AI for Animals #3: AI in factory farming (Part 2)
September 3, 2024
Welcome to the third edition of AI for Animals! This newsletter brings you the latest news and research on AI, animals, and digital minds.
Each edition also homes in on a specific topic relating to AI’s potential impacts for animals. The last edition outlined what AI could mean for factory farms; this edition digs deeper into what this could mean for the animals themselves.
If you have any questions or feedback, our email is contact@aiforanimals.org. You can subscribe to this newsletter here.
Thanks to Allison Agnello, Madeleine Archer, Yolanda Eisenstein, and Constance Li for their contributions to this month’s edition!
Max Taylor
Index
AI in factory farming (Part 2)
AI could benefit farmed animals – but at what risk?
All major animal agriculture producers have a spiel about how they value the health and wellbeing of the animals in their supply chain. In reality, they typically only care about these factors to the extent that they maximize the animals’ productivity (and, to a lesser extent, their own public image). As such, we should always be skeptical of any welfare claims that companies make surrounding their deployment of AI; their priority is to keep their consumers and shareholders happy, not their animals.
That said, AI could deliver real, tangible benefits. AI systems could continuously monitor animals to ensure they are well-fed and have enough to drink. They could signal to farmers when animals are sick or injured, so the farmer could quickly care for them or euthanise them.
Environmental sensors designed to predict disease outbreaks could indirectly prevent the suffering and early death of many animals. While such applications are unlikely to actually make the animals’ lives worth living, they could still mean a whole lot less suffering for a whole lot of animals.
But some pretty major risks accompany those benefits. Broadly, welfare implications depend on four major uncertainties:
Do these systems actually monitor welfare, or just health and productivity?
How is welfare defined?
How well can systems measure welfare?
What will producers actually do with those measurements?
For example, AI systems might only monitor health problems relevant to productivity, or only identify and communicate the most severe welfare problems. Even the most sensitive systems will be subject to inevitable errors or unsophisticated programming, leading some health and welfare problems to go unnoticed. Farmer-animal interactions could be reduced to those too complex or sensitive for AI systems to perform, such as routine mutilations, further weakening their bond with the animals.
At the industry level, as outlined in the last edition, large, intensive systems stand to benefit most from AI. This could discourage producers from transitioning to less intensive systems. It could also drive monopolization of the market by the larger companies that can afford the technology, cementing their power to influence consumers, government, and other major stakeholders. AI could also facilitate advanced breeding techniques and genetic modifications aimed at pushing farmed animals even further to their biological limits for the sake of productivity.
But perhaps the most significant risk is AI’s cost-cutting potential. The animal agriculture industry operates on razor-thin margins. If producers can use AI to widen those margins, even by a fraction, this could be enough to keep them financially viable for decades or centuries longer than they would have been otherwise.
This may be particularly true for new sectors like aquaculture and insect farming. These industries haven’t yet exploited all the potential cost-cutting tricks of more established sectors, such as the intensive chicken farming industry. They’ve not yet invested in nearly the same level of infrastructure as other animal farming sectors, giving them more opportunity to build AI systems into their housing from the get-go, rather than retrofitting later. Much less is known about the animals themselves, and their behaviors are harder for humans to understand than those of birds and mammals. In aquaculture, it’s simply much harder for humans to directly monitor and control things underwater. AI tech could greatly help to overcome these obstacles. In fact, AI-assisted, fully automated insect farms are already commercially available, and it seems that shrimp farms could soon be headed the same way.
Heavy regulation is needed
So what do we do? An outright ban on the use of AI systems in animal farming doesn’t seem politically feasible, and would mean forgoing some significant welfare gains. The most promising avenue is probably to advocate for meaningful regulation of AI systems in farming that actually prioritizes animals’ wellbeing. In this vein, Virginie Simoneau-Gilbert and Jonathan Birch have outlined four ‘Goldilocks Principles’ for regulating AI in farming. Under these principles:
AI must not be used as an excuse to regress to less humane farming systems.
Information gathered by AI systems must be made freely available.
Companies must be held accountable if they fail to deal with any welfare problems identified by those systems.
AI must supplement farmers’ work, not replace it.
Championing such principles with decision-makers would be a good start, potentially supplemented by more ambitious asks. For example, any producer using AI could be mandated to also use it to less profitable ends, such as the mandatory introduction of AI-assisted CCTV to detect potential instances of animal welfare violations, or the proactive communication of standardized welfare indicators to their website.
