AI for Animals #5: Should we try to talk to animals?
November 11, 2024
Welcome to the fifth edition of AI for Animals! This newsletter brings you the latest news and research on AI, animals, and digital minds, as well as a brief overview of a specific topic relating to AI and animals. The last edition outlined how AI could facilitate interspecies communication. In this edition, we explore the risks and opportunities this could bring for other species.
We are also incredibly excited to announce the AI for Animals 2025 Conference! This will take place March 1–2 at the University of California, Berkeley. See the link for tickets, calls for speakers, and more.
We’d love to hear your questions or feedback! Just email us at hello@aiforanimals.org. You can subscribe to this newsletter here.
Thanks to Allison Agnello, Sankalpa Ghose, Constance Li, Aditya S.K., and Santeri Tani for their contributions to this month’s edition!
Max Taylor
Index
⚡️ AI for Animals 2025 Conference ⚡️
Learn and network with leaders using AI to create a better future for all animals.
Join us for a weekend filled with exciting talks, workshops, and networking opportunities focused on how artificial intelligence and other advanced technologies are revolutionizing the way we consider and care for all sentient beings. Topics will include the future of animal agriculture, interspecies communication, sentience research, increasing animal consideration in frontier AI models, and more.
This will be a multidisciplinary conference, including a wide range of stakeholders from animal welfare scientists to AI researchers to tech developers. Sessions will cover practical applications of AI and other advanced technologies for animals and focus on furthering this emerging field through an opportunities fair, networking sessions, and workshops to develop stakeholder maps and intervention repositories.
Get your early bird tickets now!
Should we try to talk to animals?
Our last edition explored how AI might enable humans to communicate with other species (which Earth Species Project’s recent $17 million grant might well accelerate). In this edition, we ask: should we even try? To answer this question, consider the following four scenarios.
Scenario 1: Meaningful human-animal communication proves impossible.
This scenario could come about for a variety of reasons. Maybe the technology just never gets good enough. Maybe other animals’ communication is just too different from our own. Maybe we manage to understand what animals are saying and provide meaningful responses of our own, but animals fail to recognize humans (or disembodied speakers) as something worth communicating with.
Various scientists and philosophers refute the possibility of human-animal communication in principle. Many argue that animals lack the cognitive complexity, linguistic structures, and grasp of abstract concepts necessary for true language, rejecting or dismissing claims of evolutionary continuity between animal and human communication. There could well be some validity to these objections. At the same time, as illustrated in Karen Bakker’s The Sounds of Life, there has been a long history of researchers confidently asserting that certain species lack complex communication systems and various other abilities only to be proven wrong. True, we might prove unable to ‘translate’ animal communication into something auditory and linear like human language, but maybe there are other options. For example, Aza Raskin, co-founder of the Earth Species Project, suggests such translations could seek to communicate a more multisensory experience of their communication, made up of flashes of sound and color.
Scenario 2: Meaningful human-animal communication proves possible, and we use it to harm other species.
One of the first records of humans becoming aware of whale song was the 19th century sailor Captain Kelley, who had an uncanny ability to locate bowhead whales based on their vocalizations. Kelley used this skill to track the whales and harpoon them. Sadly, it seems likely that many modern humans would apply an improved understanding of animal communication in a similar way.
Already, humans use all kinds of baits, lures, and other forms of deception to make animals easier to hunt. Beyond killing them, we also use them to surprising and sinister ends. We have used bats as incendiary devices, bees to detect mines, and dolphins in military efforts. What if human-animal communication made such deception and exploitation radically more effective and widespread, including making it much easier to manipulate the billions of animals we confine and slaughter for consumption each year?
