What AI Means For Animals

There is an urgent need to expand AI ethics so that it considers nonhuman life.

Noah Campeau for Noema Magazine

Peter Singer, a professor of bioethics at Princeton University, was the 2021 recipient of the Berggruen Prize for Philosophy and Culture. His books include “Animal Liberation,” “Practical Ethics” and “The Life You Can Save.”

Tse Yip Fai has been the China strategy consultant for Mercy For Animals and is currently a research assistant for Singer’s project on the ethics of artificial intelligence concerning nonhuman animals.

The ethics of artificial intelligence has attracted considerable attention, and for good reason. But the ethical implications of AI for billions of nonhuman animals are not often discussed. Given the severe impacts some AI systems have on huge numbers of animals, this lack of attention is deeply troubling.

As more and more AI systems are deployed, they are beginning to directly impact animals in factory farms, zoos, pet care and through drones that target animals. AI also has indirect impacts on animals, both good and bad — it can be used to replace some animal experiments, for example, or to decode animal “languages.” AI can also propagate speciesist biases — try searching “chicken” on any search engine and see if you get more pictures of living chickens or dead ones. While all of these impacts need ethical assessment, the area in which AI has by far the most significant impact on animals is factory farming. The use of AI in factory farms will, in the long run, increase the already huge number of animals who suffer in terrible conditions.

AI systems in factory farms can monitor animals’ body temperature, weight and growth rates and detect parasites, ulcers and injuries. Machine learning models can be created to see how physical parameters relate to rates of growth, disease, mortality and — the ultimate criterion — profitability. The systems can then prescribe treatments for diseases or vary the quantity of food provided. In some cases, they can use their connected physical components to act directly on the animals, emitting sounds to interact with them — giving them electric shocks (when the grazing animal reaches the boundary of the desired area, for example), marking and tagging their bodies or catching and separating them.

You might be thinking that this would benefit the animals — that it means they will get sick less often, and when they do get sick, the problems will be quickly identified and cured, with less room for human error. But the short-term animal welfare benefits brought about by AI are, in our view, clearly outweighed by other consequences. When diseases are less frequent, more predictable and more controllable, factory farms can crowd more animals into confined spaces, thus increasing their profits. In fact, several AI companies openly advertise being able to pack more animals into a given space as a benefit of installing their systems.

“Factory farming is morally indefensible and needs to be replaced as soon as possible by less cruel and more efficient ways of producing food — not supported further by AI.”

In addition, when factory farms become more efficient, the price of animal products falls, leading to more demand for meat and more animals raised in factory farms. This will also make it harder for plant-based analogues and cultivated meat — meat produced from animal cells grown in bioreactors — to eventually replace factory farmed products, which could increase the scale and extend the lifetime of factory farming. This is a moral atrocity because of what it does to animals, quite apart from its disastrous consequences for the environment.

Each year, factory farming brings into existence, rears and kills more than 70 billion birds and mammals and nearly 100 billion finfish. These animals live in crowded conditions without any consideration for their welfare beyond what is essential for profitability. Male piglets often have their testicles ripped out and their sensitive tails cut off without anesthesia. To prevent the dominant chickens pecking the weaker birds to death in unnaturally crowded conditions, the ends of their beaks — a sensitive part of their anatomy, filled with nerve endings — are often cut off with a hot blade, without anesthesia. Male chicks of the laying breeds have no commercial value and are ground up or suffocated immediately after birth.

We could go on. But clearly, factory farming is morally indefensible and needs to be replaced as soon as possible by less cruel and more efficient ways of producing food — not supported further by AI.

History Repeating Itself

We can better understand this dynamic by looking at an earlier farming innovation that was touted as benefiting animals but ultimately enabled their mistreatment. In the 1930s, Bayer advertised an antibacterial called Prontosil to livestock producers in Britain, and their early success attracted other companies and countries to follow suit. In 1948, Merck & Co. obtained an official license in the U.S. to add sulfaquinoxaline to poultry feed, and the practice of lacing feed with antibiotics soon spread to Japan, China and the Soviet Union.

Antimicrobial drugs (AMDs) quickly became popular due to their ability not just to cure diseases, but to prevent them as well. The preventative use of AMDs lowered the fatality rate of farmed animals and reduced costs and risks for animal-raising operations. They were even found to make animals grow faster — some suggested that the drugs reduced the total amount of energy the animals used for fighting pathogens.

It might have seemed that these drugs helped both humans and animals. But some producers realized that the drugs made it possible to raise more animals in the same amount of space. “Historically, contagious diseases had limited how many animals could be held in any single flock or herd,” Wessels, a farm that’s been around since the 1920s, reports on its website. “Too many animals, and the entire group could be wiped out by disease. Antibiotics changed all of that. Quickly, the size of poultry, swine, dairy and cattle feedlots rose dramatically and per-unit production costs fell.” This allowed animals to be kept in ever more crowded and filthy situations, unsuited for their social and behavioral needs. High profits drove more farms to adopt these practices, and the use of AMDs is now standard in factory farming across the globe.