These ‘Goldilocks Principles’ are useful because they are sufficiently general to cover a wide range of future applications, ready to be adapted to whatever form the animal farming/AI landscape takes. This is necessary when AI is moving so quickly and so unpredictably. Animal advocates would do well to take a similar approach with our own advocacy, continually monitoring producers’ use of AI and adapting our strategies to ensure they are held to account. If we don’t, AI might end up being the best thing that ever happened to the factory farming industry.
In our next edition, we’ll explore how AI could facilitate human-animal communication, and what this might mean for our relationship with other species.
📚 Resources
For more information on AI in factory farming, check out:
The dangers of AI farming (Virginie Simoneau-Gilbert and Jonathan Birch, Aeon)
We should campaign to restrict AI use in animal agriculture (Zachary Brown, Before Porcelain)
Twelve Threats of Precision Livestock Farming (PLF) for Animal Welfare (Tuyttens et al., Frontiers in Veterinary Science)
The Hive Community Slack has several channels dedicated to discussion of AI and animals, including #c-ai-discussion for broad discussions and #s-ai-coalition for project collaboration.
If you want to dig deeper, the aiforanimals.org website has a list of relevant articles, papers, and other materials giving an overview of the AI and animals space.
🌏 Opportunities
The European AI Office has launched a consultation on trustworthy general-purpose AI models under the EU AI Act. Respond to the consultation by September 18 to give your opinion about what should be added to the Act.
Food Systems Innovation has launched a Request for Proposals ($10-25k) on “Creating Benchmark Datasets and Common Task Frameworks for Alternative Protein Development.” The application deadline is October 4.
Open Paws is currently doing research to facilitate effective campaigning against the most harmful uses of AI in factory farms and slaughterhouses, and they’re looking for assistance from volunteers. To find out more, follow the conversation on the Hive Slack community or get in touch via their website.
Meta is seeking proposals that use their newest LLM model, Llama 3.1, for economically and socially impactful projects. During their open call for applications, eligible organizations from around the world can apply online for an award of up to $500,000 USD to support their projects.
🚨 Updates
AI for Animals hosted our second meetup in San Francisco, featuring speakers and pitches in addition to our core speed networking activity. We now plan on having these meetups monthly!
Sankalpa Ghose spoke about how quantified empathy (QE) can be used to change the way animal consideration is accounted for in human systems. He reported getting valuable feedback and promising leads from people connected with AI labs who were interested in further advancing his project, animalfriendly.ai.
Bill Meyer announced Food System Innovation’s new request for proposals, which aims to set the foundational framework for AI/alt-protein research similar to what ImageNet did for computer vision. His collaborators Noa Weiss and Anna Thomas also participated in the active Q&A session.
Constance Li, founder of AI for Animals, gave a ‘lightning talk’ about AI’s potential impact for animals to a crowd of AI Safety researchers at a MATS Program event. The example of antibiotic development enabling dense indoor chicken farming was used to emphasize how technology has historically contributed to the intensification and exploitation of animals. This was the first introduction to the intersection of these two topics for many and was the only talk about animal welfare at the event.
Constance also gave a presentation to leaders in the animal advocacy movement the following day on why AI matters for animals and how the current field is developing.
🗞️ News & Research
🗣️ Understanding animals
AI Animal Communication Breakthroughs Could Revolutionize Our Relationship With Animals (Sentient Media)
Recent AI breakthroughs are enabling projects like the Earth Species Project to decode animal communication by analyzing vast datasets of vocalizations and behaviors, with efforts focusing on species such as crows, sperm whales, and rodents. These advances could reshape animal rights by providing concrete evidence of animal distress or needs, but they also carry risks of exploitation, such as enhancing factory farming practices by using AI to manipulate animal behavior.
Cetacean conversation: AI could let us talk to whales. Experts question if that's a good idea (Salon)
Project CETI is using AI to analyze sperm whale codas by matching sounds with behaviors, aiming to decode their communication and recreate it using AI-generated sequences. However, experts caution that AI-generated whale sounds could confuse or disrupt whale communication, and some question whether these codas truly function as complex language that AI can effectively translate.
[Video] How AI could help us talk to animals (Vox)
Researchers studying animal communication have made significant progress using AI to decode complex vocalizations in species like African elephants. However, validating these findings remains a significant challenge.
