This is a mechanical insect. This blog post is about a machine-learning algorithm studying existing insect common names to generate new ones. We’re mixing metaphors here a bit, but you should read on anyway. (Photo credit: Flickr/midorisyu, CC BY 2.0)

Insects get a lot of nicknames out in the world, whether it’s a farmer grumbling about a crop pest or a child marveling at a backyard bug. And these names vary from region to region. Early in the 20th century, the Entomological Society of America’s forebears recognized this as a point of confusion and established the Common Names of Insects and Related Organisms List, which ESA continues to manage (and add new names to) today.

In case you were wondering, no, ESA does not have a magical insect-naming machine at headquarters. Rather, insect common names come from the real world, proposed to and reviewed by the Standing Committee on Common Names of Insects and then recommended for approval to the ESA Governing Board.

But … what if there were an insect-naming machine?

Thanks to a researcher with a hobby tinkering with artificial intelligence, we now have something of an answer to that question.

I trained a neural network to invent new insect species, thanks to a dataset from @EntsocAmerica. Look out for the lick beetle. https://t.co/L90f6qYPeE pic.twitter.com/MqqqJBMP8V

— Janelle Shane (@JanelleCShane) February 1, 2018

Janelle Shane, Ph.D., is a research scientist in optics who likes to experiment with neural networks, and she shares her often amusing results on her blog, AI Weirdness. Late in 2017, Shane downloaded the full ESA common names list for insects and fed it to her neural network, which she describes as “a kind of machine learning program that imitates the way the human brain learns.”

Of course, the 2,200-plus names in the ESA list are all human-generated, originated in people’s minds and honed by the vagaries of human communication. What would an artificial intelligence make of that?

The results, as Shane shares, are in some cases amusing and in others, well, surprisingly plausible. The “waterpillar” and “magicking ant” are good for a laugh, but the “black pine needleminer” and “redbacked elm shortworm” are, if not ready to be slapped on an existing species right away, at least true to form.

Shane posted a longer list of machine-imagined insect names, accompanied by some fun illustrations of these new fantastic creatures, over on her blog yesterday. Click on over and check it out.