It is fair to say that AMDs were a big part of the reason humans shifted from small-scale animal raising operations to factory farming. That negative consequence outweighed, by far, all the short-term welfare benefits AMDs had initially provided for the animals. In a similar way, the involvement of AI in factory farming is likely not only to make factory farming worse for animals but also to make it harder to replace with less cruel, more environmentally friendly, safer ways of producing protein. To adapt the Spanish philosopher George Santayana’s famous saying: Those who cannot remember our past mistreatment of animals are condemned to repeat it.

Noah Campeau for Noema Magazine
A Modern Case Study

China’s fish farming industry, to take one example, is primed to be revolutionized by AI. The world’s first treatise on how to raise fish in non-natural systems, Master Taozhu’s “Book of Fish Raising” ( 養 魚 經 ), was written in China more than 2,400 years ago. It discusses the raising of carp, covering details such as pond design, breeding methods and expected profits. It was so forward-looking — in the technological sense, not the ethical — that the book still describes the basic principles of how most people in China, and perhaps in other developing Asian countries, raise their fish.

The ongoing use of older technologies leaves fish farms such as these vulnerable to fluctuating temperatures, poor water quality and diseases. Until recently, even if the operators invested in and know how to use modern monitoring equipment, they still needed to check it frequently, day and night. And people within the aquaculture industry in Asia frequently cite the difficulty of passing along their techniques to the next generation. As a result, those who raise fish often face high mortality rates in their ponds.

AI would change all that. The equipment, which may be connected to a central computer or cloud server, can train image recognition algorithms with machine-learning techniques to identify parasites, monitor the health of the fish and whether they are under- or overfed and automatically stop, decrease or increase feeding. Diseases can not only be identified but even predicted, reducing human error and saving the operators time. Moreover, AI can also address the issue of passing on expertise by learning from human operators who are proven to outperform others in the industry and by capturing as data what these outperformers do but cannot explain verbally.

However, these AI improvements aren’t going to change the basic business model of factory farming: pack in as many fish as possible, with as little food as possible, with no regard for their wellbeing. These AI improvements won’t prevent farmers from killing fish in a torturous way — cutting, descaling and skinning them while they are still alive, which is common practice. Instead, the AI simply increases their efficiency and profits.

An Ethical Approach

What, then, should the AI industry do? First, AI companies, scientists and institutions should refrain from supporting or participating in AI projects that directly help the factory farming industry. Instead, they might consider opportunities to use AI to assist the development of plant-based meat and cultivated meat, both of which have huge potential to reduce or even replace the demand for factory farmed products and both of which are already benefiting from AI. 

For example, NotCo, a plant-based alternative company, uses machine-learning natural language processing to find combinations of plant ingredients that are likely to match the tastes of animal products. Equinom, a foodtech company that only serves plant-based products, uses AI to find the right variety of seed to grow the crops used for plant-based products, so that they will have the desired taste, texture and nutrition profile. On the cultivated meat side, Animal Alternative Technologies uses AI to model how parameters such as nutrients, flow rates, pH and temperature affect efficiency and production costs.

“The AI industry will hire exactly zero employees who are nonhuman animals, so they will have no representatives if humans do not stand up to represent them.”

Some may object that we cannot expect AI companies to forego the economic benefits of involvement in a large and entirely legal industry — but it is not clear that this involvement will benefit those companies in the long run. Factory farming already has a negative image in many countries, and both the animal movement and the environmental movement are strongly opposed to it. Many wealthy investors — and some governments — are putting substantial sums into trying to replace factory farming. Temasek, which is 100% owned by the Singaporean government, invested in plant-based companies and a platform to support alternative proteins. They also invested in Upside Foods, a cultivated company, together with the government-owned Abu Dhabi Growth Fund. The Danish government has said that it will invest over $100 million into a new Plantefonden, or Plant Fund. Some famous investors are putting their money into alternatives to meat, too, such as Li Ka-shing, Bill Gates, Prince Khaled bin Alwaleed Al Saud and Natalie Portman.

Plant-based and cultivated meats will not only be more ethical, more sustainable, more environmentally friendly and better for public health — they also have the potential to be more economical. And if factory farming is, like the tobacco and coal industries, an industry “against the tide of history,” involvement with it could hurt the image of large tech companies and reduce their other business opportunities. 

Concerns about the ethics of AI have already given rise to initiatives with names like “AI for humanity,” “AI for humans” and “AI for social good.” We would love to see decision-makers expand this thinking to be more inclusive, with initiatives such as “AI for animals” or, better still, “AI for all sentient beings.” More groups should follow the lead of the Montreal Declaration for a Responsible Development of AI, which states: “The development and use of artificial intelligence systems must permit the growth of the well-being of all sentient beings.”

There is an urgent need to expand AI ethics so that it considers nonhuman life. And it is not only AI companies and AI scientists who are responsible for what is happening. There are several other human stakeholders, including philosophers working in AI ethics, NGOs, policymakers and lawmakers. Consumers should also understand and take responsibility for the consequences of their consumption choices.

Animals, unlike most humans, cannot participate in the design of AI. They cannot have their own social movements or tell us that a particular AI is harming them, and they will never design AI to benefit themselves. The AI industry will hire exactly zero employees who are nonhuman animals, so they will have no representatives if humans do not stand up to represent them. We need to do it — it’s on us all.