artificial intellicence

In my work with artificial intelligence, bodies and identities are dissolved. By painting AI generated images based on my own body I take apart and examine the construction of a self and the (female) body.

I almost always use myself as a model. During the summer of 2021 I trained a neural net, a Generative Adversarial Network (GAN), with the goal of generating images to use as originals for new works. The neural net was trained using a data set consisting of a mix of my earlier works and the photos of me that I would normally use to paint from. Thus yielding a new variant of the body I usually depict.

I'm also interested in the implicit bias written in to the programming code, and assumptions of the social and political neutrality of data and machine learning.

*here’s looking at you, kid,* 20x25 cm, 2021

The generative network was trained using the PyTorch implementation of StyleGAN2-ADA [1] which is a modified version of the original StyleGAN architecture described in [2], with image quality improvements described in [3] and an adaptive discriminator augmentation mechanism to better handle limited training data described in [4].

[1] NVIDIA Corporation.
StyleGAN2-ADA — Official PyTorch implementation. 2021.
https://github.com/NVlabs/stylegan2-ada-pytorch

[2] Tero Karras, Samuli Laine, Timo Aila.
A Style-Based Generator Architecture for Generative
Adversarial Networks. CoRR, abs/1812.04948, 2018.
https://arxiv.org/abs/1812.04948

[3] Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten,
Jaakko Lehtinen, Timo Aila.
Analyzing and Improving the Image Quality of StyleGAN.
CoRR, abs/1912.04958, 2019.
https://arxiv.org/abs/1912.04958

[4] Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine,
Jaakko Lehtinen, Timo Aila.
Training Generative Adversarial Networks with Limited
Data. CoRR, abs/2006.06676, 2020.
https://arxiv.org/abs/2006.06676