Where Art & Technology Meet
Veera Jussila is an artist and technologist whose work explores algorithmic decision-making and networks of communication. Her practice focuses on machine learning as a form of knowledge production. By using small, curated datasets, she studies the edges and logics of deep learning systems and reimagines them. Drawing from her background as a feature writer, Veera’s work often looks into existing tech narratives and seeks to dissemble them. She has graduated with an MA in Computational Arts from Goldsmiths, University of London. Veera also holds a Master’s degree in Social Sciences from the University of Helsinki. Veera was one of the computational artists from Goldsmiths who was selected to work at a collaborative In-grid residency with Arebyte Gallery during summer 2020.
In industry use, deep learning datasets can consist of millions of items (images, text fragments, etc.). This is to optimize the predictions and to make the model less “biased”. However, the human element is always there, as we have been reminded of in the case of racist image labels in ImageNet dataset. Small datasets addresses questions about big data and automation. Seeing what goes “wrong” with 300 images helps us to understand the complexity of using 30 000 images. If we want to teach a robot to speak ”like a human”, we should collect a small dataset to first face what we actually consider to be human. This can be a fruitful artistic project by itself.