Rising sea levels, extreme temperatures, and the growing human population are all contributing to the mass extinction of thousands of plants, insects, and animals, at alarming rates. As many as one million species face extinction in the next several decades, with biodiversity loss a major factor. Artificial intelligence (AI) could be the key to their (and our) survival.
Graham Taylor, a Canada CIFAR AI Chair at the Vector Institute, is using machine learning for biodiversity. He is collaborating on BIOSCAN, an $180 million global initiative that aims to revolutionize our understanding of the world’s biodiversity with the help of thousands of researchers from over 30 countries. BIOSCAN is led by Paul Hebert, a Canada Research Chair and Director of the Centre for Biodiversity Genomics at the University of Guelph.
Working with the Biodiversity Institute of Ontario at the University of Guelph, Taylor is building an inventory of life on Earth using DNA barcoding.
DNA barcoding is a groundbreaking technique used to identify the diversity of species in our ecosystem based on their DNA. It could potentially course-correct human-related land degradation and enable researchers to develop tools to protect ecosystems against loss. The process of collecting and identifying thousands of organisms is tedious and time-consuming for humans. However, machine learning algorithms could significantly reduce the time it takes in a task where every second counts.
Taylor is taking the project a step further by using computer vision to supplement the DNA information with visual analysis, a process that will improve the robustness of the organism detection and enable biodiversity researchers to examine organisms at the genome level.
“Right now we’re facing a massive loss of information. Consider the consequences of losing these genomes, each one is one thousand times more detailed than the longest book ever written,” says Taylor, an associate professor at the University of Guelph, a Canada Research Chair in Machine Learning, and academic director, NextAI. Taylor attributes that fact to Paul Hebert.
Driven by advances in deep learning, computer vision has seen wide success in consumer applications, now deployed in phones, cars and smart appliances and used by millions of people. It’s starting to see uptake in human and animal health applications. Taylor anticipates computer vision will play a larger role in improving the health of the planet.
Barcoding has revolutionized biodiversity efforts in the past ten years. Taylor is hoping to do the same with visual identification. It would give researchers the ability to identify tens of thousands of species within seconds.
“AI systems give us superhuman abilities in terms of species analysis for biodiversity. If we can build machines to accurately identify insects and other individuals, it will go a long way in advancing our knowledge of biodiversity loss and supporting our efforts towards conservation and protection,” he says.
Taylor is working on a number of biodiversity projects, including the LIFEPLAN, a global scale biodiversity monitoring effort which involves automating various formats of data, such as images, DNA samples, audio recordings of animal sounds, collections of fungal spores, and more.
The project was inspired by earlier research using computer vision to identify moths for agricultural pest monitoring. Collaborating with BIO’s John Fryxell, he’s also using computer vision for the re-identification of insects for ALUS, a project that helps farmers produce valuable ecological services on Canadian farmland.