Sometimes it feels like we’re inching closer to a future like the movie Her, where humans spend their time wandering around, constantly chatting to their own personal AI assistants. Today, services like Siri, Alexa, and Google Assistant are getting better at responding to some types of requests, but because they can misunderstand questions and aren’t cognizant of their surroundings, they’re often super awkward to use in public spaces around other people. But a patent filed by Apple and made public today suggests Siri may someday detect when you’re whispering at it, and in turn know when to whisper back.
Apple has sought to patent “a digital assistant that is capable of detecting a whispered speech input and providing a whispered speech response.” In the document, Apple provided a few scenarios when a whispering assistant might come in handy, like “while studying in a library where speaking loudly may be prohibited” or “while working at a cubicle with other co-workers surrounding the user.” The application also cites “protecting the user’s privacy” as a reason why Siri might need to better detect and understand a whispering user, and why it’d be necessary for Siri to quietly whisper back.
Note that Apple itself is using the term “whisper.”
The filing describes how a device, like an iPhone or an Apple Watch, might detect the difference between a whisper and regular speech. It would determine whether the speaker was whisper by doing things like measuring an input’s amplitude and frequency patterns. It also describes how an assistant like Siri might adjust its own tone based on the user’s tone, so it could actually whisper back—not just adjust its volume. We’re essentially talking about an assistant that would respond more like a normal human does. If someone quietly asks you, “where’s the restroom?” You don’t shout back, “HELLO, I FOUND TWO RESTROOMS NEARBY!”
Just because Apple has filed a patent for this technology doesn’t guarantee it’ll wind up inside Siri soon, or ever. Apple, like all tech companies, regularly files patents to protect ideas that sometimes never make it into consumer products. In 2009, Apple patented a “sponge-like” iPhone dock. In 2012, Apple got a patent for 5D technology (there are only four known dimensions).
Patent filings, however, do indicate some degree of interest in a given technology, as well as a desire to protect it. It’s one of the few ways secretive companies like Apple disclose unreleased concepts to the public. And despite this filing, it’s possible other companies, like Amazon or Google, could release similar assistants with their own whisper-detection methods.
Asked if this sort of technology could possibly be released any time soon, Alex Rudnicky, a professor at Carnegie Mellon University’s Language Technologies Institute, told Gizmodo that it “could come out any time because the component technologies are pretty straightforward,” Rudnicky said. “The ways that they’re doing this is kind of textbook speech processing,” but “working up an algorithm to decide when to go into whisper mode” would be more complicated. “I can’t really say why nobody’s come up with this solution. But once you hear it, it sounds really reasonable. Why not?”
You can view the full document below.