Why I’m Not Afraid of A.I. Art (and why you shouldn’t be either)

The latest controversy in art and design is the use of neural networks to generate images based on written descriptions. Some are championing these tools while others are vehemently against their usage. My current third-year topic studio is using these tools to determine their efficacy as reference, iteration, editing, explicitness, and novelty. We have begun to discover inherent biases in the images generated that could have damaging effects on how we design the built environment. Regardless of how you feel about these tools, here are some reasons why we should take advantage of them and why you shouldn’t fear A.I.

In conjunction with my Fall 2022 third-year undergraduate landscape architecture topic studio, the zeitgeist surrounding AI generated art is less than optimistic. Critics fail to see the opportunities related to high-craft, high-iteration style tools while champions fail to recognize the inherent biases of internet-trained neural networks. In a short Instagram post on September 27, 2022 I addressed some of these issues as they pertain to the ongoing work and research being conducted in my current design studio and what this means for the larger professions of art and design. These issues are accompanied with imagery that I generated using the MidJourney neural network that attempt to demystify the baseless fears of robo-superiority and a detrimental shift in the population of image crafting.

Technopanic embodies the spirit of ‘the good old days’

*A Socialist Poster Depicting the Printing Press*

With every new invention or expansion in technology, there are always those who fear the loss of certain values, morals, or ways of being that they equate with newness and progress. The printing press was going to undermine the church, the telephone would be used to invade peoples’ privacy, render engines would limit designers’ ability to draw, and AI will replace hard-working artists with soulless robots. None of these turned out to be true as the narrowmindedness of these claims comes from a place of fear, tradition, and status-quo. Let’s not be afraid of new technologies and embrace progress by understanding a craftsperson is not defined by their tools.

A Robot Does Not Have a Résumé

Neural networks have the ability to create imagery that rivals that of ‘human-generated’ imagery. By understanding how neural networks work, these cannot exist without human intervention and thus are generated by humans. Some code has the ability to augment and write itself, but we have yet to see an example of an AI image generation neural network creating an AI image generation neural network. Midjourney will not show up to your workplace, present a portfolio, and place you out of a job. If AI becomes part of an art or design workflow, it does so as a tool at the hands of a human. Let’s not fear what we do not understand but take the time to learn how these tools can work for us.

Code Does Not Usually Run Without Being Told to Do So

*A Small Car Driving a Bigger Car to the Grocery Store*

Just as AI does not have the capacity to apply for a job in the same way a human does, the code itself is not written in a way to remove the human element of clicking ‘generate’ or iterating on a prompt for generation. AI is not creating new AI and self-driving cars are not driving larger self-driving cars. Even if we were to imagine a world where these things are true, these systems operate in an anthropocentric world where changes, maintenance, and additional advancements are done so through human intervention. Let’s not be influenced by sensational media and embrace technologies that have in fact existed for decades.

This post was written before the release of Google Lab’s Pitchfork, which allows code to rewrite itself.

New Technologies Embrace a Mentality of Abundance

*A Designer Coding a New Program for 3D Modeling*

A studio artist that has spent decades honing their craft has also spent time evaluating other related works and has the ability to ingest, reason, and synthesize these works into their processes. There is not a single neural network that can replace this experience and an air of low self-esteem undermines these achievements. A skilled artist has the ability to recognize work that is visually similar to their own without discrediting their own creations. In the design profession, neural networks have the capacity to re-define the role of designers in a way that opens up the ivory tower. Many individuals have been barred from the art and design world by means of arbitrary qualifications that often have little to do with one’s ability to think. The world is more productive, interesting, and beautiful when everyone is allowed to participate. Let’s not gatekeep our professions with pedigrees and contacts but allow everyone’s voice to have a stage.

Art Requires Iteration, not a Secret Formula

*Well, I Guess its Back to the Drawing Board*

In my years of teaching computer-based graphics and digital design, there are always those that expect digital tools to behave the same way for each person. Students will become frustrated when they follow each button press, click, and numerical parameters and their work does not resemble my example. The phenomenon taking place here is not ambiguous instruction, but skewed expectations and a lack of understanding that there is no difference in output between a pencil and 3D software. We understand that drawing takes time to develop facility, but somehow lose this when it comes to computers; the touch of an artists is always necessary when creating imagery regardless of tools, techniques, formats, or methodologies used. The design world now understands the work necessary to produce a production quality render and rarely thinks that a push of a single button creates award winning work – innumerable decades of collective experience go in to creating spaces and the images of spatial conditions. The same is true for AI art: while highly legible and openly accessible, many of the images created will not be considered ‘art’ and those that are take just as long to produce without using these tools. Let’s not be enamored with ‘fast art’ and instead apply our valuable experiences to a tool that makes our imaginations tangible.

