There’s hardly anything that we haven’t tried putting machine learning to work at, and usually they take to it like a duck to water. That’s not the case here, with a neural network attempting to learn how to give paint colors those fancy names like “spring rain” and “harbor mist.” How would you like to paint your house in a nice “dorkwood” or “stanky bean”?
This unsuccessful attempt to replace the creatives in the paint-naming space was carried out by Janelle Shane, a researcher and flautist who occasionally “plays with neural networks.”
The network was given a list of 7,700 paint names and their corresponding RGB values, and set to work trying to suss out the hidden connections that govern these interesting yet occasionally obscure appellations.
At first things were promising, and the system seemed to be putting together a rudimentary logic, although it wasn’t using the “real” words for these colors.
Could the missing “a” in the third entry signify the purplish hue? Could “Caae” correspond to the overall mid-level brightness? It isn’t intelligible to us, but perhaps this is the beginning of a robust internal artistic grammar?!
A few more runs through the data and the network was ready and willing to produce highly creative and original names for colors, far more adventurous than your garden-variety “greige” and “royal purple.” Let us survey the results.