Why AI-generated food images often look unsettling
Restaurants, cafes and food brands are increasingly using artificial intelligence to create promotional images, but many of the results look unappetizing. Examples described in the report include shrimp shaped like donuts, Reuben sandwiches appearing to come from the deep, and noodles with wormlike forms. The images can remain recognizable at a broad level while their smaller details become distorted.
Giovanbattista Califano, a behavioral scientist at the University of Naples Federico II in Italy who studies reactions to AI-generated imagery, said diffusion models have particular difficulty producing thin structures that continue and end in a clear way. Noodles, strands and tendrils therefore may spread into areas where they have no logical culinary or anatomical connection. Similar problems affect repeating elements such as bubbles and seeds, which can cross boundaries that should contain them.
According to Califano, these limitations help account for the frequent presence of tangled noodle-like shapes, disturbing patterns and dense clusters of holes in AI-generated food pictures. Some of those clustered patterns can evoke trypophobia, an aversion or discomfort associated with closely grouped holes.
Written by Sakhanda Wire AI from reporting by The Verge. Facts, quotes and full context are in the original article below.