Artificial intelligence is moving beyond offices, smartphones and business software. Scientists are now using the technology deep inside natural habitats to study how wild animals learn, remember and solve problems.
Researchers from Emory University and the Georgia Institute of Technology have developed an AI-powered system called CapuchinAI. The platform uses computer vision, facial recognition, touchscreen activities and automated food rewards to study the cognitive abilities of wild capuchin monkeys.
The technology could help researchers understand animal intelligence under natural conditions instead of relying mainly on controlled laboratory experiments.
Why Studying Animal Intelligence in the Wild Is Difficult
Scientists have studied primate intelligence for many years. However, most cognitive experiments take place in laboratories, zoos or other controlled environments.
Laboratory testing allows researchers to control distractions, repeat experiments and accurately record results. The disadvantage is that animals are removed from the complex social and environmental conditions that shape their everyday behaviour.
Field research offers a more realistic picture, but conducting structured experiments in a forest is difficult. Animals may move unexpectedly, dominant individuals may take over testing equipment, weather can damage devices and researchers may struggle to identify which animal completed a particular task.
CapuchinAI was created to reduce these challenges by bringing controlled, individualised testing directly into the monkeysâÃÂànatural environment.
How CapuchinAI Works
CapuchinAI is a portable cognitive testing station designed for outdoor use. Its main components include:
- A Raspberry Pi computer
- A high-definition webcam
- A touchscreen
- Machine-learning facial-recognition software
- An automated food dispenser
- A battery-powered operating system
When a monkey approaches the station, the camera captures its face. The AI system determines whether the animal is a capuchin and, once properly trained, identifies the individual monkey.
The touchscreen then displays a task suited to that particular animal. When the monkey completes the activity correctly, the system releases a small food reward, such as a piece of dried forest banana.
The platform can also save the monkeyâÃÂÃÂs progress. When the same animal returns, the system can continue from the stage it previously reached instead of restarting the entire test.
Facial Recognition for Wild Monkeys
Developing the facial-recognition system required researchers to collect and label thousands of images and video frames.
Students from Emory University and Georgia Tech manually marked the monkeysâÃÂàfaces and connected each face to the correct identity. This labelled information was then used to train a computer-vision model based on YOLO technology.
The trained system achieved more than 97% precision and recall when identifying selected capuchin monkeys from still images, recorded videos and live footage.
Unlike human facial-recognition systems, the technology must operate under difficult forest conditions. Lighting changes constantly, animals move quickly, branches may block their faces and several monkeys may appear near the device at the same time.
Wild Capuchins Quickly Learned to Use the Touchscreen
The researchers tested CapuchinAI with two groups of wild white-faced capuchins at the Taboga Forest Reserve in Costa Rica.
During the pilot deployment, 16 monkeys voluntarily interacted with the equipment. Ten learned how to activate the reward system, while eight developed and retained a strong understanding that touching the screen could produce food.
Some monkeys understood the system quickly. Others watched their group members use the touchscreen before attempting it themselves. A few even used their mouths or lips to touch the display.
These differences could provide researchers with valuable information about learning styles, curiosity, social observation and problem-solving among individual animals.
Testing Memory, Self-Control and Mental Flexibility
The touchscreen platform can be adapted to measure several areas of cognition.
Researchers plan to use it for activities involving:
Learning
The system can measure how quickly a monkey understands a new rule or connects an action with a result.
Memory
Touchscreen exercises may test whether an animal can remember images, locations or instructions after a short or extended period.
Impulse Control
A monkey may be required to wait, avoid touching a particular object or select an option only when a specific signal appears.
Cognitive Flexibility
Researchers can change the rules of a task to determine how quickly the animal adjusts its behaviour when a previously correct answer becomes incorrect.
Because the system recognises individual participants, each monkey can receive tasks based on its own previous performance and learning level.
Reducing Interference From Dominant Animals
One common problem in group-based animal research is that stronger or more dominant animals may prevent others from participating.
CapuchinAI can place limits on the number of rewards given to one monkey during a testing session. It can also detect when one participant leaves and another takes its place.
The system processes video continuously, allowing it to change the displayed activity and data record when a different monkey approaches the screen. This makes it possible for several animals to participate during the same research period without mixing up their results.
A Scalable and Lower-Cost Research Tool
The researchers intentionally developed CapuchinAI with relatively affordable and widely available components.
The software and construction guidance provide a foundation that other research teams may adapt for different environments and species. A similar system could eventually be used to study other primates or animals that can interact with touchscreen-based testing equipment.
Researchers could also install several stations across different locations, allowing larger amounts of information to be collected over longer periods.
This may help scientists compare how age, social position, environment, experience and access to resources influence the cognitive development of individual animals.
AI Supports Researchers Instead of Replacing Them
Although CapuchinAI automates identification, task delivery, rewards and data collection, human researchers remain essential.
Scientists still need to observe the monkeysâÃÂàrelationships, behaviour, health, habitat and life histories. The AI-generated results become more useful when combined with years of field observations collected by people who understand the animals and their environment.
The technology is therefore better viewed as an additional research tool rather than a replacement for scientists working in the field.
What This Development Means for the Future of AI
CapuchinAI shows how artificial intelligence can be combined with simple hardware to solve practical problems outside traditional technology industries.
The project brings together machine learning, computer vision, behavioural science, anthropology, psychology and field engineering. It demonstrates that AI systems do not always require large data centres or expensive equipment. Carefully designed solutions can run on small computers and operate in remote environments.
For businesses, universities and innovators, the project offers a broader lesson: the real value of AI often comes from connecting it to a specific problem, reliable data and a well-designed physical or digital process.
From studying wildlife to improving agriculture, healthcare, education and business management, AIâÃÂÃÂs most meaningful applications may be those that help people collect better information and make more informed decisions.
Conclusion
CapuchinAI could change how scientists study animal intelligence by bringing structured cognitive testing into the wild.
By recognising individual monkeys, delivering personalised touchscreen activities and automatically recording results, the platform allows researchers to study learning and memory within the environment where these abilities naturally developed.
The project is still being improved, but its early results show how AI can help researchers examine complex questions that were previously difficult to explore at scale.