In the Panamanian rainforest, a rare moth lands atop an automated camera system, attracted by its solar-powered UV light. Click! The motion sensors activate, and the device snaps a photo of the insect, adding it to a growing catalogue of records.
That’s when David Rolnick’s machine-learning algorithms get to work, analyzing the photos to identify known species.
As climate change pushes more species towards extinction, it’s increasingly important to keep tabs on insects, says the McGill University computer scientist, who’s also a member of the Mila – Quebec Artificial Intelligence Institute.
“Half of all species are insects, so if you care about biodiversity, you care about insects,” Rolnick points out. That includes moths, which pollinate many food crops and feed other animals.
It’s why he worked with entomologists to develop the AI-enabled “camera trap” — a tool to help resource-strapped experts monitor species on a wide scale, quickly analyze changes in their distribution, and zero in on which populations most need protection.
“These algorithms can effectively identify thousands of species at a given location,” says Rolnick. “They don’t get tired. They can identify millions of photographs and pick out the ones which are most interesting.”
Now, Rolnick is building more algorithms that can flag potential new species and suggest ways to categorize them. But tracking insects is only one of the projects keeping him and his global team of 30 AI researchers busy.
Using machine learning to track crop growth in real time
Rolnick’s team is also working with partners at NASA Harvest and the European Space Agency’s WorldCereal project to understand how shifting weather patterns are impacting agricultural yields.
“Some of the biggest impacts of climate change are faced by the world’s farmers,” he explains. Rolnick and his team are developing AI tools to track crops in real time by analyzing satellite imagery, addressing numerous technical challenges along the way to make the software fast, accurate and applicable to all kinds of global crops.
Other projects include building better climate change models, optimizing electrical grids and identifying chemical catalysts that can help produce fuels such as hydrogen from electricity.
Creating practical, scalable AI solutions
According to Rolnick, satisfaction comes from using computer science innovations to tackle meaningful societal issues. For him, it starts with finding problems where existing algorithms can’t do the job and then working with end users to develop new ones.
“It’s not about coming up with something and giving it to somebody and saying, ‘here you go,’” he explains. “It’s about co-designing it with them from the ground up.”
His goal is to create practical tools that can be easily deployed anywhere and used on a massive scale. That’s where he expects CFI-funded equipment to make a big difference, providing a platform to test and run different solutions.
“It doesn’t stop at the research,” Rolnick says. “It is also a question of how to build tools that people will actually use.”
The CFI makes it possible for us to build these algorithms at scale and have them be relevant to users across the world. It would not be possible to be operating at the scale we do without this support.
— David Rolnick, McGill University
The research project featured in this story also benefits from funding from CIFAR, Mitacs and the Natural Sciences and Engineering Research Council of Canada.
