So all the images need to go through replica filters but ai makes it trivially easy to make substantially different images from a single prompt so now we need an ai to infer the 'meaning' of an image. It all sounds like great news for chip makers and power generators.