Is it possible to save the whales with AI? • The registry

Bright yellow buoys running AI software have been deployed to discourage cargo ships from running over nearby whales.

Collisions with cargo ships are the number one killer of whales, said Douglas McCauley, a professor of ocean science at the University of California, Santa Barbara The registry in this week. California is home to some of the busiest US ports, and whale collisions (in which the mammal invariably fares worse) are on the rise as shipping traffic increases. McCauley estimates that up to 80 endangered whales are killed along the US West Coast each year.

McCauley leads the Benioff Ocean Initiative, a philanthropic initiative funded by billionaire Salesforce founder and co-CEO Marc Benioff and his wife Lynne. The group is working with the nonprofit Marine Mammal Center to expand Whale Safe, an AI-powered system designed to warn cargo ships of incoming whales. Slowing down ships reduces the risk of deadly collisions with the animals.

Each buoy carries a computer and uses an underwater microphone to listen for whale sounds. The audio is fed into an AI algorithm running on board that can detect the clicking and screeching of certain whale species. When the software identifies these sounds, they are relayed to the base for scientists to review and record in a log. The data used to train the system was collected by Ana Sirovic, a professor at the Norwegian University of Science and Technology who has collected decades of audio recordings of whales.

“Different populations of whales have different dialects,” Professor McCauley told us. “For this AI to work in California, the AI ​​had to be specially trained on the calls of these California whales. The buoy recognizes blue, fin and humpback whales – because these are the three endangered whales in our region. These signals are then transmitted via satellite every two hours, and a scientist will check the software’s results once a day to ensure it is correctly identifying whale sounds.

Separate machine learning models analyzing the water conditions and records of whale sightings are used along with the audio detection logs to estimate the locations of these animals. A warning will be sent to cargo ships urging them to slow down if they are navigating in an area predicted to be a whale hotspot based on all this information. The AI ​​acoustic model is able to detect blue and fin whales at a distance of up to 25 kilometers; However, humpback whales are calmer and can be found within five kilometers of the buoy.

Each Whale Safe buoy costs about $250,000 to build and $200,000 to maintain. The first was used in a pilot study two years ago to evaluate technology on the Santa Barbara Channel near the ports of Los Angeles and Long Beach. Now the Benioff Ocean Initiative and the Marine Mammal Center have installed a second near San Francisco.

“Whales have lived on our planet for 50 million years and are beautiful, majestic and powerful pieces of the ocean,” said Professor McCauley. “We want to make sure that when the oceans get busier with more human trade, they’ll stay with all these things for 50 million years. They deserve a place on our planet.”

The large marine mammals are also critical to maintaining a healthy ocean and have a major impact on producing oxygen, recycling nutrients and supporting food chains.

“Whale Safe is dedicated to saving the incredible mammals that have ruled the oceans for millions of years,” said Benioff said in an opinion.

“This is a triple win for the planet – we’re saving whales, fighting climate change and promoting community health by reducing air pollution. We need more solutions like these emerging from alliances between science and business.”

The team hopes to deploy more AI-powered whale rescue buoys in other coastal areas of North America, such as Seattle, Vancouver and San Diego, and believes the technology could one day be used in other wildlife hotspots around the world, such as Sri Lanka. ®