Doctor Dolittle reaches for artificial intelligence. Will we understand the language of animals?

There are already “major joint initiatives, such as: Vocal Interaction among Humans, Animals, and Robots; Inter-species Internet; and finally, an initiative for translating whales, which recently provided a detailed research plan using AI to study sperm whale communication.” A research group is working on studying the vocal repertoire of the extinct-in-the-wild Hawaiian crow. Biologists not only dream of restoring this endangered species to its former glory but also – quite literally – “renewing its culture.”

Communication between the animal and human worlds is the foundation of many fairy tales and fables. In those stories, animals speak with a human voice (which also happens in the Polish Christmas Eve traditions, doesn’t it?), and humans, especially the chosen ones, understand the language of animals. How would an everyman have reached the glass mountain, killed the dragon, and rescued the princess without the help of the wolf and the bear, whom he had saved before? How would a protagonist have become a prince if he didn’t understand his cat, who needed boots and a hat for a decent life? Examples like these abound.

These tales are driven by the hidden human desire to tap into the wisdom of animals and to be capable of manipulating them. Nothing facilitates manipulation as much as communication, where one partner is smarter than the other. And we consider ourselves exceptionally intelligent animals – after all, we can speak, and they cannot. The marvel of man.

By the way, real communication happens when both sides understand each other. Are we also looking for ways for animals to understand us without it being mere taiming? Of course, having senses often sharper than ours, they read our signals, of which we are often unaware, such as scent or body temperature, influenced, for example, by the level of human stress. This illustrates the complexity of the issue that science is grappling with.

Four legs good, two legs bad

Since the first domesticated dog wagged its tail at us, we have believed that these pets are communicating something to us and also communicating among themselves. Not everything can be attributed to instinct, and certain things (e.g., “now I am sitting on the eggs, and you fetch food, and then we swap”) need to be negotiated in real-time. Therefore, it’s not surprising that science has been seriously engaged in studying animal languages and attempting to understand them for years. Nowadays, this interest is mainly explained by the necessity to protect endangered species and ensure animal welfare.

At least that’s how specialists from the University of St. Andrews in Scotland, led by Christian Rutz, explained it in a recent issue of Science. They write that the goal is to “initiate a cultural shift that will lead to greater respect for many species with whom we share planet Earth.”

Why are scientists trying to resurrect the Tasmanian Wolf?

Predators are necessary in all kinds of environment, says biology professor Tomasz Sulej

see more
Of course, one could make the argument that most of us envision well-being as having unlimited access to things that are either “fattening or otherwise unhealthy and harmful to the environment.” However, due to the lack of communication, we cannot determine whether animals are wise, desire goodness and beauty, rather than mere pleasures, or if they are more similar to us. Therefore, we assume, influenced by our upbringing with Hugh Lofting’s The Story of Doctor Dolittle (and for those who haven’t read it, I highly recommend it, because it’s excellent literature), that animals are good, wise, and altruistic. Even more so than most of us. Hence, we do not blame the good doctor for shifting his focus from treating humans to helping quadrupeds. So, four legs good, two legs bad.

Our brain, through some form of analysis of sound frequency over time, immediately and flawlessly determines whether it is dealing with: a) sounds of nature (wind rustling, but also bird songs, hoofbeats, barking, howling, meowing, mooing, etc.), b) human speech, or c) music. The first two types of sounds quickly become background noise in our brain, unless, for example, we are birdwatchers distinguishing bird species by their calls or when, say, the roar of a lion induces significant stress. Human speech, on the other hand, is processed quite differently and in a completely different area of the brain’s cortex. Thus, our brain is not particularly prepared to process signals from animals as language.

  Communication involves not only words but also intonation, “melody,” and the whole non-verbal realm: gestures, facial expressions, body language, “dance,” that is, rhythmic movements, and even scents. We have assumed, and have been studying this for a long time, that animal communication is largely non-verbal.

Now let’s return to the text from Science, summarising several years of research in which human efforts have been accompanied by artificial intelligences. Nothing has progressed as rapidly as AI, especially language models (known as chatbots). They have proven to be quite good at categorising styles and generating new statements in specific, often unique styles. They can read between the lines and analyse poetry (though their figurative analysis is still quite weak – they always prefer a literal interpretation). Additionally, they are capable of quickly analysing terabytes of data from all possible angles. That, however, is not enough. We still need some kind of Rosetta Stone – a fragment of ancient Egyptian literature, the discovery of which had a groundbreaking significance in deciphering hieroglyphs.

Revival of crow culture

It is true that if animals (which – I repeat – often differ significantly from us in sensory perception) have a completely different world of imagination, and thus concepts and messages, no current artificial intelligence will be of much help. Even if it were to translate, we might not perceive the meaning. How to avoid anthropocentrism in translation, nobody knows. However, we assume that artificial intelligence has a better chance of being non-human in this regard than our own human intellect.