Such harms need not be intentional. Aza Raskin raises the possibility of inadvertently starting a “Whale QAnon” by attempting to speak to whales before we fully understand what we’re saying, potentially communicating stark warnings when we were only trying for an introductory greeting. Even with the best of intentions, such clumsy communication could confuse animals, disrupt their mating habits, or divert their migration patterns in detrimental ways. Or more mundanely, interspecies communication could give rise to countless pseudo-ecotourism cash-grabs, satisfying tourists’ thirst for novelty while disrupting the animals and trashing their habitats. (This also illustrates the critical need for robust licensing of any such technology.)
Scenario 3: Meaningful human-animal communication proves possible, leading us to become less speciesist as a society and better able to cater to animals’ interests.
In the most utopian version of this scenario, interspecies communication finally leads humans to realize that animals are sentient individuals with their own interests and motivations. This entails a radical upheaval of society. Most people are no longer comfortable with the idea of animals being confined and killed for their consumption. The factory farming industry is gradually dismantled; in the meantime, veterinarians are much better able to understand and cater to animals’ needs. Governments use our new communication skills to alert animals to threats such as roads, railways, skyscrapers, and shipping routes. AI researchers can much better discern animals’ values and interests, making it more feasible (albeit still highly complex) to incorporate these into AI alignment efforts. Since researchers are much better able to estimate the capacity of pleasure and suffering among different species, animal welfare interventions become more tractable and animal advocacy funding is allocated to where it is most needed. Now that certain species are able to give or withhold assent, humans decide that we have a duty to represent them in our legal and political systems, with seismic consequences.
Scenario 4: Meaningful human-animal communication proves possible, but nothing really changes.
In this scenario, we manage to successfully communicate with one or more non-human species, but there’s no fundamental change to our relationship with other species. Maybe people see it as a hoax or a cheap gimmick. Maybe we manage to communicate with a handful of species, like whales and dolphins, and our moral circle widens just enough to fit them inside, but we continue to exploit and neglect all others where communication proved impossible. Maybe everything significant the animals say is already obvious, and we ignore them just as we currently ignore their self-evident interests. (We don’t need AI to tell us that the average mother pig wouldn’t choose to spend months stuffed in a gestation crate, or that chickens would rather not grow so abnormally large that they collapse under their own weight.)
We can’t just assume that successful interspecies communication will automatically lead to the world laid out in Scenario 3. Humans are evidently hopeless at extending admiration for one species to others, and at actually translating that admiration into positive action. Much conversation around human-animal communication focuses on what this would mean for us: broadening our horizons, enhancing our understanding of the world, deepening our appreciation of nature. That’s all great, but we can’t let this lapse into self-indulgent wonder. Similarly, we don’t want this to be a solely conservation-based exercise: we have to consider not only what this means for species and ecosystems, but what this means for individuals. If most of us end up marveling at media coverage of talking crows, while behind the scenes, Tyson and JBS are using the same bird communication devices to more closely control the movement and behaviors of the billions of chickens in their supply chains, interspecies communication efforts will have been a failure.
We should also avoid fixating on human-animal ‘translation’. We should think about all the ways we could use AI to understand animals better, and avoid fixating solely on translating their vocalizations into human language. Most of the benefits of this work could come from just improving our insights into animals’ lives, behaviors, motivations, and distinct personalities – without speaking back.
In this light, Project CETI’s recent partnership with New York University’s ‘More Than Human Life’ (MOTH) Project seems particularly exciting. This partnership aims to ‘tackle the legal and ethical implications of advances in animal communications, as well as take advantage of the field of new opportunities to advance protections and legal claims for non-humans’. It highlights the risks of such work and the critical importance of robust standards, guardrails, and benchmarks to ensure that this work ends up benefiting animals, rather than further harming and exploiting them. This is the kind of thoughtful consideration we need if we’re to ensure that this work ends up being to the benefit of our interspecies interlocutors, rather than just our own fleeting entertainment.
Next month, we’ll explore broader applications of AI to help animals living in the wild.
