Artificial intelligence is the dismantling of the animal. Should we try to speak again?

Tribal, trill, hide, wow, screaming. Animals speak all kinds of roads, however, humankind has scratched the surface of how to communicate with each other and the rest of the living world. Our race has trained some animals – and if I asked the cats, the animals also trained us – but we did not really comfortable the code to communicate between the species.

Increasingly, researchers publish animals artificial intelligence To accelerate our investigations into communication with animals – both inside the species and the branches on the tree of life. While scientists cut off the complex communication systems of animals, they approach to understand what creatures say – and perhaps even how to speak. But while we try to bridge the linguistic gap between humans and animals, some experts raise correct concerns about whether these capabilities are appropriate – or whether we should even try to communicate with animals at all.

Using Amnesty International to break the animal language

Towards the introduction of the package – or should I say pod? – CETI ProjectWhich used machine learning to analyze more than 8000 sperm whale “CODAS” – the organized clicks that you recorded Dominica sperm whale project. Researchers Detect Contestating and integration structures in whale clicks, and naming features such as “Rubato” and “decoration” to describe how whales skillfully adjust their acoustics during the conversation. These patterns have helped the team create a kind of vocal alphabet of animals – a crossing and organized system that may not be a language as we know, but it reveals a level of complexity that researchers were not aware of before. The CEI project is also working on Ethical instructions For technology, a decisive goal Looking at the risks From the use of artificial intelligence to “speak” to animals.

At the same time, Google and Wild Dolphin Project It was recently presented DolphingmmaLLM, trained for 40 years of dolphin boxes. Just as ChatGPT LLM for human inputs – which makes visual information like research papers, photos and responses for relevant information – deals with Dolphememma Dolphists ’voice data and expects what is vocal after that. Dolphememma can create a dolphin-like sound, and the primary researchers’ system is used in two directions, a distance measurement to increase hearing (suitable, chat), and an interface based on the smartphone used by dolphins to request elements such as scarves or seaweed-basic work mode for future dialogue.

Dennis Herzeng, founder and director of the Wild Delphine project, who led the development of DolPHingemma, in cooperation with Google DeepMind, said in an email to Gizmodo. “We will spend this fall time taking well -known dolphin boxes, and let us see any repetitive patterns that they find,” such as the audio used in flirting and discipline on the mother contract.

In this way, Herzeng added, that artificial intelligence applications are two parts: researchers can use to explore the natural sounds of dolphins and to better understand animal responses to the human tradition of dolphins, which are produced by the artificial intelligence chat system.

Expansion of the AI ​​Tools for Animal

Outside, researchers find that human speech models can be re -used to decode earthly animal signals. A team led by the University of Michigan Wav2VEC2-used a form to identify speech trained on human voices-to identification Dogs, races, breeds, and even individual identities based on their limits. The pre -trained human model has outperformed the trainees of only dog ​​data, indicating that the structures of the human language model can be amazingly effective in deciphering animal communication.

Of course, we need to consider the different levels of the development targeted by these artificial intelligence models. Determine whether the dog bark is aggressive or playful, or whether male or female – this may be easier for the model to define it, for example, the precise meaning coded in the whale of sperm whale. However, each study of scientists can be applied to understand how to understand the tools of artificial intelligence, as it is currently present, on such a wide field – and giving artificial intelligence an opportunity to train itself to become a more useful part of the researcher’s tool group.

And even cats – they are often seen as isolation – appear to be more communicating than they allowed. In the 2022 study of the University of Paris Nanner, Cats showed Clear signs to get to know their owner’s voice, but then, the connectors responded more when they spoke directly to “Cat Talk”. This indicates that cats are not only concerned with what we say, but also how we say – especially when he comes from someone who knows him.

Earlier this month, a pair of researchers in the ropes I found evidence Animals have a group of four “waves”, or physical gestures, which they make for each other, as well as for human operation of squid waves. The group plans to apply an algorithm to classify the types of waves, follow the movements of creatures automatically, and to understand the contexts in which animals express themselves faster.

