Deep generative models are a successful and powerful technology which is inherently limited in its ‘creative’ abilities by its uni-dimensional objective of perfect distribution learning. Yet, in spite of their limitations, Generative Adversarial Networks in particular are being used and abused as artwork production engines. Focusing on the aspect of novelty, the PhD research of Sebastian Berns explores avenues of improvements for generative models in a computational creativity setting. One goal is to develop learning algorithms which can generate artefacts that exhibit novelty. Potential applications range from the automatised production of video game assets to the digital arts.
Sebastian Berns has made many attempts throughout his life at writing a profile about himself. He has yet to find a good balance between an informative account of a person’s life and the impossibility of capturing its many facets in a limited number of words. Through his background in communication design and many years of experience as an independent graphic designer (primarily on cultural and artistic projects in Barcelona, México and Venezuela) he is familiar with playful self-reflective self-promotion. Since transitioning into computation through a master’s degree in artificial intelligence (UPC BarcelonaTech) he is increasingly conscious of adapting a more serious scientific public profile. In the spirit of novelty, this is an attempt to unite the two.