Neural Abstractions by Mario Klingemann at Fellowship, London

We’re excited to share the release with Fellowship: Neural Abstractions by Mario Klingemann, opening September 25 in London. Presented as part of the solo exhibition Mario Klingemann: Early Works, 2007–2018 at Fellowship, the collection highlights one of the artist’s most influential early series.

Mario Klingemann, Neural Abstraction #8, 2016-2017

Klingemann is among the first artists to work extensively with machine learning and neural networks as a creative medium. A self-taught programmer since the 1980s, he has spent over a decade exploring algorithms not just as tools, but as collaborators in the artistic process — a practice that has shaped the foundations of today’s AI art. He turned to machine learning technologies such as PPGN, pix2pix, and CycleGAN at a time when these tools were still unfamiliar to most artists. Instead of aiming for realistic results, he emphasized the flaws of early networks. He developed his own method, “trans-enhancement,” a process that exaggerated rather than corrected the glitches of the system. This approach gave his work a unique visual style that balances between abstraction and figuration.

These early experimentations form the basis of Mario Klingemann: Early Works, 2007–2018, an exhibition presented by Fellowship . Instead of a traditional retrospective, the show brings together four foundational series, SketchMaker, Chicken or Beef, CycleGAN Makeover, and Neural Abstractions, showing how he experimented with the aesthetics and mechanics of neural networks, exposing their inner workings.

Mario Klingemann, Neural Abstraction #9, 2016-2017

 Among these, one of the most influential projects is Neural Abstractions, which he created in 2016–17 with the use of Plug-and-Play Generative Networks (PPGN). Klingemann trained classifiers on his own hand-selected datasets that included old illustrations, vintage album covers, and even erotic photographs. Then, he let the system attempt to generate images that would “please” those classifiers. Neural Abstractions was first partially shown in January 2024 at the Kate Vass Galerie booth at NOISE Media Art Fair, Istanbul.

Visually, the resulting images are full of ambiguity. Shapes appear and fade, sometimes suggesting figures, sometimes just colours and textures. They connect to traditions of abstract painting in their focus on painterly texture and composition, but they are born from the unique constraints and possibilities of neural networks. Glitches and distortions are key elements in the images that give the works a dreamlike, almost hallucinatory quality.

Neural Abstractions is on view in the exhibition Mario Klingemann: Early Works, 2007–2018, presented by Fellowship in London from September 25 to October 9, 2025. Alongside the exhibition, Fellowship will also release a drop of the series on October 2, 2025, featuring 60 works.

Mario Klingemann, Neural Abstraction #3, 2016-2017