Can AI create something new, Max Kuwertz?
The art director speaks about his new book on prompting, AI & authorship, and why he no longer makes photo pitches with Midjourney sketches.
I've been off the radar for a few weeks, but I'm back with another episode of the interview series. This time I spoke with Berlin-based art director Max Kuwertz. Max is currently putting together Spells, a comprehensive collection introducing a new generation of creatives working with AI. The book gives us a peek into their toolbox, and introduces a mindset of skill-sharing and collectivity to the highly contested field.
Max studied Integrated Design, Communication Design and Industrial Design in Cologne, New York and Hong Kong and Contextual Design in Eindhoven. He started at design studio Meiré und Meiré, did branding for MetaDesign and interactive installations for Random Studio. He founded and edited ROM Magazin with Sascha Bente and Khesrau Behroz, and did graphic design for K41 and Standard Deviation, a techno club and record label in Kyiv. He has been working with Gen AI for two years and has built up an extensive practice of photorealistic AI imaging based on Midjourney.
For dots per inch, we talked about the idea of turning the project into a book, the beauty of chance, and working on AI commissions at a time when the industry is still an unregulated mess.
To read the interview in your browser, click here
FS Spells is a compendium of AI image pioneers and their prompts - and you are also one of them. Your photorealistic images are quite impressive. I was pretty sure you are a photographer experimenting with AI. Does that surprise you?
MK Well, people often ask me why my images look the way they do. I think it's because in the past decade, I have worked in many different fields and built up a large vocabulary and knowledge base of visual references. That's your main tool, when you're an art director and you pitch your ideas to corporate clients, you do it by putting together mood boards or sketching out things in Photoshop. The common denominator of all my jobs has always been that I enjoy creating things and, I am good at describing and presenting my ideas. However, I can't draw well and have rudimentary knowledge of 3D programs, so I always had to communicate exactly what I wanted to the creative team I was working with. I guess that's one of the reasons I find it easy to write prompts now.
FS The initial hype around Gen AI is over - many AI accounts on social media are deserted, many creatives don't want to hear about it anymore. Why is it different for you?
MK When I encounter something new, the first thing I ask myself is: what will it enable me to do? I always aim for novelty in my projects. And because I am not part of any particular visual discipline - I am not an illustrator, or a photographer, or a visual artist - I have never felt that something is being taken away from me. I was in the opposite position: I thought, wow, this is something that makes me independent.
FS How do you use Gen AI in your own practice?
MK At first, I did not see it as a tool to optimize my processes. My goal was to create something that could only be done with AI. In my personal work, I try to create images that are very similar to photographs, yet somehow different. These differences could be due to concept or composition, the objects depicted could be unrealistic or illogical, or the textures could be non-existent. My intention is for the images to contain surreal elements without being obviously AI. I like working with AI because the end result often surprises me. That's why I'm so addicted to it.
FS The Spells volume is called Pioneers. What makes the artists in the book pioneers?
MK The book features artists from five disciplines: design, photography, art, fashion, and architecture. For the most part, they are not typical "AI artists" but rather people from other creative disciplines who are experimenting with AI. For me, they are pioneers because they see the creative potential in the technology and are not afraid of it. For example, my friend Julio, a digital and sound artist, is doing some pretty interesting things with AI. His artist name is 1v14c1111ve. I also really like the work of Amos Fricke, the photographer, who has a still life project called Human Artifacts, or Brutal Integral, who do speculative architecture. But that's my personal taste. I tried to cover a broad spectrum, from different countries, different genders, some things are dark, some are cute. The only thing you don't have in there is the typical Midjourney kitsch.
FS Why a book?
MK The main reason Spells is a book is that I wanted to avoid the copy-paste moment. You have to type the prompt, and in the process you automatically start changing things. Maybe you think, I don't want this in red, I want this in blue. Or you change an object. That's a USP that I think is easily overlooked: every single one of those prompts can generate an infinite number of images. Of course, there are also insights such as how they prompt, what stylistic references they use, which cameras are referenced, the order in which the words are listed, and how long a prompt is. But the true revelation is that the best images simply have a good idea. And the style is what gets the maximum out of that idea and creates a really good picture.
FS Writing prompts is not really like speaking, it is more like a systematic word salad. What makes a good prompt?
MK Any image can be described with words. And when you start working with Gen AI, it's also a good start to have ChatGPT or Midjourney describe images for you. This will give you an idea of how the programs understand images. But the interesting results tend to happen when there's that moment of surprise or chance where you don't know exactly how something is going to be interpreted by the program in the end. That's the great thing about the prompts in Spells: you have a prompt and you see the image. And then you enter it at home and the program will give you four completely different images, it will not repeat the result you see in the book.
FS It's interesting that you say that because there's this idea that with Gen AI you have control over every detail.
MK You can control the results quite precisely these days, but it's not that interesting. There's a sweet spot, like in any work of art: if a picture is completely unambiguous, it's boring. A good prompt always contains things that are ambiguous. Many visual artists work in the same way: look at Gerhard Richter's Rakel paintings, they have a very strict framework. But within that framework there is a lot of chance. It's the same with prompts: you build a framework with a theme, a composition, a style, maybe you specify one or two elements - but then you let chance happen. It involves a lot of curating. I create many images and then I choose paths that I iterate on. Also, I work with in-painting to swap elements. It gets more and more detailed, from rough to fine.
