Janek Simon, Katarzyna Nestorowicz, and Marcin Nowicki are talking. Alek Hudzik and Arek Kowalik from Mint Magazine are asking the questions. This article is from the first print edition of Mint Magazine, published on January 15, 2025.

MINT: Director Harmony Korine has named AI as the second director of his latest film Baby Invasion.

JANEK SIMON: This gesture is some marketing populism.

M: And you, Noviki, called artificial intelligence a creative partner during the Dream of the Machine exhibition in Krakow.

MARCIN NOWICKI: For us it is not just a tool, like a chisel or Photoshop. More value can be given to this system. When working with artificial intelligence, you can expect something that you didn't assume before the project. It can show some broader perspective. We used GPT Chat when writing the script for the film, and we also use visual systems like Midjourney or Stable Diffusion - these programs are involved in our creative process. Our goal was not to get exactly the visual effect we wanted, quite the opposite. We wanted to enter the unexplored region, where the title Machine offers us imperfections, errors, so-called hallucinations, so that we can follow them. This is a partnership for us. 

JS: I don't notice that much difference between AI and earlier stages of technological development. For me, they are still just tools and I don't see the rationale for anthropomorphizing them. All the models we rely on when working with artificial intelligence are libraries that have undergone algorithmic operations. That is, if we want to consider a GPT Chat or Stable Diffusion as a co-author, exactly on the same principle we could, for example, require that we consider the Jagiellonian Library as a co-author of a scientific text. Attributing to such models the qualities of the human mind, intelligence or consciousness is, to me, no different from attributing them to the computer that asked us to put the next floppy disk in the disk drive. Of course, this may change in the future.

MN: Maybe this partnership is born in the moment of communication - one that is not distinguished from human communication. Today one talks to AI as one would to another artist.

JS: What does that mean?

MN: When we were working on the exhibition in Krakow, we spoke with Chat PG on the script for the film. We asked him what he would compare the famous move 37 in AlphaGo's duel with Lee Sedol to. He replied that it was as if a whole new species had appeared in the garden. For me, it is precisely such games of association that are the birth of a new poetics. The path from mathematical movement and logic game to plant metaphor.

KATARZYNA NESTOROWICZ: Working with chat influences the approach to working with people. We, on a daily basis, operate as a graphic design studio. From clients with an order we get a draft, to which we respond, in AI chat this is called a prompt. We, too, could be considered a tool in the hands of the client, and yet often, even at exhibitions where we don't show works, but create their identities, for example, artists call us collaborators. In my opinion, this is the essence of this discussion. After all, when we talked to actors during the exhibition in Cracow, we also prompted our command in a certain way, and they carried it out, sometimes improvising, surprising us. Artificial intelligence can do just that - surprise. Our role has a lot in common with that of an interpreter. We bridge different worlds. We are looking for a way to get out of patterns and liberate ourselves from our own opinions. That's why collaboration is so important to us, including with AI. Now we often work in project teams, with architects or sociologists. This process involves us paying attention to something that we ourselves had not thought of. AI is part of this process for us.

M: Janek, you've been using generative technologies even before we started widely talking about the impact of artificial intelligence on everyday life. How does this manifest itself in your practice?

JS: In the Meta Folklore project, I threw a wide range of non-academic sculptural representations of the human body into the model I was training. These were images from all over the world. The program tried to extract common features for this set of images. From this I created physical 3D models and sculptures. What the model produced was an intermediate step for me, an inspiration to create universal folklore that doesn't refer to any particular place.

M: It's interesting, because before our conversation we asked Chat GPT about what threats artificial intelligence poses to art. Homogenization and globalization of art were just among the answers.

MN: Artificial intelligence systems strip things of their contexts. They optimize everything and for art this is a dangerous phenomenon. That's why the generated images and films just look boring after the initial delight. Our work with artificial intelligence is more about entering regions, including aesthetic ones, that we would not be able to discover on our own.

M: Janek, and what challenges do you see in using artificial intelligence? 

JS: There are several ethical issues that come to mind. The first is the problem of consent to use images to train models. This is a new category of use that is not defined either by law or by any ethical practice. A gray area has been created. It certainly needs to be regulated, but there is a risk that it is a little too late to do so. The second issue is bias. If in the training dataset “doctor” is always a white male, the AI model learned from them will generate only such images of doctors. This is quite well researched, and there are numerous known examples of discrimination embedded in AI models used, for example, in recruitment or credit scoring. Importantly, some aspects of this discrimination have been inherited from research tools created even before AI appeared. One example is WordNet, a word classification system developed in the 1950s for research in cognitive psychology. The biases inherent in this database permeated the world of AI, as it was the basis for the ImageNet training collection - a collection of more than a million images with categories assigned to them, images commonly used in AI research. 

M: You created your own models. You fed the algorithm yourself. Why?

JS: I like to do things myself and have full control over them, it also has an ethical dimension for me. It took me about three months to find and process in Photoshop the 12,000 images needed to train the GAN model, which is the basis of the Meta Folklore project. Here lies the limit of what you can do on your own, without capital or academic institutions. 12,000 you are still able to do on your own, but 100,000 no longer. This enormous amount of time has to be spent if you don't want to compromise with the corporations that decide the shape of these software products.

For me, the main motivation for work (and life) is curiosity. I identify more with the figure of the para-scientist than the romantic artist who experiences something there in inspiration. I do research, so much so without discipline.

M: Why is it so difficult to talk about the role of AI in the arts?

KN: There is a lack of language with which we can talk about this change. This is not only true in the world of culture and art. It's simply that nowadays certain terms are being redefined all the time. This is not a threat, but rather a challenge, which means that we quickly have to somehow adapt our vocabulary to describe what is happening in the area of relations with artificial intelligence.

MN: It is not just about anthropomorphising this technology. We are not interested in giving it a human character, but in understanding that it is simply something else that we cannot yet define.

M: Perhaps this is why the discussion of AI in art is not very popular in the Polish art world.

JS: It is more a question of the Polish art world not being interested in technology. The art market favours specific formats, institutions are more focused on identity stories, specific politics or climate change.

M: Globally, however, artists associated with generative art are more present. 

MN: This is true. Not only do they engage in the creation of this art, but they are also bent on testing the boundaries of the human-non-human. Holly Herndon, an American artist living in Berlin, reflects on the morality of these systems. She creates voice models using samples of her own voice, which she makes available to other artists for artistic use.tania.

KN: She decided to create a platform through which her voice becomes a tool. That is, it can be seen as an artistic practice.

Similar processes have been taking place since the first primitive computers began to be used. The limitations of these tools we use - film grain, pixels, noise - affect the aesthetics. For me, however, generative art reaches far beyond aesthetics. It asks what human creativity is and what humanity is. This, in turn, influences, for example, our attitude to manual work, to the uniqueness of imperfect tools. We would rather have an imperfect glass made by hand than a perfect one made by a machine in a factory.  

JS: The overproduction of mass-generated content will result in a return to the techniques that distinguish humanity.

KN: I look at it realistically. I'm not going to stop this process by myself, even if I unite with a larger group of people. So I don't even think about stopping it.  

JS: Instead, one can strive to ensure that this technology is introduced in an equitable manner.

MN: Marx described the final stage of capitalism as a human-independent interconnected system of machines that do all the work. This is how we can imagine the future of work being done by neural networks.

JS: And it's called Fully Automated Luxury Communism!

M: Amen.