Kansas

How a Kansas Entomologist is Using AI to Save the World’s Hidden Bees

Brian Spiesman is an entomologist at Kansas State University and creator of the bee-tracking app, BeeMachine.

Four images of Brain Spiesman, his Beemachine app and his lab above a red and blue gradient line
2 min read

As an entomologist, I've spent a lot of time thinking about bees. We’re all aware of their plight, but we often focus on the well-known species like bumblebees and honeybees. They’re large, easy to study, and relatively charismatic. Yet thousands of tiny, solitary bee species remain overlooked, largely because they’re tricky to identify, and funding tends to go to the more economically important honeybee species, which play a vital role in global food production.

This lack of knowledge is a real problem. We know habitat loss, pesticides, and disease are devastating bumblebee populations. But without identifying and tracking other bee species, we can’t grasp the full extent of the problem.

I faced this firsthand a few years back while working on a bee project. We had hundreds of unidentifiable specimens that required expert analysis, and the process took months, creating a serious bottleneck. And with fewer and fewer experts around, it's only going to get worse.

I realized AI might be the answer. I started gathering data sets and adapting pre-trained models within Google’s TensorFlow so they would start to recognize hard-to-categorize bees, and it just grew from there. That's how the BeeMachine, an AI-powered app and website for identifying bee species, was born.

Brian Spiesman, Entomologist, Kansas State University

The app is pretty straightforward. You take a picture of a bee, and the AI gives you the top three species predictions. We're now using Google Colab for model training, which is super convenient because there’s a Gemini coding assistant that can help us quickly troubleshoot any time we have a bug in the code. We're also exploring adding an AI assistant for the app, powered by Gemini and BERT, for users’ bee-related queries.

I’m teaching my students how to use these tools in Google Colab, too, through a course called Digital Entomology, where I’m basically replicating what I've learned in this process with BeeMachine. At the core of it, I want them to learn the basics of how to apply AI in the field, and they’re already starting to take it further in their own work.

Beyond training future entomologists, we really want to develop this citizen science component. An exciting new addition is a community feature where people can share their sightings. That way, people can see where their favorite bee is living and how it’s doing. Two years since launching the app, we have over 6,200 users, more than 20,000 sightings, and I’m aiming to nearly double our species identification capacity.

I think we’re at a spot where we're not limited by what the AI models are able to do. We're more limited by the inputs we have available to train them, so that's really what we're working on right now.

Brian Spiesman, Entomologist, Kansas State University

Ultimately, it's about getting people excited about bees. We want to get people out in nature, appreciating these pollinators, and collecting data that scientists can use for large-scale monitoring. There’s this massive citizen science effort where thousands of people in North America collect data about birds, and it’s been invaluable for bird research and conservation. I’m hoping we can get something like that going for bees. With help from AI, we can reimagine what conservation looks like, bridging the gap between scientific discovery and public action, and cultivating a more biodiverse future.