When belugas communicate with each other, certain vocal cues reveal their social structures. When confronted by a predator, meerkats use a complicated set of alarm calls depending on how close and how dangerous the predator is. And in the sky, birds also communicate danger to other members of their flock.
Now, scientists are harnessing the power and rise of artificial intelligence (AI) to create a program capable of deciphering these forms of animal communication, and more, by analyzing huge data sets. The non-profit organization Earth Species Project (ESP) is working on creating this tool. A project with an “enormous potential for change,” he says.
Coding of animal communication
Biologists around the world have spent decades decoding animal communication. This is a time-consuming process that often involves recording sounds or observing other signals animals make to communicate, such as the “dance” honey bees perform to tell each other where to find nectar.
Some biologists record the sounds of animals for hours and combine them with observations of their behavior. For example, a lion may growl in warning, while a lizard may wave its hand or puff out its dewlap.
In some cases, these observations are supported by relatively new technologies, such as biologgers: devices that, once attached to animals, record everything from sound and acceleration to heart rate, among other things.
ESP went live in 2017, after advances in machine learning introduced new possibilities in decoding communication. The first article, published in Scientific Reports, described a tool that would isolate a voice from a recording of several people speaking.
It’s a tricky task, known to scientists as the “cocktail party effect.” From there, the organization began matching behavioral observations with communication signals by analyzing data stored in biologgers.
ChatGPT for animals
Their mission is not necessarily to unravel the “language” that a particular species uses to communicate, but rather to develop a tool that biologists can use to recognize the signals that animals emit in certain situations.
In part, the tool could work like a ChatGPT for animals, says Sara Keen, an acoustic engineer and neurobiologist at ESP. Basically, it is an algorithm capable of accurately predicting what communication signal will come next based on the behavior of an animal, or vice versa.
Although the process may take some time, Keen points to the tremendous advances in speech recognition software in recent years: this type of software was pretty bad 15 years ago, but now people rely on their Amazon Alexa for all sorts of things. Work on animal communication just has to catch up, he says.
Huge potential for change
While impressive in itself, creating a basic tool for understanding the communication of all animals also has huge implications for conservation and human-wildlife interactions. The ESP team is already working with whale researchers on a project that analyzes the sound of beluga whales in the St. Lawrence River (Canada) with important implications in this regard.
Similarly, a better understanding of non-human communication may reveal the ways in which certain matriarchal herds of elephants transmit information. Knowing that some elephants talk alike or differently from one another could be important in improving the success of reintroduction projects, they say.
Weigh the pros and cons
Poachers, for example, already use recordings of mating sounds to lure colorful songbirds like the yellow cardinal into traps. It is possible that a better understanding of animal communication helps bad agents to better exploit nature, hence the importance of mitigating negative consequences.
On the other hand, the potential benefits for conservation remain enormous: from predicting mass whale strandings to revealing the impact of offshore wind turbines, among others.
Edited by Erick Elola with information from Discover Magazine and the Earth Species Project.
