Tracking endangered birds used to mean sending teams of biologists to scramble around remote islands and craggy cliffs to peer down burrows with special optical fiber optic scopes.
Not only was that difficult and dangerous, it was inefficient. Even the best supplied expedition can only spend a week or two in the field.
Not any more.
Now, thanks to deep learning, biologists can alight on remote islands with just a dozen sensors, leave, and return months later to scoop up vast quantities of data.
The data — far more than any one human can analyze — is then crunched by deep learning systems developed by Conservation Metrics, based in Santa Cruz, Calif., and trained with NVIDIA GPUs.
The results are uncanny. Biologists can now use sensitive microphones track how often endangered birds run into obstacles over many months. They’ve even used the technology to find — and protect — birds that were once thought extinct.
“Deep learning has allowed us to move from projects that include five sensors deployed for a month to 150 sensors deployed for six to eight months,” Matthew McKown, CEO of Conservation Metrics, said in a conversation with tech journalist Michael Copeland in the latest episode of our AI Podcast. “That allows us to track really rare events, and get a better conservation outcome.”
To hear the full story — and the tale of how deep learning helped rediscover, and protect, a rare species of kestrel on a remote island in the Pacific, tune into the latest edition of the AI Podcast.
AI Podcast: How AI Can Improve Brain Tumor Treatment
And if you missed our podcast last week, it’s worth a listen: we spoke with Dr. Bradley Erickson, a Mayo Clinic neuroradiologist, who uses AI to predict tumor genomics using MRIs. His method could give doctors easier access to invaluable genetic information. Information that could predict how fast a tumor will progress, and if it will respond to specific drugs and other treatments.