📚 Resources
For more information on AI and human-animal communication, check out:
The work of the Earth Species Project, Project CETI, and Interspecies Internet
Toward understanding the communication in sperm whales (Andreas et al., iScience)
Cetacean conversation: AI could let us talk to whales. Experts question if that's a good idea (Salon)
ESP Technical Roadmap (Earth Species Project)
The Hive Community Slack (sign up to join) 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 AI for Animal 2025 Conference is actively inviting participation:
S-Risk Fundamentals is a new online course that will fully launch in early 2025. Express your interest in joining the 12-week fellowship, which includes project-based work and career mentorship.
🚨 Updates
The Large Language Model (LLM) Benchmarking channel in Hive Slack has been reactivated – contribute by joining the channel. If you’re not already a member of the Hive Slack, you can join here.
The Earth Species Project (ESP) has received $17 million in grants to advance AI-driven animal communication research, aiming to create multimodal foundation models capable of decoding diverse species' languages and supporting conservation efforts.
🗞️ News & Research
🗣️ Understanding animals
How Scientists Started to Decode Birdsong (The New Yorker)
Recent research reveals a remarkable level of nuance in bird communication, particularly among socially complex species like greylag geese and fairy wrens, which use specific calls and even “family dialects” to convey detailed social and environmental information. Advances in AI-powered spectrogram analysis are further aiding researchers in decoding these vocalizations, suggesting that bird calls might serve functions comparable to human language in their social complexity and context-specific adaptability.
How bioacoustics and AI can help study animal populations in the forest and beneath the waves (Phys.org)
Researchers at the University of Copenhagen are combining AI with bioacoustics to monitor wildlife by identifying species-specific sounds, enabling efficient biodiversity assessments without habitat disturbance. This approach has shown high accuracy in complex environments, such as dense forests and underwater, and is being applied to endangered species like the Hawaiian monk seal and various bird and frog species, signaling a transformative tool for conservation.
Proceedings of the 4th International Workshop on Vocal Interactivity in-and-between Humans, Animals and Robots (VIHAR 2024)
Abstracts from around 20 papers presented at the Vocal Interactivity in-and-between Humans, Animals and Robots 2024 conference.
🐔 Chicken farming
Funding boost for UK projects to improve poultry welfare (Poultry News)
The UK government is granting £4.5M to researchers exploring AI’s applications in poultry production. The FeedFlow and NetFLOX360 partnership projects seek to use AI models to alert farmers instantly when they spot concerning changes in flock behavior or health metrics, rather than waiting for routine inspections to discover issues.
Strengthening poultry research collaboration between Britain and China (Poultry World)
The Pirbright Institute and Chinese research institutions have signed a five-year agreement to collaborate on avian disease research, emphasizing AI, genomics, and single-cell biology to improve poultry health. This partnership builds on previous joint efforts, aiming to enhance disease prevention and control across key poultry-producing regions in both countries.
£2.4m investment for Farmlay Eggs (Poultry News)
Farmlay Eggs is investing £2.4m in new egg grading equipment that uses AI-based computer vision technology to analyze high-resolution images of eggs. The company hopes this will allow it to nearly double its egg grading productivity compared to its current equipment.
🐮 Cow farming
AI-powered £1.2m project aims to transform UK beef farming (Farmers’ Guide)
The £1.2 million BeefTwin project in the UK aims to address environmental, economic, and welfare challenges in beef farming using AI-powered Digital Twins, which virtually represent each cow to track feed efficiency, emissions, and health. By integrating real-time data, simulation, and machine learning, the project seeks to improve farming practices for lower emissions, enhanced animal welfare, and greater profitability, offering a sustainable model for the future of beef farming.
Drones, AI making cattle counting a dream (Rural News)
PGG Wrightson’s new SkyCount system uses drones and AI to perform livestock counts apparently with 99.9% accuracy, minimizing labor and avoiding disruptions to farm routines. Developed with Microsoft Azure and Inde, this tool is now available across New Zealand.