We Wrongfully Use the Words ‘Art’ and ‘Craft’ Interchangeably

*An Art Gallery Where No One is Looking at the Paintings*

We cannot discuss AI’s legitimacy in the art world without talking about the art world itself. Having been educated at prestigious art schools myself, I am privy to the exclusivity of the field – especially regarding the academic sides of these professions. The draconic notion of low art and high art still pervades the academy and demands legitimacy for ‘non-producing artists’ while discrediting makers and traditional artists as ‘craftspeople.’ There are obvious differences between the way we make art and the way we talk about art and hopefully these approaches iterate upon each other, but one is not more important or more prestigious than the other. I believe that image-making without discourse is considered “craft” while image-making while negotiating discourse is considered “art.” Let’s descend from the ivory tower and have actual conversations with those who view our work regardless of if they do it “wrong” – every reading is valuable.

Access to a High-Quality Camera Does Not Make You a Photographer

*Advertisement for the Newest Smart Phone*

The early 2010’s brought an abundance of personal devices with high-quality built-in cameras and a platform for sharing photos; these are of course smart phones and Instagram. The early days of Instagram was a frontier of legitimate photographers fighting for likes against mindless shutter-mashers. That is to say that the ability to take a photo does not enter that photograph into the discourse of photography. This is still true today as many photos taken by smart phone owners are rarely reviewed let alone thought about before capturing. When we apply this to AI image generation platforms and their general accessibility, the bulk of the images created are visual noise that do not have the capacity to be discussed in the same way as intentional imagery. Art comes from the artist, not the paintbrush. Let’s not assume that a hammer makes you a builder but discover how to use the hammer and when to choose a saw.

Traditional Creative Skills are Necessary for Novel Techniques

*Many Objects Scattered Throughout a Room Where Not a Single One is Discernable*

When learning how to draw with traditional art media such as graphite, ink, charcoal, or paint, it is understood that the bulk of facility comes with years of practice. It is also understood that there is a difference between virtuosity with one’s selected tool and fluency in representation. Somewhere along the way however, when the tools increase in complexity, we fail to separate the notions of ‘using the tool’ and misunderstanding what the tool can do for us; forgetting that advanced visualization software is in fact a tool. To subvert these fallacies, it is necessary to develop traditional creative skills before diving into something that is often reduced to a sequence of buttons or commands that will somehow create art; learn to sketch before you learn to render. Let’s not forget the importance of putting ink to paper, even in a world where the tools we use are advanced beyond our wildest dreams.

It Has Allowed My Students to Focus on Meaning Over Craft

*An Incredibly Intricate Façade on a New Building*

During the AI-based studio I wrote for the Fall Semester of 2022, there have been many discoveries that I hadn’t initially expected. The first is that text-to-image neural networks work exceedingly well as a writing tool. These tools such as Midjourney, Stable Diffusion, Craiyon, etc. are quite literal in showing us what we ask for. If the output images do not match the input text, it is far more likely that the text is lacking explicitness rather than the neural network not understanding what we have fed to it. This is a great analogue for descriptive, word-based narratives that are meant to paint a mental picture; if the computer can’t imagine it then the text clearly needs some editing. The second, and arguably most important, discovery we have made is that these neural networks have given some of my more timid students the confidence to produce. In design studios, I often find that students treat everything as precious and sacred, when in reality design is messy and iterative. Some students are so afraid to put pen to paper that it impacts their productivity in a way where we never fully see their thoughts and ideas represented in the same way they might see themselves. With the speed and marginal disconnect from the output itself, students have been freed from their own lack of self-confidence to truly explore visual representations of ideas they have never explored before. Let’s not assume that students who are not as practiced in image making are also less practiced in thinking and feeling and give them the tools to express themselves without being compared to virtuosity and exclusive representational standards.