Private companies (such as Google) also enter into the verb. Last weekThe largest search engine in China, BAIDU, patented with the IP administration in the country proposing to translate animal audio (specifically CAT) into human language. The speed and dirty technology is that it will take advantage of a set of data from Kitty, then use the AI ​​model to analyze the data, determine the emotional state of the animal, and remove the clear human language message that your pet was trying to transmit.

An international translator for animals?

These studies together represent a major shift in how scientists approach the communication with animals. Instead of starting from scratch, research teams build humans designed – and progress that would take a much longer time. It can be the ultimate goal (read: Be able to) Be a kind of Rosetta stone for the animal kingdom, supported by artificial intelligence.

“We have achieved well in analyzing the human language only in the past five years, and we started mastering this practice of transferring trained models to one collection and applying them to new data,” Sarah Kane, an environmental scientist and electrical engineer in the terrestrial species project.

The Earth Types project plans to launch its pioneering model in the vocal language of animal sounds, NaturelmThis year, an illustration of Nature-ADIO He lives. With input data from all parts of the tree of life – as well as human speech, environmental sounds, and even music detection – the model aims to become a transformer of human speech to its analogues on animals. The model “The transfer of the promising field of human speech to communication with animals” shows, the project States“Supporting our hypothesis that joint representation in artificial intelligence can help decipher animal languages.”

Kane added: “A large part of our work is really trying to change the way people think about our place in the world.” “We make great discoveries about communicating with animals, but in the end we find that other species are complex and accurately as we are. And that the revelation is very exciting.”

The moral dilemma

In fact, researchers generally agree on the promise of artificial intelligence -based tools to improve the collection and interpretation of animal communication data. But some feel that there is a collapse in communication between that scientific familiarity and the audience depicting how to apply these tools.

“I think there is a lot of misunderstanding in covering this topic-that machine learning in one way or another can create this contextual knowledge of nothing. As long as you have thousands of hours of audio recordings, some magic learning of the magic machine can press this meaning.” This will not happen. “

Rotz added: “The meaning comes through contextual explanatory comment, and this is the place that I think is really important to this field as a whole, in this period of excitement and enthusiasm, not to forget that these comments comments come from the basic behavioral environment and the experience of natural history.” In other words, let’s not put the horse before the cart, especially since the cart – in this case – is what works on the horse.

But with great strength … you know the clichies. Basically, how can humans develop and apply these techniques in a scientific way that illuminates and reduces harm or disturbance on their animal topics? Experts have put forward Ethical standards The handrails to use technologies that give priority to the well -being of creatures as we approach – well, wherever the technology is.

As artificial intelligence advances, talks on animal rights must develop. In the future, animals can become more active participants in those conversations – the idea that legal experts are exploration your The practice of thinkingBut one day can become a reality.

“What we need desperately – from developing the aspect of machine learning – is the formulation of these meaningful cooperatives between machine learning experts and animal researchers,” Rotz said, because only when you put two of us together, you stand an opportunity. “

There is no lack of data for nutrition in artificial intelligence models, starting with the screams of the perfect wild dog to the evil Snails’ paths (yes, really). But exactly how we benefit from the information we do from these new methods requires comprehensive Looking at ethics Participate in “speaking” with animals.

A Modern paper Regarding ethical concerns to use artificial intelligence to communicate with whales, six areas have identified major problems. These include privacy rights, cultural and emotional harm to whales, experimental formation, technological solution (excessive dependence on technology to repair problems), gender bias, and limited effectiveness to preserve the actual whale. This last issue is especially urgent, given the number of whale groups already Under a serious threat.

It seems that we are about to get to know more about the ways in which animals interact with each other – in fact, the curtain can also withdraw visions on how to learn, social communication and act in their environments. But there are still great challenges to overcome them, such as asking ourselves how to use the strong technologies that are currently being developed.