Ada
The Ada project is a series of images that are generated by an algorithm and painted by artists. Born from a computer, the digital images are the building blocks that painters use to bring the image into the physical world. Contrary to traditional paintings where there is one original and many prints, the original images in Ada exist digitally and can be shared, copied, and reproduced in their unaltered original form. There are many originals, but only one copy of each piece. It is this copy - a human’s rendering of what a computer dreamt up - that is the true piece of art.
This project is in the early stages of creation and is very much still an idea. I am working on training the generative model on a dataset of female artists to combat the male bias of art imagery. Machine learning models perpetuate the biases of the data that is used to train them. This ethical problem is at the core of machine learning: when models are trained using historical data, how do we account for the biases in representation and content?
I combat the male dominated art data by feeding my model additional works by female artists. Furthermore, I will work exclusively with local (New York area) painters who are female-identifying or non-binary artists for the commissioned paintings.
model-generated images from a partially trained model
the image content and style is diverse