So, is there a chance of understanding the language of crows and elephants, which, according to common belief, are exceptionally wise, as well as other creatures? The authors of the paper in Science claim that there is quite a good chance, and rapid progress is apparent. As they write, a dictionary for decoders is being constructed: “Some types of signals can only be produced in specific circumstances, eliciting specific behavioural responses; a classic example is the savanna baboon (Chlorocebus pygerythrus), which emits an alarm call when it spots a predator, causing other group members to seek shelter. Establishing such correlations allows for formulating hypotheses about signal functions, which can then be experimentally tested (e.g., through controlled playback).”
“Dolittle” (2020). Photo by PAP/Photoshot – Universal Pictures / The Hollywood Archive.
Indeed, one could criticise this approach, as humans are telling the algorithm what they believe animals do based on their observations. However, animals might be doing something entirely different from what we perceive. This problem is intended to be solved by the so-called self-supervised deep learning methods. These methods do not require annotated datasets or pre-defined functions potentially relevant to communication. They build basic models similar to ChatGPT based on all observations and their conditions. Without any AI translation, as seen in the picture. Of course, this approach favours species for which we have complete life histories recorded, thus benefiting popular primates and laboratory rodents. Birds are somewhat less favoured, but they still receive preferential treatment. However, there can be surprises for the audience regarding what researchers have recorded using research funds.

As the Scottish authors inform, they are already working on “major joint initiatives, such as: Earth Species Project; Communication and Coordination at Different Scales; Vocal Interaction among Humans, Animals, and Robots; Inter-species Internet; and finally, the CETI (Cetacean Translation Initiative), which recently provided a detailed research plan using AI to study communication in sperm whales (Physeter macrocephalus).” The research group that published the article in Science is working on studying the vocal repertoire of the extinct-in-the-wild Hawaiian crow (Corvus hawaiiensis). AI is meticulously comparing the vocalisations of the last individuals, which found refuge in the San Diego Zoo Wildlife Alliance, with historical recordings. Biologists not only dream of restoring this endangered species to its former glory but also – quite literally – “reviving its culture,” of which these sounds would be a manifestation.

However, as is often the case with AI learning, the amount of necessary data is enormous. “Large amounts of audio and video data are stored in community archives (such as the Macaulay Library or xeno-canto), collected by passive recording arrays, or can be extracted from the internet,” the authors of the publication inform.

Similarly, when searching the cosmos for signals from intelligent beings, science is turning to amateurs for help. If we want it to be like in fairy tales where humans understand the language of animals, then we must mobilise ourselves and provide data. “We need creative solutions for collecting data on wild animals,” say the researchers. Fortunately, miniature or even microscopic recording devices with GPS, micro-cameras powered by perovskite solar cells, and other such inventions can already be purchased online, and the development in this field is inversely proportional to the size of these devices. We hope that AI will discover details and correlations that escape us completely, as demonstrated in a recently published study in eLife analysing the differences in vocalisations of zebra finches (Taeniopygia guttata), a small bird native to Australia.

Bird Radio

Specialists would like animal communication to become the subject of genetic experiments (discovering the gene sets that condition it) and for its evolutionary nature to be considered. It is incredibly fascinating, as, for example, in birds, the ability to learn vocalisations evolved independently in three closely unrelated groups: hummingbirds, parrots, and songbirds, all of which belong to the sparrow family.

“If we want to decipher animal conversations, we need to know who is conversing with whom and under what environmental and social conditions,” explained Professor Sonja Vernes in Science. She is an expert in the vocal communication of bats and works at the University of St. Andrews and the Max Planck Institute for Psycholinguistics. But is that really enough? Humans often struggle to understand each other, despite possessing the ability to speak. When we add to that how frequently our communication involves boasting about our ignorance, spreading falsehoods, or engaging in manipulation, trust in it as a source of knowledge does not increase. Now, we are trying to understand animal communication. But what if they are not particularly enlightened either, not revealing to us unknown secrets or not being straightforward and “speaking their minds”? Will we need lie detectors and fact-checking for them? Advances in the scientific understanding of animal communication raise ethical questions, such as the permissibility of initiating conversations with wild animals, eavesdropping on them without their knowledge, etc.

“If we manage to identify communication signals associated with suffering or avoidance, passive acoustic monitoring systems could be used to eavesdrop on how ‘happy’ or ‘unhappy’ animals are in the wild,” muses the main author of the article in Science. Perhaps not realising how much this resembles the Big Brother, she is focused on supporting conservation efforts and biodiversity. “But what kind of fairy tale is this? Perhaps it is all possible! True, though, I will put it in the realm of fairy tales for now.”

– Magdalena Kawalec-Segond

TVP WEEKLY. Editorial team and jornalists

– Translated by jz
Main photo: In the movie “Dr. Dolittle 2” (2001, produced by Twentieth Century Fox Film Corporation), Eddie Murphy is seen conversing with a bear. Photo by PAP/Photoshot.
See more
Civilization wydanie 22.12.2023 – 29.12.2023
To Siberia and Ukraine
Zaporizhzhia. A soldier in a bunker asked the priest for a rosary and to teach him how to make use of it.
Civilization wydanie 15.12.2023 – 22.12.2023
Climate sheikhs. Activists as window dressing
They can shout, for which they will be rewarded with applause
Civilization wydanie 15.12.2023 – 22.12.2023
The plane broke into four million pieces
Americans have been investigating the Lockerbie bombing for 35 years.
Civilization wydanie 15.12.2023 – 22.12.2023
German experiment: a paedophile is a child's best friend
Paedophiles received subsidies from the Berlin authorities for "taking care" of the boys.
Civilization wydanie 8.12.2023 – 15.12.2023
The mastery gene
The kid is not a racehorse.