The tension surrounding the admission of art made with generative AI is akin to that experienced with the advent of digital cameras. Even if generative AI technology is inevitable, as happened with digital camera technology, we still would benefit from understanding what the transition is about.
Scouring the internet reveals AI exhibits and forums are not new. Public lectures have also contributed to understanding how this tool fits into the history of art. The Nasher Museum even experimented with AI being the curator for an art exhibit. But other venues have resisted art made with generative AI out of fear of litigation and an inability to judge how they were made. The tension surrounding the admission of art made with generative AI is akin to that experienced with the advent of digital cameras. Even if generative AI technology is inevitable, as happened with digital camera technology, we still would benefit from understanding what the transition is about ─ its pros and cons.
What is needed is combining an AI gallery art exhibit with a conversation among the participating artists exploring the strengths and weaknesses of AI art and how it fits into their toolsets. This approach moves us beyond the often sharp-edged debate and the stereotypes about generative AI. Such an exhibit would not be about “see how pretty my image is.” It would be about seeing and understanding the creative range of the art that emerges.
The Escondido Arts Partnership has now exhibited an AI art show I am AI, Are You? The artists in this exhibit have had a long history in making and exhibiting their art, primarily in digital art and photography. Since we have been immersed in digital media for several decades, it is understandable their affinity for experimenting with another digital tool ─ albeit having a quantum difference in concept, technology, and results. The artists then gathered for an online discussion to explore how AI entered into each other’s artmaking. The artists in this enterprise included Don Archer, Stephen Burns, Eric Johnson, Greg Klamt, Kazmier Maslanka, Joe Nalven, Jack Quintero, Larry Vogel and Jill Rowe.
What did they find interesting in making sense of generative AI as they crafted their imagery? Is it like fractal art? Artists have been entering numbers into mathematical algorithms since the late 1980s to generate kaleidoscopic images. Now, generative AI uses words as text prompts to create images such as the one used for this exhibit. Numbers are occasionally paired with magical qualities, but when it comes to daily use, numbers have a limited assigned meaning. Contrast that with words. Words generally have connotative meanings. They’re messy. Putting in a text prompt is far more likely to generate outrageous images that require AI modes to put in safety policies.
Further, generative AI models are trained on the images of thousands and thousands of others. Not so for fractal programs. Should we understand AI images as luck? Images that develop from expanding the canvas and letting generative AI fill in an expected what-should-be-there is now there can be seen as a matter of luck. It’s not really there. The same with text-prompts that yield an array of visual interpretations. They are not really there. These images were not intended by the artist as Mona Lisa’s smile was intended by Leonardo DaVinci. So, AI images can take on the semblance of luck.
Other artists view the results of generative AI as surprise ─ something they would not have expected, nor planned by themselves, but now pushing the artist to think “why not?” These surprises can be a new point of departure; perhaps they will motivate the artist to take a different path or simply re-energize the artist to do art. For others, the surprises are just luck and dispiriting, and a reason to abandon this tool.
Where is the “me” in images that use generative AI? If one begins this discussion from the perspective of the U.S. Copyright Office, AI images can’t be copyrighted because they lack human authorship. In effect, there is no “me” in generative AI art. However, copyright is not the arbiter of art, but of legal protection to those it deems worthy within a certain time period. Most of art generated by humans has not and cannot be copyrighted since it was created before copyright. Those protections were matters of secrecy. Other works that were copyrighted are no longer since they have fallen outside the period of time granted. AI is different; it is excludable material. And it is not created by humans.
Copyright is not the only metric to judge art. There are curatorial, commercial, cultural, historical and other frameworks. We can choose to find the “me” in generative AI art by considering the creative process. Arthur Koestler saw creativity in science, humor and art as, in part, one of mature or informed play. One can leap between different frames of endeavor and understanding as play; this is not the play of children, but of those who are knowledgeable about their field. In this broad sense, artists can juxtapose ideas and tools in creative ways; they can play with paintbrushes or algorithms.
Lest the conversation among these artists be misunderstood, the conversation delved into complex workflows; it was not limited to the bare bones stereotype of text-prompt to image. The range of art created in part or in whole with generative AI went from the text-prompt to image to hybrid images to multiple platforms that required bringing the image into editing programs for assembly. One artist wondered whether creativity will always be located in humanity. Similar to the question about artificial general intelligence and whether AI will have emergent capabilities that places such models onto the same platform in which humans “think,” we don’t have the answer to that now.
The pragmatics of collaborative artmaking The art exhibited in I am AI, Are You? drew on a collaboration with or inspiration from generative AI. The artists were compelled to go beyond the broad recipe of how they made their art. If one thinks of making pizza, something more familiar to us, we can say we want olives, mushrooms and extra cheese; but the chef has to actually know the steps and timing for crafting it. The artists using AI proceeded into the steps, often tedious, of tweaking this and that part of the image, generally with multiple iterations, erasures, do overs, and fitting in different tools to arrive at their final image. And, no one had the same workflow. It was within this highly varied workflow that artists incorporated generative AI into their image making. Some artists knew exactly what they wanted at the start of their image making; others had no idea where they were going. Not unexpectedly, the creative process is a different way of being and of becoming for each artist. That is a far more difficult metric than copyright. It is, in fact, the human factor that the copyright office can only anoint, but cannot measure.
Do artists who use generative AI for making art need to know the underlying technology? This question was not included at the I am AI exhibit inasmuch as one of the artists could have been an AI theoretician or technologist. Nor was this question given much attention at the discussion tied to the exhibit. This blind spot may ultimately be irrelevant since the practice of art leans into the use of technology, both existing and new, rather than having the invention become the art object. Harold Cohen, the inventor of AARON and its practitioner of algorithmic art is a rare exception. There will be more AI exhibits and more conversations about what it is and what artists are doing with it. This is only the next step.
Joe Nalven is a San Diego-based digital artist. He is the author of Going Digital: The Practice and Vision of Digital Artists (Thompson, 2005). He has also written about how the San Diego County Fair is struggling with AI in competitions.
Joe Nalven