3 Questions for Nora Al-Badri

“Babylonian Vision” embodies many aspects that lie at the heart of my artistic practice: It is a piece that is rooted in institutional critique and even more so in a critique of power. It is anticolonial and equally it cautiously attempts to devise an emancipatory and positive future. Moreover, the work subversively appropriates AI systems, including by means of a form of hacking.

That may initially sound a little abstract but let me try and explain it briefly. Or rather, start with a few words on what motivated me to make this piece: The special thing about “Babylonian Vision” is to my mind the autobiographical element. Part of my family lives in Iraq, but I have never been able to go there as it is not yet sufficiently safe. So, for me, technology is also an opportunity to stay in contact with and experience the (knowledge) systems of my family and ancestors. Iraq is in many respects a rich country: It is rich in histories, art, and knowledge, and I was fed up with only hearing news or discussions about destruction *by the West* and the wars and the sanctions.

I therefore created a piece with the help of generative AI and initially established what I call a “dataset of the South”; in this way I also reappropriated for myself the artifacts that had been stolen albeit at the digital level. The set ran to almost 10,000 images of Mesopotamian, neo-Sumerian and Assyrian objects (all of them held captive in museums in the Global North), and I trained them with a generative adversarial network (GAN) in order to give rise to future possible new objects. As the museums refused to make their data available, even after I officialy requested them, I scraped their databases [the editors: scraping is an automated or manual selection and extraction of data from Web sites or programs in order to store the data and subsequently process it].

To this day, many AI systems are black boxes meaning we as users cannot discern what data were used to train them. This is highly problematic but in my work it is also and precisely emancipatory as the museums cannot prove that I really used their data. I am delighted that I received such a positive response in Iraq to this piece.

From my point of view, generative AI is only interesting for art if the piece in question is conceptual in nature, and also concerns itself with the AI systems themselves. But not exclusively. These systems are always represented as somehow futuristic, and yet all machine intelligences function with data that belong to the past. They are part of the past the moment they have been inputted. Is it not then logical to use archaeological data? Moreover, in 2020 (and not much has changed since then) there was actually only one thing that neuronal networks were really good at doing, namely pattern recognition. For me, it was marvelous to see how by means of GANs I could bring entire pictorial worlds of an era in time and a region, in this case Babylon, to life and could shape them. Of course, there are speculative and abstract insights into the search for the pictorial idiom, shape, and pattern of an epoch within a specific spatial context, namely that of Babylon.

These synthetic images are a kind of living memory, and the generated image is at the same time the artifact itself. Despite the work’s digitality, the materiality it crucially important as the original images input are reproductions of material objects from our past. If we consider MI as a technology that depicts and processes part of our collective memory, then it makes sense to apply it to our cultural “big data” of the past and generate new images as traces and circulating pictorial worlds. Innate to the input images in these databases themselves are time and memory (patina, fragments). Subsequently, new synthetic images arise as living memory. The image generated is at the same time the artifact itself.

Six years later there are quite different generative models and opportunities available, but interestingly enough it is still always the GANs that can best generate ‘latent space’ without “hallucinations”. What is great is that generative AI has become more accessible, but sadly few users make the effort to train their own models, understand them, and mindfully manage the energy and resources required …

“The Post-Truth Museum” is essentially a continuation of and advance on “Babylonian Vision”, albeit with slightly different AI-based techniques (Deepfake, Natural Language Processing, etc.).

The piece is an attempt to design a just, anticolonial, and contemporary museum. To that end, I invited some of my friends and collaborators, and they have all for years been tussling to devise a critique of the discourse on restitution and reparations, to write slightly different texts. Namely their vision for the museums, and these are then performed by three museum directors. I therefore again attempted using technology to change or even invert the power relationship for a moment. The museums (in this case, the Louvre, the British Museum, and Stiftung Preußischer Kulturbesitz) have appropriated the material objects of entire cultures. This is also something termed ‘identity theft’. In the Deepfake, I used images and sound to appropriate the identities of the museums and/or their representatives in order to show a different possible future: a reconciliatory, healing, and just one in which couscous can also be eaten in museums…

The genocide in Gaza started while I was on maternity leave. For me, that was a turning point. A moral crisis of our present age. And I knew that my next work would address the “crime of crimes”. For me as an artist and incidentally also as a mother (my children are just boisterously running around me, and I have for the last five minutes been trying to find the right word) this is “unavoidable”. That we are at present witnessing several genocides or the preparation of genocides, and you need think only of Sudan, Congo, Myanmar, China, etc., nothing else would make sense at the moment and would be to shirk responsibility. Incidentally, Gaza is the first AI-assisted genocide. I am also interested here in the topic of censorship, particularly in Germany, as that likewise affects artists and scholars (genocide and Holocaust researchers). AI is of course sharply censored as regards the topic of genocide. You can ask the European Mistral model for various genocides and get a completely different answer than if you prompt the US ChatGPT or the Chinese Qwen model. At the latest since Elon Musk’s Grok model last year “hallucinated” the genocide of white people in South Africa, based on conspiracy theories by Trump, Musk, and other fascists, I want to challenge these systems on the topic of genocide. Instead of GENAI (generative AI) I will opt for “GENocideAI”! More soon.