While intensively generating images with artificial intelligence, particularly the Stable Diffusion image model, I often found myself limited by computational power. From insufficient memory to outdated frameworks, these obstacles tethered abstract computation to the physical world, exposing how materiality plays a critical role in the implementation of AI. One crucial aspect of my argument is how art production is intrinsically linked to commodity materials. Typically, discussions surrounding AI focus on intangible elements such as algorithms, data, and computational power. However, behind these intangible components lies a deeply material reality. AI is not merely code; it relies heavily on silicon, electricity, water, and human labor. It is anchored in extensive networks involving resource extraction, semiconductor manufacturing, global supply chains, and energy-intensive data centers.
Throughout my studies, new AI models continually emerged, each requiring increasingly advanced hardware, notably the latest graphics cards. The global semiconductor shortage during the pandemic further complicated acquiring necessary electronic components, impacting my ability to build media installations and significantly increasing hardware investment costs. Although many AI models are accessible online, massive data centers belonging to large corporations remain home to countless graphics cards operating relentlessly.
NVIDIA, which monopolizes the high-end graphics card market, maintains proprietary control over the CUDA framework, further restricting open research into AI technology. Such developments prompted me to delve deeper into the geopolitical narratives surrounding semiconductors, from the EU Chips Act and Taiwan's Silicon Shield to symbolic moments like NVIDIA's CEO sharing tofu pudding with TSMC's founder, Morris Chang. Investigating these histories and geopolitical dimensions of semiconductor production has encouraged reflections on my identity within the rapidly evolving landscape of technological power dynamics in AI.