Aesthetics and neural network image representations
This work addresses how aesthetic properties emerge from neural network structures, which is incremental for understanding AI-generated art.
The study analyzed generative neural network image spaces and found that perturbing parameters away from photo-realism often yields 'artistic renditions' of objects, with deep modifications leading to symbolic representations, all without exposure to human-made art.
We analyze the spaces of images encoded by generative neural networks of the BigGAN architecture. We find that generic multiplicative perturbations of neural network parameters away from the photo-realistic point often lead to networks generating images which appear as "artistic renditions" of the corresponding objects. This demonstrates an emergence of aesthetic properties directly from the structure of the photo-realistic visual environment as encoded in its neural network parametrization. Moreover, modifying a deep semantic part of the neural network leads to the appearance of symbolic visual representations. None of the considered networks had any access to images of human-made art.