It sounds crazy, but if recent research turns out to be legitimate, we might soon be able to talk to dolphins via a two-way dolphin-whistle-to-English translator.
Researchers have been trying to decode and understand the various clicks and whistles made by dolphins since the 1960s, but mid-way through last year, after decades of experimentation, they made a huge breakthrough. For the first time ever, scientists were able to instantly translate a dolphin whistle into its corresponding English word.
It happened back in August of 2013, when Denise Herzing, founder of the Wild Dolphin Project in Jupiter, Fla., was swimming in the Caribbean and listening to the dolphin pod she had been tracking for the past 25 years. Using a special listening/translation system called CHAT, Herzing suddenly heard the word “sargassum” (a genus of seaweed) come through her headset.
The specific whistle for “sargassum” was actually a word that Herzing and her team had invented in dolphin-speak. Since the late ’90s, they’ve been using artificial sounds that mimic real dolphin noises, and teaching them to the pod in hopes that the dolphins would eventually adopt them and incorporate them into their own communications. When the whistle came out, it was picked up by a pair of hydrophones, recognized/translated instantaneously with CHAT ( short for Cetacean Hearing and Telemetry), and then spoken back to her in English.
In addition to listening for these kinds of invented whistles, Herzing and her team hope to figure out how to translate the natural communications of dolphins as well. The CHAT system (developed by Georgia Tech professor and Google Glass project lead Thad Starner) is designed with a pair of finely-tuned hydrophones that can pick up the full range of dolphin sounds — many of which are imperceptible to human ears. Once recorded, the software sifts through all the different whistles, and uses pattern discovery algorithms to pinpoint language features. It starts by labeling noises that deviate from an assumed average state, and then groups ones that are similar to one another –sets of clicks or whistles with a distinct sonic signature– until all potentially meaningful patterns are extracted.
The secret to success here is repetition. Over time, if dolphins are exchanging information using these noises, then their behavior wouldn’t be completely random. There’s likely to be some discoverable patterns that could be recorded, codified, and eventually translated. With the help of today’s sophisticated information processing tools, figuring out those patterns is easier than ever. Starner’s algorithms have already discovered eight distinct components from a sample of 73 whistles, and they’ve begun to match certain parts of those whistles to specific dolphin-to-dolphin interactions. The research is still coming along, but it’s extremely promising, and could very well yield a working two-way dolphin-to-human translator in the next few years.
- A.I. perfectly predicted last year’s Super Bowl score. What happens to betting?
- Deep learning vs. machine learning: what's the difference between the two?
- Text away: 20 free texting games you can play with your friends
- Brain scans and A.I. confirm that dogs are great at recognizing our emotions
- Replaced by robots: 8 jobs that could be hit hard by the A.I. revolution