Overgrowth

A feedback loop built on DeepDream. Each pass takes what the network found in the previous image and offers it back as evidence for the next. A weak recognition strengthens, repeats, and recruits the surrounding texture until the image becomes crowded with the things the network was already prepared to see.

Nothing new has to enter the loop. The abundance comes from recurrence: interpretation becomes input, then returns as stronger interpretation. The representational habits of a system trained on other people's images become literal growth, spreading until the prior has remade the image in its own likeness.