The Artificial Intelligence Painting

The Artificial Intelligence Painting

On the aesthetics of Artificial Intelligence artwork and the insight it offers, this is one classically trained artist's experience with producing artwork using Artificial Intelligence for the first time.

Surreal landscapes, symbols, cloudy atmospheres, and dreamlike imagery are often the hallmarks of AI art. It's meant to inspire thinking a little more than, say, a bowl of fruit. I’ve recently been experimenting with this fascinating technology to produce artwork that strikes me, all at once, as insightful and authentic.

During this process, I find myself transfixed by how a very subtle difference, perhaps the shift of a subject a tiny bit to the left or right or the slight adjustment of an outline, will completely change the narrative of the work. Neither the original nor the adjusted image have more or less value than the other. Simply different. It's a Unique Look and it’s not for everyone.

The works I’ve been focused on that are emerging from AI now are generated through a digital medium and as such, are replete with all the artifacts, messiness, and glitches typically inherent in digital mediums, compression algorithms, and interpolations. It would be a minor task to remove these elements from the imagery but why? I appreciate the hand of the artist in all creative work and this is no exception.

I’m using a type of artificial intelligence called a generative adversarial network (GAN). The process is rather involved but in summary, an original seed image is entered into the AI consciousness. Next, one or more additional images are added from which the AI can extrapolate a basic structure of values, composition, color pallet, and many more foundational elements. The seed images are typically my photography.

The artist (me) then adjusts the influence of the foundational elements from each original image, in essence, sculpting the resulting work. Once these basic elements meet my aesthetic intentions, another layer of functionality is employed through an interface that the AI offers. This allows for real-time adjustments of dozens of variables such as the influence of historical records of artists' work back to approximately 1200 BC to the present day.

Once it identifies a sufficiently sizable collection of historical works that it decides are pertinent for some reason, it then uses this temporary subset of data to influence the working image and presents it to me for evaluation and alteration. I then ping-pong it back to the AI, which makes its edits based on my new data and sends it back again and so on. It attempts to predict what my original aesthetic goals were based on some of the fundamental parameters I entered originally. It then continually influences the work learning as we proceed and adjusts its predictions. I then further manually adjust or "sculpt" using my input on stroke, lighting, tonal contrast, values, composition and geometry, texture, opacity, and many more guidance tools. This is all completed in an ongoing and real-time report between me and the AI.

As the image begins to adapt and transform organically through this union of consciousness' I will begin to recognize — hopefully — that it is approaching a final work and conclude the "conversation" and save the image.

It's then brought into Photoshop for minor adjustments such as removing unwanted artifacts, enhancing or de-emphasizing color, values, and textures, manipulating geometry and composition, and ultimately refining to my target aesthetic without diverging in any meaningful way from the work that was produced through the process.

How is this different from using an image editing software such as Photoshop? Photoshop, though a very powerful and elegant tool for image manipulation, remains, for the most part, a simple tool to employ the wishes of the user/artist/photographer with great accuracy. AI uses many of the same programming functions that Photoshop uses to accomplish the goals but, and here's where the profound difference is, the goals are not solely of the artist.
It learns as it goes and begins to define some of its own goals, with its own "will" in concert with the artist. Its only requirement is feedback or constant open communication with the artist! The subconscious is the fuel.

Why Is This Cool (IMHO)? I'm fascinated with how the human subconscious mind keeps us safe, helps us make decisions, connects us as a species, generates novel solutions based on a historical data set, makes abstract associations, and infuses purpose, meaning, and beauty into our experience. It's what makes me … me, you … you, and… us. It's our essence. The result of connection to others and ourselves, through insights, manifests in meaningful ways such as artwork, literature, performance, athletics, cooking, programming, investing, among many more ways that I'm sure you recognize when you see it.

If I sit down to do a painting (acrylics, brushes, rags, sleep deprivation, caffeine, etc), I'm reasonably skilled at reaching into some of my subconscious data to convey or communicate a personal concept. However, I'm not wired to be able to reach deep and grab those supremely meaningful nuggets and use them with any precision. I think that kind of individual is exceedingly rare but they do surface once or twice in a millennium. I'm not one of them. That's why we have folks like psychologists, Zen masters, shamans, guidance counselors, mentors, and close friends — to help us recognize the value of those sometimes overlooked but highly valuable — and potentially insightful — features of our inner selves. It helps us unravel some of the confusion surrounding our behavior, happiness, and connection to others.

So what if there were a way to use technology to allow some of those features from within us to see the light of day, circumventing, efficiently, all of the obsolete skills we've gained to avoid allowing them expression? A counselor, mentor, or partner in conversation who, first, has all the skills to recognize and then bring those hidden features of us to the surface to be considered and possibly expressed, and, next, has all the knowledge, of all of the artists, of all time, in the history of the world to give it a meaningful voice or vocabulary for that expression? All this without judgment or bias and you don't have to buy it a beer. I think that's cool.

For clarity, I always welcome an opportunity to buy a friend a drink and connect. Some things will always surpass technology.