AI-driven natural language processing is getting better and better at understanding languages spoken in every country around the world. It doesn’t have to stop at land-based languages, however — as Swedish-based language startup Gavagai AB is keen to prove.
Working with researchers from the KTH Royal Institute of Technology, Gavagai AB is currently involved in a four-year project, designed to gather as much “dolphin language data” as possible, and to use this to decode the chatter of everyone’s favorite aquatic mammals.
In theory, learning what a dolphin is saying is no different to understanding other languages. Research has suggested that dolphins communicate in a language not entirely dissimilar to our own, relying on sentences composed of individual words in which the order dictates meaning. They even pause to let one another speak. Decoding this conversation relies on the same ability to draw links between sounds and meaning as would be the case with any other language — only with what one imagines includes a higher propensity to talk about tasty types of fish.
Such a system could even conceivably be used to reverse-engineer software allowing humans to communicate with dolphins by translating messages into the appropriate lines of Flipper chat. How far this will get by 2021, when the project ends, remains to be seen, but this promises to be of interest to both AI practitioners and zoologists alike.
Gavagai AB CEO Lars Hamberg stressed to Digital Trends that there are no obvious commercial applications for dolphin translation, although it’s certainly interesting to think of possible use-cases. The U.S. Navy, for instance, has previously used military dolphins for tasks including rescuing lost swimmers and locating underwater mines. This training could presumably be made easier, and other tasks added, if a more direct line of communication between man and dolphin is made possible.
But Hamberg stressed that understanding dolphins is far from the only thing Gavagai AB is working on. Right now, he is in Silicon Valley speaking with investors about raising additional capital to continue developing the Gavagai AB technology.
“The dolphin research has gone haywire in terms of publicity, but the truth is that it’s one of many, many projects that we’re doing,” Hamberg told us. “Over the last seven years, we’ve spent about $9 million working on implementing this research into an industrial scale implementation that can understand 45 different languages. On a lot of benchmark measures, we outperform all of our rivals. We’re now gaining a lot of interest, having toiled away under the radar for a long time.”
In other words, click click whistle click whistle. (That’s dolphin speak for “keep watching this space.”)
- Facebook is using Minecraft to train AI assistants, one block at a time
- How messaging app Line thrives on a culture of chaos, not cuteness
- Neuro-symbolic A.I. is the future of artificial intelligence. Here’s how it works
- Google Translate now offers improved translations in offline mode
- Who is Hikari-chan? She is the mind-blowing future of A.I. in your home