AI Epistemic Overhang

Paired with the familiar application overhang (AI can do things society cannot integrate or adopt), the epistemic overhang is AI discovering, knowing or disclosing things that human science — and human self-understanding more broadly — has a hard time modeling and incorporating. Coined in the Ends of Science studio project (Zhu, Freudenheim, Sekalala), which treats AI as an experimental "model superorganism." Bratton's key move is to insist these overhangs are not pathologies to be aligned away but "the deeper point of AI": the space in which unanticipatable Copernican shifts will land, and which must be defended rather than closed by Reflectionism.