DIFFUSED STATES continues an ongoing exploration of counterfactual technological histories, imagining a moment when artificial intelligence developed not as a centralised, corporate-driven infrastructure, but as a set of small, distributed systems embedded within local communities. Rather than scaling towards planetary computation, this speculative trajectory follows a slower, more situated evolution—one in which algorithms adapt to the languages, histories, and needs of their immediate context.

The project draws on the history of low-resource and locally maintained computing, from community-run bulletin board systems to peer-to-peer networks sustained by recycled hardware. These precedents inform an imagined shift towards AI systems that are modest in scope, but deeply embedded in the cultures that maintain them. In this scenario, data does not flow through opaque corporate pipelines, but diffuses through neighbourhood workshops, shared servers, and informal archives.

The title refers to multiple senses of “diffusion”: the technical process of image generation; the social circulation of tools and knowledge; and the physical dispersal of small-scale devices across different environments. The work combines AI-generated imagery with human-written captions, forming a dispersed narrative around how such systems might emerge—from grassroots tinkering to collectively governed infrastructures and hyper-local training data.

By working with AI models trained on biased and incomplete datasets, DIFFUSED STATES engages with both their limitations and their possibilities. Instead of treating these models as definitive or universal, the project approaches them as partial and provisional—open to reinterpretation, modification, and play. In doing so, it aligns with the idea that technology, even when shaped by exclusionary structures, can be repurposed towards shared imagination, local autonomy, and slower forms of innovation.

Diffused States was presented as part of Data Fluencies: Tributaries, curated by Roopa Vasudevan, which brought together artists exploring the ways data can be resisted, rerouted, or reimagined outside dominant narratives.