Experimenting with Tiny Datasets



In late March 2022, while sharing dinner with some fantastic FabLab folks in Southern Vermont, Neil Gershenfeld posed the idea of Locally-Made-Meaningful-Machines. My first response was, who wouldn’t want that?  Then, I realized the UVM Art + AI genetic algorithm was indeed a Locally-Made-Meaningful-Machine. Could we “bring it home” even more by downsizing our datasets from thousands to double digits? Would the genetic algorithm give us better results if we gave it fewer choices? If humans made more of the decisions?

These compositions were made by our genetic algorithm with tiny datasets. You clan click on these images to view larger ones. The gallery above will reveal animations.

Grid Prints

Original Stars

Below are six of the original founding works that make up the Athena Dataset, these were made by Jenn. Our genetic algorithm creates new compositions using their parts, (pictured above).



, ,