Farm animals are talking and we are listening
Old McDonald should be listening to what his overcrowded, over-medicated, and overlooked animals say...
Meanwhile, a team from Queen Mary University in London and the University of Roehampton are, thanks to a deep learning AI that can tune in to what farm animals are saying. It can tell whether cows and pigs are happy and well-fed, or sick and distressed - all from their moos and oinks.
The Queen Mary team hope that farms across the world will use the tech to automatically adjust settings like air ventilation, temperature, and lighting based on the collective calls of the animals. “There’s loads of information we can be gathering from them,” says lead author Michael Mcloughlin.
The work comes out of the growing field of bioacoustics: the study of how animals produce, transmit, and receive sounds. Pigs, chickens, and cattle are some of the chattiest farm animals, and perfect for recording the ways their calls change depending on the situation.
To analyse the calls, the researchers had to turn audio signals into complex numerical variables based on measurements of amplitude and frequency. The machine learning algorithm processes these. “You speak into your phone and it immediately understands what you’re saying. These same types of techniques can be used in machine learning,” Mcloughlin said.
The AI can theoretically infer population size, as well as the location, size, weight, and health of each animal. Another of Mcloughlin’s goals is to differentiate between the emotional states of large numbers of animals at once. “We don’t just talk about things like disease in animals, we also talk about their emotions.”
Widespread implementation of this is tricky though, not only for the software developers, but also because the definition of “good” animal welfare is up for interpretation. “What one person regards as being good animal well-being, another person may very well be like ‘no it’s not good enough’, you know,” Mcloughlin said.
The threshold for well-developed welfare may not be well defined, but there are definitely issues associated with factory farms. The enclosed and cramped shelters that hold these animals create a closely-knit circle for Salmonella, E.coli, and other bacteria to circulate through. A commonly-adopted practice by farmers in some countries to avoid animals becoming unwell is to feed them large doses of antibiotics. But this is fuelling a global antibiotic resistance crisis. Unhealthy animals also create unhealthy meat, which can transmit diseases to the consumers and also certainly tastes less appetising.
In future, the team hope to spread awareness about the potential of “automated bioacoustics”. One of the present challenges is the lack of an open source online database of animal sounds. To create a database that correctly labels recordings for deep learning software to use, requires expertise and specialists that understand the process.
The road to automated farming is being laid, but the team will have to spend the upcoming years analysing more audio to properly define the “good” wellbeing animals deserve to have in all of Old McDonald’s farms...