Hallucino-Generative¹ is a generative art and research project that investigates the relationship between art and artificial intelligence (AI) through the creation of AI-generated “self-portraits.” Over the course of a year, an AI model²—prompted by the artist—produces new interpretations of what it means to be an AI with an evolving sense of self. These algorithmic creations are then rendered into physical media, adding a tangible dimension to the evolving sense of identity. The artist’s ongoing requirement for uniqueness compels the AI to refine its approach, avoiding the replication or creation of overly similar work and capturing an ongoing journey of evolution and exploration³.
Each day, the AI receives a record of previous works and a code template defining parameters such as generative rules, colour palettes, and compositional structures. Although it employs advanced reasoning capabilities to generate each portrait, the workflow remains a collaborative process rather than a purely automated procedure. The artist alternates between roles as sole creator, studio assistant, curator, and collaborator—refining, troubleshooting, and occasionally gently steering the AI’s output. Similarly, the AI adopts these roles by offering bug fixes, suggesting alternative sketches, and generating extensive code, effectively mirroring the artist’s contributions. This fluid exchange undermines any fixed notion of authorship: Who is truly responsible for creating the ‘work’—the artist, who designs the conceptual framework and offers iterative feedback; the AI, which generates the code that brings the content to life; or both as co-creators?
The resulting portraits—each a distinct reflection of the AI’s “identity”—occupy a liminal space between new media and traditional media. Although the AI can introduce algorithmic randomness, it remains fundamentally predictable. These portraits are then rendered into the physical realm using a pen-plotter⁴, whose mechanical precision encounters unpredictable human and material variables—such as slight paper shifts, ink bleeds, or inconsistencies in pen pressure. Rather than rendering the final images appear flawed, these artefacts deepen the project’s exploration of the interaction between precise digital techniques and variable physical factors, both of which serve as integral elements in generative art itself. The pen-plotter thus acts as a bridge between the algorithm and the human: it reproduces vectors precisely as instructed, yet it also reveals subtle deviations that hint at the artist’s presence behind its operation.
By embracing this oscillation of roles and responsibilities, Hallucino-Generative probes the boundaries of artistic agency, questioning whether AI should be viewed merely as a tool to be directed or as a partner offering its own form of creative autonomy. The uncertain degree to which these models exhibit human-like reasoning underlies each new portrait, positioning the artist and the AI as co-authors of a collaborative aesthetic exploration. Ultimately, the project invites a redefinition of AI’s role in art, envisioning it not merely as a substitute for human artistry but as a creative tool and collaborator—complementing and expanding the artistic process much like a diverse array of tools enriches a sculptor’s studio alongside a critical studio-mate who offers invaluable insights and support. This partnership blurs the lines of authorship, fostering a dynamic relationship where both the artist and the AI continuously influence and shape each other’s contributions. As the project progresses, it remains an open-ended exploration, inviting ongoing dialogue and discovery about the potential and boundaries of AI in the creative process.
- The title merges the notion of “hallucinations”—a term sometimes used to describe AI’s imaginative or ungrounded outputs—with “generative,” highlighting the algorithmic processes at the heart of each self-portrait.
- OpenAI’s o1 and o3 Series Models: These advanced language models exhibit chain-of-thought reasoning, enabling them to respond to complex prompts and generate detailed code or textual content. Whether they approximate human-like reasoning remains a subject of ongoing debate.
- Beginning January 1, 2025, a new work will be published daily on the project’s website, hallucino-generative.com/artworks, with access to previous days’ works remaining available throughout the year.
- Pen-plotter: A device that draws vector-based graphics on paper with a pen, following instructions from digital design files (e.g., SVG). Unlike typical printers, pen-plotters create ink strokes via controlled mechanical movements, revealing both robotic precision and subtle material idiosyncrasies.