Subprime Images, by Hito Steyerl: Unveiling the Future of Photography
In a thought-provoking interview with Katrina Sluis of the Computational Culture Lab at ANU's School of Art and Design, Hito Steyerl, a leading theorist and artist, delves into the intersection of machine learning, artificial intelligence (AI), and photography's evolving role. Steyerl examines how AI-generated imagery is transforming aesthetics and the politics of visual culture. Her insights reveal how photography, once a tool for documenting reality, now intertwines with digital systems of data extraction, algorithmic bias, and speculative finance.
The Rise of Subprime Visibility
Steyerl challenges our understanding of photography in the AI era, where images are no longer solely products of the human eye and camera but results of complex, automated processes. Her exploration of derivative images and subprime visibility highlights profound shifts in image production, circulation, and valuation within the digital economy.
From Photography to Derivative Imagery
Introducing the concept of derivative images, a term coined by Jonathan Beller, Steyerl situates photography within the realm of financial speculation. Much like financial derivatives, AI-generated images become speculative bets on visual culture. These images are composites, statistically averaged from vast datasets often scraped without consent from countless online sources.
In traditional photography, the relationship between subject, photographer, and viewer maintained a sense of integrity and authorship. Steyerl argues that AI image generation disrupts this relationship, producing what she calls “subprime visibility.” Similar to subprime mortgages that bundle risky loans, AI-generated images amalgamate diverse, sometimes conflicting data, resulting in a structural mediocrity. These images, often hollow and statistically derived, lack the depth and intentionality associated with traditional photographic practices.
The Politics of Image Training and Data Extraction
Steyerl examines the politics of AI-generated imagery, exposing the labor and data infrastructures that enable these processes. Central to her argument is the exploitative nature of data extraction. Data, often scraped without permission, becomes the fuel for AI models like Stable Diffusion. Images and text from platforms, creators, and users are harvested en masse, then refined and repackaged to generate new visual content.
She also highlights the labor of “ghost workers” — underpaid, often invisible workers tasked with cleaning, annotating, and processing data for machine learning. This hidden workforce, many from disadvantaged backgrounds, forms the backbone of AI image generation. The “subprime” quality of the images not only reflects the datasets but also the exploitative labor systems that produce them.
The Collapse of Meaning in AI-Generated Images
Steyerl raises concerns about the potential collapse of meaning in AI-generated imagery. As models are increasingly trained on their own output — a recursive loop of AI-generated content feeding back into the machine — the probability space narrows. This results in what Steyerl describes as “machine dementia,” where the diversity and richness of image production diminish to a set of statistical probabilities. This leads to repetitive, shallow visual outputs that fail to resonate with human viewers as traditional photographs do.
Her notion of “model collapse” parallels the 2008 financial crisis, where derivative markets failed catastrophically due to flawed assumptions and speculation. Similarly, AI-generated imagery risks a collapse in cultural and aesthetic value. The endless recycling of existing data threatens to erode trust in images, as they increasingly reflect algorithmic biases rather than human vision or experience.
Toward a New Aesthetic Paradigm
Despite her critical stance, Steyerl doesn’t entirely dismiss the potential of AI-generated images. She hints at the possibility of a new “thermodynamic” paradigm of image-making, moving away from traditional optical and electronic frameworks. This shift emphasizes the energy, heat, and entropy involved in data processing and image generation. In this new aesthetic paradigm, images are not just representations but part of a larger, energy-consuming system of information production.
Steyerl observes that AI image generation has sparked widespread protests, particularly from artists and content creators whose work is appropriated without consent. She views these protests as a form of recursive action, where the technologies used to produce and distribute images also become catalysts for resistance. The social impact of these technologies lies not only in their output but in the counteractions they provoke, as artists and communities fight to reclaim control over their creative labor and data.
Conclusion: Rethinking Photography in the Age of AI
Steyerl’s insights invite us to reconsider photography as a dynamic, evolving practice caught in the crosscurrents of technological change, labor exploitation, and financial speculation. In the AI era, creating an image involves navigating vast, opaque networks of data, algorithms, and labor, where images are produced by systems rather than individuals.
In this context, the aesthetics of photography must evolve to address the ethical, political, and ecological implications of AI image generation. Steyerl’s concept of “subprime visibility” offers a framework for understanding how new forms of visual production undermine image integrity, while her focus on the political economy of data underscores the urgent need for resistance and collective action.
As we advance further into the digital age, Steyerl’s work reminds us that photography is no longer just about seeing — it's about understanding the hidden structures that produce what we see.