🐔 Chicken farming
Unlocking the next generation of genetics – 2024 Cobb Research Initiative (CRI) funding announced
The 2024 Cobb Research Initiative seeks to advance poultry genetics through collaborations with researchers, including leveraging AI and precision farming. This year’s recipients include the University of Queensland’s project ‘Digital Twins for Better Broilers’ and Prophet AI’s ‘Virtual Data Powered AI for Digital Health Phenotype Tracking’.
🐮 Cow farming
[Podcast] Prof Ben Hayes on how AI is driving genetic progress in beef (The Weekly Grill)
University of Queensland genetics researcher Prof Ben Hayes discusses how AI is helping drive genetic and genomic selection in the beef industry.
AI-enhanced real-time cattle identification system through tracking across various environments (Mon et al., Nature)
This study presents an AI-driven cattle identification system based on back pattern features captured by overhead cameras. The system demonstrates high accuracy in tracking and identifying cattle across varied farm environments, and could offer producers a cost-effective, low-maintenance solution for real-time, contactless cattle monitoring.
🐷 Pig farming
How do swine producers and veterinarians think precision livestock farming could help farms? (Michigan State University)
Swine veterinarians see more potential in precision livestock farming (PLF) than producers, particularly for addressing respiratory and tail-biting issues, with 75% recommending PLF to their clients. Both producers and vets agree that PLF is most needed for reducing piglet mortality, lameness, and post-weaning issues. Veterinarians tend to prioritize health and performance, while producers consider broader operational factors when deciding on PLF adoption.
🐟 Aquaculture
Google’s aquaculture AI spin-off to expand to Australia, Chile with new funding (Undercurrent News)
Tidal, Google's aquaculture AI spin-off, has secured new funding to strengthen its presence in Norway and expand to Australia and Chile. Tidal’s AI-driven underwater technology, already deployed in over 250 pens globally, helps salmon farmers track growth, monitor fish welfare, and optimize feeding through autonomous systems, aiming to solve key industry challenges.
Digital aquaponics factory ensures intelligent fish farming and vegetable growing in SW China (People’s Daily Online)
At a digital aquaponics factory in Chongqing, China, AI-driven systems manage fish farming and vegetable growing with minimal human intervention. Automated feeders and smart algorithms precisely control feeding and monitor water conditions, while predictive AI models ensure accurate water quality management, significantly improving efficiency and reducing farming costs.
🐑 Animal farming: General
Precision livestock tech has hurdles (The Western Producer)
Adoption of precision livestock farming is hindered by economic risks and limited commercial availability. Farmers are hesitant to invest in these technologies without clear evidence of a return on investment, and the complexity of some precision systems further complicates their implementation. Producers are also concerned about who controls the data generated by these systems and how it is used.
KPM Analytics Unveils AI-Powered At-Line Vision Inspection System for Meat and Poultry (Quality Assurance and Food Safety)
KPM Analytics has introduced the AI-powered TheiaVu WD-300, an at-line vision inspection system designed for the meat and poultry industry. This system uses AI to analyze product attributes such as thickness, fat distribution, and cook color, as well as detect defects like blood spots and foreign materials, providing a faster, more consistent, and objective alternative to traditional manual inspections.
🐬 Wild animals
[Podcast] Using AI for facial recognition — or fin recognition — for whales and dolphins (WBUR)
A short podcast about how AI is helping researchers track whales and dolphins, including an interview with Ted Cheeseman, founder of Happy Whale.
Scientists are using AI to save rare African forest elephants in the Congo Basin. Here's how (Discover Wildlife)
In partnership with IBM, scientists are training AI to identify individual African forest elephants by analyzing unique features like tusks and wrinkles from camera-trap images. This AI-driven approach will improve the accuracy of population estimates, which is essential for better conservation efforts and community compensation programs in areas where these endangered elephants are present.
AI technology keeps 6,000 deer from rail routes (BBC News)
AI technology installed by the UK’s Network Rail has successfully prevented nearly 6,000 deer from crossing railway tracks in England. The system uses AI-powered sensors and cameras to detect deer approaching the tracks, triggering alarms that divert them and reduce collisions.
Colorado High Schoolers Invent a Thermal AI Dashcam to Reduce Collisions with Deer (Outdoor Life)
Colorado high school students created an AI-powered dashcam that uses an infrared camera and custom AI to detect deer and other wildlife on roads, alerting drivers with visual and auditory signals. This device aims to reduce wildlife-vehicle collisions by providing earlier warnings, especially in low-light conditions.
