Silly me thinking the AI is doing edge detection, identifying objects, and populating a graph of semantic content, i.e., doing AI. No, it’s just comparing billions of bitmaps. No wonder it’s fooled by twiddling the LSBs. I’m much relieved.