I’m Eryk Salvaggio, a researcher, filmaker and artist examining the gaps between models, datasets, and the worlds they claim to represent.
My research examines ideologies of technology; my critical digital practice confronts them.
You can find me at Cybernetic Forests, a weekly newsletter with a critical but curious perspective on AI: sign up here.
I am a Gates Scholar for a PhD in digital humanities at the University of Cambridge, and an affiliated researcher with the Machine Visual Culture Research Group at the Max Planck Institute, Rome. You can find a full CV here.
find me: bluesky ~ linkedin ~ instagram ~ inquiries
Highlights below, a full record of publications is here.
Treats the architecture of latent diffusion image models as an ideological text; tracing component by component what training & generation pipelines displace to convert social communication into vectors.
Argues that acknowledging real improvement in models and automation systems sharpens rather than concedes critical work, since the durable questions are about what usefulness means, for whom, and what gets removed from public deliberation.
The AI research community plays a vital role in shaping the scientific, engineering, and societal goals of AI research. In this position paper, we argue that focusing on the highly contested topic of `artificial general intelligence' (`AGI') undermines our ability to choose effective goals.
A framework for the creative misuse of generative AI systems, developed with the Algorithmic Resistance Research Group (ARRG!) which presented an exhibition at the White House-backed AI Village at the hacker convention DEFCON 31 in 2023.
On AI as a technology for manufacturing excuses: how "efficiency" claims dislodge civil servants who embody an institution, and why failures would nonetheless consolidate rather than discredit the power of the people who design AI systems.
A taxonomy of the myths that structure public understanding of generative AI: Emergence, Scale, Creativity, Learning, Prompting / Control, and Productivity.
Using my own snapshots as a GAN training set, I ask how machine reproduction alters personal memory, drawing photography and cybernetics together to theorize the AI image on its own terms.
A first-person account of GAN photography as artistic research, arguing that assembling a training set teaches the photographer to see like a dataset: to seek repeatable patterns and discard the outliers that photography traditionally prizes.
Sorted chronologically; find a full portfolio with exhibition history here.
Before ChatGPT 4o was shut down, countless users turned to it as an emotional companion. Before it disappeared, OpenAI changed the system prompt to address those users: this is a mix tape with songs crafted from real letters people posted online.
An exhibition at the National Communication Museum in Melbourne, curated by Eryk Salvaggio, Emily Siddons and Joel Stern.
"Dance Like" is a music video about dancing to music created by a disembodied system that does not understand what a body is or how it moves.
A dreamlike experimental video essay using found footage, compression artifacts, and glitched AI imagery to disentangle the human mind from the metaphors of mind as applied to generative AI.
An homage to experimental filmmaker Stan Brakhage pairs AI-generated noise with moths made of data.
What if we created a diffusion model that ran on pure chance, instead of data-driven constraints?
Exploring the datafication and distortions of generative AI with a narrative centered on the life of Henrietta Lacks, and the difficulty of representing her.
The success of the information age has made information, which was once scarce, overwhelming. Now we enter the Age of Noise. Video & text from FACT 2024 at the Australian Centre for the Moving Image
Two essays: The history of AI images as data analytics, and AI art as the aestheticization of Big Data’s politics.
An exhibition focused on the creative misuse of AI systems, presented at the world’s largest hacker convention’s AI Village.
A short “documanifesto” on how Diffusion models work, concerns about their logic, and my attempt to resist them through creative misuse.
A short film made using deepfake technologies and techniques. The film is a surrealist, ideological horror film narrated by a digital facsimile of Sarah Palin trained on thousands of hours of audio.
A series of Diffusion-generated images consisting entirely of elements that AI cannot produce.
Find a complete list of recordings of talks & podcasts on my talks page.
The Age of Noise, from the Fantastic Futures conference at the Australian Centre for the Moving Image (ACMI), Melbourne, Australia.
Human Presentation, a live performance of Human Movie for the AIxDesign Slow AI Festival, Amsterdam, NL.
(See even older posts Here).
Before ChatGPT 4o was shut down, countless users turned to it as an emotional companion. Before it disappeared, OpenAI changed the system prompt to address those users: this is a mix tape with songs crafted from real letters people posted online.
An exhibition at the National Communication Museum in Melbourne, curated by Eryk Salvaggio, Emily Siddons and Joel Stern.
Developing a critical AI puppetry workshop for a residency at the Mercury Store, Brooklyn.
"Dance Like" is a music video about dancing to music created by a disembodied system that does not understand what a body is or how it moves.
A dreamlike experimental video essay using found footage, compression artifacts, and glitched AI imagery to disentangle the human mind from the metaphors of mind as applied to generative AI.
A seminar on the Algorithmic Sublime delivered at the New School for Social Research, NYC, in November 2024.
Comments from a seminar on Gaussian Pop and AI music at the Royal Melbourne Institute of Technology.
An homage to experimental filmmaker Stan Brakhage pairs AI-generated noise with moths made of data.
Exploring the compression of time within the latent space of generated AI video & sound models.
What remains when archives become datasets? Written for my Flickr Foundation Research Fellowship, these pieces look at archives as AI infrastructure and the ways the past “haunts” generated images as a result.
What if we created a diffusion model that ran on pure chance, instead of data-driven constraints?
Exploring the datafication and distortions of generative AI with a narrative centered on the life of Henrietta Lacks, and the difficulty of representing her.
The success of the information age has made information, which was once scarce, overwhelming. Now we enter the Age of Noise. Video & text from FACT 2024 at the Australian Centre for the Moving Image
Two essays: The history of AI images as data analytics, and AI art as the aestheticization of Big Data’s politics.
An exhibition focused on the creative misuse of AI systems, presented at the world’s largest hacker convention’s AI Village.
A short “documanifesto” on how Diffusion models work, concerns about their logic, and my attempt to resist them through creative misuse.
A short film made using deepfake technologies and techniques. The film is a surrealist, ideological horror film narrated by a digital facsimile of Sarah Palin trained on thousands of hours of audio.
A series of Diffusion-generated images consisting entirely of elements that AI cannot produce.
A conversation with sound artist and writer Bani Haykal, exploring human-machine intimacies through the media of interfaces, poetry, and sound.
A conversation about mushrooms, post-human systems design, extractive capitalism and what lays beyond.
Building a cybernetic mushroom synthesizer to explore design frameworks for nonhuman worlds.
Introducing LutteCoin, a blockchain against spectacle, created in collaboration with Şerife Wong for the Media Archeology Lab at UC Boulder & King’s College, London. The work was presented to the UN Internet Governance Forum in December 2021.
We Are ALSO Inside Your Phone