A poem speaks things – all 3D scans from Sketchfab – into place: mutter matters.
Matter Mutter: Talking Things explores the intersection of computation and architecture. It delves into how the mechanics of machine learning invite revisiting how we see and think of things and objects, both technical and digital. The essay proposes the concept of the 'vectorial object'; far from being a mere byproduct of unsupervised machine learning processes, it allows the trained model to associate a set of seemingly ordinary 3D scans—including vegetables and infrastructural cabinets—with thousands of poems. The hyperdimensional vector arises from words: titles, descriptions, and—sometimes contradictory—classifications created by a deep neural net evaluating rotating renders. An interface with the trained model allows poetry to lend each thing a plethora of potential poems, making matter 'mutter' as much as 'mutter' matters. This interplay invites reflection on questions around naming and identity, the semantic construction and attribution of meaning, discussions on alternative, non-analytical approaches to artificial intelligence, and speculation on how we might experience and build in such an augmented world. Matter Mutter was published as a full article in 2023 in Scroope 32, the Cambridge Architecture Journal.
