AI and the Creative Process

The Creative Process of Working with AI: From an Artist’s Perspective

Introduction:

The dawn of artificial intelligence (AI) in the creative realm has set the stage for passionate debates on the essence of creativity, craftsmanship, originality, and even the notion of authorship. I invite you to consider a new scenario: "When a human guides a machine to create, does it diminish the artistic integrity of the creation?" Better yet, "Does a piece of art need the physical touch of the artist to be deemed 'art'?"

Conceptual Art and Context:

As someone who finds solace in the realms of creativity, albeit more in the conceptual space than in the studio currently, I often wonder how these technological advancements change our perspective on what we consider as art. And despite not recently making brush strokes on canvas in my studio, I still perceive myself as an artist, even if many of my mental compositions haven’t reached tangible form. This raises another interesting question – is an artist defined by creating physical artifacts? Or can one be an artist by nurturing creative ideas that never take tangible form? To this latter question a layman may say no, but I imagine a great number of art historians, art theorists, and many artists would answer with a definitive “yes.”

Artists have always pushed boundaries when it comes to what may be considered art. With each new push forward, art has deconstructed itself to where it could be something quite unexpected—like Marcell Duchamp’s “Fountain,” which was just a prefabricated object, manufactured, but took on new meaning due to its contextual environment. This deconstruction can even go further where the artwork exists solely as thought without any tangible form ever existing; constructed not by physical materials but by an idea, or concept. An artform known as “conceptual art”, and not to be confused with “concept art”.

One notable conceptual artist is Yoko Ono, known for her "instruction paintings" that often exist only as written instructions for a work of art, allowing the reader to imagine the artwork in their mind. An example is her piece "Lighting Piece" from 1955, which is merely a set of instructions reading, "Light a match and watch till it goes out." The artwork is the act of following these instructions, not a physical object that can be displayed in a gallery.

If you’re like most, you might be uncertain how an artwork might be considered art without any physical properties. Let me try to explain: a comparison can be made to the concept of a song. Imagine a hit song you like, for example, "Imagine" by John Lennon. This song can exist in multiple forms—it could be on vinyl record, as a digital file, or even as ink on a sheet of music. Yet none of these is the song itself; they are just physical manifestations of it. The 'song' is a variable blend of composition, context, and the experience of the listener. And even without these physical forms, the song can still exist—something you can keep, stored away in your memory to be experienced and even shared, if you’re brave enough to sing it out loud. If you would like to see an example of this concept that helped me wrap my head around it, I would encourage you to explore Joseph Kosuth's "One and Three Chairs."

Art and Artificial Intelligence:

Yet, despite the fluid nature of art's definition throughout history, creations generated through collaboration with AI seem to be viewed differently, by a seemingly vast number of people. Novel creations, perhaps, yet not quite art. But I also remember this being true when digital media entered the scene. And yet noted contemporary artists such as Kara Walker and Bill Viola have successfully incorporated the digital media into their artwork, and thus helping to cement the medium into the art arena.

Even among my artist friends and non-artist friends alike—I would say the vast majority regard art generated with the help of AI as something different; seen as creative perhaps, but not art in the traditional sense. And I can understand why they might.

For example, AI generated images, such as those created with DALLE-2 or Midjourney, are hitting mainstream attention at an alarming rate. And I’ve observed, as many have, I’m sure, where even the most casual user will refer to their AI generated image as “art”. Simply because they've produced an aesthetically pleasing image, perhaps. As a friend recently observed, “Seems anybody and everybody is an artist, now”. The ease, speed, and immediacy of the creative process when working with AI is startling. But is that immediacy, that speed and ease, lines art shouldn’t cross? I would imagine Picasso might say otherwise, even if in theory.

Let me be clear, after a lengthy bit of tight-rope walking— I do not think every image generated by AI, even at the hands of very skilled prompt-engineers, is art. But I absolutely believe that we can create art in collaboration with AI. The mystery and the potential, I think, resides somewhere in the alchemy of concept, intention, context, and perhaps most importantly, timing. Getting everyone to agree is something else, entirely. An example of how this might get murky is the controversy that arose when a piece of AI-generated art, "Portrait of Edmond de Belamy" by the French art collective Obvious, sold at Christie's for $432,500 in 2018, sparking heated debates about authorship and originality.

My Practical Perspective:

I realize that I’ve spent a considerable amount of time discussing art theory, rather than focusing directly on my perspective of working with AI, as promised in the title. However, understanding how we create with AI and how we define these creations forms the foundation of my view in this new territory. Exploring the theoretical aspects of AI in art isn't merely a tangent—it's the ever-present Jiminy Cricket that helped guide my adventure. And hoping that I’ve provided a baseline of how I think art and AI might intersect, let me now share my personal journey navigating this intersection. And I promise, the practical, hands-on perspective I'm about to offer isn't quite as steeped in verbosity or nuance as all that art theory stuff, before.

For just over a year, I've been using AI to develop concepts and generate images. Most of which I would unabashedly define as design. But there is also work I would define as art, believing its merit right alongside one of my paintings or sculptures. And along the way, I learned some things about the process of working with AI. Here are my thoughts, and personal opinion, for what it’s worth.

AI is an extraordinary partner. But it's no magic wand. The depth, complexity, and nuance of AI-generated output are vastly dependent on the user's input and guidance. And while “anyone” might have the AI generate an image of a brown dog on a green lawn, capturing the 'essence' of our companionship with our faithful friend is another story. Even if one wishes to approach it in a more traditional manner, there are still limitless possibilities when crafting the aesthetic elements such as lighting, color harmonies, mood, and even the energy of brushstrokes, if painting is the style you’re aiming for. This more specialized approach requires a solid understanding of art history/styles, art processes, and the time to hone your working relationship, and prompt engineering with AI.

