The materiality of computation has only recently come into focus.[1] The recent boom of large language models has pushed that awareness further, drawing attention to the remarkable levels of energy required to sustain contemporary artificial intelligence.[2] As AI becomes ubiquitous and ecological crises intensify, it is increasingly understood that digital infrastructures are extractive, consuming vast amounts of energy and critical minerals while reorganising ecologies, space and labour. So-called ‘efficient’ computational models are still geological in their impact, with consequences that unfold across terrain, labour, circulation and bodies.
AI often appears instantaneous. Digital infrastructures and the minerals required to construct them remain largely invisible, fostering an illusion of immateriality and separation from the Earth. But every AI system is routed through mineral behaviour.[3] For instance, cobalt from the Democratic Republic of Congo is used in AI processors and copper cables that carry electricity to data centres.[4] The way these minerals conduct electricity, hold heat — their weight and resistance — all shape how digital computation works, participating in every algorithmic operation.
Beneath our feet, stone is organised through temporal processes of pressure, fracture and compression. Geological layers form with distinct densities and capacities to store or release energy. The underground is dynamic, shaped by physical properties as well as cultural memory, collective trauma, myth and traditions. Strata have structured both the material and the immaterial: patterns of settlement, migration, labour, and imagination. Acknowledging the spatial and material impact of AI leads us to question what kinds of relations we are building with the ground.
Architecture and spatial practice have long negotiated between bodies and environments. As the urban theorist Kevin Lynch suggests, the body is often the first receptor of spatial experience, capable of organising cities through perception.[5] In mining, this perceptual intelligence emerges directly from the body. The ground is encountered through strain, imbalance, vibration and resistance before it is ever measured or modelled.
In our practice — which combines spatial research, architecture, movement and dance — stones actively shape movement. Their weight alters rhythm, their texture redirects trajectories and their resistance can slow us down or invite us to try new paths. Mineral properties enter into dialogue with bodies, participating in the design of movement as they do in the design of circuits. By contrast, artificial intelligence largely operates as disembodied cognition, separated from sensation, fatigue and effort, qualities through which bodily intelligence emerges.[6]
Extraction as Choreography
The mineral base of AI comes from places with long, uneven histories of extraction, labour and care. To understand how computation is grounded, we approach it through bodily practices that bring minerals into circulation.
In Lubumbashi, the mining capital of the Democratic Republic of Congo, cobalt and coltan are extracted through manual labour that reshapes human movement into geological engagement. Space is produced through action as workers descend into narrow shafts, dig open pits and carry heavy loads. Each gesture is calibrated by density, fracture and instability in the earth. Bodies become adaptive infrastructures, evolving in response to depth and resistance.[7] This is a form of spatial intelligence without metrics or optimisation, an algorithm written through bodies and geology.
Artist and photographer Sammy Baloji has documented how extraction reorganises bodies and pathways in the mining regions surrounding Lubumbashi. Mémoire (2004-06) depicts this through the careful layering of colonial archival photographs of mine workers onto contemporary images of industrial ruins, illustrating how mining structured movement as discipline rather than choice.[8] Bodies are depicted mid-gesture, carrying loads, standing in formation and laboring under supervision, inscribing the industrial landscape shaped by repetitive, imposed motion linked to mineral production. Baloji’s work demonstrates that mining is a system that conditions bodies to move in specific ways, leaving traces that influence how history is experienced.
Play as Method and Counter-Infrastructure
If extraction scripts movement through material constraint, play offers another potential. In 2002, Francis Alÿs orchestrated When Faith Moves Mountains (2002), a work in which five hundred volunteers used shovels to displace a sand dune outside Lima by about four inches. This collective labour had no productive output. Geological time was compelled into human rhythm, for no reason other than the doing itself. In Alÿs’ La Roue (2002), set in Lubumbashi, a child rolls a tyre through unpaved streets — pushing it uphill with effort — then crawls inside to roll down at speed. The hill itself is a pile of mining waste, a residue of extraction and exploitation. The child transforms this extractive landscape into a space for play where the ground is not dominated but participated with.[9]
In these extractive territories, play unfolds alongside labour; the same terrain that demands gruelling physical work also invites improvisation and invention outside of productivity. These examples of mineral play also show how play is a way of thinking with material resistance. The tyre’s trajectory is determined by the irregularities of the ground. The game has rules, but they are responsive to the terrain. The ground computes the action. Play makes a mineral logic legible, with matter instructing movement and constraints generating possibility.
