Seeing vs. Sensing: How Humans and Machines Read a Photograph
A photograph looks like one thing, but it is read in two entirely different ways. A person sees a scene — a story, a mood, a history, a set of relationships. A machine sees an array of numbers, and returns labels drawn from whatever it was trained on. Neither is reading the “same” image.
A human reads for meaning
When you look at a photograph, you're not decoding pixels — you're inferring a world. Who are these people, and how do they relate to each other? What just happened, or is about to? What's outside the frame? Does it make you feel something? You bring your own history to it, and no two viewers read it quite the same way. Ambiguity isn't a failure here; it's often the point.
A machine reads for classification
A computer vision model never sees a "scene." It sees a grid of numbers — brightness and color values — converted into statistical patterns matched against categories it was trained on. It doesn't know what a photograph is of in any felt sense; it outputs whatever label its training data taught it to associate with similar patterns, each with a confidence score. Change the training data, and the same photograph gets read completely differently. There's no ambiguity tolerance, no narrative — only the categories it has.