FAQs

How were these images created?

The AI-generated images posted to my Flickr photostream were made either with Stable Diffusion or Midjourney. Each image has been placed in a folder corresponding to the primary program with which it was generated.

What is the division of labour between you and the AI in the creation of these images?

I decide what to create. The AI does most of the heavy lifting. Sometimes it’s like commissioning an artist to create a piece and leaving them to deal with the technical details in their own way. At other times, it’s like art direction. There are various degrees of involvement, constraint, and control.

Creating an image may involve little more than feeding a text prompt to the AI repeatedly until it finally spits out an image I like. Or it may require several different text and image prompts, multiple passes from one AI program to another, local alteration of individual image elements, collages, working with multiple layers in Photoshop, hours and hours of work.

Are your posts based on real photographs?

Not usually. But they are often based on some of my older AI-generated images.

Does creating art with AI amount to intellectual property theft?

There are partisans on both sides. My personal opinion is that while it is possible to unlawfully infringe on someone else's intellectual property while creating AI art with Stable Diffusion and Midjourney, this is by no means an inherent part of the generation process.

Are the subjects in these images fictional?

None of the images posted are of living persons. Given, though, that there are close to 8 billion humans on the planet, a generated person may coincidentally resemble a real one. All such images will be removed if brought to my attention.

I do, however, sometimes prompt for long-dead historical figures (like Mary Queen of Scots or Madame de Pompadour). My goal on these occasions is stylistic evocation rather than verisimilitude.

Do you consider the potential for perpetuating harmful stereotypes through your work?

The nature and size of my audience are such that this isn’t a major concern. It should in any case be possible for adults to explore visual tropes without having to be hypervigilant about the possible effects this exploration may have on overly sensitive, unseen, and quite possibly imaginary observers.

Are you concerned about the impact of these images on societal beauty standards and individual self-perception?

I get asked this a lot. The short answer is 'no'. Unrealistically beautiful women and unrealistically handsome men are staples of everyone’s media diet. What needs to change is not the proliferation of these images but our relationship to them. It doesn’t matter where the bar is placed - as long as there are “beauty standards”, there will always be literally billions of women for whom those standards will be unattainable at some point in their lives.

Is Flickr an appropriate platform for AI-generated imagery?

AI-generated images are permitted on Flickr, as long as they’re appropriately categorized and consistent with the site’s content policy - as mine are.

That said, the site works best as a forum for the sharing of photographs by people who’ve actually taken them, and I would fully support the introduction of a filter mechanism to exclude AI images from users’ feeds unless they opt in to seeing them.

Why do you choose to create AI-generated images over traditional photography?

At first, it was curiosity - exploring the new tech. Now it’s because the reward-to-effort ratio for my AI-generated content far exceeds that of my photography.

Have you considered the benefits of returning to photography with real models?

I still occasionally shoot with real models, mostly “content” shoots for my Patreon. I am, however, increasingly reluctant to share photographs of real people on Flickr.

Can you create custom images upon personal requests, such as those of a spouse or partner?

Creating such images (especially if they are NSFW) requires the consent of the persons to be depicted. I am unlikely to generate images of spouses or partners, though I may be theoretically open to creating images of the person making the request (if I am reasonably satisfied that they are in fact the person they claim to be).

What ethical considerations do you take into account when determining the apparent age of the subjects in your images?

When I specify a subject's age in a prompt, it's always 18+ (usually 22-26). But AI subjects have no personal histories and no “real” age. There is no objective biographical truth about them capable of independent verification. With actual people, you can always point out that though your model may look like a teenager, she is really 22, 25 or 27. That defence doesn’t work quite as well with AI. Judgments of this sort are necessarily subjective.

I am, in this regard, responsive to the opinions of others. If I post an image and the consensus is that the fictional subject’s apparent age lies too close to the borderline, I delete it, even if I do not share this conviction.

How do you navigate copyright issues when your AI-generated images closely resemble existing artworks or photographs?

Hasn’t happened yet - and given the nature of the images I generate, it isn’t likely to happen. But of course, if this situation were to arise, I would delete the offending image without hesitation.

How do you respond to critiques that your work might contribute to objectifying women?

That’s a difficult one. I suppose my response is that the images I create are, in fact, visual objects and are meant to be treated as such. They do not depict actual people who might have feelings to hurt or other human traits worth valuing. They are, in fact, alternative foci of interest. When we cast our objectifying gaze (on these entities that happen in this case to actually be objects), we are not objectifying women.

But does permitting or encouraging this attitude toward AI art have pernicious effects on the culture at large? Does it foster habits of mind and speech that inevitably bleed into the real world? That’s a variant of an old and contentious question: how does our media diet affect our behaviour? In what direction does the arrow of causation point? Does watching violent movies make me more violent or might they be release valves, outlets for urges that might otherwise find expression elsewhere? If I cheer enthusiastically as a fictional villain is dismembered, do I debase myself? Am I thereby contributing to a culture of operatically choreographed violence and villain dismemberment?

How do you ensure diversity and inclusivity in your AI-generated art?

Angela Carter once described pornography as “art with work to do.” While what I create isn’t pornography, the vast bulk of it falls under the rubric of, to use the currently fashionable term, “content.” It has to work in a certain way, and there are definite criteria for success and failure.

In this context, diversity and inclusivity apply only to the extent that it would be unwise of me to allow any prejudices or preconceptions I may have to unduly restrict the range of subjects I explore when generating images. To this end, I make extensive use of wildcards that, for instance, insert a randomly selected nationality into a prompt when I need a subject. Some selected items work better than others, of course, both based on my subjectivity and on the expressed preferences of my audience. Successful terms and subjects are reused. Unsuccessful ones are modified or discarded.