StyleGAN experiments, 2020
"Strange Attractions is when your outlook on life changes, that it may seem to your cognitively dimming self, that there are some awesome things happening around the world. That seems to be just about the safest thought that you can have. If the safest possible thought is also the right or most interesting one to be having, that's how you can get excited about flying over the Great Barrier Reef, finding Earth's oldest fossils on an old underwater mountain in Montana or looking at the newly discovered habitability of a deep, hot lake on Mars.
If you find that you can't find excitement for the same reasons you used to, then we must stop being ambivalent about that. You can't help being attracted to danger if you do not like the environment."
(this paragraph of text was generated with a text-based generative model GPT-2: https://talktotransformer.com/)
Strange Attractions are imagined creatures living in the latent space of the neural network StyleGAN︎︎︎. Instead of using it to generate photorealistic human faces, we challenge the algorithm and explore the area outside of its intended use. Strange Attractions rise from the neural network's interrupted face generation process and reveal the network's attempt to recall a human face. There comes an absolute bewilderedness from these neural encounters' erratic beauty, vivid colors, and stunning composition. The hallucinated meta faces are carrying the strangest emotions. From quirky looks to abstract ghosts, Strange Attractions present a wide range of visual possibilities behind the process of image synthesis, while stressing the importance of freedom in experimentation with open, generative ML/AI models.
As a tribute to ThisPersonDoesNotExist.com︎︎︎and the website's idea of open access to the generated outcomes, Strange Attractions are available in the same way at StrangeAttractions.xyz︎︎︎. People can also try to apply Strange Attractions on their own face with the Instagram facial filter (search “Strange Attractions”, or find via account https://instagram.com/lenkahamosova).
Strange Attractions are results of random vectors injected into various internal structures of StyleGAN. We also tried using StyleGAN2, but using it did not bring results of such vibrant colours. Finally, the static images were animated using the First Order Motion Model for Image Animation https://github.com/AliaksandrSiarohin/first-order-model.
The project underlines the importance of free artistic experimentation with ML/AI models and favors the deliberate imperfections and surprising abstractness of photo-realistic-oriented generative models. This is possible only because of their code's accessibility and the possibility to alter it as one pleases. When the adolescent years of GANs are over, and AI models for creating synthetic media will be turned into a product - either as a software or a service - artists will lose access to the infinite creative potential of AI and will be left with a limited set of predefined tools. With this project, we want to show what kind of unexpected gems can be hidden outside of an already defined model's use and provoke other artists and neural explorers to wander off the "safe" path of image synthesis.
in collaboration with Pavol Rusnak︎︎︎
- are featured in AI Art Gallery (as part of NeurIPS Workshop on Machine Learning for Creativity and Design 2020: http://www.aiartonline.com/
- are currently exhibited in DOX - Centre for Contemporary Art, Prague as part of the Vicious Circle exhibition: https://www.dox.cz/en/whats-on/datamaze-vicious circle
- took part in CtrlZ.ai - an off-site event of 2020 ACM Conference on Fairness, Accountability, and Transparency (FAT*) in Barcelona (Jan 2020)