I miss crappy AI art

In the last few years AI image generation has gone from something very rudimentary and nigh-on unusable for commercial purposes to the go-to option for small businesses' signs and menus from SF to Rome—this sucks. Somehow, as the models have gotten better, they've left something behind. I find myself thinking more and more wistfully of DALL·E 2, an image generation model from 2022 that conceptually shouldn't hold a candle to the current state of the art. So why do I miss it?