Another way to look at it, if text is the only thing holding these things from making enormous amounts of ads and other digital content for whatever purpose, solving the text problem makes it go from taking a few seconds of human input per image to literally 0. If it takes literally 0 human input it asymptotically increases how much slop they produce to a ridiculous extent.

And the end game of that is going to be having completely unique ads constantly generated by AI that are then A/B tested against one another with generations of ads being phased out based on engagement levels and used to train the next wave of ads. Humans will optimise these ads over and over and over and over and over and over and over. The smallest mutations with each generation as it gets more and more and more optimal at reaching a target market.

Hyper optimised advertising at a level that is unachievable with human editors with permutations of ads that only the fever dreams of an AI can fart out because a human would literally never consider sticking 3 random thumbs and a pinkie sticking out the side of a coca cola can that for some unknown bizarre reason people will click on again and again.