Paul Jensen, a microbial-systems biologist at the University of Michigan in Ann Arbor, wanted to see whether artificial intelligence (AI) tools, such as large language models, could synthesize research on different microorganisms. So he located all the relevant papers on a bacterial species that his laboratory studies — an organism that causes tooth decay called Streptococcus sobrinus — but discovered that he’d already read all of them, a few dozen in total.

Many microbiologists will be in a similar position — or worse. In a study posted on the preprint server bioRxiv last week¹, Jensen found that just ten bacterial species account for half of all publications, whereas nearly three-quarters of all named bacteria don’t have a single paper devoted to them.

“We’ve learnt a lot about a small number of species,” Jensen says. But “for a lot of bacteria, there’s nothing for a language model, for an AI to read”. The microbes important for human — and Earth’s — health are especially understudied, researchers say.

Read more: click here