A conversation with Andrea Brewster about her dialog with AI
By Peter Honig
Peter Honig: There seems so much anxiety expressed in the media recently about “computers doing things,” misbehaving or taking things from us… but I am much more interested in what you – the artist – do; your influences, the choices you make, and the steps involved in the creation of this impressive body of work.
My understanding is that one of the main tools you are using is Midjourney, a form of AI developed by a company of the same name based in Silicon Valley. The program responds to text prompts, sort of like reading in reverse. You write the description, Midjourney sources imagery from the World Wide Web in response to your “prompts”, and then creates a composite image. It is impressive technology that debuted in mid-July of last year.
Andrea Brewster: For me, Midjourney is primarily a tool, a very powerful tool, that allows me to do things that I couldn’t do before. However, I am most definitely in the driver’s seat, as the tool doesn’t do anything on its own and so I’m the one guiding it in the direction of my own vision. It is similar to being an art director or working with a fabricator. My process involves prompting the AI, receiving images, deciding which ones I like, then maybe prompting again if I’m not happy with those images. Then I combine and recombine images, and re-prompt, until I come up with something that I’m happy with. For this show, Enchanted Garden, I generated about 1400 images to arrive at 12 that I was really happy with.
P.H.: Having had a chance to explore the program myself, I was struck by how adept it is at creating imagery, but at the same time I found it very difficult to control… to get output that matched my expectations. In a sense one is working with, or against the program (and the intent of its creators) and that raises all sorts of issues about authorship, ownership, and the nature of creativity. How would you respond to those who might argue that the AI is the creator of these images?
A.B.:I don’t feel that there’s too much a question of authorship, because ultimately, it’s really about my vision as an artist, which is a hodgepodge; a patchwork quilt of all the images I’ve seen, all the experiences I have, what I’ve read, what I’ve thought about etc…etc… The digital tool is just another way of drawing out those influences and choices and presenting them to the world. I’d argue that this is what all artists do, no matter what their medium or method. Midjourney is a tool, just like a knife is a tool. I can use that knife to make a meal, to do surgery or to kill someone. It is the human factor that drives the usage of the tool, not vice versa.
P.H.:I am sure you have very specific observations about the choices the programmers made in terms of how you prompt the program and how the program responds to your input. These must reflect specific ideas about “art”. Do you play against them or try to exploit their strengths/ biases?
A.B.:Because this tool is so new, I feel that it’s important that artists jump in and play around with it. I feel that we will bring a different perspective to the conversation than programmers, computer geeks or even graphic artists. For instance, Midjourney is moving more and more in the direction of photographic realism. So, I guess this is a specific idea about art/ “good art” that is being built into the software. I personally don’t find that direction so terribly interesting, so I’m trying to see if I can push the tool in another, more impressionistic, more suggestive, more abstract direction. Right now, the program has enough flexibility to allow for these types of explorations, it will be interesting to see if it keeps this ability.
I’m more engaged with trying to get it to do what I want. It’s often like working with a small child, who doesn’t really listen when you ask it to do something. If the program doesn’t understand what you’ve asked for in the prompt, then it just ignores that item and goes on to the next word in the prompt. Although, it doesn’t even seem to be as systematic as that. But this unpredictability is also what makes working with it so compelling, because you never quite know what image you will get. It’s somewhere between spinning a roulette wheel and casting a magic spell. Sometimes you hit the jackpot and other times you end up with a mess, just like the Sorcerer’s Apprentice.
P.H.: What do you think about the controversy regarding biases being expressed in the imagery and that Midjourney is stealing from artists.
A.B.:Of course, the program has certain biases built into it, because of the biases that are part of its data set, the image data that it sources to create/ piece together images. This data set/ training data is an area, where there is a lot of controversy. There are several AI image programs besides Midjourney, Dall-e and Stable Diffusion being the two main ones. It is my understanding that all of them have utilized vast collections of images from the Internet, to teach the program how to connect words to images, so that when I type c-a-t I get or .
It seems like, the creators of the programs sort of assumed that the Internet was a vast, cultural library to be mined and so they sort of “forgot” to ask anybody if it was okay that they used these images.( I’m not really sure how or who they could have even asked) Of course, most of the images are copyrighted and they also didn’t leave a door open for folks to opt out of their project if they wanted to…So the ethics and legality of all this is something that we’re still working out…of course it may end up being a moot point, now that we’ve opened up Pandora’s box….
