A group of researchers is investigating the potential of artificial intelligence (AI) to interpret the facial expressions of animals, aiming to gain insights into their emotional states. Recent studies indicate that various research teams are employing AI models combined with computer vision to evaluate emotions, as well as detect indications of pain and distress in livestock. This application represents a specialized use of the technology, which, given sufficient high-quality data, could allow a large language model to assess in real time whether an animal requires help.
Researchers Using AI to Decipher Animal Emotions
A report from Science.org reveals that multiple research initiatives are exploring the reliability of AI in monitoring the welfare of farm animals. One noteworthy project is the Intellipig system, developed collaboratively by researchers at the University of the West of England Bristol (UWE) and Scotland’s Rural College (SRUC). Detailed information about the project is available on a web page on the university’s site, currently in its beta testing phase.
According to the latest updates, the Intellipig system has been deployed on farms where AI monitors hundreds of pigs. Each morning, a photograph is taken of every pig before feeding, and through AI identification, tailored food is provided for each animal.
Moreover, the AI system conducts an independent analysis of facial data to identify any signs of discomfort or distress. If any issues are detected, the AI notifies the farmer, who can then assess whether intervention is necessary. Researchers have also indicated that the technology has the potential to evaluate the emotions of the animals over time.
The practice of interpreting animals’ health from their facial expressions and behaviors is not groundbreaking, as a considerable body of literature exists on the topic. Professionals trained in animal behavior have utilized this knowledge to discern when animals may be unwell or emotionally distressed.
However, the report highlights that relying on human assessment is not feasible over the long term, as it typically takes an average of 100 seconds for trained individuals to identify various facial muscle movements. In contrast, AI can accomplish this task nearly instantaneously and on a much larger scale. Researchers are not advocating for the development of a fully autonomous AI system at this stage, recognizing that the technology is still in its early phases.
