Exploring Artificial Intelligence

So, generative AI.

I'm nervous about addressing this because the use of the tool is so very controversial.  Concerns about the economic and social upheaval it may inflict on artists, fueled by charges of plagiarism and piracy, make it a delicate subject to approach.

Never the less, like the printing press and the camera, AI is here and its probably not going away.  Whatever ethical and social issues it creates, those issues are borne of the reality that it is a very powerful tool.  Understanding it and exploring its use seem like necessary adaptations, even if one doesn't use it in one's art.

There's also the philosophical implications of AI art, or really, digital art in general.  Consider Magritte's work The Treachery of Images.  A painting of a pipe is not, in fact, a pipe.  But what is a painting of a wholly invented digital object or 3d space?  If I were to model a pipe in Blender, render it into a photorealistic image, and then use that image as a reference for a painting, what would I have painted?

This is not a not pipe?

AI (and digital art in general) make images even more treacherous than Magritte could have imagined.  In a world where machines are now able to produce such images, arguably at the expense of human artists who unwillingly contribute their work, how is an artist to proceed?

How could an artist utilize this tool in their process without relying on it completely as the totality of their process?  How could it be used in a way that is cognizant of the economic and social concerns as well as the intellectual property issues associated with the tool?

That's the basis for my exploration here.

For me, at least, I've identified two possible applications:  AI as a tool for generating reference images and AI as a tool for stylistic polish.

Neither of these applications replace the artist's input.  Rather, they inform the artist's hand in much the same way as any other tool of observation, be it a ruler or a reference photo.  They utilize AI as a very powerful and sophisticated tool of the artist, rather than replacing the artist completely.

Neither crosses the line into plagiarism.  Whatever the source of the images the AI draws upon, and whatever process it subjects them to, the final results of the process is far more dependent on the artist than the AI.  The outcome are images that are obviously and demonstrably the outcome of the artist's hand, as opposed to that of the tool or some hapless victim of internet image scraping.

ARTIFICIAL INTELLIGENCE AS REFERENCE IMAGE

One way in which AI could be useful is by simply providing inspiration for the composition of a piece.   In this instance, the artist might prompt the AI for a particular type of piece and then use the result as a reference image for their own work.

This it seems to me is no more dishonest than the referencing of an image clipped from a magazine or copied from google.

Technically I've used the AI generated image on the left to produce the hand-drawn image on the right, leaning on it heavily for the composition.  It would be difficult to say that the image on the right depends on the image on the left, though, or that any part of the AI generated composition isn't a well established and commonly used' element of this kind of image.

ARTIFICIAL INTELLIGENCE AS FINISHED IMAGE

The other way in which AI could be useful is by producing variations on the artist's own vision.  In this case, the artist could provide the AI with an image they've produced themselves, and prompt the AI to modify the image in some way.

A very quick, rough image by my own hand on the left, and the more polished, finished result produced by the AI based on my image.  Again, there's not much question here that my work is the source of the AI's produce, and much like a camera, the AI has taken a considerable amount of time out of the process by advancing it considerably towards a finished work in a matter of seconds.  From here I could re-engage with my own hand, although it's passable as it is for some purposes.

I'm admittedly more conflicted about this approach. The artist's hand disappears somewhat, leaving the AIs interpretation of their work rather than the work itself. Again, though, I look to the camera. A photographer shoots a hundred pictures of the same subject with small variations and then curates that mass of results to select the one that best represents their vision. The photographer hasn't drawn anything: They've selected a subject, adjusted and aimed their tool, and then edited the outcomes.

Using AI in this way is essentially the same process, I think. I struggle with it, though, I suspect due to the conceits of considering myself a studio artist  rather than a photographer

Is it any less legitimate than the use of a camera to produce highly accurate, illusionistic images?

CONCLUSIONS

AI is obviously a powerful and versatile tool that has some legitimate uses.  Misgivings about plagiarism and data scraping can be mitigated by confining the use of the tool to specific steps in the workflow (such as the examples of conceptualization and refinement).  By doing so, the AIs influence over the final piece is regulated in the same manner that the use of any reference material or artistic inspiration is limited in more traditional art, leaving the artist to take advantage of the tool's efficiencies without surrendering their agency to the machine completely.

NEXT STEPS

The next exploration would be in combining the two techniques in one work:  Using AI to generate a reference, working myself from that reference, and then running it through AI to produce a more final piece.

I'm also curious about multiple iterations of the process by taking that more final piece, working from that as a reference, and so on.