In Yellowstone National Park, a network of recorders records day and night. This device captures the howling of thousands of wolves as they drift through the valley.
That audio is now being fed into an artificial intelligence model built to do something no one has ever done before. The goal is to identify individual wolves by the unique characteristics of their voices.
This research is by the Colossal Foundation, the nonprofit arm of Colossal Biosciences.
The Dallas company is best known for its dire wolf and other extermination projects. The organization’s conservation team believes that emerging technologies can solve problems that cannot be solved using traditional methods.
Listen to keystone species
The Yellowstone effort is a partnership between Colossal Foundation, Yellowstone Wolf Project, Yellowstone Forever, and Grizzly Systems.
The team deployed 48 autonomous recording devices throughout the park.
The four wolves are now wearing audio-recording collars that combine sound with movement and location data. The researchers examined over 7,000 unique howling events and used them to train the model.
Custom-built software listens to a constant audio stream. They can pick out individual howls or choruses in a herd, and even announce the sound of gunshots.
The system clusters calls by acoustic fingerprints and extracts details such as group size and the presence of puppies.
The classifier was now able to distinguish between individual howls and choral howls with over 92% accuracy.
A method that started with birds
This approach was born out of birdsong. Birds are a natural starting point because generations of birdwatchers have trained their ears to identify species by their calls.
Colossal is the first to introduce bioacoustics to a rare bird called the long-billed pigeon, also known as the manumea. It is one of the closest living relatives of the dodo.
The model achieved 95% accuracy in distinguishing that single bird from all other species in the ecosystem, the researchers said. That training was carried over to the wolves.
The same machine learning approach that learned one rare bird adapted to the gray wolf.
The next frontier is the red wolf, widely considered to be the most endangered wolf in the world.
From gray wolf to red wolf
Colossal has conducted extensive research on red wolves, and bioacoustics opens new avenues for red wolf research. The research team wants to see if AI can tell the difference between red and gray wolves based on sound alone.
Additionally, researchers hope to determine varying degrees of ancestry within individual red wolves and even within a single animal. This will transform what was once a genetics task into something far less invasive.
Researchers may one day identify red wolves from their howls rather than blood samples.
Bioacoustic tools may eventually have applications beyond research to conflict prevention.
Gray wolves often come into conflict with humans over livestock. Tools that detect the presence of wolves near a herd could help ranchers proactively avoid conflicts.
Beyond the barking animals
This method is not limited to howling animals. Whales and dolphins, with their underwater songs and echolocation, were early targets.
In principle, almost any vocalizing species can be studied in this way.
Vocalizing animals rarely produce simply small datasets, making the task difficult, but not impossible.
Researchers even suggest that elephants’ infrasound calls may be detectable through the ground.
build the next generation
Technology is only as strong as the people who build it. This year, Colossal partnered with the nonprofit organization Conservation Nation to launch its first conservation fellowship.
Conservation Nation works to bring youth from underrepresented communities into fields where they have traditionally been underrepresented.
The program began with two roles. One is the AI for Conservation Fellowship. The other is the Research and Strategy Fellowship.
The research and strategy positions attracted applicants from major universities. The role of AI has also reached international partners, including the Mauritius Institute of Biotechnology.
The team says they received more than 100 applications for the two roles combined. This number is part of a much larger group of more than 1,400 applicants for Colossal’s internships overall this year.
What distinguishes friends
Fellows were selected for more than just their technical skills. The team looked for someone who shared the company’s desire to try something no one had done before.
Candidates who already knew the project stood out. They came eager to talk about howling wolves and discovering rare species.
The research strategy fellow, a doctoral student in conservation genomics, spent a week in Dallas. The fellow then settled into four months of remote work with the bioinformatics team.
Recruited through a partnership with Mauritius, the AI fellow will help launch a new red wolf bioacoustics project and contribute to research on long-billed pigeons.
Although much of the work will be done remotely, the team will bring researchers to Dallas for lab immersion and connect them with mentors.
Different models for conservation
The team says the group has something in common. They want to help envision a future that no one can yet envision.
Many of them are “AI natives,” in the words of conservation staff, bringing new tools and faster ways of working to established teams. That energy fits into the company’s larger philosophy.
Colossal frames itself not as an organization in the field, but as an accelerating force. Identify local partnerships and provide technology and funding to advance work already underway.
This approach is designed to build trust, as this work will be done through local stakeholders and grassroots groups in each region.
long term bet
Matt James, Colossal’s chief animal officer, explains that this interaction is a long-term gamble, not a short-term cost.
He argues that these programs are investments rather than expenditures, another bucket of funds intended to pay dividends in the future.
James likens the effort to the Skunk Works conservation program. This field tends to be considered small, partly because it is chronically underfunded.
The hope is to inject imagination and optimism and encourage the ambitious thinking that the biodiversity crisis demands.
“While we believe wholeheartedly in traditional conservation, we also acknowledge that it cannot keep up with the pace of destruction we are seeing today,” James said.
“We cannot continue to deploy existing measures and expect new results.”
—–
Like what you read? Subscribe to our newsletter for fascinating articles, exclusive content and the latest updates.
Check us out on EarthSnap, the free app from Eric Ralls and Earth.com.
—–
