Rechercher
  • anielambednarek

AI and Animal welfare

Dernière mise à jour : 20 avr.


While AI advances raise ethical concerns for humans, there is growing enthusiasm for technological progress with regard to animals. Indeed, whereas the former enjoy legal personality and fundamental rights, the latter do not have any legal protective status (or only a weak one compared to the legal framework provided to individuals), so that the question of a potential infringement of their dignity or their privacy does not arise. Nor does the issue of possible discrimination seem particularly problematic. Animals do not have to worry about algorithmic bias or the threat of mass surveillance. As the legal framework is not the same for humans and animals, the questions concerning the ethics of AI do not arise in the same terms. The stakes are even radically different: for individuals, AI conflicts with ethics, whereas for animals, technological progress seems to promote and reinforce ethics. In short, the logic is the opposite. Here are some recent examples of the promise of AI in the field of animal welfare. The contributions of new technologies to the improvement of the living conditions of wild animals, farm animals, laboratory animals and family pets will be discussed.

AI and wild animals

Wild animals have been largely overlooked in animal welfare law, as their suffering is not taken into account in legislation aimed at improving the living conditions of animals. However, advances in AI could save many lives by improving the effectiveness of the fight against poaching and the trade in the products of this prohibited practice.

AI to help animals targeted by poachers

When it comes to wildlife protection, there is an AI designed to help animals targeted by poachers. Two companies, Neurala and Air Shepherd, which specialize in visual AI and predictive analysis, have teamed up to protect elephants and rhinos that are being slaughtered in Africa. This new technology, based on predictive algorithms, helps to combat the poaching that kills tens of thousands of animals every year and precipitates the decline of wild species. The efficiency of this AI system deserves to be underlined because the combination of predictive analysis and visual AI makes it possible to multiply the performance of drone surveillance by ten. Its importance is all the greater when the animals in question are at risk of extinction. Poaching, which causes suffering to the animals, also leads to illegal wildlife trafficking that has devastating effects on the environment. Researchers in the US have also developed an AI to protect primates threatened by animal trafficking through facial recognition. The technology, called PrimNet, is based on a database of thousands of photos of monkeys and is proving to be a gentler method of identifying and locating endangered primates than microchips. The software uses a neural network to distinguish animals by their eyes or fur. It is 90% reliable. One of the objectives of this AI system is also to investigate the capture of these animals to prevent further trafficking.


AI fighting the illegal wildlife trade

The illegal trade in endangered wildlife species is one of the most lucrative crimes in the world. Large criminal networks use transport and financial systems to illegally move animal products and generate profits through laundering. Microsoft has developed an AI research project, Seeker, to detect illegal wildlife products at ports and airports and prevent traffickers from transporting and benefiting from them. The AI system that Microsoft has developed with UK Border Force is designed to dismantle these increasingly sophisticated criminal networks and stop the illegal trade. The model is based on an algorithm that can be trained on any species in a relatively short time (2 months). When Seeker identifies an illegal item, it immediately alerts security officers. Seized items can be used as evidence in criminal proceedings. The performance of this machine learning algorithm is remarkable when it comes to ivory objects (70% success rate on scanned products). The complexity of this illegal trade is such that the progress of AI in this area can only be welcomed. By providing a better understanding of criminal strategies, Seeker significantly improves the detection rates of illegal trafficking in endangered species.


AI and farm animals

In Europe, the objective of animal welfare is becoming increasingly important in national and EU law, particularly for farm animals, whose suffering during practices such as transport and slaughter is addressed in a number of texts. Although they are now recognized as sentient beings by legislation, farm animals have still not been elevated to the rank of sentient beings, even though scientific progress support this view. Indeed, in recent years, AI has distinguished itself by its ability to identify and recognize, better than human beings, the emotions of certain animals, such as sheep or pigs.

AI detects sheep suffering

Research has shown that sheep are highly intelligent social animals with cognitive performance equivalent to that of monkeys. The fact that sheep have a high capacity to recognize and remember faces tends to undermine some of the traditional prejudices associated with human intelligence: just like human intelligence, animal intelligence is capable of recognizing faces, and so are some artificial intelligence systems. According to researchers at Cambridge, there is a resemblance between the facial movements of sheep and those of humans. A scientific team has developed an AI that can detect the suffering of sheep and assess its intensity. The system is based on the Sheep Pain Facial Expression Scale, an instrument developed in 2016. According to the first published studies, the algorithm has an accuracy of 80%. In the future, this technological tool could be used on farms to detect animal suffering at an early stage and treat them as soon as possible. By using this new technology, farmers can hope to slow the progression of the most contagious and painful diseases. In France, a similar project has been carried out on goat farms: an AI system has been tested to monitor the sleeping and feeding times of these animals in order to alert the farmer if the slightest anomaly is detected.

AI translates pig vocalisations

While we are familiar with the cries, grunts and squeals of pigs, few humans are able to accurately determine the emotions that these sounds express. Decoding them can be a valuable tool for farmers who want to improve the welfare or reduce the suffering of their animals. Researchers have developed a system capable of recognizing pig vocalisations (more than 7000 sounds collected from hundreds of pigs) and translating their emotions - positive and negative - to help farmers in difficult situation. The AI's performance is remarkable - in terms of accuracy - when it comes to emotional valence and the situations in which these sounds are emitted. Besides alerting farmers in case of emergency (fights between pigs, piglets crushed by their mothers), AI allows to promote welfare situations for these animals (infrastructures, toys). While acoustic monitoring systems already exist to monitor the health of pigs, researchers are now seeking to combine physical and mental health measures to improve the general well-being of pigs on farms.