Artificial intelligence helps Aussie farmers target weeds, livestock illnesses and pests (ABC News)
In livestock farming, AI enhances animal welfare and operational efficiency by monitoring animal health and behavior, often using sensors and cameras to detect issues like illness or impending calving before human detection is possible. Advanced robotic milking systems, such as Lely's Zeta and Melbourne University’s three-robot dairy, allow cows autonomy in milking frequency, optimizing health outcomes and managing feed with precision.
🐷 Pig farming
AI decodes oinks and grunts to keep pigs happy (The Pig Site)
European researchers have developed an AI algorithm that interprets pig vocalizations, identifying emotional states based on sounds to improve animal welfare on farms. This tool could alert farmers to negative emotions in pigs and may eventually enable consumers to make welfare-informed choices through farm labeling, with positive and negative emotions indicated by distinct vocal patterns like short grunts or high-pitched squeals.
$1M NIFA grant to transform pig husbandry with AI (National Hog Farmer)
A research team led by Michigan State University has received a $1 million grant to develop AI-driven solutions to monitor piglet and sow behavior, aiming to reduce piglet preweaning mortality. The project uses computer vision to track piglet nursing patterns and teat function, providing data to improve husbandry practices and support research in areas like animal welfare and nutrition. The findings will help standardize practices and enhance piglet survival rates, with the potential for broader application across the pork industry.
Proactive Pig Project to use AI data to help manage barn conditions (National Hog Farmer)
The Proactive Pig Project (P3), funded by a $1B NIFA grant and led by NC State's Wang-Li, aims to use AI to revolutionize barn environment management by analyzing individual pig biometric data. The project brings together 11 researchers across three universities to develop an offline AI system that will proactively adjust barn conditions based on real-time pig stress levels and individual needs, potentially addressing the $900M in annual losses the US swine industry faces from suboptimal growth conditions.
🐟 Aquaculture and fishing
Power of AI could boost seafood industry's reputation (Seafood Source)
AI in the seafood industry is being used to monitor fish health by detecting diseases in aquaculture pens, track environmental conditions, and optimize feed usage. AI-driven systems also analyze trawling data in real time to reduce bycatch by releasing non-target species from nets. This technology is also applied in supply chain tracking to verify sustainability claims, all of which the seafood industry hopes will boost its reputation.
Can AI transform how shrimp farms count their stock? (Seafood Alliance)
AI technology is enhancing the accuracy of shrimp population counts, enabling farms to maintain optimal stocking densities and improve overall productivity while reducing labor costs and feeding errors. By using advanced sensors and real-time data analysis, companies like SincereAqua are addressing critical challenges in shrimp farming, resulting in healthier shrimp and more efficient operations.
How AI is changing commercial fishing and aquaculture (The National Fisherman)
AI is transforming commercial fishing and aquaculture, with applications ranging from automated feeding and sorting in fish farms to monitoring sea lice in salmon and tracking unregistered fishing vessels. In aquaculture, AI enhances feed efficiency and health monitoring, while regulators like NOAA are exploring AI for automated fish surveys and video analysis to improve monitoring accuracy and reduce human effort in resource-intensive tasks.
Strategic partnership announced for Farm in a Box and ReelData AI (The Fish Site)
Farm in a Box and ReelData AI have partnered to develop modular, land-based aquaculture systems that integrate ReelData's AI to optimize feed efficiency, growth rates, and biomass estimation for precision fish farming. This collaboration combines Farm in a Box's scalable, pre-fabricated systems with advanced AI, aiming to make land-based aquaculture more reliable, efficient, and accessible for a global market.
AI transforms fisheries with stocking, processing, and profit boosts (National Fisherman)
AI is increasingly being used in fisheries to improve efficiency and product quality, from stock assessments and bycatch monitoring to onboard processing systems. Companies like Carsoe are incorporating AI and machine learning to optimize processing steps and standardize product quality across vessels, while firms like ThisFish are using AI for supply chain traceability and quality control.
