Understanding orangutan speech: AI breakthroughs reveal complex language

The dream of understanding animal speech is coming true—thanks to technological progress—before our very eyes. After years of experiments with monkeys using sign language or communication with dolphins, the time has come for a new generation of translators. These are algorithms that study the complex language of orangutans.

Orangutans are among the animals whose way of communicating is the most complex. Research conducted on them has led to questioning the notion of the uniqueness of human speech.

"Humans use lips, tongue, and jaw to articulate voiceless consonants, activating the vocal cords (folds) in the larynx with exhaled air for vowel articulation. Orangutans can also produce both types of sounds simultaneously," notes Dr. Adriano Lameira from the University of Warwick.

The discovery of anatomical mechanisms responsible for the communicative abilities of some monkeys—especially orangutans—has also led to a revision of views on the origin of human speech.

The beginnings of communication through controlled, conscious sound production may be as old as the common ancestors of humans and great apes. This is indicated, among other things, by research on orangutans, which communicate in an unusual way—they can, which is rare in the animal world, produce two sounds at once.

"Humans rarely produce voiced and voiceless sounds simultaneously. An exception might be beatboxing. The mere fact that humans anatomically are capable of beatboxing raises questions about where this ability came from in us. Now we know that the answer may lie in the evolution of our ancestors," notes Dr. Madeleine Hardus of the Pongo Foundation.

While the ability to produce certain sounds has a genetic basis, researchers believe that the development of language itself was primarily influenced by social conditions. This is indicated by studies confirming that it is the community that shapes the vocal abilities of orangutans.

But what exactly do Bornean monkeys talk about? Although there have been successful attempts in the past to understand and even have conversations with different mammals, a breakthrough is happening before our eyes thanks to artificial intelligence.

New research published in the journal PeerJ Life & Environment sheds light on the complexity of Bornean orangutan vocalizations.

"Our research aimed to unravel the complexity of long orangutan conversations, which play a crucial role in their long-distance communication in the dense rainforests of Indonesia. Over three years, we collected hundreds of long conversation recordings, revealing a fascinating array of vocal diversity," says Dr. Wendy Erb from Cornell University in the USA.

To decipher orangutan conversations, researchers used traditional sound analysis methods along with machine learning tools such as Uniform Manifold Approximation and Projection (UMAP - Manifold Embedding Space Clustering).

After recording over 3,800 hours of recordings, they examined 117 long calls from 13 orangutans separated from the rest of the group. Based on this, they identified as many as 1,033 distinct sounds, whose sounds can be distinguished by both algorithm and human ear.

Although the content of the messages is not yet known, scientists note that the orangutan language appears to exhibit much greater "potential complexity" than previously thought. Moreover, another study indicates that some monkeys deliberately use distorted sounds, which researchers likened to slang.

This confirms the richness of sounds conveyed during a single call, during which monkeys—researchers believe—are able to convey not only simple information, e.g., about food or threats, but also complex, detailed messages.