While working on my project to create abstract depictions of human pathogenic bacteria (see here and here), some of my prompts to Stable Diffusion didn't produce an image in the style that I wanted. But they still produced something interesting. For most of them, I have no idea what part of the prompt the AI latched onto or what similar word the AI went with. There were, however, a few themes that came up often. For example, if a bacterial genus named ended in –ella, the images often featured a woman. I guess it has figured out that –ella is a feminine diminutive. It also drew a lot of flowers, birds, and insects. My guess is that the training data set had a lot of flowers, birds, and insects captioned with their scientific names.
This was supposed to be Gardnerella. Clearly the AI thought of gardens.
This was supposed to be Gordonia. This one is actually easy. There is also a genus of flowers called Gordonia.
This was supposed to be Gordonibacter.
This was supposed to be Granulibacter.
This was supposed to be Grimontia.
This was supposed to be Haemophilus. I was so astonished by the things the AI was returning, that I kept going to see what I would get.
This was supposed to be Hafnia.
This was supposed to be Helicobacter.
This was supposed to be Isoptericola.
This was supposed to be Jonesia. At least some of these appear to be the former NFL player, Thomas Q. Jones.
This was supposed to be Kerstersia.
This was supposed to be Kingella.
This was supposed to be Klebsiella.
This was supposed to be Kocuria.
This was supposed to be Kytococcus.
This was supposed to be Laclercia.
This was supposed to be Leifsonia.
This was supposed to be Leminorella.
This was supposed to be Leptospira.
This was supposed to be Leptotrichia.
This was supposed to be Listeria.
This was supposed to be Mannheimia. The AI definitely focused on red, but I have no idea why.
This was supposed to be Mobiluncus. The first part of the name must have made the AI think of automobiles.
This was supposed to be Moellerella.
This was supposed to be Moraxella.
This was supposed to be .
This was supposed to be Myroides.
This was supposed to be Neisseria.
This was supposed to be Neorickettsia.
This was supposed to be Nocardia.
This was supposed to be Ochrobactrum.
This was supposed to be Odoribacter.
This was supposed to be Oerskovia. It appears that the AI thought of a Slavic country.
This was supposed to be Oligella.
This was supposed to be Olsenella.
This was supposed to be Oribacterium.
This was supposed to be Orientia. This time the AI surprised my by not drawing anything oriental, which I thought would be an obvious interpretation of the genus name.
These illustrations were drawn using Stable Diffusion 2.1.
