Raw clips from an enormous dataset used for AI training — salvaged, selected, and sequenced (manually, lovingly,) to create a montage about action, reaction, labor and churn. It’s the beautiful and the banal, a meandering and confusing #corecore edit, or a deeply existential, meta “How it’s Made” episode where the episode itself is the product being made.
But mostly, it’s me trying to capture the overwhelming feeling of watching these clips. For an AI system, this footage is only valuable in the aggregate, as one tiny piece of data that can be collapsed into a legible, context-free average. It’s mostly useless on its own. But as I dive into these clips, scrolling through thousands and thousands of them, all captured by real living people, in their own particular time, place, and context, it often feels like I’m mainlining all the beauty in the world at once.