Eryk Salvaggio and Tristan Dot discuss the possibilities and challenges of AI art.
At the moment, AI is telling us a story about what we should do with AI art, but what could we do if we took a step back and steered the system in a more creative direction that was fairer, less biased toward certain groups, and less exploitative of other artists? Can we tell a different story?
Eric Salvaggio
Eric Salvaggio [2025] and tristan dot [2022] Both are interested in the questions raised by AI art, but approach it from different perspectives. Before joining the University of Cambridge, Tristan studied both Art History and Machine Learning. His PhD is in English and he takes a critical approach to AI imagery by bringing together technical and aesthetic expertise. “I want to understand what is happening technologically, economically and politically with AI images,” he says, adding that he is comparing current digital visual culture to 19th century designers who lived in a similar era of technological and artistic exploration.
Eric, who holds a PhD in digital humanities, says his approach is based more on artistic practice than machine learning. “While the technical elements of AI can often be opaque, I know enough about them to raise questions about them, but I look at it through the lens of the humanities and art-making,” he says. When it comes to archives, for example, he says genAI tends to work in reverse to the normal process in the humanities: taking data, lumping it together, “blurring” it back together without adding context. He wants to explore the role that the arts and humanities can play in recontextualizing data and highlighting ongoing processes.
He calls it “filling in the blanks.” “We tend to call the parts of the AI production process that we don't understand creativity as art,” he says. “What I see are gaps that AI fills inadvertently. I want to look at how we bring consideration back into that process.”
AI art
Tristan and Eryk both created their own AI art. Tristan has used his art to explore academic questions. For example, we used CCTV images to explore political questions about surveillance culture. Eric has been involved in experimental art since he was 15 years old and plans to pursue a Ph.D. Create a framework to test assumptions about the use of generative AI in policy, pedagogy, and design.
For Tristan, art is about moving away from the usual way of looking at things and exploring new ways of interpreting reality. But after that initial burst of creativity, he feels Big Tech has adopted AI art in a simplistic and standardized way, limiting its aesthetic potential.
He sees similarities with the 19th century in that AI and genAI moved from small models with more creative possibilities to larger, more fixed formats. Similarly, photography in the 19th century became increasingly industrialized and less experimental, although it had a variety of influences on painting and art. Textile designers face many of the same questions as today's AI artists around ecology, copyright, colonialism and repetition, he added. He is also interested in broader contextual issues, and says that discussions with Gates Cambridge Scholars and others at Cambridge helped him reframe his ideas about the production of images.
Eryk recalls working on a small generative adversarial network. [GANs] In the past. Guns is A type of machine learning algorithm that can generate new, never-before-seen data that is similar to existing data. he remembers. He fed the GAN with thousands of photos that he personally collected. AI is now processing vast amounts of data, and there is a more distant relationship between artists and data, “mimicking commodification culture,” he says. “GANS was like an artistic scene where individuals came together to work on niche projects,” he says. “Everyone now has access to the same tools, and it's becoming more fixed. You can't color outside the lines. You're being blocked from certain kinds of learning, and art is being scooped up and commodified.”
But it's not too late to change that, he says. Eryk is interested in discovering how AI can be used not only to recreate art based on traditional data inputs, but also to explore new opportunities for storytelling and expression. “I'm interested in the question of why we need to use AI in art production,” he says, adding that it's how AI can liberate artists.
“Right now, AI is telling us a story about what we should do with AI art, but what could we do if we took a step back and steered the system in a more creative direction that was fairer, less biased toward certain groups, less exploitative of other artists? Could we tell a different story?”
*Artist Eric Salvaggio pollen series [top picture and above left] generates “noise” and “pollen” images using an AI image generator trained only on public domain material. A visualization of the learned features of Tristan's neural network after fine-tuning the fiber pattern is shown in the right photo above.