The marine tech startup taking a dive into aquaculture (The Fish Site)
MarineSitu develops rugged, low-cost marine monitoring systems using advanced cameras, sonar, and machine learning to improve biomass estimation and health monitoring in aquaculture. Their technology is designed for harsh marine environments, and they are expanding into aquaculture with support from government grants and investment from Hatch Blue's accelerator program.
Introducing Seremoni: A New AI-Powered Standard In High-Quality Seafood (Forbes)
Seremoni automates the humane Japanese ike jime fish dispatch method using AI-powered robotics, enhancing fish quality and shelf life. By offering a more sustainable and humane fishing process, Seremoni aims to disrupt the American seafood market, starting with black cod and black sea bass, while also improving profitability for fishermen.
Fish and microchips: how AI can help Europe's fishers and protect the seas (EuroNews)
An EU-funded project called EveryFish is using AI on fishing vessels to automate catch logging, identifying fish species, size, and weight to help fishers meet regulatory requirements and protect fish stocks. The technology includes onboard catch scanners and surveillance systems that aim to reduce overfishing and improve sustainable practices across European fisheries.
🐑 Animal farming: General
Fortifi Using Artificial Intelligence to Streamline Meat Processing (Powder & Bulk Solids)
Fortifi’s partnership with Völur uses AI to analyze data from automated meat processing equipment in real time. The AI optimizes cutting and processing by adjusting decisions based on inventory levels and the specific characteristics of each carcass. This ensures that each cut maximizes meat yield, reduces waste, and supposedly lowers the environmental impact by improving efficiency and minimizing CO2 emissions during production.
Polish Proteine Resources Secures €1.4M for Its AI-Powered Insect Farming (The Recursive)
Proteine Resources uses AI for automated monitoring of insect farming conditions, ensuring optimal environments for larvae growth. The AI-driven system enhances operational efficiency by adjusting variables such as temperature, humidity, and feeding schedules, enabling sustainable, scalable production of high-protein insect-based products for animal feed and human consumption.
The Impact of Media Literacy for Meat Production (IPS)
The rise of generative AI has exacerbated misinformation regarding meat production, leading many to underestimate its environmental impact, with 74% of Americans believing it poses little threat. Analysis of social media reveals that a significant portion of users spread false narratives, with 78% disparaging plant-based alternatives and disputing scientific evidence on the detrimental effects of animal agriculture.
🦘Wild animals
AI turns human-wildlife conflict into ‘human-wildlife coexistence’ in GB, Pakistan (Dawn)
In Gilgit-Baltistan, Pakistan, AI-based trail cameras installed by WWF have reduced human-wildlife conflicts by detecting snow leopards near livestock and alerting communities, lowering retaliatory killings and promoting coexistence. This technology, which provides real-time alerts, has significantly protected livestock, contributed to conservation, and mitigated economic losses for local herders, offering a replicable model for other endangered species facing similar threats.
Human impact on fish habitats detected with AI (Universität Bern)
Researchers at the University of Bern have used ‘explainable’ AI to assess human impacts on freshwater fish habitats in Switzerland, revealing that nearly 90% of these habitats are negatively affected by human activities, such as artificial banks and barriers that hinder fish migration. This study highlights "shadow distributions"—potential habitats degraded by human interference—and aims to guide biodiversity protection efforts by identifying key conservation areas and measures for sustainable habitat restoration.
AI method captures ecotourism photos to monitor remote animal species (Phys.org)
Researchers used an AI-based model to analyze ecotourism photos of Adélie penguins, enabling tracking of colony movements and population size over time. This method offers insights into penguin responses to climate change, aiding ecological monitoring of remote species.
Advancing biodiversity monitoring with a network of automated wildlife cameras (Phys.org)
Researchers have developed an automated wildlife monitoring system using AI-powered cameras in a Dutch nature reserve. The system uses solar-powered cameras to continuously capture images, which are wirelessly transmitted to a database and analyzed by AI models to identify species like rabbits, foxes and deer. This AI-driven approach reduces manual image analysis time and is projected to be 43% cheaper than traditional monitoring methods over a 10-year period.
🍔 Alternative proteins
Three hot start-ups designing and creating sector-challenging proteins (Food Navigator)
Biomatter, MicroHarvest, and Vivici are start-ups using AI and fermentation to develop functional and scalable proteins for food and other industries. Biomatter combines generative AI with physics for enzyme design, MicroHarvest focuses on rapid biomass fermentation for sustainable protein production, and Vivici uses precision fermentation to create whey proteins for diverse food applications.
A protein-prediction AI just won the Nobel Prize. What does that mean for the future of food & our planet? (Noa Weiss, LinkedIn)
AlphaFold, Google’s Nobel Prize-winning AI model, has revolutionized protein structure prediction and is poised to significantly impact food tech by enabling the discovery of plant-based proteins that mimic animal-based functionalities. With AlphaFold 3 now predicting molecular interactions, this technology could be applied to sustainable food production, such as cultivated meat and precision fermentation.
Scientists have built an AI-powered 'electronic tongue' (Live Science)
A newly developed AI-powered "electronic tongue" can assess food safety and freshness by detecting chemical ions, providing a method to determine the quality of beverages and identify harmful substances in water. This innovation enhances decision-making processes in food safety by mimicking human taste perception, thus offering a more accurate and efficient means of evaluating consumables.
Nordic Upcycled Mycelium Startup Enters Japan’s Increasingly Attractive Alternative Protein Market (Green Queen)
Norwegian Mycelium (NoMy) has launched MycoPrime, an AI-driven upcycling service in Japan that uses data science to optimize food waste reduction efforts. The service helps agrifood companies find optimal ways to use their sidestreams, minimize waste and emissions, and develop sustainable fungi-based products through co-located production facilities. By leveraging AI and data analytics, MycoPrime aims to de-risk and scale up sustainable upcycling solutions in Japan's food industry.
🤖 Digital minds
New report: Taking AI Welfare Seriously (Eleos AI)
A new report from Eleos AI and the NYU Center for Mind, Ethics, and Policy argues that some near-future AI systems may possess consciousness or robust agency, raising significant ethical concerns around AI welfare. The authors recommend that AI companies acknowledge AI welfare as a serious issue, assess AI systems for signs of consciousness, and develop policies to handle potential moral obligations toward AI.
AI Consciousness and Public Perceptions: Four Futures (Ines Fernandez et al.)
The potential moral implications of AI consciousness depend on a) whether AIs could achieve consciousness and b) whether society would recognize them as conscious. This paper categorizes future scenarios into four quadrants based on those two questions, with each scenario carrying unique ethical risks such as AI suffering, human disempowerment, geopolitical instability, and moral degradation.
We should prevent the creation of artificial sentience (Richard Parr, EA Forum)
The creation of artificial sentience should be carefully regulated, as leaving it as an unregulated free-for-all could lead to a moral catastrophe and massive suffering. Potential policy options include voluntary codes of conduct, government-mandated regulation, and banning or delaying the creation of artificial sentience or suffering, with the strongest arguments in favor of at least a temporary ban or moratorium.
🦕 …and more
AI gives voice to dead animals in Cambridge exhibition (The Guardian)
Cambridge University’s Museum of Zoology is using AI to animate the narratives of preserved animals, enabling them to “speak” and describe their life experiences, aiming to foster empathy and concern for biodiversity among visitors. This month-long project provides animals with accents and personalities, adapting responses to visitor age and language, and is being studied to assess if personalized animal "voices" can reshape public attitudes toward conservation.
📨 That’s it for this edition - please feel free to get in touch at hello@aiforanimals.org with any ideas and feedback!