P.H.:So much conversation about the “rise of AI” focuses on fear- based ideas. In terms of artists, the idea that someone would be “put out of a job” seems to abound. But implicit in this idea is that art is primarily an economic activity and also a field of competition, rather than a form of communication.
A.B.: Will AI cause some artists to lose their jobs? Probably, but this happened when photography was introduced, it happened when we got Photoshop/ Illustrator too. I think we have to pivot; we have to adapt…of course we humans are mostly pretty good at that. Again, what are we going to do with the tool, make something aesthetic and conceptually interesting or just make endless images of hyper realistic, screaming, buxom, Viking women?
P.H.: To some extent the culture of programming must have different “values” and goals than that of the “Art World”…and I am curious what you have observed as to how this plays out interacting with the technology.
A.B.:I can’t tell you too much about the history of AI development, although I understand that it has been going on for a number of years already. Just like with any new invention or discovery, I’m sure that there have been numerous contributions by hundreds of programmers that have all built on one another. So, I’m not sure that there really is just one author for these programs. Honestly, it’s a bit too geeky for my teeny, tiny artist’s brain to understand all the talk about diffusion models, generative adversarial networks, machine learning, CLIP, neural networks etc.…I’m really just taking the magician’s wand and running with it.
P.H.: With all the images in the world to choose from and all the variables involved in the program, I am impressed by how your images seem such a logical visual outgrowth of your earlier work, the vellum drawings, distorted paper, knotting, and the 3D printing pen work from last year’s exhibit “The Fragile Balance that Lies Beneath“. In studying these new images, I found myself feeling as if I was looking at something I had seen before, only to not be able to place where…and it took several days for me to register that these were not depictions of real flowers – although they felt like it. In revisiting the statement for that exhibition, I was struck by how that idea was perfectly captured in the phrasing:
“…. deliberately ambiguous, a fantastical synthesis of biological imagery that ask the viewer to question whether the form is animal, botanical, or cellular; subtly shifting focus between a sense of familiarity and the otherworldly. Through an intuitive process of combining and recombining, Brewster allows elements to organically grow and develop in response to each other; creating “hybrids” that reference nature, but are not direct imitations of it. The resulting forms straddle the line between science and art; blurring the distinction between imagination and reality, representation and abstraction.”
The degree of control you are able to exert over the Midjourney software interface is pretty startling. It is relatively easy to produce something outlandish or bizarre, but it doesn’t seem easy to modulate precisely….
A.B.: I guess the fact that my new AI pieces look like or relate to my older pieces, just means that I was actually there doing something. However, I find that in working with Midjourney, you have to be open to the randomness which is currently inherent in the way that it works. I wouldn’t be surprised if they are working on eliminating this quality, so that “what you prompt is what you get”. But actually, I think that would move it farther away from a creative type of process. Because I find that often it’s a random sort of thought or idea that is the one that propels me to ask “what if” or to try something new.
P.H.:The use of AI in art does seem however to redefine what constitutes labor in some very interesting ways and I think this is very relevant to a discussion of your work in relation to the historic work of Jan van Huysum, who took up to a year to complete a single floral composition. His work would be defined by extreme physical dexterity in terms of being able to observe and model hundreds of floral forms in a scientific manner. Your work engages in a dialogue with work of that genre, but is the result of a very different age and working process.
Both seem to be products of cultures of abundance – his compositions flourished in the prosperous conditions of the Netherlands serving a newly rich merchant class expanding trade with a worldwide network. Your work blossoms in close proximity to the wealthy environs of the Bay Area, also the hub of a world wide web of trade – in images, technology, and a revolution in cognitive thinking and redefining labor and creativity models.
A.B.:I agree that the concepts of labor and specifically craft, are becoming less well defined. For instance, what does it say about labor when I can create 1400 images in the space of a few months! This is especially poignant for me, because much of my earlier work was highly labor intensive, often verging on the obsessive (although I personally felt it was just meditative) So, it’s quite a new experience for me to be able to create works and explore ideas more quickly. Of course, there are good things about both approaches. And maybe it’s the question of my looming mortality, which is propelling me to want to work faster.