AI can turn the tide on organised environmental crime in Africa (Enact Africa)
AI is being used to combat organized environmental crime in Africa by processing large datasets to map offender movements, identify criminal patterns, and enable real-time responses. Technologies like TrailGuard and Skylight use AI to detect poachers and illegal fishing activities, while Digital Earth Africa monitors land use changes to uncover illegal mining, significantly improving the efficiency of law enforcement in tackling these crimes.
AI can detect wildfires and monitor endangered birds, according to Hong Kong experts, start-up (South China Morning Post)
Hong Kong experts and start-ups are using AI to enhance conservation efforts, such as detecting wildfires and monitoring endangered bird populations. Robotics Cats, a local AI start-up, uses advanced surveillance cameras and deep learning to automate wildlife tracking and habitat management, earning recognition from the World Economic Forum.
🍔 Alternative proteins
Redefine Meat expands its reach into the EU (Food Manufacture)
Redefine Meat is a plant-based meat company that uses AI and 3D printing technology to create plant-based products that mimic the texture and structure of real meat. It has expanded its presence in the European market by securing listings with major retailers such as Coop in Switzerland, Jumbo in the Netherlands, Velivery in Germany, and Monoprix in France.
Equinom’s AI-Driven Manna Technology Now Commercially Available for High-Protein Pea Production (Vegconomist)
Israeli agri-food-tech company Equinom has launched its AI-powered Manna™ technology, which optimizes high-protein yellow pea production by analyzing biochemical parameters and predicting optimal processing settings. This innovation, which has increased protein content in yellow peas from the typical 21-25% to up to 75%, reduces processing needs and environmental impact, offering significant cost savings and efficiency improvements for the plant-based food industry.
Vegan cheese that tastes like cheese? These startups may have cracked the code. (Grist)
New Culture uses precision fermentation to produce dairy casein, the protein responsible for cheese's stretch and melt, allowing them to create a cow-free mozzarella that closely replicates traditional dairy cheese. Climax Foods uses AI and machine learning to analyze plant-based proteins and fats, identifying combinations that mimic the texture and flavor of cheeses like blue cheese and feta.
AI and food-grade bioreactors may improve margins on cultivated meat commercialization (Food Navigator USA)
To scale cultivated meat production, the industry must transition from costly pharmaceutical-grade bioreactors to more affordable food-grade alternatives, which currently don't exist. AI and machine learning can optimize bioprocesses by analyzing sensor data in real-time, while public-private partnerships, like the Bezos Earth Fund's AI for Climate initiative, are crucial for funding scalable solutions and reducing costs.
Planted Leverages AI for Sustainable Growth Using Microsoft Copilot (Microsoft)
Planted is using Microsoft's AI technology, including Azure OpenAI and Copilot, to optimize its manufacturing processes by improving data-driven product development and streamlining tasks. AI assists in precise coding for data management, automates error handling, and enhances the efficiency of processes like fermentation and product innovation, helping Planted increase productivity and scale its operations sustainably.
🐶 Companion animals
Could AI save your pet's life? (BBC News)
At the Royston Veterinary Centre, AI tools like Vetscan Imagyst help analyze samples such as blood, urine, and tissue in a matter of hours, enabling quicker treatment decisions and reducing the need for invasive procedures. While AI is proving to be a valuable tool for improving animal care, it must be used responsibly, with veterinary surgeons retaining the final say in diagnosis and treatment.
Can AI Raise a Cat? (Kinship)
In the Cat Royale project, AI-powered robotic arms interacted with cats by analyzing their responses to various activities, such as playing with toy birds, dangling feathers, or offering treats. The AI learned which activities each cat preferred, adjusting its interactions to prioritize those games, but human involvement was still crucial for monitoring the cats' welfare and troubleshooting any issues with the robotic system.
👽 …and more
New Virtual Reality Experience Puts Players at the Mercy of AI-Powered Aliens (PETA)
PETA's new virtual reality experience, When They Came for Us, uses OpenAI technology to create an immersive scenario where users must convince AI-powered aliens to set them free, simulating the experience of animals trapped in exploitative industries.
📨 That’s it for this edition - as always, please feel free to get in touch at contact@aiforanimals.org with any ideas and feedback!