The race to translate animal sounds into human language

In 2025, we will see AI and machine learning being used to make real progress in understanding how animals communicate and answer a question that has always puzzled humans: “What do animals say to each other?” The most recent one Coller-Dolittle PrizeOffering cash prizes of up to half a million dollars to scientists who “crack the code” is a sign of optimistic confidence that recent technological developments in machine learning and large language models (LLMs) are bringing this goal within reach.

Many research groups have been working on algorithms to understand animal sounds for years. Project Ceti, for example, has decoded this Click on the movements of sperm whales and the songs of humpback whales. These modern machine learning tools require extremely large amounts of data, and until now such amounts of high-quality and well-annotated data have been lacking.

Consider LLMs like ChatGPT, which have training data that includes all text available on the Internet. Such information on animal communication was not accessible in the past. It’s not just that human data corpora are many orders of magnitude larger than the kind of data we can access from animals in the wild: more than 500 GB of words were used to train GPT-3, compared to just more than 8,000 “codas”. ” (or vocalizations) for Project Ceti’s recent analysis of sperm whale communication.

In addition, we already have this when working with human language knowledge what is said. We even know what constitutes a “word,” which is a huge advantage over interpreting animal communication, where scientists rarely know whether, for example, the howl of a particular wolf means something different than the howl of another wolf, or whether the wolves are a Consider howling as something somehow analogous to a “word” in human language.

Nonetheless, 2025 will bring new advances, both in the amount of animal communication data available to scientists and in the type and power of AI algorithms that can be applied to that data. Automated recording of animal sounds has become easily accessible to any scientific research group, and low-cost recording devices such as AudioMoth are becoming increasingly popular.

Extensive data sets are now being put online because the recorders can be left in the field and listen to the calls of gibbons in the jungle or birds in the forest around the clock for long periods of time. There have been cases where such large datasets could not be managed manually. Now new automatic recognition algorithms based on convolutional neural networks can go through thousands of hours of recordings, pick out animal sounds and group them into different types according to their natural acoustic properties.

Once these large animal datasets become available, new analysis algorithms will become possible, such as using deep neural networks to find hidden structures in sequences of animal vocalizations that may resemble the meaningful structure in human language.

However, the fundamental question remains unclear: what exactly do we want to do with these animal sounds? Some organizations, such as Interspecies.io, have a clear goal of “transforming signals from one species into coherent signals for another.” In other words: too translate Animal communication in human language. However, most scientists agree that non-human animals do not have their own language – at least not in the way that humans do.

The Coller-Dolittle Prize is a bit more ambitious, looking for a way to “communicate with an organism or decipher its communication.” Deciphering is a slightly less ambitious goal than translating, considering that animals may not actually have a language that can be translated. Today we don’t know how much or how little information animals transmit to each other. By 2025, humanity will have the potential to surpass our understanding of not only how much animals say, but also what exactly they say to each other.