P.H.:I do think there is a tremendous amount of conceptual labor involved. Can you further define the process, in terms of how explicitly you cite specific historical references, such as the paintings of the Dutch still life artists?
A.B.: Of the Dutch Flower Painters, I was particularly inspired by the work of Rachel Ruysch I hadn’t realized that there were any women doing these works, so I was interested to find out that she was actually quite a successful artist in her day. As I mentioned in my statement, I was also very interested to find out that the Dutch flower paintings were often collages/ montages of flowers that the artists had only seen in scientific reference books and that they combined the flowers together for aesthetic effect. This resulted in bouquets of flowers that don’t even bloom at the same time of year! This seemed reminiscent of what is now possible with digital processes and AI, that we can combine together almost anything and do it seamlessly.
I am also intrigued by the theme of transience that was often a subtext in the Dutch paintings, that flowers fade and die, just as we do. Of course, with the artificial flowers that I’ve created, there’s more the concept of frozen in time. Just like other artificial flowers, whether made of wax, paper, plastic or silk, they are in a sense forever beautiful, in a moment of perpetual stasis, which I’d argue makes one actually notice the changeability and mutability of the physical world even more.
P.H.:How do you tone down specific references, such as Dutch paintings, in order to abstract and integrate them into a coherent whole?
A.B.:Honestly, I’m still learning the art of prompt crafting, and so I tend to include a lot of descriptors and references in my prompts, most of which Midjourney chooses to ignore. So, my experiments usually go through a number of iterations before I come across something that I’m interested in. Yes, I have made mention of art historical styles, but I tend to combine them together say, Art Nouveau and the Baroque, in order to see how Midjourney might mesh them together. I find myself going with and responding to the flow of what is happening, in a improvisatory sort of way. It’s more like a conversation…I start on a topic, Midjourney responds, I respond to the response, then it responds to my response etc.…etc.…And just like any good conversation, you may end up on a topic that is quite far distant from where you began.
PH: Have you used any other digital tools or image editing programs?
A.B.:Yes, I have used image editing programs too, but more along the lines of improving color saturation, removing artifacts etc.…For instance in Midjourney Version 3 it often adds in random text, like fake signatures or marginalia, which I found distracting, so I removed these using the image editing software. But, I could easily see that further complexity could be added by layering multiple images together… I’ve only just begun to scratch the surface of what is possible.
P.H.:It does occur that the idea of working in the realm of language prompts to source visual imagery is an upsetting prospect for many artists who have been trained to consider visual thinking and the language centers as representing oppositional vectors. Of course that distinction may represent ideas about creativity and thought processes that may be soon thought to be outdated.
A.B.: Maybe it’s because I’m an artist, but when you say “apple” I can easily “see”/envision an image of an apple. So, maybe the crafting of prompts is sort of reversing the process, going from image back to text, which then gets converted into image again. But, maybe for non-artists, starting with text actually opens up a whole window of possibilities for them, because they haven’t really developed the facility for envisionment nor the abilities of craft.
P.H.: Where do you think all this might be taking us?
A.B.: I have no idea where this might lead us…Will it create more artists? Will it increase society’s appreciation of what artists and creative thinkers do? Will it eliminate the need for artists altogether, when everybody can make art with a few words and a mouse click?
It’s all so new and there is so much power in these AI tools for both good and ill, that it’s a bit daunting. There’s definitely a sense that this will take us to places “where no man has gone before”. I was just remembering back to the first time one of my friends showed me the Internet. Of course what he showed me bears absolutely no resemblance to what we have now, but I think we had a sense that this could be something big, something earth shaking, something that could be a game changer.
I think AI will be the same for the next generation, for the next 30-40+ years. Maybe we can learn from the mistakes that we made with the rise of the Internet. I’m more worried that we’ll rush into the “brave new world” of AI with such speed, that many folks won’t have time to catch up and adapt and that many of us will be left in the dust beside the robotic roadside. But, thankfully I probably won’t be around to see the full extent of the damage that we end up creating.
P.H.: I really appreciate your taking the time to speak with me.
A.B.: It was my pleasure.