The tooth-billed pigeon, which is sometimes referred to as “the little dodo” of the Pacific, is a variety of fowl indigenous only on the islands of Samoa. For more than a decade, it was also thought to be lost, if not outright extinct, after its last confirmed sighting occurred in 2013. That changed in 2025, though, when the Colossal Foundation—the conservation arm of the Texas technology company—introduced a new device to Samoa.

Described as a bioacoustics array derived from sound censors and a 360-degree camera mounted on top, this “classifier” is an AI-powered machine that has only begun to be deployed over the past 13 months at sites and national parks around the world. And according to Matt James, the chief animal officer of Colossal Biosciences as well as the executive director of the Colossal Foundation, it pinged the tooth-billed pigeon 43 times on its first deployment in Samoa alone.

“We know it’s out there,” James smiles while sitting across a conference room in Dallas. “So the next step [was] to get teams out there, go get eyeballs on them. And they’ve now seen them for the first time a month ago. It’s the first time in 13 years the bird’s been seen by a human, so the next step is can you begin to grab these animals and put them back into human care, so we can create captive breeding and release programs?”

That last bit might prove especially prudent considering one of the reasons the bird is virtually extinct is because an invasive breed of feral cat has flooded the island. It’s hardly an ideal situation, but according to the Colossal officer, it’s a chance to “bring them to human care, remove the cats, begin to build the population, and put them back in the wild, all because an AI tool is able to find it where nobody else could find it.”

It’s also a sample of work that might be less flashy than the de-extinction project that’s made Colossal famous after they brought a version of the dire wolf back from prehistory last year, but it has profound and immediate implications for species already here, from this relatively large pigeon in the South Pacific to elephants in Africa, and the much debated American gray wolf inside the continental U.S.

When we catch up with James, it’s during the recent grand opening of the Colossal Biosciences headquarters and laboratories in Dallas. We previously toured the campus last year when Colossal dominated social media with images of baby dire wolves—and Colossal CEO Ben Lamm remains adamant that the goal of having a woolly mammoth calf walking the earth again by 2028 remains on track. But it’s clear the technology company is at a pivot point following the last 12 months. The closest thing to a real-life Jurassic Park is now open to school tours, the world’s first permanent BioVault dedicated to preserving endangered or extinct genetic genomes is being established in Dubai, and as James muses, “[If] ’25 was about growth… ’26 is about delivery. It’s time.”

The foundation’s  “bioacoustics classifier” and the AI system it implements seems like one such device that can deliver sooner. Already its systems are being implemented in Yellowstone National Park in Wyoming, where it is being used to monitor gray wolves, and in the Samburu National Reserve in Kenya, where it is intended to study and protect the African elephant.

“We’ve developed it and we just give it to our conservation partners for free,” James explains. “The hardware is a piece that we will buy for them and then deploy it.” In Yellowstone, they’ve already deployed 55 classifiers, which create a grid system monitoring five kilometers each. “They’re really far spaced out,” James continues, “you don’t need a ton of them to do that, and so for $55,000, you essentially can cover all of the world’s largest national park.”

First deployed at Yellowstone in January 2025, the devices work by remotely connecting with a cell network which uses bioacoustics that record the sound of an animal’s cry to build and analyze spectrogram data, deciphering what is, say, a gray wolf instead of a coyote. The goal is for Colossal AI engineers to match that with a machine-learning tool that is able to individually distinguish a specific elephant or gray wolf, or pigeon, from another with a single drone image. According to James, when the bioacoustics classifier was first tested, it had a 96 percent accuracy rate, and that score has only gone up since it’s been refined.

The applications for such tools also expand far beyond just finding hard-to-seek species. The plan is to enhance the study of living animals, as well as mitigate potentially dangerous encounters with humans.

“So we now have a tool that basically can look at drone imagery, track elephants, and say we know exactly who this elephant is,” James notes. “What it begins to create is ethogram data, so it’s telling you behaviorally what they are doing. Right now they’re muddying, they’re flapping their ears, they’re all directing their attention to one individual in the herd. So [in our] partnership with Save the Elephants in Samburu, that’s giving them a lot of data to work on management.”

He continues, “But if you think about one of the biggest issues with elephant conservation and wolf conservation, it’s conflict with humans. If you can begin to identify, basically, the troublemakers in each group, you can create early warning systems for cattlemen in Montana or for crop co-ops in Kenya and say, ‘Hey, just so you know, there’s a troublemaker in your area, you should be deploying your mitigation tools tonight and avoid some of the conflict that results in elephant or wolf death.’”

The technology has so far not actually been used to mitigate conflict between ranchers and gray wolves in the American West—one of the continued flashpoints which contributes to the ongoing debate about delisting the gray wolf as an endangered species in 44 states—yet the intent is there to both improve relations with those worried about wolves attacking livestock, as well as a general public that might still view the timber wolf as a nuisance or monstrous beast.

“I’m hoping it creates an opportunity to understand wolf language,” says James. “When we start talking about language, it’s a great way to anthropomorphize a species that’s severely persecuted, wrongfully persecuted. And if you can create some compassion for this species, hopefully you’re reducing some of the persecution. Cattlemen probably don’t really care for that part, but if we can say, ‘Hey, it’s a great conflict mitigation tool,’ they care about that. [Meanwhile] the mass public says, ‘Okay, well, maybe wolves aren’t as bad as we thought they were,’ and I think that’s the winning combination.”

The technology already exists, and in fact appears to be the tip of the spear in introducing AI tools into conservation efforts. After all, James muses that Abhishek Jana, a senior scientist working in the Colossal Foundation’s AI unit, retrofitted the bioacoustics tech from being a “bird classifier” to a wolf-based device in a single weekend. “We literally said on a Friday, ‘Hey, can you make a wolf classifier?’ And on Monday, I had an email where he said ‘try this.’”

It makes the mind wonder about the implications a decade from now.

The post Colossal Is Using New AI Tools That Might Just Save the Gray Wolf appeared first on Den of Geek.