Culture, Ecology on Wednesday October 2, 2024. Please register through the Centre for Drones and Culture, The University of Cambridge, site 

• Visual essay chapter "Imaginational Metaveillance, Creative Painting Practice and the Airborne Drone", in Drones in Society: New Visual Aesthetics  Palgrave MacMillan.
  
  
• In November 2024, I am presenting at Artificial Visionaries two day workshop, The University of Queensland. Humanitix Registration   
  
  
• In March 2025, a major solo exhibition DRONE: Ghosts and Shadows at the University of Southern Queensland Art Gallery, Toowoomba, Australia. 

Cheers,

Kathryn