AI can rapidly find similarities across millions of images, suggest dates and styles, and point researchers toward overlooked detail. This is one part of looking: comparison. But sensing bodily scale, taking responsibility for a historical claim, or having personal memory unexpectedly triggered cannot be reduced to classification.nnThe central question is not whether a machine ‘understands art’, but how much authority we grant its answer.
A system trained on existing archives may amplify old exclusions; a confident tone may erase uncertainty. Its best role is not automated critic delivering verdicts, but an accountable tool that cites sources, offers alternatives and returns people to the work itself.
