In 2025, we will witness the deployment of artificial intelligence and machine learning technologies to make real progress in understanding how animals communicate with one another, in an attempt to answer a question that has puzzled humans since the dawn of time: “What are animals talking about?” The Coller-Dolittle Prize—which offers cash rewards of up to half a million dollars to scientists who succeed in “decoding”—is a clear indicator of growing confidence that recent developments in machine learning and large language models (LLMs) are putting this achievement within our reach.
Many research teams have spent years working to create algorithms that decode creature sounds. For example, Project CETI seeks to analyze the “clicks” of sperm whales and the songs of humpback whales. However, modern machine learning tools require massive amounts of data, and high-quality, large-scale data of this kind remains scarce to this day.
Data shortage versus technological abundance
To understand the challenge, it is enough to consider the difference between what language models like ChatGPT have access to and animal communication data. ChatGPT is trained on the vast amount of text available online, while the number of reliable audio models available for field studies on sperm whales, for example, is no more than 8,000 calls—a fraction of what Project CETI is analyzing.
In addition, in the case of human language, we already know what is being said and we know the concept of a “word” itself, which facilitates the analysis process. In the animal world, scientists often do not know whether a specific wolf howl means something different from another wolf’s howl, or whether those howls are equivalent to “words” in the human sense.
Larger data and more advanced software
Despite the challenges, 2025 holds great promise, both in terms of the growing volume of animal audio data and the technical capabilities of artificial intelligence and animal algorithms. Almost every research team can now use low-cost recording devices such as AudioMoth, enabling the collection of massive amounts of animal data around the clock.
Large groups of these recordings have already begun to emerge, covering the sounds of gibbons in jungles or forest birds throughout the year. Dealing with these huge amounts of data was once nearly impossible, but computer vision algorithms and convolutional neural networks (CNNs) are now able to examine thousands of hours at lightning speed, extracting animal sounds to classify them according to their natural acoustic characteristics.
Once this vast data becomes available, new opportunities arise to analyze it, such as using deep neural networks to discover the underlying structure in animal tone sequences and calls—which may be somewhat similar to the semantic structures of human language.
What do we do with these sounds?
The core dilemma remains determining the ultimate goal of understanding these sounds. Some entities, such as the Interspecies.io organization, seek to “translate animal signals into interspecies-understood signals,” meaning translating animal communication for humans. However, most scientists hesitate to say that animals have a “language” like ours, even if they possess meaningful sounds.
The Coller-Dolittle Prize comes in with a more flexible approach, aiming to find a mechanism that allows us to “communicate with or decode the communication of living creatures.” We may not be able to translate this communication literally if animals do not have a language in the human sense, but we can still understand the basic meanings and intentions behind those sounds.
So far, we do not know the exact amount of information animals exchange among themselves, or the nature of the “vocabulary” they use. But in 2025, thanks to artificial intelligence and animals combined, we will have a golden opportunity to make a qualitative leap in our understanding of the extent and true nature of their communication complexity. This year may finally see the emergence of answers based on massive data and advanced analysis, answering a question that has long occupied our minds: “What are living creatures saying to each other?”