The dark side of this story? AI undoubtedly has potential to change long established practices and even replace certain creative jobs. This is scary, especially if you’ve based your career and identity on being creatively skilled, as I and many others have. But I also can see that AI has potential to catalyze a renaissance in creativity, visual communication, and in learning. And yet, this foundation couldn’t exist without the wealth of experience the global community has contributed. This brings up another important consideration—that being how AI is trained using existing contributions and styles of a vast number of creators as a training ground. This expansive playground is what allows the AI to reproduce or even combine artists styles. While questions about originality and authorship have always nipped at artists heels (looking at you, Picasso), this is something on another level.

Tips for New Users of AI:

Even so, one can still utilize the best of AI while being mindful of the foundation on which it was built. And if you’re looking to start some creative projects utilizing AI, I can share some insights I gleaned from my adventure.

AI Can be a Brilliant Collaborative Partner:

Working with AI can be like having an exceptionally agreeable colleague with limitless imagination. And while harnessing AI's full potential might be dependent upon the crafting of your prompt and the approach of your goal, the AI is likely able to teach you along the way.

Approach it like a Dance:

Collaborating with AI is like a dance—a responsive movement where both partners contribute. You might need to lead the dance by providing input but stay nimble when AI offers its interpretation and suggestions. And when working with images, refine your steps with iterative prompts.

Embrace Serendipity:

In art, serendipitous moments often lead to remarkable creations. When working with AI, unexpected outcomes become part of the creative process. Try to be open to some unexpected surprises. These moments might allow you to explore uncharted artistic territories.

Find Your Balance Between Control and Chaos:

Artists often grapple with maintaining control over their work while embracing spontaneity and chaos. This balance becomes even more compelling when collaborating with AI.

Share Credit in Your Collaboration:

When presenting works created with AI, I think it's crucial to recognize that these are joint efforts—a fusion of human creativity and machine intelligence—which led to the outcome. Acknowledging this also allows for better understanding with your audience.

Conclusion:

My journey of working with AI in the creative process has brought me face-to-face with many philosophical and practical questions. What does it mean to create? How is creativity evolving with technology? Who is the artist in the age of AI? I don't profess to have definitive answers. However, I am certain that the narrative around art and AI is far from over; it is continually evolving, just like the technology itself. The essence of the journey, I believe, lies in engaging with these questions and in embracing the uncertainty. After all, isn't that what art is all about—pushing boundaries, exploring unknowns, and embracing new paradigms?

To echo Picasso's famous words, "Every act of creation is first an act of destruction." As AI disrupts our traditional notions of creativity and authorship, it also paves the way for a new kind of art—an art form that blends human intentionality with machine interpretation, giving rise to creations that are as intriguing as they are unpredictable. And isn't that the very essence of creativity—a dance between the known and the unknown, the controlled and the chaotic? As we navigate this new landscape, it's important to remember that we, as creators, hold the reins. The onus is on us to work with AI responsibly, ethically, and, above all, with respect.

Note: This perspective as been written in collaboration with the AI, ChatGPT-4. If you’re curious to see the work that I and ChatGPT-4 (Chat) have done together, I encourage you to visit the website, akaart.studio, dedicated to documenting our journey.

The Creative Process of Working with AI: Sidecar

John Hammer

A visual journal highlighting the journey.

When I gained the early-beta access to DALLE-2 and Midjourney (v.1, at that time) last summer, I was exceptionally keen to see what was possible. In the beginning, I primarily focused on DALLE-2, then with the upgrade of Midjourney to v.2, I began to work more heavily with that AI.

Because I tend to be a rather curious cat, I sometimes approach new things in my particular way. Before I made my very first image in DALLE-2, I spent a considerable amount of time planning how I was going to get started. I knew that I didn’t want to read how others might be doing it. For whatever reason I wanted to go into this as fresh and as unbiased as possible.

I also wanted to establish a baseline of what the AI was capable of. I thought a good way to begin would be to ask DALLE-2 to create something that might be more challenging due to the subtlety required in maintaining separate elements in the image, such as texture, volume, and lighting. My first image and challenge was for the AI to create an image of a white cat sleeping on a white bed in a white room with soft morning light.

John Hammer, DALLE-2

The white cat, white bed, white room, challenge.

First, I was surprised by the AI’s ability to understand my request and generate an image that reflected the elements I thought would be challenging. However, I also felt disappointment. Despite multiple versions of the prompt, the AI struggled to generate specific facial features in detail. This challenge persisted for quite a while, with nearly flawless results only recently achieved - even with human faces.

John Hammer, DALLE-2

The concept of "longing' expressed using water.

I also began to experiment with form and material. How might the AI interpret requests that would force it to decide what an abstract concept would look like, such as vanity, or empathy, or longing, if only allowed to use an unconventional material, such as water?

I then began working to see how I could take one thing, like a jellyfish, to serve as the design model for multiple design concepts, such as architecture, moisture collectors, site-specific sculpture concepts, and even fashion and fabric design. At this juncture I was also working heavily in Midjourney.

John Hammer, DALLE-2

Machinery based upon the jellyfish form.

John Hammer, DALLE-2

A concept for a kinetic jellyfish sculpture for the Kansas praire.

John Hammer, DALLE-2

A jellyfish inspired raincoat.

The results came with a realization. There was no way I was going to get ‘exactly’ what I had in my mind’s eye, but the tradeoff was something that I was far happier with; the AI is more of a creative collaborator than a tool. Treating it and expecting it to behave, at least in the creative arena, as a tool leads one to expect repeatable results. This is never the case with AI image generation, even if using the exact same prompt. And to navigate this I elected a more flexible and collaborative approach. Surprises can be wonderful gifts.