Play thus forms a counter-infrastructure to extraction: temporal, non-optimised and collectively negotiated. It resists logics of maximisation, allows failure and rejects predetermined outcomes. In extractive zones, play enables care where care is structurally denied. Children playing near mines negotiate with extraction, inventing rules from the same debris and terrain that shape labour.
Building on these observations, we develop movement scores as analogue algorithms, spatial instructions that arise from negotiating with materials instead of executing predefined code. Weight, resistance, pull and texture are instructional tools. A stone held in the hand becomes a collaborator, the mass determining how quickly you can move, how long you can hold a position, and where fatigue enters the body. We are not looking for a fixed outcome. Movement emerges through reciprocal adjustment with matter as co-author, mirroring the often invisible narratives of geological substance.
This approach reframes intelligence as the calibration of relations between bodies and matter, and acknowledges that stones possess agency and that resistance constitutes a form of information. We propose a geology of computation that rejects the fiction of immateriality. Geological AI proposes a shift from ground as a resource for extraction to ground as a collaborator in emergent forms of intelligence. It challenges dominant narratives of AI as a purely technological solution by foregrounding embodied and spatial intelligence.
If computation can learn from the conditional logic of play — where action emerges through feedback with material constraints instead of fixed optimisation goals — AI can begin to acknowledge its material impact and open the possibility of becoming slower and more relational. This implies more inclusive systems that value local and Indigenous knowledge, feminist and slow AI, and smaller models grounded in situated relations rather than large-scale optimisation. It suggests development that includes withdrawal as a valid action.
This perspective extends beyond art or speculation. Current debates regarding the formalisation of artisanal and small-scale mining (ASM) reveal a tension between industrial models of efficiency and local forms of geological knowledge.[10] ASM practices are often dismissed as inefficient compared to large-scale, machine-driven extraction. However, they are deeply attuned to local conditions, embedded within community life and capable of distributing economic opportunity more equitably. Viewed through the lens of geological play, these practices — when carried out with attention to safety and environmental responsibility — offer ways to negotiate with geology rather than override it. They show that extraction does not need to be maximised to be meaningful; that optimisation can be redefined as the capacity to sustain relations between land, labour and community over time.
----
[1] Jussi Parikka, A Geology of Media, University of Minnesota Press, Minneapolis, 2015, p. 4-6.
[2] Emma Strubell, Ananya Ganesh and Andrew McCallum, ‘Energy and Policy Considerations for Modern Deep Learning Research,’ in Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 34 No. 09: Is. 9, EAAI-20 / AAAI Special Programs, pp. 13693–13696.
[3] Pitron Guillaume, The Rare Metals War: The Dark Side of Clean Energy and Digital Technologies, Bianca Jacobsohn (trans.), Scribe Publications, London, 2020.
[4] Amnesty International, This Is What We Die For: Human Rights Abuses in the Democratic Republic of the Congo Power the Global Trade in Cobalt, Amnesty International Ltd, London, 2016.
[5] Kevin Lynch, The Image of the City, MIT Press, Cambridge, MA, 1960, p. 2.
[6] Shaun Gallagher, How the Body Shapes the Way We Think: A New View of Cognition, Oxford University Press, Oxford, 2005.
[7] Yamba Amisi Mwana et al., L'activité minière au Katanga et la perception de ses impacts à Lubumbashi, Kolwezi, Likasi et Kipushi, Presses Universitaires de Liège, Liège, 2018.
[8] Sammy Baloji, Mémoire, photographic series, 2004–2006.
[9] Francis Alÿs,The Nature of the Game, Claes Gerard-Jan and Stéphane Symons (eds.), Leuven University Press, Leuven, 2023.
[10] Célestin Banza Lubaba Nkulu et al., ‘Sustainability of artisanal mining of cobalt in DR Congo,’ Nature Sustainability, 1, 2018, p. 495–504.
----