Mario Klingemann (b. 1970) is a pioneer of digital and AI-driven art, an artist who has spent over a decade pushing the boundaries of creative algorithms. His practice has been pivotal in defining the aesthetic of AI art, blending technical ingenuity with artistic intuition. Notably, this exhibition offers collectors a unique opportunity to acquire seminal works from a practice that continues to challenge and redefine what art can be.
In Early Works, Klingemann investigates the aesthetics and mechanics of AI-generated imagery, using machine learning not simply as a tool but as a creative collaborator. Long before “AI art” entered the mainstream, Klingemann was leveraging neural networks in unconventional ways to coax emergent forms, textures, and structures from data. A defining feature of his early work is his embrace of the imperfections inherent in early generative models: the glitches, surreal distortions, and convolutional artifacts that might otherwise be seen as errors. Rather than striving for seamless photorealism, he amplifies these artifacts through a custom upscaling process he calls “trans-enhancement,” yielding a distinct, hallucinatory quality. The resulting visual language is uniquely AI-born, images that hover in a liminal space between the organic and the abstract, hinting at familiar forms yet ultimately unmoored from conventional reality.
A crucial element of Klingemann’s practice is his innovative approach to dataset curation. He meticulously assembles training sets from a wide range of sources, from historical archives and album covers to explicit photographs, crafting bespoke “taxonomies” that imbue his algorithms with a particular visual sensibility. His experiments with early neural techniques such as PPGN, pix2pix, and CycleGAN predate the widespread adoption of GAN art, marking him as one of the first artists to explore generative faces and manipulated image synthesis. The techniques and datasets developed for one project often serve as the foundation for another, resulting in a body of work that evolves through a continuous, iterative feedback loop.
Visually, Klingemann’s work occupies a fascinating territory between abstraction and figuration. His compositions often hint at recognizable subjects - human figures, animals, or objects - yet remain elusive, shaped by the uncanny logic of neural networks. There is a palpable tension in his imagery, an interplay between clarity and ambiguity, structure and dissolution, which imbues his work with a dynamic sense of movement and transformation.
Early Works is not a retrospective in the traditional sense; it is a focused survey of four foundational series that chart Klingemann’s evolving dialogue with the machine. By exposing the inner workings of machine learning while celebrating its unpredictability, he has forged a new visual language, one that is both computationally innovative and deeply human in its curiosity and experimentation.
The exhibition features works from four series - SketchMaker (40), Chicken or Beef (40), CycleGAN Makeover (40) and Neural Abstractions (60).
SketchMaker (2007–2014)
SketchMaker is the genesis of Klingemann’s journey – an early generative art project built in Adobe Flash around 2007 that marked his first foray into machine creativity. In this work, Klingemann asked a radical question: Can a computer autonomously create images that “look like art”? His answer took the form of a procedural sketch generator driven by evolutionary logic. The system randomly assembled simple drawing and image-processing actions, “atomic” operations like drawing shapes, inverting colors, blurring, and rotating, into complex sequences, essentially coding a digital ape at a typewriter, blindly throwing paint onto a canvas. Each press of the spacebar yielded a new, unpredictable composition, a visual “DNA” strand of layered operations.
Crucially, SketchMaker did not rely solely on random generation; it featured a feedback loop to refine its output. Klingemann devised a “judgement module”, a custom classifier trained on examples of art versus non-art, that would analyze each new image and score its aesthetic resemblance to known artworks. This evolutionary process allowed the software to iterate toward compositions that possessed an uncanny art-like quality, effectively enabling the program to learn a rudimentary sense of taste long before deep learning became commonplace.
Conceptually, SketchMaker introduced motifs that would define Klingemann’s practice: glitch aesthetics, chance-driven creativity, and the transfer of agency from artist to algorithm. The visuals often emerged as jagged, abstract collages of color and form - sometimes chaotic, sometimes eerily harmonious - embodying the unpredictable nature of generative processes. By intentionally embracing the imperfections and “errors” of the system, Klingemann explored the beauty of machine misbehavior as an aesthetic frontier, posing the provocative question: “Who’s the artist now?”
Though created with pre-neural techniques, SketchMaker foreshadows the AI art movement to come. It stands as a historical keystone, a self-contained generative artist system that anticipated many of the questions and possibilities later realized through deep learning. In its glitchy, evolutionary sketches, one can already see the seeds of Klingemann’s later works, where algorithmic intuition for form and a rebellious embrace of computational spontaneity would flourish.
Neural Abstractions (2016–17)
Neural Abstractions represents Klingemann’s breakthrough into true AI-assisted image generation, using early deep neural networks to summon surreal new visions. Created in 2016–17, these works grew out of experiments with Plug-and-Play Generative Networks (PPGN) – an innovative technique that predates GANs, allowing an AI to generate images by maximizing the outputs of a classifier rather than by direct synthesis. Instead of training a network to draw per se, Klingemann ingeniously trained a suite of custom classifier models (on datasets he curated himself) and then set a generative process to “please” those models. In practice, he fed the network a target concept – not via text prompt, but via a classifier node such as “portrait” or “map” – and the system iteratively adjusted a noise image until the classifier was highly activated.
