Watch your pet for any length of time, and it’s almost impossible not to wonder: “What are they thinking?” One day, thanks to artificial intelligence and machine learning technologies, that might not be so hard to answer. There are numerous ground-breaking attempts to use algorithms to decode animal language and to translate it into human. So one day, our pets might actually be able to talk to us.
“I think it’s the right time, with the right data and with the right expertise, to possibly solve this problem,” says Professor Michael Bronstein. He’s part of an initiative called Project CETI (Cetacean Translation Initiative), which in April announced a five-year initiative using machine learning technologies and state-of-the-art robotics to try to translate sperm whale songs. It’s one of the biggest ever projects attempting to devise an animal translator. And it comes from a group of computer scientists, Harvard fellows and biologists whose previous collaborations have seen them use machine learning to analyse sperm whale songs and identify which whale was speaking with 94% accuracy.
“CETI is a once-in-a-lifetime opportunity,” said Bronstein in a presentation about the project. He added: “And I would say without exaggeration that it is probably the craziest moonshot that I have ever participated in.”
It’s the latest AI animal translation initiative in a field that has been making huge advances in recent years. In 2017, a start-up for a pet translator was launched. by biology professor Con Slobodchikoff, who has previously spent 30 years studying how to talk to dogs (prairie dogs) and used AI to create a dog translator. He also developed an algorithm that decodes dog vocalization and lets them be understood in English.
In 2019, University of Washington scientists created Deepsqueak – a type of AI that can decode the language of mice and identify rat and mice squeaks so that what they’re experiencing during lab trials can be better understood. It follows up research in Egyptian fruit bats which applied an algorithm to their voices and learned to tell the difference between them arguing over food and getting grumpy over who had the best resting spot. It’s even started to hit the market (albeit in rudimentary form) thanks to a cat app from one of the team that worked on Amazon Alexa’s voice recognition software, which tries to let you figure out how to talk to cats by functioning as a cat translator. With the help of AI and machine-learning technologies, you can understand cats’ behaviour and translate their meows instantly through the collar.
These leaps forward towards speaking with animals have partly been driven by the rapid rise in processing power and the rapid rise in algorithm sophistication. Advances in video and audio recording devices also play a part, as – just with humans – in order to understand what’s being said, it’s also important to understand the context in which it’s said. CETI (the Cetacean Translation Initiative) will accomplish this partly through using underwater robotic fish, AI, and drones to monitor sperm whales.
What does the future of this field hold?
A non-profit animal decoding initiative called Earth Species Project claim that by the end of this decade they “expect to achieve significant progress” in translating between humans and animals. Within five years, Project CETI expect to try to speak with animals, or in their words, be: “attempting to communicate with whales in their ‘language’.” The only slight downside? At least initially, it might be something that’s based around them speaking to us than the other way around. According to Gruber: “It’s not about us talking to animals… we want to know what they’re saying.”
Ultimately, some species might have more interesting things to communicate than others. “With the cat app, I’m not sure what they’d have to say,” offers Slobodchikoff. “A lot of times it might just be: ‘you idiot, just feed me and leave me alone’.” For people hoping for AI pet translators, at least that would be a start. In addition to that, AI and machine learning could make things easier for farmers by identifying animal behaviour and recognizing their pain.
All of this is possible with AI language tools and advanced machine learning that enable humans to speak with animals and make a wonderful bond with them. This poses the question, what will AI and machine learning be like in half a century?