According to a 2017 study, approximately 80+ endangered whales are killed by ship strikes on the US West Coast every year. Whale Safe, a collaborative project between the Benioff Ocean Science Laboratory and UCSB, are using AI to try and reduce these numbers:
Each buoy carries an onboard computer that records whale sounds using an underwater microphone. An AI algorithm then detects the clicks and squeals of specific whale species before transmitting the data via satellite every two hours, where a scientist will review the software’s results to check the whale sounds are being correctly identified. The location of the animals is then calculated by analyzing water conditions (via a separate machine learning model) and local records of whale sightings. If a ship is located within the calculated path of the animals, a notification can be issued to warn the crew.
Notably, the AI was trained on whale calls from Californian whales only so if their plans to expand are to be successful, they’ll need to add some more diverse calls or it won’t work. The joys of machine learning: get useful data, and more of it if you wanna scale!
(via The Verge)