This New AI Translates Dog Barks Into English
The idea of some magical device that translates our pet's communications into human speech is not new—only up until now, it's been the realm of fiction. Remember the baby translating device that drove a wedge between Homer Simpson and his estranged brother, Herb? Remember the device worn by Doug the dog on his collar in the Pixar film Up?
Now it seems that this technology is almost here. Dr. Con Slobodchikoff, a professor at Northern Arizona University who has spent more than 30 years studying how animals communicate, is now working on an instrument that will use AI to learn and translate animal noises and facial expressions into simple English.
First noting that prairie dogs warn others of predators using high-pitched wails, Slobodchikoff then mapped the complex series of patterns that prairie dog calls take on. Different combinations of calls can indicate incredibly nuanced things, he found, including colors and appearances.
"I thought, if we can do this with prairie dogs, we can certainly do it with dogs and cats," Slobodchikoff told NBC news. Hence, he's working on an algorithm that turns the noises made by dogs and cats into English. Last year, he founded a company called Zoolingua that is solely focused on developing an instrument to translate pet sounds, facial expressions, and body movements. Their website, active as of October 2017, has yet to publish any additional reports or findings.
"In human language terms, the species of predator can be considered a noun, and the descriptions can be considered adjectives or adverbs," Dr. Slobodchikoff told Psychology Today. "Also, the rate at which the chirps were produced correlated with the speed of travel of the predator, so that was analogous to a verb in human language. In effect, a single alarm call was comparable to a human sentence."
"While most people are amazed at prairie dogs having descriptive terms for size, shape, color, and type of predator, to me the most striking discovery was finding that all of these calls were composed of phonemes, just like human words are composed of phonemes," he went on to explain.
"A phoneme in human language is the smallest unit of sound, and phonemes are assembled into morphemes, or the smallest units of meaning, which are then assembled into words. We found the exact same thing in prairie dog alarm calls: each call was composed of the same phonemes, but they were assembled in different orders, just like the phonemes in human words. The journal where we published this did not like our use of the term "phonemes" so we had to change it to 'acoustic structures.' "
We've got to be careful with AI when language is concerned though. Last July, Facebook shut down its AI initiative after it began designing its own language.
More in the realm of Dr. Slobochikoff's research, Facebook is also training AI to read human expressions. If it sounds like this isn't too far off from Apple's facial recognition software, remember, that's a voluntary feature that users can choose to opt out of. As a tool for understanding animals without the ability to verbally communicate, we're all for the myriad of ways that AI can deepen our understanding of how they live and behave. Turning that tech onto humans, or worse yet, weaponizing it, remains a terrifying prospect.