Rejecting the generic categories bundled with the original PPGN code (which skewed towards dogs and everyday objects), Klingemann plugged in his own bespoke categories. He hand-sorted tens of thousands of images – from vintage album covers to 19th-century book illustrations to even erotic photographs – into conceptual groupings, training neural nets to recognize these themes. In doing so, he effectively hacked the generative process to follow the peculiar contours of his personal datasets. The resulting images, which he termed “Neural Abstractions,” are richly ambiguous. Hints of familiar shapes, a silhouette that could be a reclining nude or the glint of an eye, dissolve into painterly washes of color and texture. The AI seems to grope toward forms without ever fully resolving them, offering a “base understanding of composition and interesting colors” while birthing proto-figures that hover on the brink of recognition.
Because the raw outputs were modest in size and clarity, a limitation of 2016-era networks, Klingemann devised a novel post-processing step that would become a hallmark of his style: transhancement. This custom technique fed the low-resolution neural outputs into a secondary pix2pix-based model that upscaled and hallucinated extra detail, layer by layer. Unlike traditional upscaling which merely smooths or interpolates, transhancement exaggerates the AI’s own artifacts, creating “beautiful convolutional artifacts” and razor-sharp edges that amplify the image’s uncanny character. By repeatedly doubling an image’s size, Klingemann transformed a 256px blurry apparition into a large, crisply freakish tableau.
The artifacts - glitchy striations, bizarre textures - are not flaws but the very soul of the work, imbuing it with an otherworldly, hallucinatory quality. In Neural Abstractions, one discerns a machine dreaming aloud, as if it were trying to reconstruct reality from only half-remembered shapes. This series solidified Klingemann’s status as a pioneer of AI art, establishing an aesthetic of ambiguity and unpredictability that remains “uniquely computational yet deeply human.”
CycleGAN Makeover (2017)
In CycleGAN Makeover, Klingemann turns the emerging tool of CycleGAN toward questions of portraiture, identity, and machine bias. Created in 2017, this series explores the AI-mediated transformation of gender in photographic images - a kind of automated drag act performed by neural networks. CycleGAN, a new generative model that learns to translate images from one domain to another without paired examples, provided the flexibility to reimagine gender. Seizing on this potential, Klingemann trained CycleGAN on two carefully prepared domains: one of female portraits and one of male portraits. By teaching the system to convert an image from one set to the other, he created a model capable of morphing faces across the gender divide. The process was akin to an AI “makeover”: feed in a photo of a woman, and out emerges a version with masculine features; feed in a man, and the network feminizes his appearance.
What makes this undertaking compelling are the cultural and representational questions it raises. The AI was not explicitly instructed on what “male” or “female” means, it deduced patterns from its training data, encapsulating prevalent visual stereotypes. The resulting portraits mirror societal biases as the machine subtly adjusts features - smoothing or lightening skin, altering jawlines, and modifying hair or stubble - to produce transformations that are startling and often off-kilter. A face may emerge with features that almost pass as plausibly male or female, yet remain ambiguous with mismatched eyes or an eerie smile. Klingemann embraced these unexpected outcomes as artistic serendipity, applying his transhancement technique to magnify odd details and produce ghostly, uncanny portraits that flicker between genders.
These images carry a haunting, archival quality. By mining open archives of 19th-century photography to compile a dataset of vintage faces, Klingemann allowed modern AI to “repaint” anonymous figures of the past. Conceptually, the series engages with themes of fluid identity and the algorithmic gaze. In an era when algorithms increasingly mediate self-presentation, his work questions what it means when a machine reconstructs you according to its learned biases. CycleGAN Makeover stands as a fascinating study in machine bias and creativity, a visual exploration that blurs the line between male and female, provoking us to reconsider our preconceptions.
Chicken or Beef (2018)
In the whimsically titled Chicken or Beef, Klingemann ventures into the realm of figuration and flesh, using AI to deconstruct the human body into abstract, meat-like forms. This 2018 series evolved as a side project of his “Imposture” experiments, where he trained GANs on images of nude bodies to explore how machines perceive the human figure.
Klingemann’s process began with a pix2pix model that learned to render a human form from a stick-figure pose, using a dataset of tens of thousands of full-body photographs curated specifically to capture unclothed, full-length poses. Instead of feeding the network conventional poses, he generated random, often impossible configurations, such as contorted mannequins with extra limbs or absurd postures beyond human capability. The AI dutifully attempted to render a human form from these nonsensical skeletons.
The outcome was a series of grotesque yet compelling images: fleshy, pinkish-beige masses that vaguely suggest torsos, limbs, or faces but never resolve into a discernible figure. These biomorphic forms are simultaneously anatomical and abstract, an arm melting into a torso, skin-like textures merging with void, placing the viewer in a perceptual limbo: are they seeing a contorted human, an animal carcass, or simply digital noise? The title Chicken or Beef wryly reinforces this ambiguity, evoking the trivial choice of an in-flight meal while humorously equating human forms to meat options.
Klingemann intentionally dialed up the strangeness to elicit outputs that are more abstract and less recognizable than his earlier Imposture series. Commentators have noted that his flesh-like distortions recall the Surrealist disarticulations of Hans Bellmer’s doll figures and the painterly grotesques of Francis Bacon. Each image emerges as the AI’s hallucination of a human in the absence of a coherent blueprint, a feedback loop of the algorithm’s imagination producing bizarre, creaturely shapes.
There is also a dialogue with the contemporaneous work of Robbie Barrat, a fellow pioneer of AI art who around 2018 was using GANs to generate nude figure studies. In fact, Barrat’s series of AI-generated nude portrait NFTs, given away at Christie’s 2018 Art+Tech Summit on 17th July 2018 and later dubbed the “Lost Robbies”, became iconic early examples of AI art’s cultural and market value. Klingemann’s Chicken or Beef, unveiled on the exact same day, stands as a companion statement.
