Answers are grounded in this paper and cite their sources.
If the modern house was once conceived by architect Le Corbusier as “a machine for living in,” its images have become machines for training artificial intelligence (AI) in. Larger than any architectural reference-book, AI training datasets now possess significant collections of digitized building interiors and domestic spaces that have been three-dimensionally (3D-) modelled, synthetically generated, or scanned from life. Within these datasets, a home is simultaneously unique and massive, scenographic and calculating, highly personal and utterly depersonalized. It is in this new aggregative context that the home becomes “invisual”: an ensemble calibrated for training machine vision rather than for human interpretation. What, and who, is being trained? I bring Haraway’s method of feminist objectivity to AI training logics, locating inequities that datasets engineer into the home and dismantling their “god trick” of distance and totality—proposing, instead, the notion of untraining as an imaginative act of un-domestication.
No takes yet. Share an insight, caveat, or question.
Amelyn Ng (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: