Move Over Dr. Dolittle, Now AI Can Talk to Animals
How We’re Learning Animal Communication Through Artificial Intelligence
Did you know that 55 percent of animal owners talk with their pets? While these are typically one-sided conversations, punctuated with pets tilting their heads in confusion, a future wherein we can effectively communicate with animals is now within reach.
For decades, since we adopted animals as pets and even before, the human species has tried to understand animal language and communication. Researchers have even gone as far as teaching sign language to monkeys. The Earth Species Project (ESP) team, however, is determined to take advanced animal communication beyond that of our primate cousins with the help of artificial intelligence.
Earth Species Project
Earth Species Project was established in 2018 by founding Twitter member Britt Selvitelle and Mozilla Labs’ Aza Raskin. After securing multi-year funding in 2021, ESP has grown its AI and animal specialists team and is diligently working to bridge the gap between animal and human communication.
On their website, the ESP team describes themselves as:
“We are thinkers, makers, artificial intelligence engineers, conservation technologists, learners, entrepreneurs and artists who care deeply about the interdependent future of our planet and its species.”
What Tech Does AI Use to Understand Animals?
In order to create an AI model capable of translating animal thoughts, behaviors, and noises into something understandable by humans, researchers have employed the use of several complex technologies and scientific disciplines.
Computer Vision: Computer vision is a field of study that focuses on enabling computers to understand and interpret visual information from images or videos, mimicking human visual perception. It involves tasks like object detection, image classification, and image segmentation, with applications ranging from autonomous vehicles to facial recognition systems.
Bioacoustics: Bioacoustics is the scientific study of sounds produced by living organisms, including animals, insects, and even humans. It explores the generation, transmission, and reception of these sounds and their role in communication, behavior, and ecological interactions. Bioacoustics research often involves analyzing and interpreting acoustic signals to gain insights into various biological processes and ecosystems.
Geometric Language Representation: Geometric language representation is a way of understanding a language using geometric shapes and structures. It involves mapping words or sentences onto points or areas in a geometric space. This approach helps analyze and manipulate language using mathematical operations, making it easier to process and interpret text for tasks like translation, sentiment analysis, and finding relevant documents.
Large Language Processing: Large language processing refers to the computational techniques and models used to process and understand vast amounts of textual data. It involves leveraging advanced natural language processing (NLP) algorithms and deep learning models to extract meaningful information, perform text classification, sentiment analysis, question answering, and language generation tasks. Large language processing is essential for various applications, including chatbots, information retrieval systems, and language translation services.
What would you say to your pet?
Step-by-Step Communication Process
The above technology is a generalization and not specific to Earth Species Project’s method of AI-animal communication. However, they clearly discuss their communication process on their website, which you can view here. Here’s a digestible rundown of their advanced communication process that’s powered by artificial intelligence and machine learning.
Data Collection: As with any AI/ML model, this process starts by collecting vast amounts of data. In the case of ESP, this means monitoring animals through multiple modalities, such as sounds, physical behaviors, and environmental interaction. This data later allows researchers and ML models to build a clear image of how animals communicate with one another and their perception of the world.
Foundation Model Training: A foundation model is an AI system that is trained on large amounts of data and can be trained for self-supervision. Focusing on biological data collection by monitoring animals, EPS’s foundation model can review and track their data sources without human input, making the analysis process much swifter.
Learned Representation: Using a geometric language representation model, the data process by the foundation model can be mapped in an understandable way that’s suited for further machine learning applications.
Decode Representation: ESP’s self-supervised ML models can analyze their data and representations to identify patterns that arise when animals send certain signals, such as a lion’s roar. The ML models then analyze how much these signals impact the individual animal's behavior and those around it. These patterns can then be used to predict animal behavior, helping to build an understanding of what animal signals mean.
Interpret: Using ML algorithms, the decoded data is then interpreted into meaningful information to provide researching data on animal behavior and communication techniques.
Communicate: The final step of the process is communication with animals, which is typically achieved through playback experiments. This involves playing a sound or displaying a particular behavior and seeing if it produces the predicted outcome, per the ML algorithm's determination.
Although this is a complex and long-winded process that will likely take many years to fully refine, it has already proven to be effective. In fact, researchers in Germany were able to communicate with honey bees using a similar process and a robotic bee. For example, they could tell bees the location of nectar by making their robotic bee do a “waggle dance” near the source, a behavior earlier identified by researchers.
Applications of AI Animal Communication
If you grew up watching Looney Toons, Scooby Doo, and Doctor Dolittle, your mind is probably spinning with possibilities. And, although you’re probably not going to be able to take your dog looking for clues (and Scooby snacks) any time soon, there are a slew of useful applications for AI animal communication. Here are just a few.
Understand Animal Needs: If you’re a pet owner, then you understand the frustration when your dog seems to needlessly whine and bark or when your cat attacks you for apparently no reason. AI-powered communication will help to solve the mystery behind your pets’ erratic behavior by allowing you to literally ask them what they want. This can help improve the quality of life for your pets and make your own life easier.
Climate Change: Climate change is happening, and the effects are already being felt in the animal kingdom. For instance, birds sing their morning song at dawn due to the moisture levels in the air setting the perfect conditions to carry their calls, meaning the atmosphere is changing the settings in which they sing and stifling their song. Understanding birdsong and replicating it with AI could help mitigate climate change's effects on birds and other wild animals.
Utility and Service Animals: Service animals require an intimate understanding of their human counterparts and in-depth training. AI animal communication could streamline the training process and allow the animals’ owners to more easily communicate their needs.
Ethical Concerns and Challenges
Being able to talk with animals brings many new opportunities and benefits for humankind, but that’s not to say it’s as advantageous for the rest of the animal kingdom. Here are a few ethical concerns and challenges to consider.
Animal Manipulation: AI animal communication could be used to manipulate animals into doing things they otherwise wouldn’t. For example, criminal organizations may be able to use animals to carry and deliver drugs under clear instructions from humans. Dangerous and exotic animals that fall into the wrong hands may also threaten human life and themselves under the direction of criminals.
Habitat Disturbances: While conducting playback experiments, there’s a risk of disturbing animals during crucial activities like foraging and hunting. In fact, it’s possible that these experiments could influence and change the way certain animals engage in essential activities. As a result, ESP is working closely with biologists to mitigate and avoid any potential risks.
Altering Animal Cultures: There are several animal species that rely on vocal signals, such as whales, dolphins, and birds. Because whale songs are learned and spread across the globe by different groups of whales, for example, communicating a human message could cause a culture change by spreading a man-made song. This could have unknowable knock-on effects within the greater whale community.
While a pessimist might worry about the above risks, which are indeed possible, an optimist might revel in the opportunities AI animal communication will bring. A world in which we can talk with animals is truly impressive and, if done correctly, could lead to prosperity for both humans and animals.
Thanks for reading.
If you enjoyed this article, please subscribe to receive email notifications whenever we post.
AI Business Report is brought to you by Californian development agency, Idea Maker.