Knowledge is partialHow is knowledge constructed? Across human history, following the dominant epistemological criteria and political narratives, it has been written down or erased, extracted or discarded, circulated or silenced.
Artificial intelligenceArtificial intelligence as a new paradigm of knowledge is built on top of this fundamental bias. It reproduces the illusion of impartiality and completeness.
Or does it reinforce it?
Knowledge on the edgeCan we search for knowledge on the edge?
Images of alternative technologies?
Elements of nature with non-Western meanings?
Cultural expressions and rituals that are not part of the dominant narrative?
Marginal names the dataset, not the cultures.
ParadoxEvery image we can find to represent those margins has already been curated in some way.
Some Western museum or archive, or person with power, decided at a given moment that this image was worth keeping and publishing. That decision is the only reason it reaches us today.
Distortion trainingNevertheless, we can train an AI model on those images. The images are already filtered, so this is not a way around the paradox. It does not replace the model, and it does not correct the archive. The point is to make the mechanism visible: to see how the model associates what it was shown.
Alpha mapsThe trained model generates the alpha maps.
SculpturesThe alpha maps are used to build the sculptures: two principal sculptures in clay, and a group of individual sculptures in clay and plant-based resin.