A Scientific Review of Zoology's Holy Grail
Talking to animals has captivated human imagination for millennia. From King Solomon’s legendary ability to understand birds to the fictional Doctor Dolittle, this fantasy reflects our deep curiosity about the countless intelligent species with whom we share our planet. Despite forming bonds with animals, true conversation remains beyond reach — a communication gap that represents one of zoology’s final frontiers.
The Wild Robot
Watching “The Wild Robot” recently sparked my imagination. The film follows Roz, a service robot shipwrecked on an uninhabited island who must adapt to her surroundings, build relationships with the local wildlife, and become the adoptive mother of an orphaned goose, Brightbill.
One of the first things that Roz learns to do is talk to animals (who all understand each other, but let’s put that aside for now), adapting to a hostile environment when nobody is there to give her commands. The movie is lovely and I highly recommend seeing it, but it also made me think: how far are we from that? Is it even slightly realistic?
Understanding Without Words
What’s often overlooked is that humans already communicate with animals daily, even without complete linguistic understanding. Pet owners recognize their dog’s different barks or their cat’s various meows, understanding when they signal hunger, excitement, or distress. Working animals like herding dogs or service animals respond to complex commands and provide feedback through behavior. Farmers develop nuanced understandings of their livestock’s vocalizations and body language.
Dog getting basic command
These interactions represent a form of mutual understanding developed through coevolution and shared lives. We may not grasp the full complexity of animal communication systems, but meaningful exchanges occur continuously through observation, repetition, and emotional connection. The question isn’t whether we can communicate with animals, we already do, but whether we can reveal new patterns and meanings in animal vocalizations and respond in their own language.
Current State of Animal Communication
Scientists have made remarkable progress in communicating with animals, though not quite at the level depicted in fiction. Perhaps the most famous examples involve great apes learning sign language. Koko the gorilla reportedly mastered over 1,000 signs and understood approximately 2,000 words of spoken English. One of Koko’s most famous moments was when she expressed sadness over the death of her pet kitten, signing ‘sad’ repeatedly. It’s moments like these that make us wonder — just how much do animals understand?
Koko the gorilla
Beyond primates, research has revealed sophisticated communication in numerous species. Marine biologists use specialized microphones to record and analyze dolphin vocalizations, discovering what appears to be a complex system of clicks, whistles, and body language. Projects like CHAT (Cetacean Hearing and Telemetry) attempt to decode dolphin communication patterns and potentially enable two-way exchanges.
The Quest for an AnimalRosetta Stone
AI systems have revolutionized human language translation, decoding everything from everyday conversations to ancient texts hidden for centuries. This same pattern-recognition capability could finally unlock the mystery of animal communication.
Just as the Rosetta Stone, an ancient tablet inscribed with the same text in three different scripts, cracked the code of Egyptian hieroglyphics, AI might become our key to understanding animal languages. The stone worked by connecting known languages to unknown ones, creating a translation bridge. Similarly, AI could link animal sounds with their behaviors, environments, and physical responses, revealing patterns and meanings that have always existed just beyond human perception.
The Rosetta Stone
As you can imagine, many scientists are interested in and actively researching this topic. I will try to give an overview on the prominent names in the field and what they are doing.
The Earth Species Project (ESP): Foundation Models for Bioacoustics
The Earth Species Project (ESP) is a non-profit organization that stands at the forefront of AI-powered animal communication research with their groundbreaking NatureLM-audio, the first and most capable large audio language model tailored specifically for animals.
Overview of NatureLM-audio
Unlike traditional approaches, ESP focuses on developing foundation models for bioacoustics that can work across species. Their research spans diverse creatures including zebra finches, Hawaiian and carrion crows, beluga whales, and elephants.
Aza Raskin, son of Macintosh project initiator Jef Raskin, is a prominent figure at the intersection of technology and ethics. Co-founder of both the Center for Humane Technology and the Earth Species Project. Raskin’s work, including his contributions to the NatureLM-audio project at ESP, is deeply rooted in his concern for the impact of technology on society and an advocate for the ethical use of technology.
Aza Raskin
Project CETI: Conversations with Sperm Whales
Project CETI (Cetacean Translation Initiative) focuses on deciphering the complex communication of sperm whales — creatures with the largest brains on Earth. The project analyzes the intricate click patterns called “codas” that these whales use to communicate in their social groups
If you have never heard that, check out the video about — Sperm Whale Clicks Underwater.
CETI takes a high-tech approach, using advanced machine learning and non-invasive robotics. Their AVATARS technology allows researchers to capture whale vocalizations without disturbing their natural behavior.