FS People tend to argue away Gen AI by saying that nothing new can come out of it, because it only contains everything old.Â
MK There are people who have been stating for 20 years that everything already exists. I don't believe that. To me, it feels more as if the camera had just been invented, and people only know painting, and now they start photographing the things they used to paint. And then they think, "Holy shit, the camera is so much faster, how is this going to end?". Fast forward 150 years later, painting is super diverse, so is photography.
It's important to understand AI as a tool, or a new medium, it's as original as Photoshop or a paintbrush. The algorithm doesn't decide for you and it's not creative. You can use this tool for very reactionary and conservative things, which is probably the easiest because the data set is from the past. But the data set in your head, your references, also refer to the past. AI can be used in the same way and is actually a much better fit for new concepts. However, like a paintbrush or camera, you need to have an idea of what you want to do with it.
FS So you would say that you are the author of your AI images?
MK Definitely! In my personal work, I try to do things that I feel add to the cultural palette rather than simply repeat something. I aim for things that are more me than AI. That way, I can present it and say it's my work. I think that's a good rule of thumb. You have to ask yourself, at what point does the work become relevant? For a lot of people working with generative AI, neither one nor the other is true. Either the work is redundant, or there is simply too little authorship in it, no message. Then it's meaningless content flicker.
FS All prompts in Spells and your own practice are based on Midjourney. To what extent are you concerned with the issue of copyright?
MK It concerns me. I see it as a great injustice that the whole thing is based on an illegal data scrape. There should be some kind of AI tax to make it fair or at least fairer. In Spells, there is the rule to not reference living artists in the prompts. It is important to respect their livelihood and not create work that resembles their art. But I think it is okay to use a reference to Pablo Picasso. Or to combine styles to make something new.
FS Have you done any commissioned work with AI?
MK Yes, I've done some commissioned work, posters for Standard Deviation, a podcast cover for Undone’s Mezut Özil podcast, and the Hitze podcast about the climate activists of the Last Generation. These are jobs where I can now offer everything from one source.
FS So optimization?
MK That was not my reason for starting! As a designer, you often have commissions where the client has no budget for photographers or illustrators. In such cases, you typically work with typography or licensed material. These were the first instances where AI became relevant to me, as it greatly expanded the possibilities within a project. This was a motivating factor for me. But now, of course, I have gotten so good at it that for some projects I can get just as good a result with AI as I could with another creative.
FS Do you have an example?
MK Yes. I was assigned to create a podcast cover, and we came up with a concept that we shot with a photographer. We had already completed the shoot when the podcast lineup changed from one host to two, so we had to revise the original concept. Because the new idea was complex, I created an AI image to sketch out what we wanted to shoot. We hired the same photographer, but mainly because the process was messed up, the result did not satisfy the client. In the end, they said what you did with AI kind of captures the vibe more. Can't we use that instead? Then I realized, of shit, if you give a draft like that to a client, it's already pretty realistic and, that influences how the client evaluates the actual result. I don't do photo sketches with AI that anymore. What I've learned is that if you're planning to work with a third party, the sketches have to have a sketch-like quality.
FS And does it bother you that your images are not protected by copyright or Urheberrecht in Germany?
MK I think that's only fair, my clients know that. After all, when I use these tools, I'm using images that I don't own, so it would be a double standard for me to say I want copyright protection. But when I was researching the book, I noticed that a lot of people didn't want to share any prompts, not one. I don't understand that. If you draw so much from this pool of images and knowledge, I think you can be transparent and give something back. Programs like Glaze and Nightshade are also important because they allow artists to choose whether or not their images are used for AI training.
FS I understand that in theory, but if someone took your image and just used it - would you still see it that way?
MK Of course not. But the question is whether a company could do that today without a serious backlash. Copyright was a difficult issue even before AI, because we live in an age where things are often recontextualized, remixed, and sampled.
FS And does that concern your commercial clients?
MK Yes, definitely. I made a movie for Samsung that was for internal use only. It's still too thin ice for them. I can't show that either.
FS Although?
MK Although, they don't own the movie either (laughs). But there are contracts of cause. When I did the Hitze podcast for rbb, they also asked if it could be used commercially, is it legal? And I said, of course you can use it, it's legal according to current jurisdiction and the terms and conditions of Midjourney. But how that will develop, no one knows.
FS As I was looking through the names in the book, one thing that struck me is that this new generation, like you, is very interdisciplinary. Take Jumoke Fernandez for example, she is very confident in her role as a young self-taught designer and has casually integrated Gen AI into her tech stack to do 2D, 3D, UI/UX, AR etc for work in fashion, game design, and advertising. It feels very natural. Would you say this is symptomatic of what is to come?
MK I think what a lot of us have in common is that we don't fetishize our tools. The camera, brush or hammer are just tools to create something with. If there's one good thing technology has done, it's allowing people to achieve more and better results. Today's programs are so simple that you can teach yourself. If we could solve the ethical problems associated with generative AI, along with all its pitfalls and downsides, what it makes possible is that anyone can create the images they want based solely on their ideas and imagination. Isn't that amazing?
***
Dive deeper:
You can see more of Max Kuwertz work on his website and instagram
Pre-order Spells with an early bird offer until Jan 23rd on Kickstarter
Learn how to protect your images from AI data scraping with Glaze and Nightshade here
Find me on Instagram @frauke.schnoor