By Brian Santo, contributing writer
Science fiction is replete with devices that automatically translate the languages of aliens for us. Some do it almost instantaneously, apparently learning on the fly; we can reasonably infer that such devices might be based on artificial intelligence (AI), complete with machine learning. Lacking aliens to converse with (or so the government wants us to believe), some AI researchers are working on universal translators for animal speech.
We might have a device that lets us talk with the animals within 10 years, according to Veritone, an AI company working on such a thing. Until then, baby steps. And among the first of those baby steps is learning to listen to the animals.
Researchers at the University of Georgia and the Georgia Institute of Technology (which publishes a periodical called — we kid you not — PoultryTech ) spent a couple of years recording chickens under different conditions and then feeding those tapes into a machine-learning system with the intent of producing an AI that chicken farmers can install to monitor the well-being and health of their chickens.
That might sound batty, but chickens are highly vocal, and you can tell when they’re happy and when they’re not. My wife’s chickens tend to be chatty in the mornings. They complain if we forget to let them out of their coop; chastise neighbors’ cats who prowl into the yard; squawk when they’re squabbling; crow when they lay eggs; raise a ruckus when the raccoons eye them too hungrily; and coo like doves when the missus comes into their midst. Each different vocalization is distinct and identifiable. I can tell if I have to jump because there’s a predator in the yard or if I can ignore the noise and keep typing because it’s just Agnes jumping over the fence to chew on the catnip growing near the garage.
The AI from the Georgia colleges apparently can, too. Not write stories about chicken translators. Not yet, anyway. Which is lucky for me, I must say. What it can do is distinguish chicken vocalizations. The AI listens to flocks to see if chickens are talking when they’d ordinarily be expected to, or if they sound happy, or sick, or frightened, or if they’re making the noises they tend to make when suffering some environmental stress, such as excessive heat, for example.
The AI has been evaluated by at least one working chicken operation, Wilcox Farms, an organic egg supplier in the Pacific Northwest. At the moment, the system is still struggling to screen out the sounds of ventilation systems and automatic feeders and other environmental noises, but the researchers are working on that, too, according to Scientific American .
And then it should be a hop, skip, and a jump to developing translators for other animals.
In a recent blog predicting such a thing within 10 years, Veritone called attention to the work of Con Slobodchikoff, professor emeritus at the department of biological sciences at Northern Arizona University, who has been using AI to analyze the vocalizations of prairie dogs. Slobodchikoff says prairie dogs have different words for different predators and for different colors (describing the pelts of other animals and even the clothing of humans).
Given the advances in human language translation achieved by Google with Google Translate and by companies like Lingmo International and Waverly Labs with earpieces for translation, Veritone believes that it’s entirely possible to have a natural language conversation with your dog in the not-too-distant future.
And finally, the allusions we are legally obligated to mention in stories of this nature:
Learn more about Electronic Products Magazine