Yup; you identify a new mode of failure! If it cannot recognize one image as identical to another, it ends up overtraining and over-generalizing on the dups...

Yup; you identify a new mode of failure! If it cannot recognize one image as identical to another, it ends up overtraining and over-generalizing on the dups.... Again, showing zero of anything resembling abstract understanding; merely a statistical correlation of blobs within each image with the text, e.g., finding the common set of pixels in each image that corresponds to "astronaut" and filtering out the rest. Obviously, this is far more useful than color-key-deletion or whole-image search. But, it isn't intelligent abstraction.