Will AI enable humans to talk to animals? – 01/09/2024 – Science

Nature is noisy. Whether in the forest, mountain or field, the sounds of animals are constant. Even in the concrete city of Berlin, you can often hear the cries of flocks of sparrows in the bushes, or the songs of nightingales that visit the German capital annually in May.

It is already known that cows moo with regional accents; monkeys make sounds specific to the threats they face; mice sing; and crickets scream for sex.

Recent research also shows that the “clicks” heard in sperm whale vocalizations are like vowels in human speech. Everything indicates that these whales communicate with vowel patterns a and i.

Increasingly, scientists have used artificial intelligence (AI) tools to study animal vocalizations, especially complex communication systems.

Tools for analyzing animal sounds

In recent years, the world has experienced a boom in understanding animal communication, in part due to AI. It allows researchers to analyze large volumes of audio data from animals in seconds — something that would take decades to be done by humans.

There are hundreds of AI tools for analyzing vocalizations from different species.

Kevin Coffey, a neuroscientist at the University of Washington, United States, helped develop DeepSqueak, an automatic learning tool that decodes rodent conversations. Machine learning is a branch of AI based on the idea that systems and computers can learn from data.

DeepSqueak identifies rodent “voices” from raw audio data, compares them to vocalizations of similar characteristics, and provides insights into the animals’ behavior.

“Rats use ultrasonic vocalizations. High-pitched sounds of 50 kilohertz (kHz) have been described as similar to laughter, but there are many types of these sounds made in different positive situations, such as playing, courting, or even during the ‘trip’ of a dose of drugs,” says Coffey.

Rats also emit 22 kHz sounds that are used in negative situations, such as when they are in pain or sick. Coffey uses these frequencies to know when an experiment is harmful to his laboratory rats.

Humans cannot hear these sounds as they are outside the frequency range that the human ear is capable of hearing, but DeepSqueak and other tools can help decode them.

Since its launch in 2018, DeepSqueak has been used to study rodent social behavior, drug use, autism, and more. The tool has also been modified for use on many other species, such as dolphins, monkeys and birds.

Coffey explains that scientists used to analyze spectrograms of ultrasonic vocalizations manually, but now AI tools allow them to automate the process, which saves time. But it is still up to humans to define the meaning of vocalizations.

“AI and deep learning tools are not magic. They are not going to suddenly translate every animal sound into English. The hard work is being done by biologists who need to observe animals in a multitude of situations and associate the sounds with behaviors, emotions, etc.”, explains the neuroscientist.

Do animals have language?

Communication is essentially done by transmitting information. All animals communicate in some way – through smells, pheromones, behaviors and vocalizations.

But the idea that animals have their own language is contested, in part because of how humans view language and its function in society.

Yes, animals have forms of language, as far as anyone knows. But the dialogues of sperm whales or the symbolic communications of monkeys do not come close to the richness of human language. Language is a particularly advanced toolkit for communication that appears to be unique to humans, according to Coffey.

Anthropologists, in turn, claim that human language is unique due to its ability to create and maintain beliefs, relationships and cultural identities. Human language also allows us to express intimate thoughts and feelings so that other individuals can understand them. It is believed that other animals cannot do this with their vocalizations and actions.

Some theories even suggest that human consciousness, the highest metaphorical version of consciousness, developed alongside the capacity for language.

“There is discussion about what exactly defines language, and whether some elements of animal communication are similar to language [humana]”, reports Coffey. “The rodents we studied are certainly very social and very communicative. The vocalizations they produce are highly varied and carry different types of information, but I still wouldn’t consider them a language.”

Beware of anthropomorphization

It is often attributed to the Austrian philosopher Ludwig Wittgenstein that his words had meaning, but the words of a parrot did not. The idea is that when humans say something, they don’t just pronounce the sounds, they also mean something.

Leaving aside the debate about whether Wittgenstein was right or wrong, some researchers claim that it is precisely this idea of ​​human meaning that leads scientists and non-scientists to project meanings into animal sounds.

For example, when scientists in Berlin discovered that rats “laugh” when tickled, observers declared that rats have a sense of humor.

Scientists responded, arguing that interpreting rats’ high-frequency chirps as humor, or even laughter, would be a form of anthropomorphization, that is, the tendency to “humanize” animals or even objects, such as robots. How do you know if the rats found tickling “funny”? There’s no way to know, say scientists. You can’t read a mouse’s mind.

On the other hand, some experts argue that humans often underestimate the capabilities of animals. As a result, we see more and more studies claiming “unprecedented” discoveries, such as fish that feel pain, intelligent octopuses or rats that laugh.

Many parrot owners will tell you that Wittgenstein was wrong: that their animal’s sounds could mean it is in pain or hungry, for example. Science has evolved since the Austrian philosopher’s writings in the early 20th century. But for Coffey, a certain degree of anthropomorphization is not all bad, as it helps humans connect with the animal world.

“Mice make structured songs with many syllables. They are much more complicated when males are calling females than when males are talking to themselves. I’ve always thought that was beautiful. But it’s good that academics remain cautious about communication and language. We will probably never tell the mice and dolphins about esports or inflation,” says the neuroscientist.

Why listen to animals?

Understanding animal communication goes far beyond human curiosity. Coffey believes AI can be a benefit to animals themselves: “Personally, I would like to improve the lives of laboratory animals and increase the rate of translatable discoveries in neuroscience. Understanding rodent communication in the laboratory is one piece of that puzzle.” head.”

Other scientists and organizations use artificial intelligence, for example, to monitor the biodiversity of animals in nature. Researchers at the University of Würzburg in Germany used microphones to record soundscapes of tropical forests and analyzed the cacophony of animal vocalizations, including insects and birds, to track the restoration of biodiversity.

The Earth Species Project (ESP) is also using AI to monitor biodiversity. The project espouses a bold belief, described on its website, that “an understanding of non-human languages ​​will transform our relationship with the rest of nature.”

The group’s ambition is to employ AI to deepen understanding of animal communication, helping humanity connect with and protect other living species.