They also gather contextual information like depth and temperature, enabling a more complete understanding of whale communication.
Their dataset could be found in the avatars-data repository.
A customized off-the-shelf drone flying to deploy a whale tag developed by Project CETI
Working with experienced whale biologists like Dr. Shane Gero, who has studied Caribbean sperm whale families for over two decades, the project emphasizes “listening to whales in their own setting, on their own terms.”
Dr. Shane Gero, doing his thing
Recent Project CETI publications demonstrate that sperm whale codas exhibit complex contextual and combinatorial structure, suggesting a sophisticated communication system. They’ve also advanced autonomous tracking technology, improving data collection, and have discovered that sperm whales use a phonetic alphabet, including vowels and diphthongs. These findings significantly deepen our understanding of sperm whale communication complexity.
Diverse Approaches Across the Animal Kingdom
Beyond the major initiatives, numerous specialized projects are expanding our understanding of animal communication through AI.
Zoolingua develops technology to interpret the facial expressions and sounds of dogs, focusing on our closest domestic companions.
DeepSqueak, originally created to analyze rodent vocalizations, has successfully expanded to study lemurs and even whales.
The Dolphin Communication Project categorizes dolphin calls to understand their vocal interactions.
Interspecies.io explicitly aims to translate communication between different species into human-comprehensible signals.
AI and Animal Communication in Israel
The quest to understand animal communication is a global endeavor, and Israel is making notable contributions to this exciting field. A significant highlight is the Coller-Dolittle Prize for Interspecies Two-Way Communication, a substantial award launched in May 2024 by the Jeremy Coller Foundation and Tel Aviv University.
This prestigious prize, offering a grand prize of $10 million, underscores the recognition within Israel’s academic community of the immense potential for AI to unlock the mysteries of animal communication and the importance of fostering such groundbreaking research.
Beyond this prominent prize, Israeli researchers are actively at the forefront of exploring animal communication through the lens of AI. Professor Yossi Yovel, a distinguished neuroecologist at Tel Aviv University, leads fascinating research into the intricate communication of fruit bats. His team at Tel Aviv University has meticulously built a large database encompassing a wealth of bat noises and corresponding video footage.
By employing sophisticated machine learning algorithms, They have trained models to recognize and categorize the subtle nuances of different bat sounds. This innovative approach has enabled the algorithm to effectively match specific acoustic signals with distinct social interactions captured on film, providing unprecedented insights into the social lives of these nocturnal creatures. You can delve deeper into their captivating work by watching this video produced by the BBC.
Prof. Yossi Yovel, doing his thing
Professor Yovel’s work, while celebrating the remarkable progress enabled by AI, also offers a valuable perspective on the inherent challenges in fully bridging the communication gap between humans and animals. He points out the fundamental differences in sensory perception and how animals navigate the world, suggesting that there might be limitations to our comprehension, even with the aid of advanced AI.
The Cocktail Party Problem
Imagine being at a loud party, trying to focus on one friend’s voice while dozens of others are talking around you. That’s the same challenge scientists face when decoding animal communication in the wild.
This phenomenon, called “The Cocktail Party Problem” first described by Colin Cherry in 1953, has been extensively studied in psychology, neuroscience, and machine learning.
Animals communicate in environments filled with wind, water, and other background noise, making it hard to attribute vocalizations to individuals or behaviors. While solutions for separating human voices have seen great advancements, the problem remains a major challenge in bioacoustics.
Cocktail Party
In 2022, researchers at ESP introduced BioCPPNet , a deep learning model designed to separate overlapping non-human vocalizations, with the ambition to “Solving the cocktail-party problem”. This model was an amazing contribution to the world of bioacoustics, but there are still limitations, particularly in the availability of large, labeled datasets, which are crucial for improving model accuracy and generalization.
Ethics and Future Implications
Even if AI enables us to interpret animal vocalizations, ethical concerns arise. Would it be ethical to manipulate animal behavior by mimicking their language? Should humans intervene in natural animal interactions? Understanding communication doesn’t equate to forming true dialogues, and misinterpretation could have unintended consequences. Moreover, many species communicate in ways fundamentally different from human speech, raising philosophical questions about what it truly means to “talk” to animals.
Despite these challenges, AI-driven animal communication research is rapidly evolving. From translating dolphin whistles to analyzing bat social calls, these projects push the boundaries of both technology and our understanding of nature.
As AI continues to bridge the gap between humans and animals, one question remains: If we could truly understand them, would we be ready to listen?
Originally published on AI Superhero
Top comments (0)