This is especially important as researchers have shown that certain categories of animals, including those destined for slaughter, have mental abilities that allow them to have subjective experiences. This capacity to experience various degrees of emotion, ranging from pain to pleasure, is translated into the term ‘sentience’.


AI and laboratory animals

Recently Swiss voters rejected a referendum proposal to make Switzerland the first country to ban medical and scientific experiments on animals. This negative result might occult the fact that, thanks to the progress of AI, there are alternative solutions which European Union is trying to promote.

An AI able to replace animals for laboratory tests

To test certain substances and verify the toxicity of products (drugs, make-up), laboratories use and kill millions of animals every year (rodents such as rats, mice and guinea pigs, but also monkeys, cats and dogs). In the United States, researchers from a Baltimore University have published promising work to put an end to these practices that cause great suffering. The algorithm developed by the scientists gave predictions that were as reliable as tests carried out on animals in the laboratory (if not more so). The development of a few tests could make it possible to assess the dangerousness of thousands of chemicals, with regard to the risks associated with inhalation or the effects on the environment. The widespread use of this algorithm, which is on the way to revolutionize toxicology, would make it possible to avoid the torture of a very large number of animals used for experimentation. For those who are not convinced by these alternative methods, it should be remembered that the effectiveness of these algorithms is likely to improve the fight against pollution and, consequently, human health, insofar as this is one of the main causes of illness and mortality in the world. Besides being ethically problematic, tests on mammals are costly and of very relative reliability in assessing the dangerousness of potentially harmful substances (due to differences between species and problems of extrapolation). Animal ethics, environmental concerns and human health require more reliable solutions for a better management of chemical risks. In this respect, advances in AI deserve special attention from public authorities.

Alternative methods promoted by the European Union

In the European Union (EU), millions of animals are used for scientific research every year. Faced with this alarming situation, the EU has set itself the objective of promoting alternative methods to laboratory tests, while ensuring strict control of the safety of chemical products. With regard to animal testing, Directive 2010/63/EU, which sets high animal welfare requirements, establishes the 3R rule, which consists of « replacing » animal testing as soon as possible, « reducing » the number of animals used and « refining » procedures by optimizing the methods used to reduce animal suffering. EU law therefore requires the use of alternative methods where they exist. The EU itself invests millions of euros in toxicology projects that do not involve tests on animals.


AI and pets

While AI facial recognition and voice recognition systems are already being tested and used for wild and farmed animals, these same technologies can also improve the relationship between humans and their pets. However, these tools raise thorny ethical issues as they can be misused. The risk-based approach advocated by the EU tends to prohibit the use of these AI systems in Europe in order to distinguish AI made in Europe from the Chinese model.

an AI that can find lost dogs and understand their language

In 2018, the startup Zoolingua proposed to use AI to develop the language of dogs. The idea is to aggregate and analyse thousands of data on dogs with the aim of understanding the sounds they make when they growl or bark. The deep learning-based project aims to develop an automatic speech recognition system. Far from being fanciful, the researchers' aim is to prevent the euthanasia of millions of dogs killed each year in the United States due to behavioral problems. Understanding animals better would improve their daily well-being and could also save their lives. While voice recognition is still in its infancy, facial recognition is already a reality. Like human fingerprints, dog’s noses are unique: each has a unique pattern of ridges and creases. This is why Asian and American start-ups are offering applications that allow owners to record their pets' profiles in order to find them in the event of a runaway. In China, the startup Megvii is able to identify animals through pre-registration of their nose in a database. The application, based on neural network technology, allows unfortunate owners to find their lost dog if they have created a profile of their companion in the software. The Chinese startup praises the performance of its AI system, claiming 95% success rates. Similarly, the American software Finding Rover allows anyone who finds a lost pet (dog or cat) to take a photo of the animal in order to find its owner. These solutions, if they become widespread, could be useful for animal shelters. However, they are problematic in Europe, as the EU is moving to ban facial recognition systems because of the dangers these technologies pose to the fundamental rights of individuals.

Possible abuses against humans

Far from merely helping owners find their dogs, AI systems based on facial recognition of animals can identify owners who are deemed negligent or "uncivilised". The Megvii startup is not limited to dog identification, as it provides the Chinese government with a system for recognizing citizens who break the law, due to the massive amounts of personal data accumulated. Since Megvii is using its technology to support the Chinese government's surveillance programme, there are concerns that the application could be used to monitor owners. Indeed, China has millions of AI systems used to control the actions of its population. It is likely that owners who break the law (picking up dog mess, wearing a leash) will be fined through this. This is what is at stake in the European AI regulation project, which seeks to strictly regulate or even prohibit certain "high-risk" AI systems (such as biometric recognition), because of its impact on fundamental rights: whether it be the identification of individuals or mass surveillance.



While recent advances in AI raise well-founded concerns about the protection of fundamental rights, prompting the EU to establish a regulatory framework that is both ethical and legal, it should not be forgotten that many AI systems have the potential to significantly improve the living conditions of many categories of animals. However, their development must not be hindered by an overly strict, rights-based and anthropocentric EU regulation.

9 vues0 commentaire

Posts récents

Voir tout